diff --git "a/test.json" "b/test.json"
new file mode 100644--- /dev/null
+++ "b/test.json"
@@ -0,0 +1,5309 @@
+{"input": "question answering is done by using Material| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "question answering", "output": "hierarchical table dataset", "neg_sample": ["question answering is done by using Material", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "HiTab : A Hierarchical Table Dataset for Question Answering and Natural Language Generation.", "forward": false, "src_ids": "2022.acl-long.78_1"}
+{"input": "hierarchical table dataset is used for Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "hierarchical table dataset", "output": "question answering", "neg_sample": ["hierarchical table dataset is used for Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "HiTab : A Hierarchical Table Dataset for Question Answering and Natural Language Generation.", "forward": true, "src_ids": "2022.acl-long.78_2"}
+{"input": "partially supervised training is used for Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "partially supervised training", "output": "question answering", "neg_sample": ["partially supervised training is used for Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": true, "src_ids": "2022.acl-long.78_3"}
+{"input": "symbolic reasoning is done by using OtherScientificTerm| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "symbolic reasoning", "output": "hierarchy - aware logical form", "neg_sample": ["symbolic reasoning is done by using OtherScientificTerm", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting hierarchical structure , we devise a hierarchy - aware logical form for symbolic reasoning over tables , which shows high effectiveness .", "forward": false, "src_ids": "2022.acl-long.78_4"}
+{"input": "hierarchy - aware logical form is used for Method| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "hierarchy - aware logical form", "output": "symbolic reasoning", "neg_sample": ["hierarchy - aware logical form is used for Method", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting hierarchical structure , we devise a hierarchy - aware logical form for symbolic reasoning over tables , which shows high effectiveness .", "forward": true, "src_ids": "2022.acl-long.78_5"}
+{"input": "partially supervised training is done by using Method| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "partially supervised training", "output": "entity and quantity alignment", "neg_sample": ["partially supervised training is done by using Method", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": false, "src_ids": "2022.acl-long.78_6"}
+{"input": "question answering is done by using Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "question answering", "output": "partially supervised training", "neg_sample": ["question answering is done by using Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": false, "src_ids": "2022.acl-long.78_7"}
+{"input": "conditional generation is done by using Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "conditional generation", "output": "partially supervised training", "neg_sample": ["conditional generation is done by using Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": false, "src_ids": "2022.acl-long.78_8"}
+{"input": "natural language generation is done by using Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "natural language generation", "output": "partially supervised training", "neg_sample": ["natural language generation is done by using Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": false, "src_ids": "2022.acl-long.78_9"}
+{"input": "entity and quantity alignment is used for Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "entity and quantity alignment", "output": "partially supervised training", "neg_sample": ["entity and quantity alignment is used for Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": true, "src_ids": "2022.acl-long.78_10"}
+{"input": "conditional generation is used for Task| context: tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables . hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics .", "entity": "conditional generation", "output": "natural language generation", "neg_sample": ["conditional generation is used for Task", "tables are often created with hierarchies , but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables .", "hierarchical tables challenge numerical reasoning by complex hierarchical indexing , as well as implicit relationships of calculation and semantics ."], "relation": "used for", "id": "2022.acl-long.78", "year": 2022, "rel_sent": "Targeting table reasoning , we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG , and largely reduce spurious predictions in QA and produce better descriptions in NLG .", "forward": true, "src_ids": "2022.acl-long.78_11"}
+{"input": "lifelong pre - training is done by using Method| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "lifelong pre - training", "output": "elle", "neg_sample": ["lifelong pre - training is done by using Method", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "To this end , we propose ELLE , aiming at efficient lifelong pre - training for emerging data .", "forward": false, "src_ids": "2022.findings-acl.220_12"}
+{"input": "emerging data is done by using Task| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "emerging data", "output": "lifelong pre - training", "neg_sample": ["emerging data is done by using Task", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "ELLE : Efficient Lifelong Pre - training for Emerging Data.", "forward": false, "src_ids": "2022.findings-acl.220_13"}
+{"input": "emerging data is done by using Task| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "emerging data", "output": "lifelong pre - training", "neg_sample": ["emerging data is done by using Task", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "To this end , we propose ELLE , aiming at efficient lifelong pre - training for emerging data .", "forward": false, "src_ids": "2022.findings-acl.220_14"}
+{"input": "elle is used for Task| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "elle", "output": "lifelong pre - training", "neg_sample": ["elle is used for Task", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "To this end , we propose ELLE , aiming at efficient lifelong pre - training for emerging data .", "forward": true, "src_ids": "2022.findings-acl.220_15"}
+{"input": "lifelong pre - training is used for Material| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "lifelong pre - training", "output": "emerging data", "neg_sample": ["lifelong pre - training is used for Material", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "ELLE : Efficient Lifelong Pre - training for Emerging Data.", "forward": true, "src_ids": "2022.findings-acl.220_16"}
+{"input": "lifelong pre - training is used for Material| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "lifelong pre - training", "output": "emerging data", "neg_sample": ["lifelong pre - training is used for Material", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "To this end , we propose ELLE , aiming at efficient lifelong pre - training for emerging data .", "forward": true, "src_ids": "2022.findings-acl.220_17"}
+{"input": "knowledge acquisition is done by using Method| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "knowledge acquisition", "output": "function preserved model expansion", "neg_sample": ["knowledge acquisition is done by using Method", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "Specifically , ELLE consists of ( 1 ) function preserved model expansion , which flexibly expands an existing PLM 's width and depth to improve the efficiency of knowledge acquisition ; and ( 2 ) pre - trained domain prompts , which disentangle the versatile knowledge learned during pre - training and stimulate the proper knowledge for downstream tasks .", "forward": false, "src_ids": "2022.findings-acl.220_18"}
+{"input": "function preserved model expansion is used for Task| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "function preserved model expansion", "output": "knowledge acquisition", "neg_sample": ["function preserved model expansion is used for Task", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "Specifically , ELLE consists of ( 1 ) function preserved model expansion , which flexibly expands an existing PLM 's width and depth to improve the efficiency of knowledge acquisition ; and ( 2 ) pre - trained domain prompts , which disentangle the versatile knowledge learned during pre - training and stimulate the proper knowledge for downstream tasks .", "forward": true, "src_ids": "2022.findings-acl.220_19"}
+{"input": "downstream tasks is done by using OtherScientificTerm| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "downstream tasks", "output": "pre - trained domain prompts", "neg_sample": ["downstream tasks is done by using OtherScientificTerm", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "Specifically , ELLE consists of ( 1 ) function preserved model expansion , which flexibly expands an existing PLM 's width and depth to improve the efficiency of knowledge acquisition ; and ( 2 ) pre - trained domain prompts , which disentangle the versatile knowledge learned during pre - training and stimulate the proper knowledge for downstream tasks .", "forward": false, "src_ids": "2022.findings-acl.220_20"}
+{"input": "pre - trained domain prompts is used for Task| context: current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow . this requires plms to integrate the information from all the sources in a lifelong manner . although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive .", "entity": "pre - trained domain prompts", "output": "downstream tasks", "neg_sample": ["pre - trained domain prompts is used for Task", "current pre - trained language models ( plm ) are typically trained with static data , ignoring that in real - world scenarios , streaming data of various sources may continuously grow .", "this requires plms to integrate the information from all the sources in a lifelong manner .", "although this goal could be achieved by exhaustive pre - training on all the existing data , such a process is known to be computationally expensive ."], "relation": "used for", "id": "2022.findings-acl.220", "year": 2022, "rel_sent": "Specifically , ELLE consists of ( 1 ) function preserved model expansion , which flexibly expands an existing PLM 's width and depth to improve the efficiency of knowledge acquisition ; and ( 2 ) pre - trained domain prompts , which disentangle the versatile knowledge learned during pre - training and stimulate the proper knowledge for downstream tasks .", "forward": true, "src_ids": "2022.findings-acl.220_21"}
+{"input": "span representation is done by using Method| context: there are two main paradigms for named entity recognition ( ner ): sequence labelling and span classification . in contrast to sequence labelling , unconstrained span - based methods tend to assign entity labels to overlapping spans , which is generally undesirable , especially for ner tasks without nested entities .", "entity": "span representation", "output": "graph neural networks", "neg_sample": ["span representation is done by using Method", "there are two main paradigms for named entity recognition ( ner ): sequence labelling and span classification .", "in contrast to sequence labelling , unconstrained span - based methods tend to assign entity labels to overlapping spans , which is generally undesirable , especially for ner tasks without nested entities ."], "relation": "used for", "id": "2022.acl-srw.9", "year": 2022, "rel_sent": "Accordingly , we propose GNNer , a framework that uses Graph Neural Networks to enrich the span representation to reduce the number of overlapping spans during prediction .", "forward": false, "src_ids": "2022.acl-srw.9_22"}
+{"input": "graph neural networks is used for Method| context: there are two main paradigms for named entity recognition ( ner ): sequence labelling and span classification . in contrast to sequence labelling , unconstrained span - based methods tend to assign entity labels to overlapping spans , which is generally undesirable , especially for ner tasks without nested entities .", "entity": "graph neural networks", "output": "span representation", "neg_sample": ["graph neural networks is used for Method", "there are two main paradigms for named entity recognition ( ner ): sequence labelling and span classification .", "in contrast to sequence labelling , unconstrained span - based methods tend to assign entity labels to overlapping spans , which is generally undesirable , especially for ner tasks without nested entities ."], "relation": "used for", "id": "2022.acl-srw.9", "year": 2022, "rel_sent": "Accordingly , we propose GNNer , a framework that uses Graph Neural Networks to enrich the span representation to reduce the number of overlapping spans during prediction .", "forward": true, "src_ids": "2022.acl-srw.9_23"}
+{"input": "radiology findings summarization is done by using Method| context: the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians . summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention . with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g. , static pre - defined clinical ontologies or extra background information ) . yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings .", "entity": "radiology findings summarization", "output": "graph enhanced contrastive learning", "neg_sample": ["radiology findings summarization is done by using Method", "the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians .", "summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention .", "with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g.", ", static pre - defined clinical ontologies or extra background information ) .", "yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings ."], "relation": "used for", "id": "2022.acl-long.320", "year": 2022, "rel_sent": "Graph Enhanced Contrastive Learning for Radiology Findings Summarization.", "forward": false, "src_ids": "2022.acl-long.320_24"}
+{"input": "graph enhanced contrastive learning is used for Task| context: the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians . summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention . with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g. , static pre - defined clinical ontologies or extra background information ) . yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings .", "entity": "graph enhanced contrastive learning", "output": "radiology findings summarization", "neg_sample": ["graph enhanced contrastive learning is used for Task", "the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians .", "summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention .", "with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g.", ", static pre - defined clinical ontologies or extra background information ) .", "yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings ."], "relation": "used for", "id": "2022.acl-long.320", "year": 2022, "rel_sent": "Graph Enhanced Contrastive Learning for Radiology Findings Summarization.", "forward": true, "src_ids": "2022.acl-long.320_25"}
+{"input": "graph encoder is used for OtherScientificTerm| context: the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians . summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention . with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g. , static pre - defined clinical ontologies or extra background information ) . yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings .", "entity": "graph encoder", "output": "graph", "neg_sample": ["graph encoder is used for OtherScientificTerm", "the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians .", "summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention .", "with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g.", ", static pre - defined clinical ontologies or extra background information ) .", "yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings ."], "relation": "used for", "id": "2022.acl-long.320", "year": 2022, "rel_sent": "Then , a graph encoder ( e.g. , graph neural networks ( GNNs ) ) is adopted to model relation information in the constructed graph .", "forward": true, "src_ids": "2022.acl-long.320_26"}
+{"input": "relation information is done by using Method| context: the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians . summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention . with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g. , static pre - defined clinical ontologies or extra background information ) . yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings .", "entity": "relation information", "output": "graph encoder", "neg_sample": ["relation information is done by using Method", "the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians .", "summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention .", "with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g.", ", static pre - defined clinical ontologies or extra background information ) .", "yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings ."], "relation": "used for", "id": "2022.acl-long.320", "year": 2022, "rel_sent": "Then , a graph encoder ( e.g. , graph neural networks ( GNNs ) ) is adopted to model relation information in the constructed graph .", "forward": false, "src_ids": "2022.acl-long.320_27"}
+{"input": "graph is done by using Method| context: the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians . summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention . with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g. , static pre - defined clinical ontologies or extra background information ) . yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings .", "entity": "graph", "output": "graph encoder", "neg_sample": ["graph is done by using Method", "the impression section of a radiology report summarizes the most prominent observation from the findings section and is the most important section for radiologists to communicate to physicians .", "summarizing findings is time - consuming and can be prone to error for inexperienced radiologists , and thus automatic impression generation has attracted substantial attention .", "with the encoder - decoder framework , most previous studies explore incorporating extra knowledge ( e.g.", ", static pre - defined clinical ontologies or extra background information ) .", "yet , they encode such knowledge by a separate encoder to treat it as an extra input to their models , which is limited in leveraging their relations with the original findings ."], "relation": "used for", "id": "2022.acl-long.320", "year": 2022, "rel_sent": "Then , a graph encoder ( e.g. , graph neural networks ( GNNs ) ) is adopted to model relation information in the constructed graph .", "forward": false, "src_ids": "2022.acl-long.320_28"}
+{"input": "table - to - text datasets is used for Task| context: the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "table - to - text datasets", "output": "text - to - table", "neg_sample": ["table - to - text datasets is used for Task", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "We consider text - to - table as an inverse problem of the well - studied table - to - text , and make use of four existing table - to - text datasets in our experiments on text - to - table .", "forward": true, "src_ids": "2022.acl-long.180_29"}
+{"input": "vanilla seq2seq model is used for Task| context: in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "entity": "vanilla seq2seq model", "output": "information extraction", "neg_sample": ["vanilla seq2seq model is used for Task", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "We first employ a seq2seq model fine - tuned from a pre - trained language model to perform the task .", "forward": true, "src_ids": "2022.acl-long.180_30"}
+{"input": "information extraction is done by using Method| context: we study a new problem setting of information extraction ( ie ) , referred to as text - to - table . in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data . the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "information extraction", "output": "vanilla seq2seq model", "neg_sample": ["information extraction is done by using Method", "we study a new problem setting of information extraction ( ie ) , referred to as text - to - table .", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "We first employ a seq2seq model fine - tuned from a pre - trained language model to perform the task .", "forward": false, "src_ids": "2022.acl-long.180_31"}
+{"input": "table generation is done by using Method| context: we study a new problem setting of information extraction ( ie ) , referred to as text - to - table . in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data . the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "table generation", "output": "vanilla seq2seq model", "neg_sample": ["table generation is done by using Method", "we study a new problem setting of information extraction ( ie ) , referred to as text - to - table .", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "We also develop a new method within the seq2seq approach , exploiting two additional techniques in table generation : table constraint and table relation embeddings .", "forward": false, "src_ids": "2022.acl-long.180_32"}
+{"input": "vanilla seq2seq model is done by using Method| context: we study a new problem setting of information extraction ( ie ) , referred to as text - to - table . in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data . the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "vanilla seq2seq model", "output": "vanilla seq2seq model", "neg_sample": ["vanilla seq2seq model is done by using Method", "we study a new problem setting of information extraction ( ie ) , referred to as text - to - table .", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "The results also show that our method can further boost the performances of the vanilla seq2seq model .", "forward": false, "src_ids": "2022.acl-long.180_33"}
+{"input": "vanilla seq2seq model is used for Method| context: we study a new problem setting of information extraction ( ie ) , referred to as text - to - table . in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data . the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "vanilla seq2seq model", "output": "vanilla seq2seq model", "neg_sample": ["vanilla seq2seq model is used for Method", "we study a new problem setting of information extraction ( ie ) , referred to as text - to - table .", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "The results also show that our method can further boost the performances of the vanilla seq2seq model .", "forward": true, "src_ids": "2022.acl-long.180_34"}
+{"input": "vanilla seq2seq model is used for Task| context: we study a new problem setting of information extraction ( ie ) , referred to as text - to - table . in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data . the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "vanilla seq2seq model", "output": "table generation", "neg_sample": ["vanilla seq2seq model is used for Task", "we study a new problem setting of information extraction ( ie ) , referred to as text - to - table .", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "We also develop a new method within the seq2seq approach , exploiting two additional techniques in table generation : table constraint and table relation embeddings .", "forward": true, "src_ids": "2022.acl-long.180_35"}
+{"input": "text - to - table is done by using Material| context: we study a new problem setting of information extraction ( ie ) , referred to as text - to - table . in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data . the problem setting differs from those of the existing methods for ie . as far as we know , there has been no previous work that studies the problem .", "entity": "text - to - table", "output": "table - to - text datasets", "neg_sample": ["text - to - table is done by using Material", "we study a new problem setting of information extraction ( ie ) , referred to as text - to - table .", "in text - to - table , given a text , one creates a table or several tables expressing the main content of the text , while the model is learned from text - table pair data .", "the problem setting differs from those of the existing methods for ie .", "as far as we know , there has been no previous work that studies the problem ."], "relation": "used for", "id": "2022.acl-long.180", "year": 2022, "rel_sent": "We consider text - to - table as an inverse problem of the well - studied table - to - text , and make use of four existing table - to - text datasets in our experiments on text - to - table .", "forward": false, "src_ids": "2022.acl-long.180_36"}
+{"input": "ordinary differential equations ( ode ) is done by using OtherScientificTerm| context: residual networks are an euler discretization of solutions to ordinary differential equations ( ode ) .", "entity": "ordinary differential equations ( ode )", "output": "higher - order solution", "neg_sample": ["ordinary differential equations ( ode ) is done by using OtherScientificTerm", "residual networks are an euler discretization of solutions to ordinary differential equations ( ode ) ."], "relation": "used for", "id": "2022.acl-long.571", "year": 2022, "rel_sent": "We first show that a residual block of layers in Transformer can be described as a higher - order solution to ODE .", "forward": false, "src_ids": "2022.acl-long.571_37"}
+{"input": "reproducibility is used for Task| context: while recent progress in the field of ml has been significant , the reproducibility of these cutting - edge results is often lacking , with many submissions lacking the necessary information in order to ensure subsequent reproducibility . despite proposals such as the reproducibility checklist and reproducibility criteria at several major conferences , the reflex for carrying out research with reproducibility in mind is lacking in the broader ml community .", "entity": "reproducibility", "output": "university - level computer science programs", "neg_sample": ["reproducibility is used for Task", "while recent progress in the field of ml has been significant , the reproducibility of these cutting - edge results is often lacking , with many submissions lacking the necessary information in order to ensure subsequent reproducibility .", "despite proposals such as the reproducibility checklist and reproducibility criteria at several major conferences , the reflex for carrying out research with reproducibility in mind is lacking in the broader ml community ."], "relation": "used for", "id": "2022.acl-tutorials.2", "year": 2022, "rel_sent": "We also provide a framework for using reproducibility as a teaching tool in university - level computer science programs .", "forward": true, "src_ids": "2022.acl-tutorials.2_38"}
+{"input": "multi - choice matching networks is used for Task| context: few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities .", "entity": "multi - choice matching networks", "output": "unified low - shot relation extraction", "neg_sample": ["multi - choice matching networks is used for Task", "few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities ."], "relation": "used for", "id": "2022.acl-long.397", "year": 2022, "rel_sent": "In this paper , we propose Multi - Choice Matching Networks to unify low - shot relation extraction .", "forward": true, "src_ids": "2022.acl-long.397_39"}
+{"input": "unified low - shot relation extraction is done by using Method| context: few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities .", "entity": "unified low - shot relation extraction", "output": "multi - choice matching networks", "neg_sample": ["unified low - shot relation extraction is done by using Method", "few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities ."], "relation": "used for", "id": "2022.acl-long.397", "year": 2022, "rel_sent": "In this paper , we propose Multi - Choice Matching Networks to unify low - shot relation extraction .", "forward": false, "src_ids": "2022.acl-long.397_40"}
+{"input": "question answering is done by using OtherScientificTerm| context: deep nlp models have been shown to be brittle to input perturbations . minimally perturbed inputs - can help ameliorate this weakness .", "entity": "question answering", "output": "counterfactuals", "neg_sample": ["question answering is done by using OtherScientificTerm", "deep nlp models have been shown to be brittle to input perturbations .", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "We focus on the task of creating counterfactuals for question answering , which presents unique challenges related to world knowledge , semantic diversity , and answerability .", "forward": false, "src_ids": "2022.acl-long.117_41"}
+{"input": "question generation model is used for OtherScientificTerm| context: deep nlp models have been shown to be brittle to input perturbations . minimally perturbed inputs - can help ameliorate this weakness .", "entity": "question generation model", "output": "counterfactuals", "neg_sample": ["question generation model is used for OtherScientificTerm", "deep nlp models have been shown to be brittle to input perturbations .", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "Using an open - domain QA framework and question generation model trained on original task data , we create counterfactuals that are fluent , semantically diverse , and automatically labeled .", "forward": true, "src_ids": "2022.acl-long.117_42"}
+{"input": "open - domain qa framework is used for OtherScientificTerm| context: deep nlp models have been shown to be brittle to input perturbations . minimally perturbed inputs - can help ameliorate this weakness .", "entity": "open - domain qa framework", "output": "counterfactuals", "neg_sample": ["open - domain qa framework is used for OtherScientificTerm", "deep nlp models have been shown to be brittle to input perturbations .", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "Using an open - domain QA framework and question generation model trained on original task data , we create counterfactuals that are fluent , semantically diverse , and automatically labeled .", "forward": true, "src_ids": "2022.acl-long.117_43"}
+{"input": "counterfactuals is used for Task| context: deep nlp models have been shown to be brittle to input perturbations . recent work has shown that data augmentation using counterfactuals - i.e. minimally perturbed inputs - can help ameliorate this weakness .", "entity": "counterfactuals", "output": "question answering", "neg_sample": ["counterfactuals is used for Task", "deep nlp models have been shown to be brittle to input perturbations .", "recent work has shown that data augmentation using counterfactuals - i.e.", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "We focus on the task of creating counterfactuals for question answering , which presents unique challenges related to world knowledge , semantic diversity , and answerability .", "forward": true, "src_ids": "2022.acl-long.117_44"}
+{"input": "counterfactual evaluation is done by using Method| context: deep nlp models have been shown to be brittle to input perturbations . recent work has shown that data augmentation using counterfactuals - i.e. minimally perturbed inputs - can help ameliorate this weakness .", "entity": "counterfactual evaluation", "output": "retrieve - generate - filter(rgf ) technique", "neg_sample": ["counterfactual evaluation is done by using Method", "deep nlp models have been shown to be brittle to input perturbations .", "recent work has shown that data augmentation using counterfactuals - i.e.", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "To address these challenges , we develop a Retrieve - Generate - Filter(RGF ) technique to create counterfactual evaluation and training data with minimal human supervision .", "forward": false, "src_ids": "2022.acl-long.117_45"}
+{"input": "retrieve - generate - filter(rgf ) technique is used for Method| context: deep nlp models have been shown to be brittle to input perturbations . recent work has shown that data augmentation using counterfactuals - i.e. minimally perturbed inputs - can help ameliorate this weakness .", "entity": "retrieve - generate - filter(rgf ) technique", "output": "counterfactual evaluation", "neg_sample": ["retrieve - generate - filter(rgf ) technique is used for Method", "deep nlp models have been shown to be brittle to input perturbations .", "recent work has shown that data augmentation using counterfactuals - i.e.", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "To address these challenges , we develop a Retrieve - Generate - Filter(RGF ) technique to create counterfactual evaluation and training data with minimal human supervision .", "forward": true, "src_ids": "2022.acl-long.117_46"}
+{"input": "counterfactuals is done by using Method| context: deep nlp models have been shown to be brittle to input perturbations . recent work has shown that data augmentation using counterfactuals - i.e. minimally perturbed inputs - can help ameliorate this weakness .", "entity": "counterfactuals", "output": "open - domain qa framework", "neg_sample": ["counterfactuals is done by using Method", "deep nlp models have been shown to be brittle to input perturbations .", "recent work has shown that data augmentation using counterfactuals - i.e.", "minimally perturbed inputs - can help ameliorate this weakness ."], "relation": "used for", "id": "2022.acl-long.117", "year": 2022, "rel_sent": "Using an open - domain QA framework and question generation model trained on original task data , we create counterfactuals that are fluent , semantically diverse , and automatically labeled .", "forward": false, "src_ids": "2022.acl-long.117_47"}
+{"input": "detecting aggressive texts is done by using Material| context: recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society . several attempts have been made to mitigate the usage and propagation of such content . however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus .", "entity": "detecting aggressive texts", "output": "m - bad", "neg_sample": ["detecting aggressive texts is done by using Material", "recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society .", "several attempts have been made to mitigate the usage and propagation of such content .", "however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus ."], "relation": "used for", "id": "2022.constraint-1.9", "year": 2022, "rel_sent": "M - BAD : A Multilabel Dataset for Detecting Aggressive Texts and Their Targets.", "forward": false, "src_ids": "2022.constraint-1.9_48"}
+{"input": "multilabel dataset is used for Task| context: recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society . several attempts have been made to mitigate the usage and propagation of such content . however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus .", "entity": "multilabel dataset", "output": "detecting aggressive texts", "neg_sample": ["multilabel dataset is used for Task", "recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society .", "several attempts have been made to mitigate the usage and propagation of such content .", "however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus ."], "relation": "used for", "id": "2022.constraint-1.9", "year": 2022, "rel_sent": "M - BAD : A Multilabel Dataset for Detecting Aggressive Texts and Their Targets.", "forward": true, "src_ids": "2022.constraint-1.9_49"}
+{"input": "m - bad is used for Task| context: recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society . several attempts have been made to mitigate the usage and propagation of such content . however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus .", "entity": "m - bad", "output": "detecting aggressive texts", "neg_sample": ["m - bad is used for Task", "recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society .", "several attempts have been made to mitigate the usage and propagation of such content .", "however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus ."], "relation": "used for", "id": "2022.constraint-1.9", "year": 2022, "rel_sent": "M - BAD : A Multilabel Dataset for Detecting Aggressive Texts and Their Targets.", "forward": true, "src_ids": "2022.constraint-1.9_50"}
+{"input": "context - dependent aggression is done by using Method| context: recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society . several attempts have been made to mitigate the usage and propagation of such content . however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus .", "entity": "context - dependent aggression", "output": "error analysis", "neg_sample": ["context - dependent aggression is done by using Method", "recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society .", "several attempts have been made to mitigate the usage and propagation of such content .", "however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus ."], "relation": "used for", "id": "2022.constraint-1.9", "year": 2022, "rel_sent": "Error analysis of the models exhibits the difficulty to identify context - dependent aggression , and this work argues that further research is required to address these issues .", "forward": false, "src_ids": "2022.constraint-1.9_51"}
+{"input": "error analysis is used for OtherScientificTerm| context: recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society . several attempts have been made to mitigate the usage and propagation of such content . however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus .", "entity": "error analysis", "output": "context - dependent aggression", "neg_sample": ["error analysis is used for OtherScientificTerm", "recently , detection and categorization of undesired ( e. g. , aggressive , abusive , offensive , hate ) content from online platforms has grabbed the attention of researchers because of its detrimental impact on society .", "several attempts have been made to mitigate the usage and propagation of such content .", "however , most past studies were conducted primarily for english , where low - resource languages like bengali remained out of the focus ."], "relation": "used for", "id": "2022.constraint-1.9", "year": 2022, "rel_sent": "Error analysis of the models exhibits the difficulty to identify context - dependent aggression , and this work argues that further research is required to address these issues .", "forward": true, "src_ids": "2022.constraint-1.9_52"}
+{"input": "meta - dataset is used for Task| context: humans ( e.g. , crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples . despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g. , a question - answering system can not solve classification tasks ) . a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it .", "entity": "meta - dataset", "output": "cross - task generalization", "neg_sample": ["meta - dataset is used for Task", "humans ( e.g.", ", crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples .", "despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g.", ", a question - answering system can not solve classification tasks ) .", "a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it ."], "relation": "used for", "id": "2022.acl-long.244", "year": 2022, "rel_sent": "Using this meta - dataset , we measure cross - task generalization by training models on seen tasks and measuring generalization to the remaining unseen ones .", "forward": true, "src_ids": "2022.acl-long.244_53"}
+{"input": "nlp datasets is done by using OtherScientificTerm| context: humans ( e.g. , crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples . despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g. , a question - answering system can not solve classification tasks ) . a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it .", "entity": "nlp datasets", "output": "crowdsourcing instructions", "neg_sample": ["nlp datasets is done by using OtherScientificTerm", "humans ( e.g.", ", crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples .", "despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g.", ", a question - answering system can not solve classification tasks ) .", "a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it ."], "relation": "used for", "id": "2022.acl-long.244", "year": 2022, "rel_sent": "The instructions are obtained from crowdsourcing instructions used to create existing NLP datasets and mapped to a unified schema .", "forward": false, "src_ids": "2022.acl-long.244_54"}
+{"input": "crowdsourcing instructions is used for Material| context: humans ( e.g. , crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples . despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g. , a question - answering system can not solve classification tasks ) . a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it .", "entity": "crowdsourcing instructions", "output": "nlp datasets", "neg_sample": ["crowdsourcing instructions is used for Material", "humans ( e.g.", ", crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples .", "despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g.", ", a question - answering system can not solve classification tasks ) .", "a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it ."], "relation": "used for", "id": "2022.acl-long.244", "year": 2022, "rel_sent": "The instructions are obtained from crowdsourcing instructions used to create existing NLP datasets and mapped to a unified schema .", "forward": true, "src_ids": "2022.acl-long.244_55"}
+{"input": "cross - task generalization is done by using Generic| context: humans ( e.g. , crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples . despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g. , a question - answering system can not solve classification tasks ) . a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it .", "entity": "cross - task generalization", "output": "meta - dataset", "neg_sample": ["cross - task generalization is done by using Generic", "humans ( e.g.", ", crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples .", "despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g.", ", a question - answering system can not solve classification tasks ) .", "a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it ."], "relation": "used for", "id": "2022.acl-long.244", "year": 2022, "rel_sent": "Using this meta - dataset , we measure cross - task generalization by training models on seen tasks and measuring generalization to the remaining unseen ones .", "forward": false, "src_ids": "2022.acl-long.244_56"}
+{"input": "task - specific instructions is done by using Method| context: humans ( e.g. , crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples . despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g. , a question - answering system can not solve classification tasks ) . a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it .", "entity": "task - specific instructions", "output": "generative pre - trained language models", "neg_sample": ["task - specific instructions is done by using Method", "humans ( e.g.", ", crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples .", "despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g.", ", a question - answering system can not solve classification tasks ) .", "a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it ."], "relation": "used for", "id": "2022.acl-long.244", "year": 2022, "rel_sent": "We adopt generative pre - trained language models to encode task - specific instructions along with input and generate task output .", "forward": false, "src_ids": "2022.acl-long.244_57"}
+{"input": "generative pre - trained language models is used for OtherScientificTerm| context: humans ( e.g. , crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples . despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g. , a question - answering system can not solve classification tasks ) . a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it .", "entity": "generative pre - trained language models", "output": "task - specific instructions", "neg_sample": ["generative pre - trained language models is used for OtherScientificTerm", "humans ( e.g.", ", crowdworkers ) have a remarkable ability in solving different tasks , by simply reading textual instructions that define them and looking at a few examples .", "despite the success of the conventional supervised learning on individual datasets , such models often struggle with generalization across tasks ( e.g.", ", a question - answering system can not solve classification tasks ) .", "a long - standing challenge in ai is to build a model that learns a new task by understanding the human - readable instructions that define it ."], "relation": "used for", "id": "2022.acl-long.244", "year": 2022, "rel_sent": "We adopt generative pre - trained language models to encode task - specific instructions along with input and generate task output .", "forward": true, "src_ids": "2022.acl-long.244_58"}
+{"input": "model out - of - distribution is done by using Method| context: by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e.", "entity": "model out - of - distribution", "output": "rational - centric framework", "neg_sample": ["model out - of - distribution is done by using Method", "by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e."], "relation": "used for", "id": "2022.acl-long.481", "year": 2022, "rel_sent": "We present a novel rational - centric framework with human - in - the - loop - Rationales - centric Double - robustness Learning ( RDL ) - to boost model out - of - distribution performance in few - shot learning scenarios .", "forward": false, "src_ids": "2022.acl-long.481_59"}
+{"input": "rational - centric framework is used for Task| context: by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e.", "entity": "rational - centric framework", "output": "model out - of - distribution", "neg_sample": ["rational - centric framework is used for Task", "by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e."], "relation": "used for", "id": "2022.acl-long.481", "year": 2022, "rel_sent": "We present a novel rational - centric framework with human - in - the - loop - Rationales - centric Double - robustness Learning ( RDL ) - to boost model out - of - distribution performance in few - shot learning scenarios .", "forward": true, "src_ids": "2022.acl-long.481_60"}
+{"input": "spurious associations is done by using Method| context: by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e.", "entity": "spurious associations", "output": "semi - factual augmentations", "neg_sample": ["spurious associations is done by using Method", "by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e."], "relation": "used for", "id": "2022.acl-long.481", "year": 2022, "rel_sent": "phrases that cause the prediction ) , human interventions and semi - factual augmentations to decouple spurious associations and bias models towards generally applicable underlying distributions , which enables fast and accurate generalisation .", "forward": false, "src_ids": "2022.acl-long.481_61"}
+{"input": "bias models is done by using Method| context: by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e.", "entity": "bias models", "output": "semi - factual augmentations", "neg_sample": ["bias models is done by using Method", "by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e."], "relation": "used for", "id": "2022.acl-long.481", "year": 2022, "rel_sent": "phrases that cause the prediction ) , human interventions and semi - factual augmentations to decouple spurious associations and bias models towards generally applicable underlying distributions , which enables fast and accurate generalisation .", "forward": false, "src_ids": "2022.acl-long.481_62"}
+{"input": "semi - factual augmentations is used for OtherScientificTerm| context: by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e.", "entity": "semi - factual augmentations", "output": "spurious associations", "neg_sample": ["semi - factual augmentations is used for OtherScientificTerm", "by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e."], "relation": "used for", "id": "2022.acl-long.481", "year": 2022, "rel_sent": "phrases that cause the prediction ) , human interventions and semi - factual augmentations to decouple spurious associations and bias models towards generally applicable underlying distributions , which enables fast and accurate generalisation .", "forward": true, "src_ids": "2022.acl-long.481_63"}
+{"input": "semi - factual augmentations is used for Method| context: by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e.", "entity": "semi - factual augmentations", "output": "bias models", "neg_sample": ["semi - factual augmentations is used for Method", "by using static semi - factual generation and dynamic human - intervened correction , rdl , acting like a sensible ' inductive bias ' , exploits rationales ( i.e."], "relation": "used for", "id": "2022.acl-long.481", "year": 2022, "rel_sent": "phrases that cause the prediction ) , human interventions and semi - factual augmentations to decouple spurious associations and bias models towards generally applicable underlying distributions , which enables fast and accurate generalisation .", "forward": true, "src_ids": "2022.acl-long.481_64"}
+{"input": "differentially - private text representation learning is done by using Method| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "differentially - private text representation learning", "output": "reparametrization trick", "neg_sample": ["differentially - private text representation learning is done by using Method", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.87", "year": 2022, "rel_sent": "How reparametrization trick broke differentially - private text representation learning.", "forward": false, "src_ids": "2022.acl-short.87_65"}
+{"input": "reparametrization trick is used for Method| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "reparametrization trick", "output": "differentially - private text representation learning", "neg_sample": ["reparametrization trick is used for Method", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.87", "year": 2022, "rel_sent": "How reparametrization trick broke differentially - private text representation learning.", "forward": true, "src_ids": "2022.acl-short.87_66"}
+{"input": "text representation learning is done by using Method| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "entity": "text representation learning", "output": "differential privacy ( dp )", "neg_sample": ["text representation learning is done by using Method", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods ."], "relation": "used for", "id": "2022.acl-short.87", "year": 2022, "rel_sent": "Our main goal is to raise awareness and help the community understand potential pitfalls of applying differential privacy to text representation learning .", "forward": false, "src_ids": "2022.acl-short.87_67"}
+{"input": "text representation learning is done by using Material| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "text representation learning", "output": "nlp papers", "neg_sample": ["text representation learning is done by using Material", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.87", "year": 2022, "rel_sent": "In this short paper , we formally analyze several recent NLP papers proposing text representation learning using DPText ( Beigi et al . , 2019a , b ; Alnasser et al . , 2021 ; Beigi et al . , 2021 ) and reveal their false claims of being differentially private .", "forward": false, "src_ids": "2022.acl-short.87_68"}
+{"input": "nlp papers is used for Task| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "nlp papers", "output": "text representation learning", "neg_sample": ["nlp papers is used for Task", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.87", "year": 2022, "rel_sent": "In this short paper , we formally analyze several recent NLP papers proposing text representation learning using DPText ( Beigi et al . , 2019a , b ; Alnasser et al . , 2021 ; Beigi et al . , 2021 ) and reveal their false claims of being differentially private .", "forward": true, "src_ids": "2022.acl-short.87_69"}
+{"input": "differential privacy ( dp ) is used for Task| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "differential privacy ( dp )", "output": "text representation learning", "neg_sample": ["differential privacy ( dp ) is used for Task", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.87", "year": 2022, "rel_sent": "Our main goal is to raise awareness and help the community understand potential pitfalls of applying differential privacy to text representation learning .", "forward": true, "src_ids": "2022.acl-short.87_70"}
+{"input": "question - schema interaction graph encoder is done by using OtherScientificTerm| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "entity": "question - schema interaction graph encoder", "output": "syntax", "neg_sample": ["question - schema interaction graph encoder is done by using OtherScientificTerm", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "S^2SQL : Injecting Syntax to Question - Schema Interaction Graph Encoder for Text - to - SQL Parsers.", "forward": false, "src_ids": "2022.findings-acl.99_71"}
+{"input": "text - to - sql parsers is done by using Method| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "text - to - sql parsers", "output": "question - schema interaction graph encoder", "neg_sample": ["text - to - sql parsers is done by using Method", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "S^2SQL : Injecting Syntax to Question - Schema Interaction Graph Encoder for Text - to - SQL Parsers.", "forward": false, "src_ids": "2022.findings-acl.99_72"}
+{"input": "syntax is used for Method| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "syntax", "output": "question - schema interaction graph encoder", "neg_sample": ["syntax is used for Method", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "S^2SQL : Injecting Syntax to Question - Schema Interaction Graph Encoder for Text - to - SQL Parsers.", "forward": true, "src_ids": "2022.findings-acl.99_73"}
+{"input": "question - schema interaction graph encoder is used for Method| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "question - schema interaction graph encoder", "output": "text - to - sql parsers", "neg_sample": ["question - schema interaction graph encoder is used for Method", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "S^2SQL : Injecting Syntax to Question - Schema Interaction Graph Encoder for Text - to - SQL Parsers.", "forward": true, "src_ids": "2022.findings-acl.99_74"}
+{"input": "question - schema graph encoder is used for Method| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "question - schema graph encoder", "output": "text - to - sql parsers", "neg_sample": ["question - schema graph encoder is used for Method", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "In this paper , we propose S^2SQL , injecting Syntax to question - Schema graph encoder for Text - to - SQL parsers , which effectively leverages the syntactic dependency information of questions in text - to - SQL to improve the performance .", "forward": true, "src_ids": "2022.findings-acl.99_75"}
+{"input": "text - to - sql parsers is done by using Method| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "text - to - sql parsers", "output": "question - schema graph encoder", "neg_sample": ["text - to - sql parsers is done by using Method", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "In this paper , we propose S^2SQL , injecting Syntax to question - Schema graph encoder for Text - to - SQL parsers , which effectively leverages the syntactic dependency information of questions in text - to - SQL to improve the performance .", "forward": false, "src_ids": "2022.findings-acl.99_76"}
+{"input": "diverse relational edge embedding is done by using OtherScientificTerm| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "diverse relational edge embedding", "output": "decoupling constraint", "neg_sample": ["diverse relational edge embedding is done by using OtherScientificTerm", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "We also employ the decoupling constraint to induce diverse relational edge embedding , which further improves the network 's performance .", "forward": false, "src_ids": "2022.findings-acl.99_77"}
+{"input": "decoupling constraint is used for OtherScientificTerm| context: the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing . the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well .", "entity": "decoupling constraint", "output": "diverse relational edge embedding", "neg_sample": ["decoupling constraint is used for OtherScientificTerm", "the task of converting a natural language question into an executable sql query , known as text - to - sql , is an important branch of semantic parsing .", "the state - of - the - art graph - based encoder has been successfully used in this task but does not model the question syntax well ."], "relation": "used for", "id": "2022.findings-acl.99", "year": 2022, "rel_sent": "We also employ the decoupling constraint to induce diverse relational edge embedding , which further improves the network 's performance .", "forward": true, "src_ids": "2022.findings-acl.99_78"}
+{"input": "task - oriented dialogue generation is done by using Method| context: we study the interpretability issue of task - oriented dialogue systems in this paper . previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans . since deriving reasoning chains requires multi - hop reasoning for task - oriented dialogues , existing neuro - symbolic approaches would induce error propagation due to the one - phase design .", "entity": "task - oriented dialogue generation", "output": "interpretable neuro - symbolic reasoning framework", "neg_sample": ["task - oriented dialogue generation is done by using Method", "we study the interpretability issue of task - oriented dialogue systems in this paper .", "previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans .", "since deriving reasoning chains requires multi - hop reasoning for task - oriented dialogues , existing neuro - symbolic approaches would induce error propagation due to the one - phase design ."], "relation": "used for", "id": "2022.acl-long.338", "year": 2022, "rel_sent": "An Interpretable Neuro - Symbolic Reasoning Framework for Task - Oriented Dialogue Generation.", "forward": false, "src_ids": "2022.acl-long.338_79"}
+{"input": "interpretable neuro - symbolic reasoning framework is used for Task| context: we study the interpretability issue of task - oriented dialogue systems in this paper . previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans . since deriving reasoning chains requires multi - hop reasoning for task - oriented dialogues , existing neuro - symbolic approaches would induce error propagation due to the one - phase design .", "entity": "interpretable neuro - symbolic reasoning framework", "output": "task - oriented dialogue generation", "neg_sample": ["interpretable neuro - symbolic reasoning framework is used for Task", "we study the interpretability issue of task - oriented dialogue systems in this paper .", "previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans .", "since deriving reasoning chains requires multi - hop reasoning for task - oriented dialogues , existing neuro - symbolic approaches would induce error propagation due to the one - phase design ."], "relation": "used for", "id": "2022.acl-long.338", "year": 2022, "rel_sent": "An Interpretable Neuro - Symbolic Reasoning Framework for Task - Oriented Dialogue Generation.", "forward": true, "src_ids": "2022.acl-long.338_80"}
+{"input": "explicit reasoning is done by using Method| context: we study the interpretability issue of task - oriented dialogue systems in this paper . previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans .", "entity": "explicit reasoning", "output": "neuro - symbolic", "neg_sample": ["explicit reasoning is done by using Method", "we study the interpretability issue of task - oriented dialogue systems in this paper .", "previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans ."], "relation": "used for", "id": "2022.acl-long.338", "year": 2022, "rel_sent": "To obtain a transparent reasoning process , we introduce neuro - symbolic to perform explicit reasoning that justifies model decisions by reasoning chains .", "forward": false, "src_ids": "2022.acl-long.338_81"}
+{"input": "neuro - symbolic is used for Method| context: we study the interpretability issue of task - oriented dialogue systems in this paper . previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans . since deriving reasoning chains requires multi - hop reasoning for task - oriented dialogues , existing neuro - symbolic approaches would induce error propagation due to the one - phase design .", "entity": "neuro - symbolic", "output": "explicit reasoning", "neg_sample": ["neuro - symbolic is used for Method", "we study the interpretability issue of task - oriented dialogue systems in this paper .", "previously , most neural - based task - oriented dialogue systems employ an implicit reasoning strategy that makes the model predictions uninterpretable to humans .", "since deriving reasoning chains requires multi - hop reasoning for task - oriented dialogues , existing neuro - symbolic approaches would induce error propagation due to the one - phase design ."], "relation": "used for", "id": "2022.acl-long.338", "year": 2022, "rel_sent": "To obtain a transparent reasoning process , we introduce neuro - symbolic to perform explicit reasoning that justifies model decisions by reasoning chains .", "forward": true, "src_ids": "2022.acl-long.338_82"}
+{"input": "fact checking is done by using Method| context: fact checking is a challenging task that requires corresponding evidences to verify the property of a claim based on reasoning . previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact .", "entity": "fact checking", "output": "heterogeneous - graph reasoning", "neg_sample": ["fact checking is done by using Method", "fact checking is a challenging task that requires corresponding evidences to verify the property of a claim based on reasoning .", "previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact ."], "relation": "used for", "id": "2022.fever-1.2", "year": 2022, "rel_sent": "Heterogeneous - Graph Reasoning and Fine - Grained Aggregation for Fact Checking.", "forward": false, "src_ids": "2022.fever-1.2_83"}
+{"input": "fine - grained aggregation is used for Task| context: previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact .", "entity": "fine - grained aggregation", "output": "fact checking", "neg_sample": ["fine - grained aggregation is used for Task", "previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact ."], "relation": "used for", "id": "2022.fever-1.2", "year": 2022, "rel_sent": "Heterogeneous - Graph Reasoning and Fine - Grained Aggregation for Fact Checking.", "forward": true, "src_ids": "2022.fever-1.2_84"}
+{"input": "heterogeneous - graph reasoning is used for Task| context: previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact .", "entity": "heterogeneous - graph reasoning", "output": "fact checking", "neg_sample": ["heterogeneous - graph reasoning is used for Task", "previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact ."], "relation": "used for", "id": "2022.fever-1.2", "year": 2022, "rel_sent": "Heterogeneous - Graph Reasoning and Fine - Grained Aggregation for Fact Checking.", "forward": true, "src_ids": "2022.fever-1.2_85"}
+{"input": "implicit stance of evidences is done by using Method| context: fact checking is a challenging task that requires corresponding evidences to verify the property of a claim based on reasoning . previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact .", "entity": "implicit stance of evidences", "output": "fine - grained aggregation module", "neg_sample": ["implicit stance of evidences is done by using Method", "fact checking is a challenging task that requires corresponding evidences to verify the property of a claim based on reasoning .", "previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact ."], "relation": "used for", "id": "2022.fever-1.2", "year": 2022, "rel_sent": "Towards the above issues , we propose a novel heterogeneous - graph reasoning and fine - grained aggregation model , with twofollowing modules : 1 ) a heterogeneous graph attention network module to distinguish different types of relationships within the constructed graph ; 2 ) fine - grained aggregation module which learns the implicit stance of evidences towards the prediction result in details .", "forward": false, "src_ids": "2022.fever-1.2_86"}
+{"input": "fine - grained aggregation module is used for OtherScientificTerm| context: fact checking is a challenging task that requires corresponding evidences to verify the property of a claim based on reasoning . previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact .", "entity": "fine - grained aggregation module", "output": "implicit stance of evidences", "neg_sample": ["fine - grained aggregation module is used for OtherScientificTerm", "fact checking is a challenging task that requires corresponding evidences to verify the property of a claim based on reasoning .", "previous studies generally i ) construct the graph by treating each evidence - claim pair as node which is a simple way that ignores to exploit their implicit interaction , or building a fully - connected graph among claim and evidences where the entailment relationship between claim and evidence would be considered equal to the semantic relationship among evidences ; ii ) aggregate evidences equally without considering their different stances towards the verification of fact ."], "relation": "used for", "id": "2022.fever-1.2", "year": 2022, "rel_sent": "Towards the above issues , we propose a novel heterogeneous - graph reasoning and fine - grained aggregation model , with twofollowing modules : 1 ) a heterogeneous graph attention network module to distinguish different types of relationships within the constructed graph ; 2 ) fine - grained aggregation module which learns the implicit stance of evidences towards the prediction result in details .", "forward": true, "src_ids": "2022.fever-1.2_87"}
+{"input": "cross - lingual emotion classification is done by using Task| context: the main challenge in english - malay cross - lingual emotion classification is that there are no malay training emotion corpora . given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level .", "entity": "cross - lingual emotion classification", "output": "english - malay word embeddings alignment", "neg_sample": ["cross - lingual emotion classification is done by using Task", "the main challenge in english - malay cross - lingual emotion classification is that there are no malay training emotion corpora .", "given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level ."], "relation": "used for", "id": "2022.wassa-1.12", "year": 2022, "rel_sent": "English - Malay Word Embeddings Alignment for Cross - lingual Emotion Classification with Hierarchical Attention Network.", "forward": false, "src_ids": "2022.wassa-1.12_88"}
+{"input": "english - malay word embeddings alignment is used for Task| context: given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level .", "entity": "english - malay word embeddings alignment", "output": "cross - lingual emotion classification", "neg_sample": ["english - malay word embeddings alignment is used for Task", "given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level ."], "relation": "used for", "id": "2022.wassa-1.12", "year": 2022, "rel_sent": "English - Malay Word Embeddings Alignment for Cross - lingual Emotion Classification with Hierarchical Attention Network.", "forward": true, "src_ids": "2022.wassa-1.12_89"}
+{"input": "zero - shot learning is done by using Method| context: the main challenge in english - malay cross - lingual emotion classification is that there are no malay training emotion corpora . given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level .", "entity": "zero - shot learning", "output": "mbert", "neg_sample": ["zero - shot learning is done by using Method", "the main challenge in english - malay cross - lingual emotion classification is that there are no malay training emotion corpora .", "given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level ."], "relation": "used for", "id": "2022.wassa-1.12", "year": 2022, "rel_sent": "Experimental results show that the performance of our model is better than mBERT in zero - shot learning by 2.4 % and Malay BERT by 0.8 % when a limited number of Malay tweets is available .", "forward": false, "src_ids": "2022.wassa-1.12_90"}
+{"input": "mbert is used for Method| context: the main challenge in english - malay cross - lingual emotion classification is that there are no malay training emotion corpora . given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level .", "entity": "mbert", "output": "zero - shot learning", "neg_sample": ["mbert is used for Method", "the main challenge in english - malay cross - lingual emotion classification is that there are no malay training emotion corpora .", "given that machine translation could fall short in contextually complex tweets , we only limited machine translation to the word level ."], "relation": "used for", "id": "2022.wassa-1.12", "year": 2022, "rel_sent": "Experimental results show that the performance of our model is better than mBERT in zero - shot learning by 2.4 % and Malay BERT by 0.8 % when a limited number of Malay tweets is available .", "forward": true, "src_ids": "2022.wassa-1.12_91"}
+{"input": "a2lctc is used for OtherScientificTerm| context: when mapping a natural language instruction to a sequence of actions , it is often useful toidentify sub - tasks in the instruction . such sub - task segmentation , however , is not necessarily provided in the training data .", "entity": "a2lctc", "output": "sub - task structures", "neg_sample": ["a2lctc is used for OtherScientificTerm", "when mapping a natural language instruction to a sequence of actions , it is often useful toidentify sub - tasks in the instruction .", "such sub - task segmentation , however , is not necessarily provided in the training data ."], "relation": "used for", "id": "2022.lnls-1.1", "year": 2022, "rel_sent": "We present the A2LCTC ( Action - to - Language Connectionist Temporal Classification ) algorithm to automatically discover a sub - task segmentation of an action sequence . A2LCTC does not require annotations of correct sub - task segments and learns tofind them from pairs of instruction and action sequence in a weakly - supervised manner . We experiment with the ALFRED dataset and show that A2LCTC accurately finds the sub - task structures . With the discovered sub - tasks segments , we also train agents that work on the downstream task and empirically show that our algorithm improves the performance .", "forward": true, "src_ids": "2022.lnls-1.1_92"}
+{"input": "sub - task structures is done by using Method| context: when mapping a natural language instruction to a sequence of actions , it is often useful toidentify sub - tasks in the instruction . such sub - task segmentation , however , is not necessarily provided in the training data .", "entity": "sub - task structures", "output": "a2lctc", "neg_sample": ["sub - task structures is done by using Method", "when mapping a natural language instruction to a sequence of actions , it is often useful toidentify sub - tasks in the instruction .", "such sub - task segmentation , however , is not necessarily provided in the training data ."], "relation": "used for", "id": "2022.lnls-1.1", "year": 2022, "rel_sent": "We present the A2LCTC ( Action - to - Language Connectionist Temporal Classification ) algorithm to automatically discover a sub - task segmentation of an action sequence . A2LCTC does not require annotations of correct sub - task segments and learns tofind them from pairs of instruction and action sequence in a weakly - supervised manner . We experiment with the ALFRED dataset and show that A2LCTC accurately finds the sub - task structures . With the discovered sub - tasks segments , we also train agents that work on the downstream task and empirically show that our algorithm improves the performance .", "forward": false, "src_ids": "2022.lnls-1.1_93"}
+{"input": "personality profiling models is done by using Method| context: this is an expensive but accurate data collection strategy . another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results . such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource .", "entity": "personality profiling models", "output": "psychometric tests", "neg_sample": ["personality profiling models is done by using Method", "this is an expensive but accurate data collection strategy .", "another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results .", "such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource ."], "relation": "used for", "id": "2022.wassa-1.35", "year": 2022, "rel_sent": "Items from Psychometric Tests as Training Data for Personality Profiling Models of Twitter Users.", "forward": false, "src_ids": "2022.wassa-1.35_94"}
+{"input": "psychometric tests is used for Method| context: machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users . this is an expensive but accurate data collection strategy . another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results . such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource .", "entity": "psychometric tests", "output": "personality profiling models", "neg_sample": ["psychometric tests is used for Method", "machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users .", "this is an expensive but accurate data collection strategy .", "another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results .", "such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource ."], "relation": "used for", "id": "2022.wassa-1.35", "year": 2022, "rel_sent": "Items from Psychometric Tests as Training Data for Personality Profiling Models of Twitter Users.", "forward": true, "src_ids": "2022.wassa-1.35_95"}
+{"input": "big five personality traits is done by using OtherScientificTerm| context: machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users . this is an expensive but accurate data collection strategy . another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results . such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource .", "entity": "big five personality traits", "output": "psychometric test items", "neg_sample": ["big five personality traits is done by using OtherScientificTerm", "machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users .", "this is an expensive but accurate data collection strategy .", "another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results .", "such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource ."], "relation": "used for", "id": "2022.wassa-1.35", "year": 2022, "rel_sent": "We investigate this approach for personality profiling , and evaluate BERT classifiers fine - tuned on such psychometric test items for the big five personality traits ( openness , conscientiousness , extraversion , agreeableness , neuroticism ) and analyze various augmentation strategies regarding their potential to address the challenges coming with such a small corpus .", "forward": false, "src_ids": "2022.wassa-1.35_96"}
+{"input": "psychometric test items is used for OtherScientificTerm| context: machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users . this is an expensive but accurate data collection strategy . another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results . such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource .", "entity": "psychometric test items", "output": "big five personality traits", "neg_sample": ["psychometric test items is used for OtherScientificTerm", "machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users .", "this is an expensive but accurate data collection strategy .", "another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results .", "such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource ."], "relation": "used for", "id": "2022.wassa-1.35", "year": 2022, "rel_sent": "We investigate this approach for personality profiling , and evaluate BERT classifiers fine - tuned on such psychometric test items for the big five personality traits ( openness , conscientiousness , extraversion , agreeableness , neuroticism ) and analyze various augmentation strategies regarding their potential to address the challenges coming with such a small corpus .", "forward": true, "src_ids": "2022.wassa-1.35_97"}
+{"input": "personality traits is done by using Method| context: machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users . this is an expensive but accurate data collection strategy . another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results . such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource .", "entity": "personality traits", "output": "in - domain training", "neg_sample": ["personality traits is done by using Method", "machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users .", "this is an expensive but accurate data collection strategy .", "another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results .", "such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource ."], "relation": "used for", "id": "2022.wassa-1.35", "year": 2022, "rel_sent": "Our evaluation on a publicly available Twitter corpus shows a comparable performance to in - domain training for 4/5 personality traits with T5 - based data augmentation .", "forward": false, "src_ids": "2022.wassa-1.35_98"}
+{"input": "in - domain training is used for OtherScientificTerm| context: machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users . this is an expensive but accurate data collection strategy . another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results . such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource .", "entity": "in - domain training", "output": "personality traits", "neg_sample": ["in - domain training is used for OtherScientificTerm", "machine - learned models for author profiling in social media often rely on data acquired via self - reporting - based psychometric tests ( questionnaires ) filled out by social media users .", "this is an expensive but accurate data collection strategy .", "another , less costly alternative , which leads to potentially more noisy and biased data , is to rely on labels inferred from publicly available information in the profiles of the users , for instance self - reported diagnoses or test results .", "such corpora of test items constitute ' small data ' , but their availability for many concepts is a rich resource ."], "relation": "used for", "id": "2022.wassa-1.35", "year": 2022, "rel_sent": "Our evaluation on a publicly available Twitter corpus shows a comparable performance to in - domain training for 4/5 personality traits with T5 - based data augmentation .", "forward": true, "src_ids": "2022.wassa-1.35_99"}
+{"input": "conversational machine reading comprehension ( cmrc ) is done by using Method| context: coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) . existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "conversational machine reading comprehension ( cmrc )", "output": "graph - combined coreference resolution methods", "neg_sample": ["conversational machine reading comprehension ( cmrc ) is done by using Method", "coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) .", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "Graph - combined Coreference Resolution Methods on Conversational Machine Reading Comprehension with Pre - trained Language Model.", "forward": false, "src_ids": "2022.dialdoc-1.8_100"}
+{"input": "graph - combined coreference resolution methods is used for Task| context: existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "graph - combined coreference resolution methods", "output": "conversational machine reading comprehension ( cmrc )", "neg_sample": ["graph - combined coreference resolution methods is used for Task", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "Graph - combined Coreference Resolution Methods on Conversational Machine Reading Comprehension with Pre - trained Language Model.", "forward": true, "src_ids": "2022.dialdoc-1.8_101"}
+{"input": "fusion model is used for Task| context: existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "fusion model", "output": "conversational machine reading comprehension ( cmrc )", "neg_sample": ["fusion model is used for Task", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "We propose two graph - combined methods , evidence - enhanced and the fusion model , for CMRC to integrate coreference graphs from different levels of the PLM architecture .", "forward": true, "src_ids": "2022.dialdoc-1.8_102"}
+{"input": "graph - combined methods is used for Task| context: existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "graph - combined methods", "output": "conversational machine reading comprehension ( cmrc )", "neg_sample": ["graph - combined methods is used for Task", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "We propose two graph - combined methods , evidence - enhanced and the fusion model , for CMRC to integrate coreference graphs from different levels of the PLM architecture .", "forward": true, "src_ids": "2022.dialdoc-1.8_103"}
+{"input": "graph structures is done by using Method| context: coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) . existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "graph structures", "output": "graph - based approach", "neg_sample": ["graph structures is done by using Method", "coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) .", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "In this study , a novel graph - based approach is proposed to integrate the coreference of given text into graph structures ( called coreference graphs ) , which can pinpoint a pronoun 's referential entity .", "forward": false, "src_ids": "2022.dialdoc-1.8_104"}
+{"input": "coreference is done by using Method| context: coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) . existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "coreference", "output": "graph - based approach", "neg_sample": ["coreference is done by using Method", "coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) .", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "In this study , a novel graph - based approach is proposed to integrate the coreference of given text into graph structures ( called coreference graphs ) , which can pinpoint a pronoun 's referential entity .", "forward": false, "src_ids": "2022.dialdoc-1.8_105"}
+{"input": "graph - based approach is used for OtherScientificTerm| context: existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "graph - based approach", "output": "coreference", "neg_sample": ["graph - based approach is used for OtherScientificTerm", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "In this study , a novel graph - based approach is proposed to integrate the coreference of given text into graph structures ( called coreference graphs ) , which can pinpoint a pronoun 's referential entity .", "forward": true, "src_ids": "2022.dialdoc-1.8_106"}
+{"input": "graph - based approach is used for OtherScientificTerm| context: coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) . existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "graph - based approach", "output": "graph structures", "neg_sample": ["graph - based approach is used for OtherScientificTerm", "coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) .", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "In this study , a novel graph - based approach is proposed to integrate the coreference of given text into graph structures ( called coreference graphs ) , which can pinpoint a pronoun 's referential entity .", "forward": true, "src_ids": "2022.dialdoc-1.8_107"}
+{"input": "conversational machine reading comprehension ( cmrc ) is done by using Method| context: coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) . existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency .", "entity": "conversational machine reading comprehension ( cmrc )", "output": "graph - combined methods", "neg_sample": ["conversational machine reading comprehension ( cmrc ) is done by using Method", "coreference resolution such as for anaphora has been an essential challenge that is commonly found in conversational machine reading comprehension ( cmrc ) .", "existing approaches based on pre - trained language models ( plms ) mainly rely on an end - to - end method , which still has limitations in clarifying referential dependency ."], "relation": "used for", "id": "2022.dialdoc-1.8", "year": 2022, "rel_sent": "We propose two graph - combined methods , evidence - enhanced and the fusion model , for CMRC to integrate coreference graphs from different levels of the PLM architecture .", "forward": false, "src_ids": "2022.dialdoc-1.8_108"}
+{"input": "entity and relation extraction is done by using Method| context: recent entity and relation extraction works focus on investigating how to obtain a better span representation from the pre - trained encoder . however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) .", "entity": "entity and relation extraction", "output": "packed levitated markers ( pl - marker )", "neg_sample": ["entity and relation extraction is done by using Method", "recent entity and relation extraction works focus on investigating how to obtain a better span representation from the pre - trained encoder .", "however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) ."], "relation": "used for", "id": "2022.acl-long.337", "year": 2022, "rel_sent": "Packed Levitated Marker for Entity and Relation Extraction.", "forward": false, "src_ids": "2022.acl-long.337_109"}
+{"input": "packed levitated markers ( pl - marker ) is used for Task| context: however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) .", "entity": "packed levitated markers ( pl - marker )", "output": "entity and relation extraction", "neg_sample": ["packed levitated markers ( pl - marker ) is used for Task", "however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) ."], "relation": "used for", "id": "2022.acl-long.337", "year": 2022, "rel_sent": "Packed Levitated Marker for Entity and Relation Extraction.", "forward": true, "src_ids": "2022.acl-long.337_110"}
+{"input": "entity boundary information is done by using Method| context: recent entity and relation extraction works focus on investigating how to obtain a better span representation from the pre - trained encoder . however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) .", "entity": "entity boundary information", "output": "neighborhood - oriented packing strategy", "neg_sample": ["entity boundary information is done by using Method", "recent entity and relation extraction works focus on investigating how to obtain a better span representation from the pre - trained encoder .", "however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) ."], "relation": "used for", "id": "2022.acl-long.337", "year": 2022, "rel_sent": "In particular , we propose a neighborhood - oriented packing strategy , which considers the neighbor spans integrally to better model the entity boundary information .", "forward": false, "src_ids": "2022.acl-long.337_111"}
+{"input": "neighborhood - oriented packing strategy is used for OtherScientificTerm| context: recent entity and relation extraction works focus on investigating how to obtain a better span representation from the pre - trained encoder . however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) .", "entity": "neighborhood - oriented packing strategy", "output": "entity boundary information", "neg_sample": ["neighborhood - oriented packing strategy is used for OtherScientificTerm", "recent entity and relation extraction works focus on investigating how to obtain a better span representation from the pre - trained encoder .", "however , a major limitation of existing works is that they ignore the interrelation between spans ( pairs ) ."], "relation": "used for", "id": "2022.acl-long.337", "year": 2022, "rel_sent": "In particular , we propose a neighborhood - oriented packing strategy , which considers the neighbor spans integrally to better model the entity boundary information .", "forward": true, "src_ids": "2022.acl-long.337_112"}
+{"input": "pretrained language models is done by using Method| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "pretrained language models", "output": "task - guided disentangled tuning ( tdt )", "neg_sample": ["pretrained language models is done by using Method", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "Task - guided Disentangled Tuning for Pretrained Language Models.", "forward": false, "src_ids": "2022.findings-acl.247_113"}
+{"input": "pretrained language models is done by using Method| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "pretrained language models", "output": "task - guided disentangled tuning ( tdt )", "neg_sample": ["pretrained language models is done by using Method", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "To address this issue , we propose Task - guided Disentangled Tuning ( TDT ) for PLMs , which enhances the generalization of representations by disentangling task - relevant signals from the entangled representations .", "forward": false, "src_ids": "2022.findings-acl.247_114"}
+{"input": "generalization of representations is done by using Method| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "generalization of representations", "output": "task - guided disentangled tuning ( tdt )", "neg_sample": ["generalization of representations is done by using Method", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "To address this issue , we propose Task - guided Disentangled Tuning ( TDT ) for PLMs , which enhances the generalization of representations by disentangling task - relevant signals from the entangled representations .", "forward": false, "src_ids": "2022.findings-acl.247_115"}
+{"input": "task - guided disentangled tuning ( tdt ) is used for Method| context: however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "task - guided disentangled tuning ( tdt )", "output": "pretrained language models", "neg_sample": ["task - guided disentangled tuning ( tdt ) is used for Method", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "Task - guided Disentangled Tuning for Pretrained Language Models.", "forward": true, "src_ids": "2022.findings-acl.247_116"}
+{"input": "task - guided disentangled tuning ( tdt ) is used for Method| context: however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "task - guided disentangled tuning ( tdt )", "output": "pretrained language models", "neg_sample": ["task - guided disentangled tuning ( tdt ) is used for Method", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "To address this issue , we propose Task - guided Disentangled Tuning ( TDT ) for PLMs , which enhances the generalization of representations by disentangling task - relevant signals from the entangled representations .", "forward": true, "src_ids": "2022.findings-acl.247_117"}
+{"input": "task - guided disentangled tuning ( tdt ) is used for Task| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "task - guided disentangled tuning ( tdt )", "output": "generalization of representations", "neg_sample": ["task - guided disentangled tuning ( tdt ) is used for Task", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "To address this issue , we propose Task - guided Disentangled Tuning ( TDT ) for PLMs , which enhances the generalization of representations by disentangling task - relevant signals from the entangled representations .", "forward": true, "src_ids": "2022.findings-acl.247_118"}
+{"input": "indicative guidance is done by using Method| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "indicative guidance", "output": "learnable confidence model", "neg_sample": ["indicative guidance is done by using Method", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "For a given task , we introduce a learnable confidence model to detect indicative guidance from context , and further propose a disentangled regularization to mitigate the over - reliance problem .", "forward": false, "src_ids": "2022.findings-acl.247_119"}
+{"input": "learnable confidence model is used for OtherScientificTerm| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "learnable confidence model", "output": "indicative guidance", "neg_sample": ["learnable confidence model is used for OtherScientificTerm", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "For a given task , we introduce a learnable confidence model to detect indicative guidance from context , and further propose a disentangled regularization to mitigate the over - reliance problem .", "forward": true, "src_ids": "2022.findings-acl.247_120"}
+{"input": "over - reliance problem is done by using Method| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "over - reliance problem", "output": "disentangled regularization", "neg_sample": ["over - reliance problem is done by using Method", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "For a given task , we introduce a learnable confidence model to detect indicative guidance from context , and further propose a disentangled regularization to mitigate the over - reliance problem .", "forward": false, "src_ids": "2022.findings-acl.247_121"}
+{"input": "disentangled regularization is used for Task| context: pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks . however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime .", "entity": "disentangled regularization", "output": "over - reliance problem", "neg_sample": ["disentangled regularization is used for Task", "pretrained language models ( plms ) trained on large - scale unlabeled corpus are typically fine - tuned on task - specific downstream datasets , which have produced state - of - the - art results on various nlp tasks .", "however , the data discrepancy issue in domain and scale makes fine - tuning fail to efficiently capture task - specific patterns , especially in low data regime ."], "relation": "used for", "id": "2022.findings-acl.247", "year": 2022, "rel_sent": "For a given task , we introduce a learnable confidence model to detect indicative guidance from context , and further propose a disentangled regularization to mitigate the over - reliance problem .", "forward": true, "src_ids": "2022.findings-acl.247_122"}
+{"input": "si and sg tasks is done by using Task| context: simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions . previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive . in recent years , pre - trained language models ( plms ) based approaches have become the de - facto standard in nlp since they learn generic knowledge from a large corpus . the knowledge embedded in plms may be useful for si and sg tasks . nevertheless , there are few works to explore it .", "entity": "si and sg tasks", "output": "probing simile knowledge", "neg_sample": ["si and sg tasks is done by using Task", "simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions .", "previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive .", "in recent years , pre - trained language models ( plms ) based approaches have become the de - facto standard in nlp since they learn generic knowledge from a large corpus .", "the knowledge embedded in plms may be useful for si and sg tasks .", "nevertheless , there are few works to explore it ."], "relation": "used for", "id": "2022.acl-long.404", "year": 2022, "rel_sent": "In this paper , we probe simile knowledge from PLMs to solve the SI and SG tasks in the unified framework of simile triple completion for the first time .", "forward": false, "src_ids": "2022.acl-long.404_123"}
+{"input": "si and sg tasks is done by using Method| context: simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions . previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive . nevertheless , there are few works to explore it .", "entity": "si and sg tasks", "output": "pre - trained language models", "neg_sample": ["si and sg tasks is done by using Method", "simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions .", "previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive .", "nevertheless , there are few works to explore it ."], "relation": "used for", "id": "2022.acl-long.404", "year": 2022, "rel_sent": "In this paper , we probe simile knowledge from PLMs to solve the SI and SG tasks in the unified framework of simile triple completion for the first time .", "forward": false, "src_ids": "2022.acl-long.404_124"}
+{"input": "probing simile knowledge is used for Task| context: simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions . previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive . in recent years , pre - trained language models ( plms ) based approaches have become the de - facto standard in nlp since they learn generic knowledge from a large corpus . nevertheless , there are few works to explore it .", "entity": "probing simile knowledge", "output": "si and sg tasks", "neg_sample": ["probing simile knowledge is used for Task", "simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions .", "previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive .", "in recent years , pre - trained language models ( plms ) based approaches have become the de - facto standard in nlp since they learn generic knowledge from a large corpus .", "nevertheless , there are few works to explore it ."], "relation": "used for", "id": "2022.acl-long.404", "year": 2022, "rel_sent": "In this paper , we probe simile knowledge from PLMs to solve the SI and SG tasks in the unified framework of simile triple completion for the first time .", "forward": true, "src_ids": "2022.acl-long.404_125"}
+{"input": "pre - trained language models is used for Task| context: simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions . previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive . in recent years , pre - trained language models ( plms ) based approaches have become the de - facto standard in nlp since they learn generic knowledge from a large corpus . nevertheless , there are few works to explore it .", "entity": "pre - trained language models", "output": "si and sg tasks", "neg_sample": ["pre - trained language models is used for Task", "simile interpretation ( si ) and simile generation ( sg ) are challenging tasks for nlp because models require adequate world knowledge to produce predictions .", "previous works have employed many hand - crafted resources to bring knowledge - related into models , which is time - consuming and labor - intensive .", "in recent years , pre - trained language models ( plms ) based approaches have become the de - facto standard in nlp since they learn generic knowledge from a large corpus .", "nevertheless , there are few works to explore it ."], "relation": "used for", "id": "2022.acl-long.404", "year": 2022, "rel_sent": "In this paper , we probe simile knowledge from PLMs to solve the SI and SG tasks in the unified framework of simile triple completion for the first time .", "forward": true, "src_ids": "2022.acl-long.404_126"}
+{"input": "shared knowledge is done by using Method| context: cognitively plausible visual dialogue models should keep a mental scoreboard of shared established facts in the dialogue context .", "entity": "shared knowledge", "output": "dialogue representations", "neg_sample": ["shared knowledge is done by using Method", "cognitively plausible visual dialogue models should keep a mental scoreboard of shared established facts in the dialogue context ."], "relation": "used for", "id": "2022.acl-short.73", "year": 2022, "rel_sent": "Can Visual Dialogue Models Do Scorekeeping ? Exploring How Dialogue Representations Incrementally Encode Shared Knowledge.", "forward": false, "src_ids": "2022.acl-short.73_127"}
+{"input": "dialogue representations is used for OtherScientificTerm| context: cognitively plausible visual dialogue models should keep a mental scoreboard of shared established facts in the dialogue context .", "entity": "dialogue representations", "output": "shared knowledge", "neg_sample": ["dialogue representations is used for OtherScientificTerm", "cognitively plausible visual dialogue models should keep a mental scoreboard of shared established facts in the dialogue context ."], "relation": "used for", "id": "2022.acl-short.73", "year": 2022, "rel_sent": "Can Visual Dialogue Models Do Scorekeeping ? Exploring How Dialogue Representations Incrementally Encode Shared Knowledge.", "forward": true, "src_ids": "2022.acl-short.73_128"}
+{"input": "anaphoric expressions is done by using Task| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "anaphoric expressions", "output": "human reading process", "neg_sample": ["anaphoric expressions is done by using Task", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "In this paper , we imitate the human reading process in connecting the anaphoric expressions and explicitly leverage the coreference information of the entities to enhance the word embeddings from the pre - trained language model , in order to highlight the coreference mentions of the entities that must be identified for coreference - intensive question answering in QUOREF , a relatively new dataset that is specifically designed to evaluate the coreference - related performance of a model .", "forward": false, "src_ids": "2022.acl-long.91_129"}
+{"input": "human reading process is used for OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "human reading process", "output": "anaphoric expressions", "neg_sample": ["human reading process is used for OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "In this paper , we imitate the human reading process in connecting the anaphoric expressions and explicitly leverage the coreference information of the entities to enhance the word embeddings from the pre - trained language model , in order to highlight the coreference mentions of the entities that must be identified for coreference - intensive question answering in QUOREF , a relatively new dataset that is specifically designed to evaluate the coreference - related performance of a model .", "forward": true, "src_ids": "2022.acl-long.91_130"}
+{"input": "word embeddings is done by using OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "word embeddings", "output": "coreference information", "neg_sample": ["word embeddings is done by using OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "In this paper , we imitate the human reading process in connecting the anaphoric expressions and explicitly leverage the coreference information of the entities to enhance the word embeddings from the pre - trained language model , in order to highlight the coreference mentions of the entities that must be identified for coreference - intensive question answering in QUOREF , a relatively new dataset that is specifically designed to evaluate the coreference - related performance of a model .", "forward": false, "src_ids": "2022.acl-long.91_131"}
+{"input": "language model is done by using OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "language model", "output": "coreference information", "neg_sample": ["language model is done by using OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We demonstrate that the explicit incorporation of coreference information in the fine - tuning stage performs better than the incorporation of the coreference information in pre - training a language model .", "forward": false, "src_ids": "2022.acl-long.91_132"}
+{"input": "fine - tuning stage is done by using OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "fine - tuning stage", "output": "coreference information", "neg_sample": ["fine - tuning stage is done by using OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We demonstrate that the explicit incorporation of coreference information in the fine - tuning stage performs better than the incorporation of the coreference information in pre - training a language model .", "forward": false, "src_ids": "2022.acl-long.91_133"}
+{"input": "coreference information is used for OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "coreference information", "output": "word embeddings", "neg_sample": ["coreference information is used for OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "In this paper , we imitate the human reading process in connecting the anaphoric expressions and explicitly leverage the coreference information of the entities to enhance the word embeddings from the pre - trained language model , in order to highlight the coreference mentions of the entities that must be identified for coreference - intensive question answering in QUOREF , a relatively new dataset that is specifically designed to evaluate the coreference - related performance of a model .", "forward": true, "src_ids": "2022.acl-long.91_134"}
+{"input": "encoder layer is used for OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "encoder layer", "output": "coreference mentions", "neg_sample": ["encoder layer is used for OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We use two strategies tofine - tune a pre - trained language model , namely , placing an additional encoder layer after a pre - trained language model tofocus on the coreference mentions or constructing a relational graph convolutional network to model the coreference relations .", "forward": true, "src_ids": "2022.acl-long.91_135"}
+{"input": "coreference mentions is done by using OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "coreference mentions", "output": "encoder layer", "neg_sample": ["coreference mentions is done by using OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We use two strategies tofine - tune a pre - trained language model , namely , placing an additional encoder layer after a pre - trained language model tofocus on the coreference mentions or constructing a relational graph convolutional network to model the coreference relations .", "forward": false, "src_ids": "2022.acl-long.91_136"}
+{"input": "coreference relations is done by using Method| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "coreference relations", "output": "relational graph convolutional network", "neg_sample": ["coreference relations is done by using Method", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We use two strategies tofine - tune a pre - trained language model , namely , placing an additional encoder layer after a pre - trained language model tofocus on the coreference mentions or constructing a relational graph convolutional network to model the coreference relations .", "forward": false, "src_ids": "2022.acl-long.91_137"}
+{"input": "relational graph convolutional network is used for OtherScientificTerm| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "relational graph convolutional network", "output": "coreference relations", "neg_sample": ["relational graph convolutional network is used for OtherScientificTerm", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We use two strategies tofine - tune a pre - trained language model , namely , placing an additional encoder layer after a pre - trained language model tofocus on the coreference mentions or constructing a relational graph convolutional network to model the coreference relations .", "forward": true, "src_ids": "2022.acl-long.91_138"}
+{"input": "coreference information is used for Task| context: machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models .", "entity": "coreference information", "output": "fine - tuning stage", "neg_sample": ["coreference information is used for Task", "machine reading comprehension is a heavily - studied research and test field for evaluating new pre - trained language models ( prlms ) and fine - tuning strategies , and recent studies have enriched the pre - trained language models with syntactic , semantic and other linguistic information to improve the performance of the models ."], "relation": "used for", "id": "2022.acl-long.91", "year": 2022, "rel_sent": "We demonstrate that the explicit incorporation of coreference information in the fine - tuning stage performs better than the incorporation of the coreference information in pre - training a language model .", "forward": true, "src_ids": "2022.acl-long.91_139"}
+{"input": "composition is used for Task| context: multilingual pre - trained language models , such as mbert and xlm - r , have shown impressive cross - lingual ability . surprisingly , both of them use multilingual masked language model ( mlm ) without any cross - lingual supervision or aligned data . despite the encouraging results , we still lack a clear understanding of why cross - lingual ability could emerge from multilingual mlm . in our work , we argue that cross - language ability comes from the commonality between languages .", "entity": "composition", "output": "cross - linguistic transfer", "neg_sample": ["composition is used for Task", "multilingual pre - trained language models , such as mbert and xlm - r , have shown impressive cross - lingual ability .", "surprisingly , both of them use multilingual masked language model ( mlm ) without any cross - lingual supervision or aligned data .", "despite the encouraging results , we still lack a clear understanding of why cross - lingual ability could emerge from multilingual mlm .", "in our work , we argue that cross - language ability comes from the commonality between languages ."], "relation": "used for", "id": "2022.acl-long.322", "year": 2022, "rel_sent": "Our main conclusion is that the contribution of constituent order and word co - occurrence is limited , while the composition is more crucial to the success of cross - linguistic transfer .", "forward": true, "src_ids": "2022.acl-long.322_140"}
+{"input": "cross - linguistic transfer is done by using OtherScientificTerm| context: multilingual pre - trained language models , such as mbert and xlm - r , have shown impressive cross - lingual ability . surprisingly , both of them use multilingual masked language model ( mlm ) without any cross - lingual supervision or aligned data . despite the encouraging results , we still lack a clear understanding of why cross - lingual ability could emerge from multilingual mlm . in our work , we argue that cross - language ability comes from the commonality between languages .", "entity": "cross - linguistic transfer", "output": "composition", "neg_sample": ["cross - linguistic transfer is done by using OtherScientificTerm", "multilingual pre - trained language models , such as mbert and xlm - r , have shown impressive cross - lingual ability .", "surprisingly , both of them use multilingual masked language model ( mlm ) without any cross - lingual supervision or aligned data .", "despite the encouraging results , we still lack a clear understanding of why cross - lingual ability could emerge from multilingual mlm .", "in our work , we argue that cross - language ability comes from the commonality between languages ."], "relation": "used for", "id": "2022.acl-long.322", "year": 2022, "rel_sent": "Our main conclusion is that the contribution of constituent order and word co - occurrence is limited , while the composition is more crucial to the success of cross - linguistic transfer .", "forward": false, "src_ids": "2022.acl-long.322_141"}
+{"input": "emotional classification is done by using Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "emotional classification", "output": "pre - trained model", "neg_sample": ["emotional classification is done by using Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "Continuing Pre - trained Model with Multiple Training Strategies for Emotional Classification.", "forward": false, "src_ids": "2022.wassa-1.22_142"}
+{"input": "pre - trained model is used for Task| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "pre - trained model", "output": "emotional classification", "neg_sample": ["pre - trained model is used for Task", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "Continuing Pre - trained Model with Multiple Training Strategies for Emotional Classification.", "forward": true, "src_ids": "2022.wassa-1.22_143"}
+{"input": "training strategies is used for Task| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "training strategies", "output": "emotional classification", "neg_sample": ["training strategies is used for Task", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "Continuing Pre - trained Model with Multiple Training Strategies for Emotional Classification.", "forward": true, "src_ids": "2022.wassa-1.22_144"}
+{"input": "lingjing team 's method is used for Task| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "lingjing team 's method", "output": "emotional classification", "neg_sample": ["lingjing team 's method is used for Task", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the contribution of the LingJing team 's method to the Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis ( WASSA ) 2022 shared task on Emotion Classification .", "forward": true, "src_ids": "2022.wassa-1.22_145"}
+{"input": "emotional classification is done by using Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "emotional classification", "output": "lingjing team 's method", "neg_sample": ["emotional classification is done by using Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the contribution of the LingJing team 's method to the Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis ( WASSA ) 2022 shared task on Emotion Classification .", "forward": false, "src_ids": "2022.wassa-1.22_146"}
+{"input": "computational approaches is done by using Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "computational approaches", "output": "lingjing team 's method", "neg_sample": ["computational approaches is done by using Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the contribution of the LingJing team 's method to the Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis ( WASSA ) 2022 shared task on Emotion Classification .", "forward": false, "src_ids": "2022.wassa-1.22_147"}
+{"input": "subjectivity is done by using Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "subjectivity", "output": "computational approaches", "neg_sample": ["subjectivity is done by using Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the contribution of the LingJing team 's method to the Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis ( WASSA ) 2022 shared task on Emotion Classification .", "forward": false, "src_ids": "2022.wassa-1.22_148"}
+{"input": "lingjing team 's method is used for Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "lingjing team 's method", "output": "computational approaches", "neg_sample": ["lingjing team 's method is used for Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the contribution of the LingJing team 's method to the Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis ( WASSA ) 2022 shared task on Emotion Classification .", "forward": true, "src_ids": "2022.wassa-1.22_149"}
+{"input": "computational approaches is used for OtherScientificTerm| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "computational approaches", "output": "subjectivity", "neg_sample": ["computational approaches is used for OtherScientificTerm", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the contribution of the LingJing team 's method to the Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis ( WASSA ) 2022 shared task on Emotion Classification .", "forward": true, "src_ids": "2022.wassa-1.22_150"}
+{"input": "deberta pre - trained language model is done by using Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "deberta pre - trained language model", "output": "continual pre - training method", "neg_sample": ["deberta pre - trained language model is done by using Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the continual pre - training method for the masked language model ( MLM ) to enhance the DeBERTa pre - trained language model .", "forward": false, "src_ids": "2022.wassa-1.22_151"}
+{"input": "masked language model ( mlm ) is done by using Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "masked language model ( mlm )", "output": "continual pre - training method", "neg_sample": ["masked language model ( mlm ) is done by using Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the continual pre - training method for the masked language model ( MLM ) to enhance the DeBERTa pre - trained language model .", "forward": false, "src_ids": "2022.wassa-1.22_152"}
+{"input": "continual pre - training method is used for Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "continual pre - training method", "output": "masked language model ( mlm )", "neg_sample": ["continual pre - training method is used for Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the continual pre - training method for the masked language model ( MLM ) to enhance the DeBERTa pre - trained language model .", "forward": true, "src_ids": "2022.wassa-1.22_153"}
+{"input": "masked language model ( mlm ) is used for Method| context: emotion is the essential attribute of human beings . perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence .", "entity": "masked language model ( mlm )", "output": "deberta pre - trained language model", "neg_sample": ["masked language model ( mlm ) is used for Method", "emotion is the essential attribute of human beings .", "perceiving and understanding emotions in a human - like manner is the most central part of developing emotional intelligence ."], "relation": "used for", "id": "2022.wassa-1.22", "year": 2022, "rel_sent": "This paper describes the continual pre - training method for the masked language model ( MLM ) to enhance the DeBERTa pre - trained language model .", "forward": true, "src_ids": "2022.wassa-1.22_154"}
+{"input": "document - level eae is done by using OtherScientificTerm| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "entity": "document - level eae", "output": "syntactic structures", "neg_sample": ["document - level eae is done by using OtherScientificTerm", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Hence , in this work , we study the importance of syntactic structures in document - level EAE .", "forward": false, "src_ids": "2022.findings-acl.130_155"}
+{"input": "syntactic structures is used for Task| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "entity": "syntactic structures", "output": "document - level eae", "neg_sample": ["syntactic structures is used for Task", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Hence , in this work , we study the importance of syntactic structures in document - level EAE .", "forward": true, "src_ids": "2022.findings-acl.130_156"}
+{"input": "structures of documents is done by using Method| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored . in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents .", "entity": "structures of documents", "output": "optimal transport ( ot )", "neg_sample": ["structures of documents is done by using Method", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Specifically , we propose to employ Optimal Transport ( OT ) to induce structures of documents based on sentence - level syntactic structures and tailored to EAE task .", "forward": false, "src_ids": "2022.findings-acl.130_157"}
+{"input": "eae task is done by using Method| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored . in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents .", "entity": "eae task", "output": "optimal transport ( ot )", "neg_sample": ["eae task is done by using Method", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Specifically , we propose to employ Optimal Transport ( OT ) to induce structures of documents based on sentence - level syntactic structures and tailored to EAE task .", "forward": false, "src_ids": "2022.findings-acl.130_158"}
+{"input": "optimal transport ( ot ) is used for OtherScientificTerm| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored . in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents .", "entity": "optimal transport ( ot )", "output": "structures of documents", "neg_sample": ["optimal transport ( ot ) is used for OtherScientificTerm", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Specifically , we propose to employ Optimal Transport ( OT ) to induce structures of documents based on sentence - level syntactic structures and tailored to EAE task .", "forward": true, "src_ids": "2022.findings-acl.130_159"}
+{"input": "optimal transport ( ot ) is used for Task| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored . in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents .", "entity": "optimal transport ( ot )", "output": "eae task", "neg_sample": ["optimal transport ( ot ) is used for Task", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Specifically , we propose to employ Optimal Transport ( OT ) to induce structures of documents based on sentence - level syntactic structures and tailored to EAE task .", "forward": true, "src_ids": "2022.findings-acl.130_160"}
+{"input": "unrelated context words is done by using Method| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored . in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents .", "entity": "unrelated context words", "output": "regularization technique", "neg_sample": ["unrelated context words is done by using Method", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Furthermore , we propose a novel regularization technique to explicitly constrain the contributions of unrelated context words in the final prediction for EAE .", "forward": false, "src_ids": "2022.findings-acl.130_161"}
+{"input": "regularization technique is used for OtherScientificTerm| context: event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger . despite the success of prior works in sentence - level eae , the document - level setting is less explored . in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents .", "entity": "regularization technique", "output": "unrelated context words", "neg_sample": ["regularization technique is used for OtherScientificTerm", "event argument extraction ( eae ) is one of the sub - tasks of event extraction , aiming to recognize the role of each entity mention toward a specific event trigger .", "despite the success of prior works in sentence - level eae , the document - level setting is less explored .", "in particular , whereas syntactic structures of sentences have been shown to be effective for sentence - level eae , prior document - level eae models totally ignore syntactic structures for documents ."], "relation": "used for", "id": "2022.findings-acl.130", "year": 2022, "rel_sent": "Furthermore , we propose a novel regularization technique to explicitly constrain the contributions of unrelated context words in the final prediction for EAE .", "forward": true, "src_ids": "2022.findings-acl.130_162"}
+{"input": "neural machine translation is done by using OtherScientificTerm| context: lexically constrained neural machine translation ( nmt ) , which controls the generation of nmt models with pre - specified constraints , is important in many practical scenarios . due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box .", "entity": "neural machine translation", "output": "vectorized lexical constraints", "neg_sample": ["neural machine translation is done by using OtherScientificTerm", "lexically constrained neural machine translation ( nmt ) , which controls the generation of nmt models with pre - specified constraints , is important in many practical scenarios .", "due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box ."], "relation": "used for", "id": "2022.acl-long.487", "year": 2022, "rel_sent": "Integrating Vectorized Lexical Constraints for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.487_163"}
+{"input": "vectorized lexical constraints is used for Task| context: due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box .", "entity": "vectorized lexical constraints", "output": "neural machine translation", "neg_sample": ["vectorized lexical constraints is used for Task", "due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box ."], "relation": "used for", "id": "2022.acl-long.487", "year": 2022, "rel_sent": "Integrating Vectorized Lexical Constraints for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.487_164"}
+{"input": "modeling constraint pairs is done by using Method| context: lexically constrained neural machine translation ( nmt ) , which controls the generation of nmt models with pre - specified constraints , is important in many practical scenarios . due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box .", "entity": "modeling constraint pairs", "output": "attention modules", "neg_sample": ["modeling constraint pairs is done by using Method", "lexically constrained neural machine translation ( nmt ) , which controls the generation of nmt models with pre - specified constraints , is important in many practical scenarios .", "due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box ."], "relation": "used for", "id": "2022.acl-long.487", "year": 2022, "rel_sent": "The proposed integration method is based on the assumption that the correspondence between keys and values in attention modules is naturally suitable for modeling constraint pairs .", "forward": false, "src_ids": "2022.acl-long.487_165"}
+{"input": "attention modules is used for Task| context: lexically constrained neural machine translation ( nmt ) , which controls the generation of nmt models with pre - specified constraints , is important in many practical scenarios . due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box .", "entity": "attention modules", "output": "modeling constraint pairs", "neg_sample": ["attention modules is used for Task", "lexically constrained neural machine translation ( nmt ) , which controls the generation of nmt models with pre - specified constraints , is important in many practical scenarios .", "due to the representation gap between discrete constraints and continuous vectors in nmt models , most existing works choose to construct synthetic data or modify the decoding algorithm to impose lexical constraints , treating the nmt model as a black box ."], "relation": "used for", "id": "2022.acl-long.487", "year": 2022, "rel_sent": "The proposed integration method is based on the assumption that the correspondence between keys and values in attention modules is naturally suitable for modeling constraint pairs .", "forward": true, "src_ids": "2022.acl-long.487_166"}
+{"input": "emotion cause pair extraction is done by using Method| context: we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset . existing methods have set a fixed size window to capture relations between neighboring clauses . however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data .", "entity": "emotion cause pair extraction", "output": "multi - granularity semantic aware graph model", "neg_sample": ["emotion cause pair extraction is done by using Method", "we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset .", "existing methods have set a fixed size window to capture relations between neighboring clauses .", "however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data ."], "relation": "used for", "id": "2022.findings-acl.95", "year": 2022, "rel_sent": "Multi - Granularity Semantic Aware Graph Model for Reducing Position Bias in Emotion Cause Pair Extraction.", "forward": false, "src_ids": "2022.findings-acl.95_167"}
+{"input": "multi - granularity semantic aware graph model is used for Task| context: we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset . existing methods have set a fixed size window to capture relations between neighboring clauses . however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data .", "entity": "multi - granularity semantic aware graph model", "output": "emotion cause pair extraction", "neg_sample": ["multi - granularity semantic aware graph model is used for Task", "we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset .", "existing methods have set a fixed size window to capture relations between neighboring clauses .", "however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data ."], "relation": "used for", "id": "2022.findings-acl.95", "year": 2022, "rel_sent": "Multi - Granularity Semantic Aware Graph Model for Reducing Position Bias in Emotion Cause Pair Extraction.", "forward": true, "src_ids": "2022.findings-acl.95_168"}
+{"input": "fine - grained semantic features is done by using OtherScientificTerm| context: we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset . existing methods have set a fixed size window to capture relations between neighboring clauses . however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data .", "entity": "fine - grained semantic features", "output": "keywords", "neg_sample": ["fine - grained semantic features is done by using OtherScientificTerm", "we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset .", "existing methods have set a fixed size window to capture relations between neighboring clauses .", "however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data ."], "relation": "used for", "id": "2022.findings-acl.95", "year": 2022, "rel_sent": "In particular , we first explore semantic dependencies between clauses and keywords extracted from the document that convey fine - grained semantic features , obtaining keywords enhanced clause representations .", "forward": false, "src_ids": "2022.findings-acl.95_169"}
+{"input": "keywords is used for OtherScientificTerm| context: we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset . existing methods have set a fixed size window to capture relations between neighboring clauses . however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data .", "entity": "keywords", "output": "fine - grained semantic features", "neg_sample": ["keywords is used for OtherScientificTerm", "we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset .", "existing methods have set a fixed size window to capture relations between neighboring clauses .", "however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data ."], "relation": "used for", "id": "2022.findings-acl.95", "year": 2022, "rel_sent": "In particular , we first explore semantic dependencies between clauses and keywords extracted from the document that convey fine - grained semantic features , obtaining keywords enhanced clause representations .", "forward": true, "src_ids": "2022.findings-acl.95_170"}
+{"input": "coarse - grained semantic relations is done by using Method| context: we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset . existing methods have set a fixed size window to capture relations between neighboring clauses . however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data .", "entity": "coarse - grained semantic relations", "output": "clause graph", "neg_sample": ["coarse - grained semantic relations is done by using Method", "we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset .", "existing methods have set a fixed size window to capture relations between neighboring clauses .", "however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data ."], "relation": "used for", "id": "2022.findings-acl.95", "year": 2022, "rel_sent": "Besides , a clause graph is also established to model coarse - grained semantic relations between clauses .", "forward": false, "src_ids": "2022.findings-acl.95_171"}
+{"input": "clause graph is used for OtherScientificTerm| context: we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset . existing methods have set a fixed size window to capture relations between neighboring clauses . however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data .", "entity": "clause graph", "output": "coarse - grained semantic relations", "neg_sample": ["clause graph is used for OtherScientificTerm", "we observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ecpe dataset .", "existing methods have set a fixed size window to capture relations between neighboring clauses .", "however , they neglect the effective semantic connections between distant clauses , leading to poor generalization ability towards position - insensitive data ."], "relation": "used for", "id": "2022.findings-acl.95", "year": 2022, "rel_sent": "Besides , a clause graph is also established to model coarse - grained semantic relations between clauses .", "forward": true, "src_ids": "2022.findings-acl.95_172"}
+{"input": "prompt - based few - shot text classification is done by using Method| context: while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored .", "entity": "prompt - based few - shot text classification", "output": "explicit and implicit consistency regularization enhanced language model", "neg_sample": ["prompt - based few - shot text classification is done by using Method", "while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored ."], "relation": "used for", "id": "2022.findings-acl.283", "year": 2022, "rel_sent": "By employing both explicit and implicit consistency regularization , EICO advances the performance of prompt - based few - shot text classification .", "forward": false, "src_ids": "2022.findings-acl.283_173"}
+{"input": "explicit and implicit consistency regularization is used for Method| context: while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored .", "entity": "explicit and implicit consistency regularization", "output": "explicit and implicit consistency regularization enhanced language model", "neg_sample": ["explicit and implicit consistency regularization is used for Method", "while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored ."], "relation": "used for", "id": "2022.findings-acl.283", "year": 2022, "rel_sent": "By employing both explicit and implicit consistency regularization , EICO advances the performance of prompt - based few - shot text classification .", "forward": true, "src_ids": "2022.findings-acl.283_174"}
+{"input": "self - training is done by using Method| context: while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored .", "entity": "self - training", "output": "explicit and implicit consistency regularization", "neg_sample": ["self - training is done by using Method", "while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored ."], "relation": "used for", "id": "2022.findings-acl.283", "year": 2022, "rel_sent": "This work revisits the consistency regularization in self - training and presents explicit and implicit consistency regularization enhanced language model ( EICO ) .", "forward": false, "src_ids": "2022.findings-acl.283_175"}
+{"input": "explicit and implicit consistency regularization enhanced language model is done by using Method| context: while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored .", "entity": "explicit and implicit consistency regularization enhanced language model", "output": "explicit and implicit consistency regularization", "neg_sample": ["explicit and implicit consistency regularization enhanced language model is done by using Method", "while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored ."], "relation": "used for", "id": "2022.findings-acl.283", "year": 2022, "rel_sent": "By employing both explicit and implicit consistency regularization , EICO advances the performance of prompt - based few - shot text classification .", "forward": false, "src_ids": "2022.findings-acl.283_176"}
+{"input": "explicit and implicit consistency regularization enhanced language model is used for Task| context: while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored .", "entity": "explicit and implicit consistency regularization enhanced language model", "output": "prompt - based few - shot text classification", "neg_sample": ["explicit and implicit consistency regularization enhanced language model is used for Task", "while the prompt - based fine - tuning methods had advanced few - shot natural language understanding tasks , self - training methods are also being explored ."], "relation": "used for", "id": "2022.findings-acl.283", "year": 2022, "rel_sent": "By employing both explicit and implicit consistency regularization , EICO advances the performance of prompt - based few - shot text classification .", "forward": true, "src_ids": "2022.findings-acl.283_177"}
+{"input": "nlms is used for OtherScientificTerm| context: to apply a similar approach to analyze neural language models ( nlm ) , it is first necessary to establish that different models are similar enough in the generalizations they make .", "entity": "nlms", "output": "linguistic phenomena", "neg_sample": ["nlms is used for OtherScientificTerm", "to apply a similar approach to analyze neural language models ( nlm ) , it is first necessary to establish that different models are similar enough in the generalizations they make ."], "relation": "used for", "id": "2022.acl-long.568", "year": 2022, "rel_sent": "In this paper , we show that NLMs with different initialization , architecture , and training data acquire linguistic phenomena in a similar order , despite their different end performance .", "forward": true, "src_ids": "2022.acl-long.568_178"}
+{"input": "linguistic phenomena is done by using Method| context: the learning trajectories of linguistic phenomena in humans provide insight into linguistic representation , beyond what can be gleaned from inspecting the behavior of an adult speaker . to apply a similar approach to analyze neural language models ( nlm ) , it is first necessary to establish that different models are similar enough in the generalizations they make .", "entity": "linguistic phenomena", "output": "nlms", "neg_sample": ["linguistic phenomena is done by using Method", "the learning trajectories of linguistic phenomena in humans provide insight into linguistic representation , beyond what can be gleaned from inspecting the behavior of an adult speaker .", "to apply a similar approach to analyze neural language models ( nlm ) , it is first necessary to establish that different models are similar enough in the generalizations they make ."], "relation": "used for", "id": "2022.acl-long.568", "year": 2022, "rel_sent": "In this paper , we show that NLMs with different initialization , architecture , and training data acquire linguistic phenomena in a similar order , despite their different end performance .", "forward": false, "src_ids": "2022.acl-long.568_179"}
+{"input": "multilingual text style transfer is done by using Task| context: we exploit the pre - trained seq2seq model mbart for multilingual text style transfer .", "entity": "multilingual text style transfer", "output": "language and task adaptation", "neg_sample": ["multilingual text style transfer is done by using Task", "we exploit the pre - trained seq2seq model mbart for multilingual text style transfer ."], "relation": "used for", "id": "2022.acl-short.29", "year": 2022, "rel_sent": "Multilingual Pre - training with Language and Task Adaptation for Multilingual Text Style Transfer.", "forward": false, "src_ids": "2022.acl-short.29_180"}
+{"input": "multilingual formality transfer is done by using Method| context: we exploit the pre - trained seq2seq model mbart for multilingual text style transfer .", "entity": "multilingual formality transfer", "output": "modular approach", "neg_sample": ["multilingual formality transfer is done by using Method", "we exploit the pre - trained seq2seq model mbart for multilingual text style transfer ."], "relation": "used for", "id": "2022.acl-short.29", "year": 2022, "rel_sent": "Besides , in view of the general scarcity of parallel data , we propose a modular approach for multilingual formality transfer , which consists of two training strategies that target adaptation to both language and task .", "forward": false, "src_ids": "2022.acl-short.29_181"}
+{"input": "modular approach is used for Task| context: we exploit the pre - trained seq2seq model mbart for multilingual text style transfer .", "entity": "modular approach", "output": "multilingual formality transfer", "neg_sample": ["modular approach is used for Task", "we exploit the pre - trained seq2seq model mbart for multilingual text style transfer ."], "relation": "used for", "id": "2022.acl-short.29", "year": 2022, "rel_sent": "Besides , in view of the general scarcity of parallel data , we propose a modular approach for multilingual formality transfer , which consists of two training strategies that target adaptation to both language and task .", "forward": true, "src_ids": "2022.acl-short.29_182"}
+{"input": "adaptation is done by using Method| context: we exploit the pre - trained seq2seq model mbart for multilingual text style transfer .", "entity": "adaptation", "output": "training strategies", "neg_sample": ["adaptation is done by using Method", "we exploit the pre - trained seq2seq model mbart for multilingual text style transfer ."], "relation": "used for", "id": "2022.acl-short.29", "year": 2022, "rel_sent": "Besides , in view of the general scarcity of parallel data , we propose a modular approach for multilingual formality transfer , which consists of two training strategies that target adaptation to both language and task .", "forward": false, "src_ids": "2022.acl-short.29_183"}
+{"input": "training strategies is used for Task| context: we exploit the pre - trained seq2seq model mbart for multilingual text style transfer .", "entity": "training strategies", "output": "adaptation", "neg_sample": ["training strategies is used for Task", "we exploit the pre - trained seq2seq model mbart for multilingual text style transfer ."], "relation": "used for", "id": "2022.acl-short.29", "year": 2022, "rel_sent": "Besides , in view of the general scarcity of parallel data , we propose a modular approach for multilingual formality transfer , which consists of two training strategies that target adaptation to both language and task .", "forward": true, "src_ids": "2022.acl-short.29_184"}
+{"input": "deobfuscation is done by using Task| context: recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution . to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches . however , existing authorship obfuscation approaches do not consider the adversarial threat model .", "entity": "deobfuscation", "output": "adversarially trained authorship attributors", "neg_sample": ["deobfuscation is done by using Task", "recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution .", "to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches .", "however , existing authorship obfuscation approaches do not consider the adversarial threat model ."], "relation": "used for", "id": "2022.acl-long.509", "year": 2022, "rel_sent": "Adversarial Authorship Attribution for Deobfuscation.", "forward": false, "src_ids": "2022.acl-long.509_185"}
+{"input": "obfuscators is done by using Task| context: recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution . to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches . however , existing authorship obfuscation approaches do not consider the adversarial threat model .", "entity": "obfuscators", "output": "adversarially trained authorship attributors", "neg_sample": ["obfuscators is done by using Task", "recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution .", "to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches .", "however , existing authorship obfuscation approaches do not consider the adversarial threat model ."], "relation": "used for", "id": "2022.acl-long.509", "year": 2022, "rel_sent": "We show that adversarially trained authorship attributors are able to degrade the effectiveness of existing obfuscators from 20 - 30 % to 5 - 10 % .", "forward": false, "src_ids": "2022.acl-long.509_186"}
+{"input": "adversarially trained authorship attributors is used for Task| context: recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution . to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches . however , existing authorship obfuscation approaches do not consider the adversarial threat model . specifically , they are not evaluated against adversarially trained authorship attributors that are aware of potential obfuscation .", "entity": "adversarially trained authorship attributors", "output": "deobfuscation", "neg_sample": ["adversarially trained authorship attributors is used for Task", "recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution .", "to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches .", "however , existing authorship obfuscation approaches do not consider the adversarial threat model .", "specifically , they are not evaluated against adversarially trained authorship attributors that are aware of potential obfuscation ."], "relation": "used for", "id": "2022.acl-long.509", "year": 2022, "rel_sent": "Adversarial Authorship Attribution for Deobfuscation.", "forward": true, "src_ids": "2022.acl-long.509_187"}
+{"input": "adversarially trained authorship attributors is used for Method| context: recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution . to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches . however , existing authorship obfuscation approaches do not consider the adversarial threat model . specifically , they are not evaluated against adversarially trained authorship attributors that are aware of potential obfuscation .", "entity": "adversarially trained authorship attributors", "output": "obfuscators", "neg_sample": ["adversarially trained authorship attributors is used for Method", "recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution .", "to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches .", "however , existing authorship obfuscation approaches do not consider the adversarial threat model .", "specifically , they are not evaluated against adversarially trained authorship attributors that are aware of potential obfuscation ."], "relation": "used for", "id": "2022.acl-long.509", "year": 2022, "rel_sent": "We show that adversarially trained authorship attributors are able to degrade the effectiveness of existing obfuscators from 20 - 30 % to 5 - 10 % .", "forward": true, "src_ids": "2022.acl-long.509_188"}
+{"input": "knowledge storage and evidence generation is done by using Method| context: question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents . however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task .", "entity": "knowledge storage and evidence generation", "output": "child - tuning approach", "neg_sample": ["knowledge storage and evidence generation is done by using Method", "question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents .", "however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task ."], "relation": "used for", "id": "2022.dialdoc-1.14", "year": 2022, "rel_sent": "With the Child - Tuning approach being designed , the knowledge storage and evidence generation avoid catastrophic forgetting for response generation .", "forward": false, "src_ids": "2022.dialdoc-1.14_189"}
+{"input": "response generation is done by using Task| context: question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents . however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task .", "entity": "response generation", "output": "knowledge storage and evidence generation", "neg_sample": ["response generation is done by using Task", "question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents .", "however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task ."], "relation": "used for", "id": "2022.dialdoc-1.14", "year": 2022, "rel_sent": "With the Child - Tuning approach being designed , the knowledge storage and evidence generation avoid catastrophic forgetting for response generation .", "forward": false, "src_ids": "2022.dialdoc-1.14_190"}
+{"input": "catastrophic forgetting is done by using Task| context: question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents . however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task .", "entity": "catastrophic forgetting", "output": "knowledge storage and evidence generation", "neg_sample": ["catastrophic forgetting is done by using Task", "question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents .", "however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task ."], "relation": "used for", "id": "2022.dialdoc-1.14", "year": 2022, "rel_sent": "With the Child - Tuning approach being designed , the knowledge storage and evidence generation avoid catastrophic forgetting for response generation .", "forward": false, "src_ids": "2022.dialdoc-1.14_191"}
+{"input": "child - tuning approach is used for Task| context: question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents . however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task .", "entity": "child - tuning approach", "output": "knowledge storage and evidence generation", "neg_sample": ["child - tuning approach is used for Task", "question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents .", "however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task ."], "relation": "used for", "id": "2022.dialdoc-1.14", "year": 2022, "rel_sent": "With the Child - Tuning approach being designed , the knowledge storage and evidence generation avoid catastrophic forgetting for response generation .", "forward": true, "src_ids": "2022.dialdoc-1.14_192"}
+{"input": "knowledge storage and evidence generation is used for OtherScientificTerm| context: question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents . however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task .", "entity": "knowledge storage and evidence generation", "output": "catastrophic forgetting", "neg_sample": ["knowledge storage and evidence generation is used for OtherScientificTerm", "question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents .", "however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task ."], "relation": "used for", "id": "2022.dialdoc-1.14", "year": 2022, "rel_sent": "With the Child - Tuning approach being designed , the knowledge storage and evidence generation avoid catastrophic forgetting for response generation .", "forward": true, "src_ids": "2022.dialdoc-1.14_193"}
+{"input": "knowledge storage and evidence generation is used for Task| context: question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents . however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task .", "entity": "knowledge storage and evidence generation", "output": "response generation", "neg_sample": ["knowledge storage and evidence generation is used for Task", "question answering ( qa ) is a natural language processing ( nlp ) task that can measure language and semantics understanding ability , it requires a system not only to retrieve relevant documents from a large number of articles but also to answer corresponding questions according to documents .", "however , various language styles and sources of human questions and evidence documents form the different embedding semantic spaces , which may bring some errors to the downstream qa task ."], "relation": "used for", "id": "2022.dialdoc-1.14", "year": 2022, "rel_sent": "With the Child - Tuning approach being designed , the knowledge storage and evidence generation avoid catastrophic forgetting for response generation .", "forward": true, "src_ids": "2022.dialdoc-1.14_194"}
+{"input": "low - resource ner is done by using Method| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "low - resource ner", "output": "masked entity language modeling ( melm )", "neg_sample": ["low - resource ner is done by using Method", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "MELM : Data Augmentation with Masked Entity Language Modeling for Low - Resource NER.", "forward": false, "src_ids": "2022.acl-long.160_195"}
+{"input": "ner is done by using Method| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "ner", "output": "masked entity language modeling ( melm )", "neg_sample": ["ner is done by using Method", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "Thereby , MELM generates high - quality augmented data with novel entities , which provides rich entity regularity knowledge and boosts NER performance .", "forward": false, "src_ids": "2022.acl-long.160_196"}
+{"input": "high - quality augmented data is done by using Method| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "high - quality augmented data", "output": "masked entity language modeling ( melm )", "neg_sample": ["high - quality augmented data is done by using Method", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "Thereby , MELM generates high - quality augmented data with novel entities , which provides rich entity regularity knowledge and boosts NER performance .", "forward": false, "src_ids": "2022.acl-long.160_197"}
+{"input": "data augmentation is used for Task| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "data augmentation", "output": "low - resource ner", "neg_sample": ["data augmentation is used for Task", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "MELM : Data Augmentation with Masked Entity Language Modeling for Low - Resource NER.", "forward": true, "src_ids": "2022.acl-long.160_198"}
+{"input": "masked entity language modeling ( melm ) is used for Task| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "masked entity language modeling ( melm )", "output": "low - resource ner", "neg_sample": ["masked entity language modeling ( melm ) is used for Task", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "MELM : Data Augmentation with Masked Entity Language Modeling for Low - Resource NER.", "forward": true, "src_ids": "2022.acl-long.160_199"}
+{"input": "data augmentation framework is used for Task| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "data augmentation framework", "output": "low - resource ner", "neg_sample": ["data augmentation framework is used for Task", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "In this work , we propose Masked Entity Language Modeling ( MELM ) as a novel data augmentation framework for low - resource NER .", "forward": true, "src_ids": "2022.acl-long.160_200"}
+{"input": "masked entity language modeling ( melm ) is used for Task| context: data augmentation is an effective solution to data scarcity in low - resource scenarios .", "entity": "masked entity language modeling ( melm )", "output": "ner", "neg_sample": ["masked entity language modeling ( melm ) is used for Task", "data augmentation is an effective solution to data scarcity in low - resource scenarios ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "Thereby , MELM generates high - quality augmented data with novel entities , which provides rich entity regularity knowledge and boosts NER performance .", "forward": true, "src_ids": "2022.acl-long.160_201"}
+{"input": "low - resource ner is done by using Method| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "low - resource ner", "output": "data augmentation framework", "neg_sample": ["low - resource ner is done by using Method", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "In this work , we propose Masked Entity Language Modeling ( MELM ) as a novel data augmentation framework for low - resource NER .", "forward": false, "src_ids": "2022.acl-long.160_202"}
+{"input": "masked entity tokens is done by using Method| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "masked entity tokens", "output": "fine - tuned melm", "neg_sample": ["masked entity tokens is done by using Method", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "To alleviate the token - label misalignment issue , we explicitly inject NER labels into sentence context , and thus the fine - tuned MELM is able to predict masked entity tokens by explicitly conditioning on their labels .", "forward": false, "src_ids": "2022.acl-long.160_203"}
+{"input": "fine - tuned melm is used for OtherScientificTerm| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "fine - tuned melm", "output": "masked entity tokens", "neg_sample": ["fine - tuned melm is used for OtherScientificTerm", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "To alleviate the token - label misalignment issue , we explicitly inject NER labels into sentence context , and thus the fine - tuned MELM is able to predict masked entity tokens by explicitly conditioning on their labels .", "forward": true, "src_ids": "2022.acl-long.160_204"}
+{"input": "masked entity language modeling ( melm ) is used for Material| context: data augmentation is an effective solution to data scarcity in low - resource scenarios . however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance .", "entity": "masked entity language modeling ( melm )", "output": "high - quality augmented data", "neg_sample": ["masked entity language modeling ( melm ) is used for Material", "data augmentation is an effective solution to data scarcity in low - resource scenarios .", "however , when applied to token - level tasks such as ner , data augmentation methods often suffer from token - label misalignment , which leads to unsatsifactory performance ."], "relation": "used for", "id": "2022.acl-long.160", "year": 2022, "rel_sent": "Thereby , MELM generates high - quality augmented data with novel entities , which provides rich entity regularity knowledge and boosts NER performance .", "forward": true, "src_ids": "2022.acl-long.160_205"}
+{"input": "gender diversity is done by using Method| context: almost all prior work on this problem adjusts the training data or the model itself . by contrast , our approach changes only the inference procedure .", "entity": "gender diversity", "output": "beam search", "neg_sample": ["gender diversity is done by using Method", "almost all prior work on this problem adjusts the training data or the model itself .", "by contrast , our approach changes only the inference procedure ."], "relation": "used for", "id": "2022.findings-acl.301", "year": 2022, "rel_sent": "We constrain beam search to improve gender diversity in n - best lists , and rerank n - best lists using gender features obtained from the source sentence .", "forward": false, "src_ids": "2022.findings-acl.301_206"}
+{"input": "n - best lists is done by using Method| context: almost all prior work on this problem adjusts the training data or the model itself . by contrast , our approach changes only the inference procedure .", "entity": "n - best lists", "output": "beam search", "neg_sample": ["n - best lists is done by using Method", "almost all prior work on this problem adjusts the training data or the model itself .", "by contrast , our approach changes only the inference procedure ."], "relation": "used for", "id": "2022.findings-acl.301", "year": 2022, "rel_sent": "We constrain beam search to improve gender diversity in n - best lists , and rerank n - best lists using gender features obtained from the source sentence .", "forward": false, "src_ids": "2022.findings-acl.301_207"}
+{"input": "beam search is used for OtherScientificTerm| context: generating machine translations via beam search seeks the most likely output under a model . however , beam search has been shown to amplify demographic biases exhibited by a model . almost all prior work on this problem adjusts the training data or the model itself . by contrast , our approach changes only the inference procedure .", "entity": "beam search", "output": "gender diversity", "neg_sample": ["beam search is used for OtherScientificTerm", "generating machine translations via beam search seeks the most likely output under a model .", "however , beam search has been shown to amplify demographic biases exhibited by a model .", "almost all prior work on this problem adjusts the training data or the model itself .", "by contrast , our approach changes only the inference procedure ."], "relation": "used for", "id": "2022.findings-acl.301", "year": 2022, "rel_sent": "We constrain beam search to improve gender diversity in n - best lists , and rerank n - best lists using gender features obtained from the source sentence .", "forward": true, "src_ids": "2022.findings-acl.301_208"}
+{"input": "intent classification is done by using Method| context: pre - trained transformer - based models were reported to be robust in intent classification .", "entity": "intent classification", "output": "pre - trained transformers", "neg_sample": ["intent classification is done by using Method", "pre - trained transformer - based models were reported to be robust in intent classification ."], "relation": "used for", "id": "2022.nlp4convai-1.2", "year": 2022, "rel_sent": "Are Pre - trained Transformers Robust in Intent Classification ? A Missing Ingredient in Evaluation of Out - of - Scope Intent Detection.", "forward": false, "src_ids": "2022.nlp4convai-1.2_209"}
+{"input": "word embeddings is used for OtherScientificTerm| context: word embeddings are powerful dictionaries , which may easily capture language variations .", "entity": "word embeddings", "output": "rare words", "neg_sample": ["word embeddings is used for OtherScientificTerm", "word embeddings are powerful dictionaries , which may easily capture language variations ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In this paper , we propose to use definitions retrieved in traditional dictionaries to produce word embeddings for rare words .", "forward": true, "src_ids": "2022.findings-acl.208_210"}
+{"input": "rare words is done by using OtherScientificTerm| context: however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "rare words", "output": "word embeddings", "neg_sample": ["rare words is done by using OtherScientificTerm", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In this paper , we propose to use definitions retrieved in traditional dictionaries to produce word embeddings for rare words .", "forward": false, "src_ids": "2022.findings-acl.208_211"}
+{"input": "embeddings of unknown words is done by using Method| context: word embeddings are powerful dictionaries , which may easily capture language variations . however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "embeddings of unknown words", "output": "definition neural network", "neg_sample": ["embeddings of unknown words is done by using Method", "word embeddings are powerful dictionaries , which may easily capture language variations .", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In our experiments , DefiNNet and DefBERT significantly outperform state - of - the - art as well as baseline methods devised for producing embeddings of unknown words .", "forward": false, "src_ids": "2022.findings-acl.208_212"}
+{"input": "oov words is done by using Method| context: word embeddings are powerful dictionaries , which may easily capture language variations . however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "oov words", "output": "define bert", "neg_sample": ["oov words is done by using Method", "word embeddings are powerful dictionaries , which may easily capture language variations .", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In fact , DefiNNet significantly outperforms FastText , which implements a method for the same task - based on n - grams , and DefBERT significantly outperforms the BERT method for OOV words .", "forward": false, "src_ids": "2022.findings-acl.208_213"}
+{"input": "define bert is used for OtherScientificTerm| context: word embeddings are powerful dictionaries , which may easily capture language variations . however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "define bert", "output": "embeddings of unknown words", "neg_sample": ["define bert is used for OtherScientificTerm", "word embeddings are powerful dictionaries , which may easily capture language variations .", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In our experiments , DefiNNet and DefBERT significantly outperform state - of - the - art as well as baseline methods devised for producing embeddings of unknown words .", "forward": true, "src_ids": "2022.findings-acl.208_214"}
+{"input": "definition neural network is used for OtherScientificTerm| context: word embeddings are powerful dictionaries , which may easily capture language variations . however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "definition neural network", "output": "embeddings of unknown words", "neg_sample": ["definition neural network is used for OtherScientificTerm", "word embeddings are powerful dictionaries , which may easily capture language variations .", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In our experiments , DefiNNet and DefBERT significantly outperform state - of - the - art as well as baseline methods devised for producing embeddings of unknown words .", "forward": true, "src_ids": "2022.findings-acl.208_215"}
+{"input": "bert method is used for OtherScientificTerm| context: word embeddings are powerful dictionaries , which may easily capture language variations . however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "bert method", "output": "oov words", "neg_sample": ["bert method is used for OtherScientificTerm", "word embeddings are powerful dictionaries , which may easily capture language variations .", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In fact , DefiNNet significantly outperforms FastText , which implements a method for the same task - based on n - grams , and DefBERT significantly outperforms the BERT method for OOV words .", "forward": true, "src_ids": "2022.findings-acl.208_216"}
+{"input": "define bert is used for OtherScientificTerm| context: word embeddings are powerful dictionaries , which may easily capture language variations . however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries .", "entity": "define bert", "output": "oov words", "neg_sample": ["define bert is used for OtherScientificTerm", "word embeddings are powerful dictionaries , which may easily capture language variations .", "however , these dictionaries fail to give sense to rare words , which are surprisingly often covered by traditional dictionaries ."], "relation": "used for", "id": "2022.findings-acl.208", "year": 2022, "rel_sent": "In fact , DefiNNet significantly outperforms FastText , which implements a method for the same task - based on n - grams , and DefBERT significantly outperforms the BERT method for OOV words .", "forward": true, "src_ids": "2022.findings-acl.208_217"}
+{"input": "ranking candidate clarification questions is done by using Method| context: we deal with the scenario of conversational search , where user queries are under - specified or ambiguous . this calls for a mixed - initiative setup . user - asks ( queries ) and system - answers , as well as system - asks ( clarification questions ) and user response , in order to clarify her information needs .", "entity": "ranking candidate clarification questions", "output": "deep - learning models", "neg_sample": ["ranking candidate clarification questions is done by using Method", "we deal with the scenario of conversational search , where user queries are under - specified or ambiguous .", "this calls for a mixed - initiative setup .", "user - asks ( queries ) and system - answers , as well as system - asks ( clarification questions ) and user response , in order to clarify her information needs ."], "relation": "used for", "id": "2022.dialdoc-1.7", "year": 2022, "rel_sent": "Our method leverages passage retrieval from background content tofine - tune two deep - learning models for ranking candidate clarification questions .", "forward": false, "src_ids": "2022.dialdoc-1.7_218"}
+{"input": "deep - learning models is used for Task| context: we deal with the scenario of conversational search , where user queries are under - specified or ambiguous . this calls for a mixed - initiative setup . user - asks ( queries ) and system - answers , as well as system - asks ( clarification questions ) and user response , in order to clarify her information needs .", "entity": "deep - learning models", "output": "ranking candidate clarification questions", "neg_sample": ["deep - learning models is used for Task", "we deal with the scenario of conversational search , where user queries are under - specified or ambiguous .", "this calls for a mixed - initiative setup .", "user - asks ( queries ) and system - answers , as well as system - asks ( clarification questions ) and user response , in order to clarify her information needs ."], "relation": "used for", "id": "2022.dialdoc-1.7", "year": 2022, "rel_sent": "Our method leverages passage retrieval from background content tofine - tune two deep - learning models for ranking candidate clarification questions .", "forward": true, "src_ids": "2022.dialdoc-1.7_219"}
+{"input": "named entity recognition is done by using Method| context: named entity recognition ( ner ) is a fundamental task in natural language processing . recent works treat named entity recognition as a reading comprehension task , constructing type - specific queries manually to extract entities . this paradigm suffers from three issues . first , type - specific queries can only extract one type of entities per inference , which is inefficient . second , the extraction for different types of entities is isolated , ignoring the dependencies between them . third , query construction relies on external knowledge and is difficult to apply to realistic scenarios with hundreds of entity types .", "entity": "named entity recognition", "output": "parallel instance query network ( piqn )", "neg_sample": ["named entity recognition is done by using Method", "named entity recognition ( ner ) is a fundamental task in natural language processing .", "recent works treat named entity recognition as a reading comprehension task , constructing type - specific queries manually to extract entities .", "this paradigm suffers from three issues .", "first , type - specific queries can only extract one type of entities per inference , which is inefficient .", "second , the extraction for different types of entities is isolated , ignoring the dependencies between them .", "third , query construction relies on external knowledge and is difficult to apply to realistic scenarios with hundreds of entity types ."], "relation": "used for", "id": "2022.acl-long.67", "year": 2022, "rel_sent": "Parallel Instance Query Network for Named Entity Recognition.", "forward": false, "src_ids": "2022.acl-long.67_220"}
+{"input": "parallel instance query network ( piqn ) is used for Task| context: this paradigm suffers from three issues . first , type - specific queries can only extract one type of entities per inference , which is inefficient . second , the extraction for different types of entities is isolated , ignoring the dependencies between them . third , query construction relies on external knowledge and is difficult to apply to realistic scenarios with hundreds of entity types .", "entity": "parallel instance query network ( piqn )", "output": "named entity recognition", "neg_sample": ["parallel instance query network ( piqn ) is used for Task", "this paradigm suffers from three issues .", "first , type - specific queries can only extract one type of entities per inference , which is inefficient .", "second , the extraction for different types of entities is isolated , ignoring the dependencies between them .", "third , query construction relies on external knowledge and is difficult to apply to realistic scenarios with hundreds of entity types ."], "relation": "used for", "id": "2022.acl-long.67", "year": 2022, "rel_sent": "Parallel Instance Query Network for Named Entity Recognition.", "forward": true, "src_ids": "2022.acl-long.67_221"}
+{"input": "multimodal sentiment analysis is done by using Generic| context: m - sena is an open - sourced platform for multimodal sentiment analysis .", "entity": "multimodal sentiment analysis", "output": "integrated platform", "neg_sample": ["multimodal sentiment analysis is done by using Generic", "m - sena is an open - sourced platform for multimodal sentiment analysis ."], "relation": "used for", "id": "2022.acl-demo.20", "year": 2022, "rel_sent": "M - SENA : An Integrated Platform for Multimodal Sentiment Analysis.", "forward": false, "src_ids": "2022.acl-demo.20_222"}
+{"input": "intermediate representation visualization is done by using Method| context: m - sena is an open - sourced platform for multimodal sentiment analysis .", "entity": "intermediate representation visualization", "output": "model evaluation and analysis tools", "neg_sample": ["intermediate representation visualization is done by using Method", "m - sena is an open - sourced platform for multimodal sentiment analysis ."], "relation": "used for", "id": "2022.acl-demo.20", "year": 2022, "rel_sent": "Moreover , we use model evaluation and analysis tools provided by M - SENA to present intermediate representation visualization , on - the - fly instance test , and generalization ability test results .", "forward": false, "src_ids": "2022.acl-demo.20_223"}
+{"input": "model evaluation and analysis tools is used for Method| context: m - sena is an open - sourced platform for multimodal sentiment analysis .", "entity": "model evaluation and analysis tools", "output": "intermediate representation visualization", "neg_sample": ["model evaluation and analysis tools is used for Method", "m - sena is an open - sourced platform for multimodal sentiment analysis ."], "relation": "used for", "id": "2022.acl-demo.20", "year": 2022, "rel_sent": "Moreover , we use model evaluation and analysis tools provided by M - SENA to present intermediate representation visualization , on - the - fly instance test , and generalization ability test results .", "forward": true, "src_ids": "2022.acl-demo.20_224"}
+{"input": "domain classifier is done by using Method| context: dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search . this requires strong locality properties from the representation space , e.g. , close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data .", "entity": "domain classifier", "output": "momentum method", "neg_sample": ["domain classifier is done by using Method", "dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search .", "this requires strong locality properties from the representation space , e.g.", ", close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data ."], "relation": "used for", "id": "2022.findings-acl.316", "year": 2022, "rel_sent": "To achieve that , we propose Momentum adversarial Domain Invariant Representation learning ( MoDIR ) , which introduces a momentum method to train a domain classifier that distinguishes source versus target domains , and then adversarially updates the DR encoder to learn domain invariant representations .", "forward": false, "src_ids": "2022.findings-acl.316_225"}
+{"input": "momentum method is used for Method| context: dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search . this requires strong locality properties from the representation space , e.g. , close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data .", "entity": "momentum method", "output": "domain classifier", "neg_sample": ["momentum method is used for Method", "dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search .", "this requires strong locality properties from the representation space , e.g.", ", close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data ."], "relation": "used for", "id": "2022.findings-acl.316", "year": 2022, "rel_sent": "To achieve that , we propose Momentum adversarial Domain Invariant Representation learning ( MoDIR ) , which introduces a momentum method to train a domain classifier that distinguishes source versus target domains , and then adversarially updates the DR encoder to learn domain invariant representations .", "forward": true, "src_ids": "2022.findings-acl.316_226"}
+{"input": "domain invariant representations is done by using Method| context: dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search . this requires strong locality properties from the representation space , e.g. , close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data .", "entity": "domain invariant representations", "output": "dr encoder", "neg_sample": ["domain invariant representations is done by using Method", "dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search .", "this requires strong locality properties from the representation space , e.g.", ", close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data ."], "relation": "used for", "id": "2022.findings-acl.316", "year": 2022, "rel_sent": "To achieve that , we propose Momentum adversarial Domain Invariant Representation learning ( MoDIR ) , which introduces a momentum method to train a domain classifier that distinguishes source versus target domains , and then adversarially updates the DR encoder to learn domain invariant representations .", "forward": false, "src_ids": "2022.findings-acl.316_227"}
+{"input": "dr encoder is used for Method| context: dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search . this requires strong locality properties from the representation space , e.g. , close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data .", "entity": "dr encoder", "output": "domain invariant representations", "neg_sample": ["dr encoder is used for Method", "dense retrieval ( dr ) methods conduct text retrieval by first encoding texts in the embedding space and then matching them by nearest neighbor search .", "this requires strong locality properties from the representation space , e.g.", ", close allocations of each small group of relevant texts , which are hard to generalize to domains without sufficient training data ."], "relation": "used for", "id": "2022.findings-acl.316", "year": 2022, "rel_sent": "To achieve that , we propose Momentum adversarial Domain Invariant Representation learning ( MoDIR ) , which introduces a momentum method to train a domain classifier that distinguishes source versus target domains , and then adversarially updates the DR encoder to learn domain invariant representations .", "forward": true, "src_ids": "2022.findings-acl.316_228"}
+{"input": "automatic method is used for Method| context: given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task .", "entity": "automatic method", "output": "pretrained language models", "neg_sample": ["automatic method is used for Method", "given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task ."], "relation": "used for", "id": "2022.acl-long.72", "year": 2022, "rel_sent": "In this paper , we propose an automatic method to mitigate the biases in pretrained language models .", "forward": true, "src_ids": "2022.acl-long.72_229"}
+{"input": "pretrained language models is done by using Method| context: human - like biases and undesired social stereotypes exist in large pretrained language models . given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task .", "entity": "pretrained language models", "output": "automatic method", "neg_sample": ["pretrained language models is done by using Method", "human - like biases and undesired social stereotypes exist in large pretrained language models .", "given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task ."], "relation": "used for", "id": "2022.acl-long.72", "year": 2022, "rel_sent": "In this paper , we propose an automatic method to mitigate the biases in pretrained language models .", "forward": false, "src_ids": "2022.acl-long.72_230"}
+{"input": "biased prompts is done by using Method| context: human - like biases and undesired social stereotypes exist in large pretrained language models . given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task .", "entity": "biased prompts", "output": "beam search method", "neg_sample": ["biased prompts is done by using Method", "human - like biases and undesired social stereotypes exist in large pretrained language models .", "given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task ."], "relation": "used for", "id": "2022.acl-long.72", "year": 2022, "rel_sent": "Specifically , we propose a variant of the beam search method to automatically search for biased prompts such that the cloze - style completions are the most different with respect to different demographic groups .", "forward": false, "src_ids": "2022.acl-long.72_231"}
+{"input": "beam search method is used for OtherScientificTerm| context: human - like biases and undesired social stereotypes exist in large pretrained language models . given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task .", "entity": "beam search method", "output": "biased prompts", "neg_sample": ["beam search method is used for OtherScientificTerm", "human - like biases and undesired social stereotypes exist in large pretrained language models .", "given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task ."], "relation": "used for", "id": "2022.acl-long.72", "year": 2022, "rel_sent": "Specifically , we propose a variant of the beam search method to automatically search for biased prompts such that the cloze - style completions are the most different with respect to different demographic groups .", "forward": true, "src_ids": "2022.acl-long.72_232"}
+{"input": "biases is done by using Method| context: human - like biases and undesired social stereotypes exist in large pretrained language models . given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task .", "entity": "biases", "output": "distribution alignment loss", "neg_sample": ["biases is done by using Method", "human - like biases and undesired social stereotypes exist in large pretrained language models .", "given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task ."], "relation": "used for", "id": "2022.acl-long.72", "year": 2022, "rel_sent": "Given the identified biased prompts , we then propose a distribution alignment loss to mitigate the biases .", "forward": false, "src_ids": "2022.acl-long.72_233"}
+{"input": "biases is done by using Method| context: human - like biases and undesired social stereotypes exist in large pretrained language models . given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task .", "entity": "biases", "output": "auto - debias approach", "neg_sample": ["biases is done by using Method", "human - like biases and undesired social stereotypes exist in large pretrained language models .", "given the wide adoption of these models in real - world applications , mitigating such biases has become an emerging and important task ."], "relation": "used for", "id": "2022.acl-long.72", "year": 2022, "rel_sent": "Experiment results on standard datasets and metrics show that our proposed Auto - Debias approach can significantly reduce biases , including gender and racial bias , in pretrained language models such as BERT , RoBERTa and ALBERT .", "forward": false, "src_ids": "2022.acl-long.72_234"}
+{"input": "iterative inplace editing approach is used for Task| context: current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "iterative inplace editing approach", "output": "text revision", "neg_sample": ["iterative inplace editing approach is used for Task", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "In this paper , we present an iterative inplace editing approach for text revision , which requires no parallel data .", "forward": true, "src_ids": "2022.in2writing-1.7_235"}
+{"input": "text revision is done by using Method| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "text revision", "output": "iterative inplace editing approach", "neg_sample": ["text revision is done by using Method", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "In this paper , we present an iterative inplace editing approach for text revision , which requires no parallel data .", "forward": false, "src_ids": "2022.in2writing-1.7_236"}
+{"input": "attribute function is done by using Method| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "attribute function", "output": "distributed representation of the text", "neg_sample": ["attribute function is done by using Method", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "At the first step , the distributed representation of the text optimizes on the fly towards an attribute function .", "forward": false, "src_ids": "2022.in2writing-1.7_237"}
+{"input": "distributed representation of the text is used for OtherScientificTerm| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "distributed representation of the text", "output": "attribute function", "neg_sample": ["distributed representation of the text is used for OtherScientificTerm", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "At the first step , the distributed representation of the text optimizes on the fly towards an attribute function .", "forward": true, "src_ids": "2022.in2writing-1.7_238"}
+{"input": "unsupervised methods is used for Task| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "unsupervised methods", "output": "text formalization", "neg_sample": ["unsupervised methods is used for Task", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "It achieves competitive and even better performance than state - of - the - art supervised methods on text simplification , and gains better performance than strong unsupervised methods on text formalization .", "forward": true, "src_ids": "2022.in2writing-1.7_239"}
+{"input": "supervised methods is used for Task| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "supervised methods", "output": "text simplification", "neg_sample": ["supervised methods is used for Task", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "It achieves competitive and even better performance than state - of - the - art supervised methods on text simplification , and gains better performance than strong unsupervised methods on text formalization .", "forward": true, "src_ids": "2022.in2writing-1.7_240"}
+{"input": "text simplification is done by using Method| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "text simplification", "output": "supervised methods", "neg_sample": ["text simplification is done by using Method", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "It achieves competitive and even better performance than state - of - the - art supervised methods on text simplification , and gains better performance than strong unsupervised methods on text formalization .", "forward": false, "src_ids": "2022.in2writing-1.7_241"}
+{"input": "text formalization is done by using Method| context: text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity . current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus .", "entity": "text formalization", "output": "unsupervised methods", "neg_sample": ["text formalization is done by using Method", "text revision refers to a family of natural language generation tasks , where the source and target sequences share moderate resemblance in surface form but differentiate in attributes , such as text formality and simplicity .", "current state - of - the - art methods formulate these tasks as sequence - to - sequence learning problems , which rely on large - scale parallel training corpus ."], "relation": "used for", "id": "2022.in2writing-1.7", "year": 2022, "rel_sent": "It achieves competitive and even better performance than state - of - the - art supervised methods on text simplification , and gains better performance than strong unsupervised methods on text formalization .", "forward": false, "src_ids": "2022.in2writing-1.7_242"}
+{"input": "commonsense reasoning is done by using Task| context: generative commonsense reasoning ( gcr ) in natural language is to reason about the commonsense while generating coherent text . recent years have seen a surge of interest in improving the generation quality of commonsense reasoning tasks . nevertheless , these approaches have seldom investigated diversity in the gcr tasks , which aims to generate alternative explanations for a real - world situation or predict all possible outcomes . diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge .", "entity": "commonsense reasoning", "output": "diversifying content generation", "neg_sample": ["commonsense reasoning is done by using Task", "generative commonsense reasoning ( gcr ) in natural language is to reason about the commonsense while generating coherent text .", "recent years have seen a surge of interest in improving the generation quality of commonsense reasoning tasks .", "nevertheless , these approaches have seldom investigated diversity in the gcr tasks , which aims to generate alternative explanations for a real - world situation or predict all possible outcomes .", "diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge ."], "relation": "used for", "id": "2022.findings-acl.149", "year": 2022, "rel_sent": "Diversifying Content Generation for Commonsense Reasoning with Mixture of Knowledge Graph Experts.", "forward": false, "src_ids": "2022.findings-acl.149_243"}
+{"input": "diversifying content generation is used for Task| context: nevertheless , these approaches have seldom investigated diversity in the gcr tasks , which aims to generate alternative explanations for a real - world situation or predict all possible outcomes . diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge .", "entity": "diversifying content generation", "output": "commonsense reasoning", "neg_sample": ["diversifying content generation is used for Task", "nevertheless , these approaches have seldom investigated diversity in the gcr tasks , which aims to generate alternative explanations for a real - world situation or predict all possible outcomes .", "diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge ."], "relation": "used for", "id": "2022.findings-acl.149", "year": 2022, "rel_sent": "Diversifying Content Generation for Commonsense Reasoning with Mixture of Knowledge Graph Experts.", "forward": true, "src_ids": "2022.findings-acl.149_244"}
+{"input": "diversity is done by using Method| context: generative commonsense reasoning ( gcr ) in natural language is to reason about the commonsense while generating coherent text . recent years have seen a surge of interest in improving the generation quality of commonsense reasoning tasks . nevertheless , these approaches have seldom investigated diversity in the gcr tasks , which aims to generate alternative explanations for a real - world situation or predict all possible outcomes . diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge .", "entity": "diversity", "output": "mokge", "neg_sample": ["diversity is done by using Method", "generative commonsense reasoning ( gcr ) in natural language is to reason about the commonsense while generating coherent text .", "recent years have seen a surge of interest in improving the generation quality of commonsense reasoning tasks .", "nevertheless , these approaches have seldom investigated diversity in the gcr tasks , which aims to generate alternative explanations for a real - world situation or predict all possible outcomes .", "diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge ."], "relation": "used for", "id": "2022.findings-acl.149", "year": 2022, "rel_sent": "Empirical experiments demonstrated that MoKGE can significantly improve the diversity while achieving on par performance on accuracy on two GCR benchmarks , based on both automatic and human evaluations .", "forward": false, "src_ids": "2022.findings-acl.149_245"}
+{"input": "mokge is used for OtherScientificTerm| context: generative commonsense reasoning ( gcr ) in natural language is to reason about the commonsense while generating coherent text . recent years have seen a surge of interest in improving the generation quality of commonsense reasoning tasks . diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge .", "entity": "mokge", "output": "diversity", "neg_sample": ["mokge is used for OtherScientificTerm", "generative commonsense reasoning ( gcr ) in natural language is to reason about the commonsense while generating coherent text .", "recent years have seen a surge of interest in improving the generation quality of commonsense reasoning tasks .", "diversifying gcr is challenging as it expects to generate multiple outputs that are not only semantically different but also grounded in commonsense knowledge ."], "relation": "used for", "id": "2022.findings-acl.149", "year": 2022, "rel_sent": "Empirical experiments demonstrated that MoKGE can significantly improve the diversity while achieving on par performance on accuracy on two GCR benchmarks , based on both automatic and human evaluations .", "forward": true, "src_ids": "2022.findings-acl.149_246"}
+{"input": "hidden state tokens is done by using Method| context: transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision . as such , improving its computational efficiency becomes paramount . one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers . prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency . however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor .", "entity": "hidden state tokens", "output": "transkimmer architecture", "neg_sample": ["hidden state tokens is done by using Method", "transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision .", "as such , improving its computational efficiency becomes paramount .", "one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers .", "prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency .", "however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor ."], "relation": "used for", "id": "2022.acl-long.502", "year": 2022, "rel_sent": "To address the above limitations , we propose the Transkimmer architecture , which learns to identify hidden state tokens that are not required by each layer .", "forward": false, "src_ids": "2022.acl-long.502_247"}
+{"input": "transkimmer architecture is used for OtherScientificTerm| context: transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision . as such , improving its computational efficiency becomes paramount . one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers . prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency . however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor .", "entity": "transkimmer architecture", "output": "hidden state tokens", "neg_sample": ["transkimmer architecture is used for OtherScientificTerm", "transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision .", "as such , improving its computational efficiency becomes paramount .", "one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers .", "prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency .", "however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor ."], "relation": "used for", "id": "2022.acl-long.502", "year": 2022, "rel_sent": "To address the above limitations , we propose the Transkimmer architecture , which learns to identify hidden state tokens that are not required by each layer .", "forward": true, "src_ids": "2022.acl-long.502_248"}
+{"input": "skimming decision is done by using Method| context: transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision . as such , improving its computational efficiency becomes paramount . one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers . prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency . however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor .", "entity": "skimming decision", "output": "parameterized predictor", "neg_sample": ["skimming decision is done by using Method", "transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision .", "as such , improving its computational efficiency becomes paramount .", "one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers .", "prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency .", "however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor ."], "relation": "used for", "id": "2022.acl-long.502", "year": 2022, "rel_sent": "The key idea in Transkimmer is to add a parameterized predictor before each layer that learns to make the skimming decision .", "forward": false, "src_ids": "2022.acl-long.502_249"}
+{"input": "parameterized predictor is used for OtherScientificTerm| context: transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision . as such , improving its computational efficiency becomes paramount . one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers . prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency . however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor .", "entity": "parameterized predictor", "output": "skimming decision", "neg_sample": ["parameterized predictor is used for OtherScientificTerm", "transformer architecture has become the de - facto model for many machine learning tasks from natural language processing and computer vision .", "as such , improving its computational efficiency becomes paramount .", "one of the major computational inefficiency of transformer based models is that they spend the identical amount of computation throughout all layers .", "prior works have proposed to augment the transformer model with the capability of skimming tokens to improve its computational efficiency .", "however , they suffer from not having effectual and end - to - end optimization of the discrete skimming predictor ."], "relation": "used for", "id": "2022.acl-long.502", "year": 2022, "rel_sent": "The key idea in Transkimmer is to add a parameterized predictor before each layer that learns to make the skimming decision .", "forward": true, "src_ids": "2022.acl-long.502_250"}
+{"input": "gpt compression is done by using Method| context: gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain . the success of gpt is mostly attributed to its pre - training on huge amount of data and its large number of parameters . despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory . this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature .", "entity": "gpt compression", "output": "kronecker decomposition", "neg_sample": ["gpt compression is done by using Method", "gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain .", "the success of gpt is mostly attributed to its pre - training on huge amount of data and its large number of parameters .", "despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory .", "this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature ."], "relation": "used for", "id": "2022.acl-short.24", "year": 2022, "rel_sent": "Kronecker Decomposition for GPT Compression.", "forward": false, "src_ids": "2022.acl-short.24_251"}
+{"input": "kronecker decomposition is used for Task| context: gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain . the success of gpt is mostly attributed to its pre - training on huge amount of data and its large number of parameters . despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory . this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature . in this work , we use kronecker decomposition to compress the linear mappings of the gpt-2 model .", "entity": "kronecker decomposition", "output": "gpt compression", "neg_sample": ["kronecker decomposition is used for Task", "gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain .", "the success of gpt is mostly attributed to its pre - training on huge amount of data and its large number of parameters .", "despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory .", "this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature .", "in this work , we use kronecker decomposition to compress the linear mappings of the gpt-2 model ."], "relation": "used for", "id": "2022.acl-short.24", "year": 2022, "rel_sent": "Kronecker Decomposition for GPT Compression.", "forward": true, "src_ids": "2022.acl-short.24_252"}
+{"input": "pre - training is used for Method| context: gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain . the success of gpt is mostly attributed to its pre - training on huge amount of data and its large number of parameters . despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory . this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature . in this work , we use kronecker decomposition to compress the linear mappings of the gpt-2 model .", "entity": "pre - training", "output": "kronecker gpt-2 model ( kngpt2 )", "neg_sample": ["pre - training is used for Method", "gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain .", "the success of gpt is mostly attributed to its pre - training on huge amount of data and its large number of parameters .", "despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory .", "this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature .", "in this work , we use kronecker decomposition to compress the linear mappings of the gpt-2 model ."], "relation": "used for", "id": "2022.acl-short.24", "year": 2022, "rel_sent": "We evaluate our model on both language modeling and General Language Understanding Evaluation benchmark tasks and show that with more efficient pre - training and similar number of parameters , our KnGPT2 outperforms the existing DistilGPT2 model significantly .", "forward": true, "src_ids": "2022.acl-short.24_253"}
+{"input": "kronecker gpt-2 model ( kngpt2 ) is done by using Method| context: gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain . despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory . this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature . in this work , we use kronecker decomposition to compress the linear mappings of the gpt-2 model .", "entity": "kronecker gpt-2 model ( kngpt2 )", "output": "pre - training", "neg_sample": ["kronecker gpt-2 model ( kngpt2 ) is done by using Method", "gpt is an auto - regressive transformer - based pre - trained language model which has attracted a lot of attention in the natural language processing ( nlp ) domain .", "despite the superior performance of gpt , this overparameterized nature of gpt can be very prohibitive for deploying this model on devices with limited computational power or memory .", "this problem can be mitigated using model compression techniques ; however , compressing gpt models has not been investigated much in the literature .", "in this work , we use kronecker decomposition to compress the linear mappings of the gpt-2 model ."], "relation": "used for", "id": "2022.acl-short.24", "year": 2022, "rel_sent": "We evaluate our model on both language modeling and General Language Understanding Evaluation benchmark tasks and show that with more efficient pre - training and similar number of parameters , our KnGPT2 outperforms the existing DistilGPT2 model significantly .", "forward": false, "src_ids": "2022.acl-short.24_254"}
+{"input": "pictographs is done by using Method| context: communication between physician and patients can lead to misunderstandings , especially for disabled people . an automatic system that translates natural language into a pictographic language is one of the solutions that could help to overcome this issue .", "entity": "pictographs", "output": "speech therapists", "neg_sample": ["pictographs is done by using Method", "communication between physician and patients can lead to misunderstandings , especially for disabled people .", "an automatic system that translates natural language into a pictographic language is one of the solutions that could help to overcome this issue ."], "relation": "used for", "id": "2022.slpat-1.6", "year": 2022, "rel_sent": "In this preliminary study , we present the French version of a translation system using the Arasaac pictographs and we investigate the strategies used by speech therapists to translate into pictographs .", "forward": false, "src_ids": "2022.slpat-1.6_255"}
+{"input": "speech therapists is used for OtherScientificTerm| context: communication between physician and patients can lead to misunderstandings , especially for disabled people . an automatic system that translates natural language into a pictographic language is one of the solutions that could help to overcome this issue .", "entity": "speech therapists", "output": "pictographs", "neg_sample": ["speech therapists is used for OtherScientificTerm", "communication between physician and patients can lead to misunderstandings , especially for disabled people .", "an automatic system that translates natural language into a pictographic language is one of the solutions that could help to overcome this issue ."], "relation": "used for", "id": "2022.slpat-1.6", "year": 2022, "rel_sent": "In this preliminary study , we present the French version of a translation system using the Arasaac pictographs and we investigate the strategies used by speech therapists to translate into pictographs .", "forward": true, "src_ids": "2022.slpat-1.6_256"}
+{"input": "multilingual task - oriented dialogue is done by using Task| context: recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems . however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g. , chinese and german . this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems .", "entity": "multilingual task - oriented dialogue", "output": "natural language processing", "neg_sample": ["multilingual task - oriented dialogue is done by using Task", "recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems .", "however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g.", ", chinese and german .", "this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems ."], "relation": "used for", "id": "2022.acl-tutorials.8", "year": 2022, "rel_sent": "Natural Language Processing for Multilingual Task - Oriented Dialogue.", "forward": false, "src_ids": "2022.acl-tutorials.8_257"}
+{"input": "natural language processing is used for Task| context: recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems . however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g. , chinese and german . this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems .", "entity": "natural language processing", "output": "multilingual task - oriented dialogue", "neg_sample": ["natural language processing is used for Task", "recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems .", "however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g.", ", chinese and german .", "this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems ."], "relation": "used for", "id": "2022.acl-tutorials.8", "year": 2022, "rel_sent": "Natural Language Processing for Multilingual Task - Oriented Dialogue.", "forward": true, "src_ids": "2022.acl-tutorials.8_258"}
+{"input": "dialogue is done by using OtherScientificTerm| context: recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems . however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g. , chinese and german . this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems .", "entity": "dialogue", "output": "multilinguality", "neg_sample": ["dialogue is done by using OtherScientificTerm", "recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems .", "however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g.", ", chinese and german .", "this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems ."], "relation": "used for", "id": "2022.acl-tutorials.8", "year": 2022, "rel_sent": "The tutorial will aim to provide answers or shed new light to the following questions : a ) Why are multilingual dialogue systems so hard to build : what makes multilinguality for dialogue more challenging than for other NLP applications and tasks ?", "forward": false, "src_ids": "2022.acl-tutorials.8_259"}
+{"input": "nlp applications is done by using OtherScientificTerm| context: recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems . however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g. , chinese and german . this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems .", "entity": "nlp applications", "output": "multilinguality", "neg_sample": ["nlp applications is done by using OtherScientificTerm", "recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems .", "however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g.", ", chinese and german .", "this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems ."], "relation": "used for", "id": "2022.acl-tutorials.8", "year": 2022, "rel_sent": "The tutorial will aim to provide answers or shed new light to the following questions : a ) Why are multilingual dialogue systems so hard to build : what makes multilinguality for dialogue more challenging than for other NLP applications and tasks ?", "forward": false, "src_ids": "2022.acl-tutorials.8_260"}
+{"input": "multilinguality is used for Material| context: however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g. , chinese and german . this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems .", "entity": "multilinguality", "output": "dialogue", "neg_sample": ["multilinguality is used for Material", "however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g.", ", chinese and german .", "this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems ."], "relation": "used for", "id": "2022.acl-tutorials.8", "year": 2022, "rel_sent": "The tutorial will aim to provide answers or shed new light to the following questions : a ) Why are multilingual dialogue systems so hard to build : what makes multilinguality for dialogue more challenging than for other NLP applications and tasks ?", "forward": true, "src_ids": "2022.acl-tutorials.8_261"}
+{"input": "multilinguality is used for Task| context: recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems . however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g. , chinese and german . this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems .", "entity": "multilinguality", "output": "nlp applications", "neg_sample": ["multilinguality is used for Task", "recent advances in deep learning have also enabled fast progress in the research of task - oriented dialogue ( tod ) systems .", "however , the majority of tod systems are developed for english and merely a handful of other widely spoken languages , e.g.", ", chinese and german .", "this hugely limits the global reach and , consequently , transformative socioeconomic potential of such systems ."], "relation": "used for", "id": "2022.acl-tutorials.8", "year": 2022, "rel_sent": "The tutorial will aim to provide answers or shed new light to the following questions : a ) Why are multilingual dialogue systems so hard to build : what makes multilinguality for dialogue more challenging than for other NLP applications and tasks ?", "forward": true, "src_ids": "2022.acl-tutorials.8_262"}
+{"input": "open - ended text generation tasks is done by using Task| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "open - ended text generation tasks", "output": "event transition planning", "neg_sample": ["open - ended text generation tasks is done by using Task", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.269", "year": 2022, "rel_sent": "Event Transition Planning for Open - ended Text Generation.", "forward": false, "src_ids": "2022.findings-acl.269_263"}
+{"input": "event transition planning is used for Task| context: the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "event transition planning", "output": "open - ended text generation tasks", "neg_sample": ["event transition planning is used for Task", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.269", "year": 2022, "rel_sent": "Event Transition Planning for Open - ended Text Generation.", "forward": true, "src_ids": "2022.findings-acl.269_264"}
+{"input": "' coarse ' plot skeleton is done by using Method| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "' coarse ' plot skeleton", "output": "event transition planner", "neg_sample": ["' coarse ' plot skeleton is done by using Method", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.269", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse ' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": false, "src_ids": "2022.findings-acl.269_265"}
+{"input": "event transition planner is used for OtherScientificTerm| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "event transition planner", "output": "' coarse ' plot skeleton", "neg_sample": ["event transition planner is used for OtherScientificTerm", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.269", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse ' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": true, "src_ids": "2022.findings-acl.269_266"}
+{"input": "skeleton is done by using Method| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "skeleton", "output": "text generator", "neg_sample": ["skeleton is done by using Method", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.269", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse ' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": false, "src_ids": "2022.findings-acl.269_267"}
+{"input": "text generator is used for OtherScientificTerm| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "text generator", "output": "skeleton", "neg_sample": ["text generator is used for OtherScientificTerm", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.269", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse ' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": true, "src_ids": "2022.findings-acl.269_268"}
+{"input": "multilingual setting is done by using OtherScientificTerm| context: incorporating stronger syntactic biases into neural language models ( lms ) is a long - standing goal , but research in this area often focuses on modeling english text , where constituent treebanks are readily available . extending constituent tree - based lms to the multilingual setting , where dependency treebanks are more common , is possible via dependency - to - constituency conversion methods . however , this raises the question of which tree formats are best for learning the model , and for which languages .", "entity": "multilingual setting", "output": "syntax injection", "neg_sample": ["multilingual setting is done by using OtherScientificTerm", "incorporating stronger syntactic biases into neural language models ( lms ) is a long - standing goal , but research in this area often focuses on modeling english text , where constituent treebanks are readily available .", "extending constituent tree - based lms to the multilingual setting , where dependency treebanks are more common , is possible via dependency - to - constituency conversion methods .", "however , this raises the question of which tree formats are best for learning the model , and for which languages ."], "relation": "used for", "id": "2022.spnlp-1.1", "year": 2022, "rel_sent": "Our best model shows the advantage over sequential / overparameterized LMs , suggesting the positive effect of syntax injection in a multilingual setting .", "forward": false, "src_ids": "2022.spnlp-1.1_269"}
+{"input": "syntax injection is used for Material| context: incorporating stronger syntactic biases into neural language models ( lms ) is a long - standing goal , but research in this area often focuses on modeling english text , where constituent treebanks are readily available . however , this raises the question of which tree formats are best for learning the model , and for which languages .", "entity": "syntax injection", "output": "multilingual setting", "neg_sample": ["syntax injection is used for Material", "incorporating stronger syntactic biases into neural language models ( lms ) is a long - standing goal , but research in this area often focuses on modeling english text , where constituent treebanks are readily available .", "however , this raises the question of which tree formats are best for learning the model , and for which languages ."], "relation": "used for", "id": "2022.spnlp-1.1", "year": 2022, "rel_sent": "Our best model shows the advantage over sequential / overparameterized LMs , suggesting the positive effect of syntax injection in a multilingual setting .", "forward": true, "src_ids": "2022.spnlp-1.1_270"}
+{"input": "real - life applications is done by using Method| context: a slot value might be provided segment by segment over multiple - turn interactions in a dialog , especially for some important information such as phone numbers and names . it is a common phenomenon in daily life , but little attention has been paid to it in previous work .", "entity": "real - life applications", "output": "slot based task - oriented dialog", "neg_sample": ["real - life applications is done by using Method", "a slot value might be provided segment by segment over multiple - turn interactions in a dialog , especially for some important information such as phone numbers and names .", "it is a common phenomenon in daily life , but little attention has been paid to it in previous work ."], "relation": "used for", "id": "2022.findings-acl.27", "year": 2022, "rel_sent": "More work should be done to meet the new challenges raised from SSTOD which widely exists in real - life applications .", "forward": false, "src_ids": "2022.findings-acl.27_271"}
+{"input": "slot based task - oriented dialog is used for Task| context: a slot value might be provided segment by segment over multiple - turn interactions in a dialog , especially for some important information such as phone numbers and names . it is a common phenomenon in daily life , but little attention has been paid to it in previous work .", "entity": "slot based task - oriented dialog", "output": "real - life applications", "neg_sample": ["slot based task - oriented dialog is used for Task", "a slot value might be provided segment by segment over multiple - turn interactions in a dialog , especially for some important information such as phone numbers and names .", "it is a common phenomenon in daily life , but little attention has been paid to it in previous work ."], "relation": "used for", "id": "2022.findings-acl.27", "year": 2022, "rel_sent": "More work should be done to meet the new challenges raised from SSTOD which widely exists in real - life applications .", "forward": true, "src_ids": "2022.findings-acl.27_272"}
+{"input": "interactive machine translation is done by using Method| context: interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account . existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right . however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd .", "entity": "interactive machine translation", "output": "bilingual text - infilling method", "neg_sample": ["interactive machine translation is done by using Method", "interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account .", "existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right .", "however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd ."], "relation": "used for", "id": "2022.acl-long.138", "year": 2022, "rel_sent": "BiTIIMT : A Bilingual Text - infilling Method for Interactive Machine Translation.", "forward": false, "src_ids": "2022.acl-long.138_273"}
+{"input": "interactive neural machine translation is done by using Method| context: interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account . existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right . however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd .", "entity": "interactive neural machine translation", "output": "bilingual text - infilling method", "neg_sample": ["interactive neural machine translation is done by using Method", "interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account .", "existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right .", "however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd ."], "relation": "used for", "id": "2022.acl-long.138", "year": 2022, "rel_sent": "In this work , we propose a novel BiTIIMT system , Bilingual Text - Infilling for Interactive Neural Machine Translation .", "forward": false, "src_ids": "2022.acl-long.138_274"}
+{"input": "bilingual text - infilling method is used for Task| context: interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account . existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right . however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd .", "entity": "bilingual text - infilling method", "output": "interactive machine translation", "neg_sample": ["bilingual text - infilling method is used for Task", "interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account .", "existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right .", "however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd ."], "relation": "used for", "id": "2022.acl-long.138", "year": 2022, "rel_sent": "BiTIIMT : A Bilingual Text - infilling Method for Interactive Machine Translation.", "forward": true, "src_ids": "2022.acl-long.138_275"}
+{"input": "bilingual text - infilling method is used for Task| context: existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right . however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd .", "entity": "bilingual text - infilling method", "output": "interactive neural machine translation", "neg_sample": ["bilingual text - infilling method is used for Task", "existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right .", "however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd ."], "relation": "used for", "id": "2022.acl-long.138", "year": 2022, "rel_sent": "In this work , we propose a novel BiTIIMT system , Bilingual Text - Infilling for Interactive Neural Machine Translation .", "forward": true, "src_ids": "2022.acl-long.138_276"}
+{"input": "translation prediction is done by using OtherScientificTerm| context: interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account . existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right . however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd .", "entity": "translation prediction", "output": "revised words", "neg_sample": ["translation prediction is done by using OtherScientificTerm", "interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account .", "existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right .", "however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd ."], "relation": "used for", "id": "2022.acl-long.138", "year": 2022, "rel_sent": "In this way , our system performs decoding without explicit constraints and makes full use of revised words for better translation prediction .", "forward": false, "src_ids": "2022.acl-long.138_277"}
+{"input": "revised words is used for Task| context: interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account . existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right . however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd .", "entity": "revised words", "output": "translation prediction", "neg_sample": ["revised words is used for Task", "interactive neural machine translation ( inmt ) is able to guarantee high - quality translations by taking human interactions into account .", "existing imt systems relying on lexical constrained decoding ( lcd ) enable humans to translate in a flexible translation order beyond the left - to - right .", "however , they typically suffer from two significant limitations in translation efficiency and quality due to the reliance on lcd ."], "relation": "used for", "id": "2022.acl-long.138", "year": 2022, "rel_sent": "In this way , our system performs decoding without explicit constraints and makes full use of revised words for better translation prediction .", "forward": true, "src_ids": "2022.acl-long.138_278"}
+{"input": "inter - learner uniformity is done by using Method| context: child language learners develop with remarkable uniformity , both in their learning trajectories and ultimate outcomes , despite major differences in their learning environments .", "entity": "inter - learner uniformity", "output": "tolerance principle", "neg_sample": ["inter - learner uniformity is done by using Method", "child language learners develop with remarkable uniformity , both in their learning trajectories and ultimate outcomes , despite major differences in their learning environments ."], "relation": "used for", "id": "2022.cmcl-1.7", "year": 2022, "rel_sent": "We conclude that while the Tolerance Principle , a type - based model of productivity learning , accounts for inter - learner uniformity , it also interacts with input distributions to drive cross - linguistic variation in learning trajectories .", "forward": false, "src_ids": "2022.cmcl-1.7_279"}
+{"input": "tolerance principle is used for OtherScientificTerm| context: child language learners develop with remarkable uniformity , both in their learning trajectories and ultimate outcomes , despite major differences in their learning environments .", "entity": "tolerance principle", "output": "inter - learner uniformity", "neg_sample": ["tolerance principle is used for OtherScientificTerm", "child language learners develop with remarkable uniformity , both in their learning trajectories and ultimate outcomes , despite major differences in their learning environments ."], "relation": "used for", "id": "2022.cmcl-1.7", "year": 2022, "rel_sent": "We conclude that while the Tolerance Principle , a type - based model of productivity learning , accounts for inter - learner uniformity , it also interacts with input distributions to drive cross - linguistic variation in learning trajectories .", "forward": true, "src_ids": "2022.cmcl-1.7_280"}
+{"input": "emotion detection is done by using Method| context: detecting emotions in languages is important to accomplish a complete interaction between humans and machines .", "entity": "emotion detection", "output": "transformer based ensemble", "neg_sample": ["emotion detection is done by using Method", "detecting emotions in languages is important to accomplish a complete interaction between humans and machines ."], "relation": "used for", "id": "2022.wassa-1.25", "year": 2022, "rel_sent": "Transformer based ensemble for emotion detection.", "forward": false, "src_ids": "2022.wassa-1.25_281"}
+{"input": "transformer based ensemble is used for Task| context: detecting emotions in languages is important to accomplish a complete interaction between humans and machines .", "entity": "transformer based ensemble", "output": "emotion detection", "neg_sample": ["transformer based ensemble is used for Task", "detecting emotions in languages is important to accomplish a complete interaction between humans and machines ."], "relation": "used for", "id": "2022.wassa-1.25", "year": 2022, "rel_sent": "Transformer based ensemble for emotion detection.", "forward": true, "src_ids": "2022.wassa-1.25_282"}
+{"input": "wassa 2022 shared task is used for Task| context: detecting emotions in languages is important to accomplish a complete interaction between humans and machines .", "entity": "wassa 2022 shared task", "output": "emotion detection", "neg_sample": ["wassa 2022 shared task is used for Task", "detecting emotions in languages is important to accomplish a complete interaction between humans and machines ."], "relation": "used for", "id": "2022.wassa-1.25", "year": 2022, "rel_sent": "This paper describes our contribution to the WASSA 2022 shared task which handles this crucial task of emotion detection .", "forward": true, "src_ids": "2022.wassa-1.25_283"}
+{"input": "emotion detection is done by using Task| context: detecting emotions in languages is important to accomplish a complete interaction between humans and machines .", "entity": "emotion detection", "output": "wassa 2022 shared task", "neg_sample": ["emotion detection is done by using Task", "detecting emotions in languages is important to accomplish a complete interaction between humans and machines ."], "relation": "used for", "id": "2022.wassa-1.25", "year": 2022, "rel_sent": "This paper describes our contribution to the WASSA 2022 shared task which handles this crucial task of emotion detection .", "forward": false, "src_ids": "2022.wassa-1.25_284"}
+{"input": "re - ranking is done by using Task| context: our proposed novelties address two weaknesses in the literature .", "entity": "re - ranking", "output": "match prediction", "neg_sample": ["re - ranking is done by using Task", "our proposed novelties address two weaknesses in the literature ."], "relation": "used for", "id": "2022.findings-acl.156", "year": 2022, "rel_sent": "We add a new , auxiliary task , match prediction , to learn re - ranking .", "forward": false, "src_ids": "2022.findings-acl.156_285"}
+{"input": "prediction errors is done by using Task| context: our proposed novelties address two weaknesses in the literature .", "entity": "prediction errors", "output": "re - ranking", "neg_sample": ["prediction errors is done by using Task", "our proposed novelties address two weaknesses in the literature ."], "relation": "used for", "id": "2022.findings-acl.156", "year": 2022, "rel_sent": "Second , previous work suggests that re - ranking could help correct prediction errors .", "forward": false, "src_ids": "2022.findings-acl.156_286"}
+{"input": "match prediction is used for Task| context: our proposed novelties address two weaknesses in the literature .", "entity": "match prediction", "output": "re - ranking", "neg_sample": ["match prediction is used for Task", "our proposed novelties address two weaknesses in the literature ."], "relation": "used for", "id": "2022.findings-acl.156", "year": 2022, "rel_sent": "We add a new , auxiliary task , match prediction , to learn re - ranking .", "forward": true, "src_ids": "2022.findings-acl.156_287"}
+{"input": "re - ranking is used for OtherScientificTerm| context: our proposed novelties address two weaknesses in the literature .", "entity": "re - ranking", "output": "prediction errors", "neg_sample": ["re - ranking is used for OtherScientificTerm", "our proposed novelties address two weaknesses in the literature ."], "relation": "used for", "id": "2022.findings-acl.156", "year": 2022, "rel_sent": "Second , previous work suggests that re - ranking could help correct prediction errors .", "forward": true, "src_ids": "2022.findings-acl.156_288"}
+{"input": "similes is done by using Method| context: simile interpretation is a crucial task in natural language processing .", "entity": "similes", "output": "pre - trained language models", "neg_sample": ["similes is done by using Method", "simile interpretation is a crucial task in natural language processing ."], "relation": "used for", "id": "2022.acl-long.543", "year": 2022, "rel_sent": "Can Pre - trained Language Models Interpret Similes as Smart as Human ?.", "forward": false, "src_ids": "2022.acl-long.543_289"}
+{"input": "simile interpretation is done by using Method| context: simile interpretation is a crucial task in natural language processing .", "entity": "simile interpretation", "output": "pre - trained language models", "neg_sample": ["simile interpretation is done by using Method", "simile interpretation is a crucial task in natural language processing ."], "relation": "used for", "id": "2022.acl-long.543", "year": 2022, "rel_sent": "In this paper , we investigate the ability of PLMs in simile interpretation by designing a novel task named Simile Property Probing , i.e. , to let the PLMs infer the shared properties of similes .", "forward": false, "src_ids": "2022.acl-long.543_290"}
+{"input": "shared properties of similes is done by using Method| context: simile interpretation is a crucial task in natural language processing .", "entity": "shared properties of similes", "output": "pre - trained language models", "neg_sample": ["shared properties of similes is done by using Method", "simile interpretation is a crucial task in natural language processing ."], "relation": "used for", "id": "2022.acl-long.543", "year": 2022, "rel_sent": "In this paper , we investigate the ability of PLMs in simile interpretation by designing a novel task named Simile Property Probing , i.e. , to let the PLMs infer the shared properties of similes .", "forward": false, "src_ids": "2022.acl-long.543_291"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: simile interpretation is a crucial task in natural language processing . nowadays , pre - trained language models ( plms ) have achieved state - of - the - art performance on many tasks .", "entity": "pre - trained language models", "output": "similes", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "simile interpretation is a crucial task in natural language processing .", "nowadays , pre - trained language models ( plms ) have achieved state - of - the - art performance on many tasks ."], "relation": "used for", "id": "2022.acl-long.543", "year": 2022, "rel_sent": "Can Pre - trained Language Models Interpret Similes as Smart as Human ?.", "forward": true, "src_ids": "2022.acl-long.543_292"}
+{"input": "pre - trained language models is used for Task| context: nowadays , pre - trained language models ( plms ) have achieved state - of - the - art performance on many tasks . however , it remains under - explored whether plms can interpret similes or not .", "entity": "pre - trained language models", "output": "simile interpretation", "neg_sample": ["pre - trained language models is used for Task", "nowadays , pre - trained language models ( plms ) have achieved state - of - the - art performance on many tasks .", "however , it remains under - explored whether plms can interpret similes or not ."], "relation": "used for", "id": "2022.acl-long.543", "year": 2022, "rel_sent": "In this paper , we investigate the ability of PLMs in simile interpretation by designing a novel task named Simile Property Probing , i.e. , to let the PLMs infer the shared properties of similes .", "forward": true, "src_ids": "2022.acl-long.543_293"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: simile interpretation is a crucial task in natural language processing . nowadays , pre - trained language models ( plms ) have achieved state - of - the - art performance on many tasks . however , it remains under - explored whether plms can interpret similes or not .", "entity": "pre - trained language models", "output": "shared properties of similes", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "simile interpretation is a crucial task in natural language processing .", "nowadays , pre - trained language models ( plms ) have achieved state - of - the - art performance on many tasks .", "however , it remains under - explored whether plms can interpret similes or not ."], "relation": "used for", "id": "2022.acl-long.543", "year": 2022, "rel_sent": "In this paper , we investigate the ability of PLMs in simile interpretation by designing a novel task named Simile Property Probing , i.e. , to let the PLMs infer the shared properties of similes .", "forward": true, "src_ids": "2022.acl-long.543_294"}
+{"input": "sentiment - aware context is done by using Method| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "sentiment - aware context", "output": "bisyn - gat+", "neg_sample": ["sentiment - aware context is done by using Method", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "Specifically , BiSyn - GAT+ fully exploits the syntax information ( e.g. , phrase segmentation and hierarchical structure ) of the constituent tree of a sentence to model the sentiment - aware context of every single aspect ( called intra - context ) and the sentiment relations across aspects ( called inter - context ) for learning .", "forward": false, "src_ids": "2022.findings-acl.144_295"}
+{"input": "syntax information is done by using Method| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "syntax information", "output": "bisyn - gat+", "neg_sample": ["syntax information is done by using Method", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "Specifically , BiSyn - GAT+ fully exploits the syntax information ( e.g. , phrase segmentation and hierarchical structure ) of the constituent tree of a sentence to model the sentiment - aware context of every single aspect ( called intra - context ) and the sentiment relations across aspects ( called inter - context ) for learning .", "forward": false, "src_ids": "2022.findings-acl.144_296"}
+{"input": "aspect - based sentiment analysis is done by using Method| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "aspect - based sentiment analysis", "output": "bi - syntax aware graph attention network", "neg_sample": ["aspect - based sentiment analysis is done by using Method", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "BiSyn - GAT+ : Bi - Syntax Aware Graph Attention Network for Aspect - based Sentiment Analysis.", "forward": false, "src_ids": "2022.findings-acl.144_297"}
+{"input": "bi - syntax aware graph attention network is used for Task| context: it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "bi - syntax aware graph attention network", "output": "aspect - based sentiment analysis", "neg_sample": ["bi - syntax aware graph attention network is used for Task", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "BiSyn - GAT+ : Bi - Syntax Aware Graph Attention Network for Aspect - based Sentiment Analysis.", "forward": true, "src_ids": "2022.findings-acl.144_298"}
+{"input": "bisyn - gat+ is used for OtherScientificTerm| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "bisyn - gat+", "output": "syntax information", "neg_sample": ["bisyn - gat+ is used for OtherScientificTerm", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "Specifically , BiSyn - GAT+ fully exploits the syntax information ( e.g. , phrase segmentation and hierarchical structure ) of the constituent tree of a sentence to model the sentiment - aware context of every single aspect ( called intra - context ) and the sentiment relations across aspects ( called inter - context ) for learning .", "forward": true, "src_ids": "2022.findings-acl.144_299"}
+{"input": "bisyn - gat+ is used for OtherScientificTerm| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "bisyn - gat+", "output": "sentiment - aware context", "neg_sample": ["bisyn - gat+ is used for OtherScientificTerm", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "Specifically , BiSyn - GAT+ fully exploits the syntax information ( e.g. , phrase segmentation and hierarchical structure ) of the constituent tree of a sentence to model the sentiment - aware context of every single aspect ( called intra - context ) and the sentiment relations across aspects ( called inter - context ) for learning .", "forward": true, "src_ids": "2022.findings-acl.144_300"}
+{"input": "learning is done by using OtherScientificTerm| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "learning", "output": "sentiment relations", "neg_sample": ["learning is done by using OtherScientificTerm", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "Specifically , BiSyn - GAT+ fully exploits the syntax information ( e.g. , phrase segmentation and hierarchical structure ) of the constituent tree of a sentence to model the sentiment - aware context of every single aspect ( called intra - context ) and the sentiment relations across aspects ( called inter - context ) for learning .", "forward": false, "src_ids": "2022.findings-acl.144_301"}
+{"input": "sentiment relations is used for Task| context: aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference . it is challenging because a sentence may contain multiple aspects or complicated ( e.g. , conditional , coordinating , or adversative ) relations . recently , exploiting dependency syntax information with graph neural networks has been the most popular trend . despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g. , the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) .", "entity": "sentiment relations", "output": "learning", "neg_sample": ["sentiment relations is used for Task", "aspect - based sentiment analysis ( absa ) is a fine - grained sentiment analysis task that aims to align aspects and corresponding sentiments for aspect - specific sentiment polarity inference .", "it is challenging because a sentence may contain multiple aspects or complicated ( e.g.", ", conditional , coordinating , or adversative ) relations .", "recently , exploiting dependency syntax information with graph neural networks has been the most popular trend .", "despite its success , methods that heavily rely on the dependency tree pose challenges in accurately modeling the alignment of the aspects and their words indicative of sentiment , since the dependency tree may provide noisy signals of unrelated associations ( e.g.", ", the ' conj ' relation between ' great ' and ' dreadful ' in figure 2 ) ."], "relation": "used for", "id": "2022.findings-acl.144", "year": 2022, "rel_sent": "Specifically , BiSyn - GAT+ fully exploits the syntax information ( e.g. , phrase segmentation and hierarchical structure ) of the constituent tree of a sentence to model the sentiment - aware context of every single aspect ( called intra - context ) and the sentiment relations across aspects ( called inter - context ) for learning .", "forward": true, "src_ids": "2022.findings-acl.144_302"}
+{"input": "pretrained language models is done by using Method| context: effectively finetuning pretrained language models ( plms ) is critical for their success in downstream tasks . however , plms may have risks in overfitting the pretraining tasks and data , which usually have gap with the target downstream tasks . such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance .", "entity": "pretrained language models", "output": "noisytune", "neg_sample": ["pretrained language models is done by using Method", "effectively finetuning pretrained language models ( plms ) is critical for their success in downstream tasks .", "however , plms may have risks in overfitting the pretraining tasks and data , which usually have gap with the target downstream tasks .", "such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance ."], "relation": "used for", "id": "2022.acl-short.76", "year": 2022, "rel_sent": "In this paper , we propose a very simple yet effective method named NoisyTune to help better finetune PLMs on downstream tasks by adding some noise to the parameters of PLMs before fine - tuning .", "forward": false, "src_ids": "2022.acl-short.76_303"}
+{"input": "finetuning is done by using Method| context: effectively finetuning pretrained language models ( plms ) is critical for their success in downstream tasks . however , plms may have risks in overfitting the pretraining tasks and data , which usually have gap with the target downstream tasks . such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance .", "entity": "finetuning", "output": "noisytune", "neg_sample": ["finetuning is done by using Method", "effectively finetuning pretrained language models ( plms ) is critical for their success in downstream tasks .", "however , plms may have risks in overfitting the pretraining tasks and data , which usually have gap with the target downstream tasks .", "such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance ."], "relation": "used for", "id": "2022.acl-short.76", "year": 2022, "rel_sent": "Extensive experiments on both GLUE English benchmark and XTREME multilingual benchmark show NoisyTune can consistently empower the finetuning of different PLMs on different downstream tasks .", "forward": false, "src_ids": "2022.acl-short.76_304"}
+{"input": "downstream tasks is done by using Method| context: such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance .", "entity": "downstream tasks", "output": "pretrained language models", "neg_sample": ["downstream tasks is done by using Method", "such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance ."], "relation": "used for", "id": "2022.acl-short.76", "year": 2022, "rel_sent": "In this paper , we propose a very simple yet effective method named NoisyTune to help better finetune PLMs on downstream tasks by adding some noise to the parameters of PLMs before fine - tuning .", "forward": false, "src_ids": "2022.acl-short.76_305"}
+{"input": "noisytune is used for Method| context: such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance .", "entity": "noisytune", "output": "pretrained language models", "neg_sample": ["noisytune is used for Method", "such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance ."], "relation": "used for", "id": "2022.acl-short.76", "year": 2022, "rel_sent": "In this paper , we propose a very simple yet effective method named NoisyTune to help better finetune PLMs on downstream tasks by adding some noise to the parameters of PLMs before fine - tuning .", "forward": true, "src_ids": "2022.acl-short.76_306"}
+{"input": "pretrained language models is used for Task| context: such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance .", "entity": "pretrained language models", "output": "downstream tasks", "neg_sample": ["pretrained language models is used for Task", "such gap may be difficult for existing plm finetuning methods to overcome and lead to suboptimal performance ."], "relation": "used for", "id": "2022.acl-short.76", "year": 2022, "rel_sent": "In this paper , we propose a very simple yet effective method named NoisyTune to help better finetune PLMs on downstream tasks by adding some noise to the parameters of PLMs before fine - tuning .", "forward": true, "src_ids": "2022.acl-short.76_307"}
+{"input": "noisytune is used for Method| context: however , plms may have risks in overfitting the pretraining tasks and data , which usually have gap with the target downstream tasks .", "entity": "noisytune", "output": "finetuning", "neg_sample": ["noisytune is used for Method", "however , plms may have risks in overfitting the pretraining tasks and data , which usually have gap with the target downstream tasks ."], "relation": "used for", "id": "2022.acl-short.76", "year": 2022, "rel_sent": "Extensive experiments on both GLUE English benchmark and XTREME multilingual benchmark show NoisyTune can consistently empower the finetuning of different PLMs on different downstream tasks .", "forward": true, "src_ids": "2022.acl-short.76_308"}
+{"input": "unsupervised simple definition generation is done by using Method| context: the definition generation task can help language learners by providing explanations for unfamiliar words . this task has attracted much attention in recent years . a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training .", "entity": "unsupervised simple definition generation", "output": "multitasking framework", "neg_sample": ["unsupervised simple definition generation is done by using Method", "the definition generation task can help language learners by providing explanations for unfamiliar words .", "this task has attracted much attention in recent years .", "a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training ."], "relation": "used for", "id": "2022.acl-long.409", "year": 2022, "rel_sent": "Multitasking Framework for Unsupervised Simple Definition Generation.", "forward": false, "src_ids": "2022.acl-long.409_309"}
+{"input": "multitasking framework is used for Task| context: the definition generation task can help language learners by providing explanations for unfamiliar words . this task has attracted much attention in recent years . a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training .", "entity": "multitasking framework", "output": "unsupervised simple definition generation", "neg_sample": ["multitasking framework is used for Task", "the definition generation task can help language learners by providing explanations for unfamiliar words .", "this task has attracted much attention in recent years .", "a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training ."], "relation": "used for", "id": "2022.acl-long.409", "year": 2022, "rel_sent": "Multitasking Framework for Unsupervised Simple Definition Generation.", "forward": true, "src_ids": "2022.acl-long.409_310"}
+{"input": "simple definition generation ( sdg ) is used for OtherScientificTerm| context: this task has attracted much attention in recent years . a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training .", "entity": "simple definition generation ( sdg )", "output": "language learners", "neg_sample": ["simple definition generation ( sdg ) is used for OtherScientificTerm", "this task has attracted much attention in recent years .", "a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training ."], "relation": "used for", "id": "2022.acl-long.409", "year": 2022, "rel_sent": "We propose a novel task of Simple Definition Generation ( SDG ) to help language learners and low literacy readers .", "forward": true, "src_ids": "2022.acl-long.409_311"}
+{"input": "language learners is done by using Task| context: the definition generation task can help language learners by providing explanations for unfamiliar words . this task has attracted much attention in recent years . a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training .", "entity": "language learners", "output": "simple definition generation ( sdg )", "neg_sample": ["language learners is done by using Task", "the definition generation task can help language learners by providing explanations for unfamiliar words .", "this task has attracted much attention in recent years .", "a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training ."], "relation": "used for", "id": "2022.acl-long.409", "year": 2022, "rel_sent": "We propose a novel task of Simple Definition Generation ( SDG ) to help language learners and low literacy readers .", "forward": false, "src_ids": "2022.acl-long.409_312"}
+{"input": "decoders is done by using Method| context: the definition generation task can help language learners by providing explanations for unfamiliar words . this task has attracted much attention in recent years . a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training .", "entity": "decoders", "output": "parameter sharing scheme", "neg_sample": ["decoders is done by using Method", "the definition generation task can help language learners by providing explanations for unfamiliar words .", "this task has attracted much attention in recent years .", "a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training ."], "relation": "used for", "id": "2022.acl-long.409", "year": 2022, "rel_sent": "We disentangle the complexity factors from the text by carefully designing a parameter sharing scheme between two decoders .", "forward": false, "src_ids": "2022.acl-long.409_313"}
+{"input": "parameter sharing scheme is used for Method| context: the definition generation task can help language learners by providing explanations for unfamiliar words . this task has attracted much attention in recent years . a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training .", "entity": "parameter sharing scheme", "output": "decoders", "neg_sample": ["parameter sharing scheme is used for Method", "the definition generation task can help language learners by providing explanations for unfamiliar words .", "this task has attracted much attention in recent years .", "a significant challenge of this task is the lack of learner 's dictionaries in many languages , and therefore the lack of data for supervised training ."], "relation": "used for", "id": "2022.acl-long.409", "year": 2022, "rel_sent": "We disentangle the complexity factors from the text by carefully designing a parameter sharing scheme between two decoders .", "forward": true, "src_ids": "2022.acl-long.409_314"}
+{"input": "named entity recognition ( ner ) is done by using Method| context: despite the advances in digital healthcare systems offering curated structured knowledge , much of the critical information still lies in large volumes of unlabeled and unstructured clinical texts . these texts , which often contain protected health information ( phi ) , are exposed to information extraction tools for downstream applications , risking patient identification . existing works in de - identification rely on using large - scale annotated corpora in english , which often are not suitable in real - world multilingual settings . pre - trained language models ( lm ) have shown great potential for cross - lingual transfer in low - resource settings .", "entity": "named entity recognition ( ner )", "output": "lms", "neg_sample": ["named entity recognition ( ner ) is done by using Method", "despite the advances in digital healthcare systems offering curated structured knowledge , much of the critical information still lies in large volumes of unlabeled and unstructured clinical texts .", "these texts , which often contain protected health information ( phi ) , are exposed to information extraction tools for downstream applications , risking patient identification .", "existing works in de - identification rely on using large - scale annotated corpora in english , which often are not suitable in real - world multilingual settings .", "pre - trained language models ( lm ) have shown great potential for cross - lingual transfer in low - resource settings ."], "relation": "used for", "id": "2022.bionlp-1.20", "year": 2022, "rel_sent": "In this work , we empirically show the few - shot cross - lingual transfer property of LMs for named entity recognition ( NER ) and apply it to solve a low - resource and real - world challenge of code - mixed ( Spanish - Catalan ) clinical notes de - identification in the stroke domain .", "forward": false, "src_ids": "2022.bionlp-1.20_315"}
+{"input": "lms is used for Task| context: despite the advances in digital healthcare systems offering curated structured knowledge , much of the critical information still lies in large volumes of unlabeled and unstructured clinical texts . these texts , which often contain protected health information ( phi ) , are exposed to information extraction tools for downstream applications , risking patient identification . existing works in de - identification rely on using large - scale annotated corpora in english , which often are not suitable in real - world multilingual settings . pre - trained language models ( lm ) have shown great potential for cross - lingual transfer in low - resource settings .", "entity": "lms", "output": "named entity recognition ( ner )", "neg_sample": ["lms is used for Task", "despite the advances in digital healthcare systems offering curated structured knowledge , much of the critical information still lies in large volumes of unlabeled and unstructured clinical texts .", "these texts , which often contain protected health information ( phi ) , are exposed to information extraction tools for downstream applications , risking patient identification .", "existing works in de - identification rely on using large - scale annotated corpora in english , which often are not suitable in real - world multilingual settings .", "pre - trained language models ( lm ) have shown great potential for cross - lingual transfer in low - resource settings ."], "relation": "used for", "id": "2022.bionlp-1.20", "year": 2022, "rel_sent": "In this work , we empirically show the few - shot cross - lingual transfer property of LMs for named entity recognition ( NER ) and apply it to solve a low - resource and real - world challenge of code - mixed ( Spanish - Catalan ) clinical notes de - identification in the stroke domain .", "forward": true, "src_ids": "2022.bionlp-1.20_316"}
+{"input": "reasoning - based machine reading comprehension is done by using Method| context: recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text . conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text .", "entity": "reasoning - based machine reading comprehension", "output": "adaptive logic graph network ( adalogn )", "neg_sample": ["reasoning - based machine reading comprehension is done by using Method", "recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text .", "conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text ."], "relation": "used for", "id": "2022.acl-long.494", "year": 2022, "rel_sent": "AdaLoGN : Adaptive Logic Graph Network for Reasoning - Based Machine Reading Comprehension.", "forward": false, "src_ids": "2022.acl-long.494_317"}
+{"input": "logical relations is done by using Method| context: recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text . conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text .", "entity": "logical relations", "output": "adaptive logic graph network ( adalogn )", "neg_sample": ["logical relations is done by using Method", "recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text .", "conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text ."], "relation": "used for", "id": "2022.acl-long.494", "year": 2022, "rel_sent": "It incorporates an adaptive logic graph network ( AdaLoGN ) which adaptively infers logical relations to extend the graph and , essentially , realizes mutual and iterative reinforcement between neural and symbolic reasoning .", "forward": false, "src_ids": "2022.acl-long.494_318"}
+{"input": "adaptive logic graph network ( adalogn ) is used for Task| context: recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text . conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text .", "entity": "adaptive logic graph network ( adalogn )", "output": "reasoning - based machine reading comprehension", "neg_sample": ["adaptive logic graph network ( adalogn ) is used for Task", "recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text .", "conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text ."], "relation": "used for", "id": "2022.acl-long.494", "year": 2022, "rel_sent": "AdaLoGN : Adaptive Logic Graph Network for Reasoning - Based Machine Reading Comprehension.", "forward": true, "src_ids": "2022.acl-long.494_319"}
+{"input": "adaptive logic graph network ( adalogn ) is used for OtherScientificTerm| context: recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text . conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text .", "entity": "adaptive logic graph network ( adalogn )", "output": "logical relations", "neg_sample": ["adaptive logic graph network ( adalogn ) is used for OtherScientificTerm", "recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text .", "conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text ."], "relation": "used for", "id": "2022.acl-long.494", "year": 2022, "rel_sent": "It incorporates an adaptive logic graph network ( AdaLoGN ) which adaptively infers logical relations to extend the graph and , essentially , realizes mutual and iterative reinforcement between neural and symbolic reasoning .", "forward": true, "src_ids": "2022.acl-long.494_320"}
+{"input": "context - option interaction is done by using Method| context: recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text . conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text .", "entity": "context - option interaction", "output": "subgraph - to - node message passing mechanism", "neg_sample": ["context - option interaction is done by using Method", "recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text .", "conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text ."], "relation": "used for", "id": "2022.acl-long.494", "year": 2022, "rel_sent": "We also implement a novel subgraph - to - node message passing mechanism to enhance context - option interaction for answering multiple - choice questions .", "forward": false, "src_ids": "2022.acl-long.494_321"}
+{"input": "subgraph - to - node message passing mechanism is used for OtherScientificTerm| context: recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text . conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text .", "entity": "subgraph - to - node message passing mechanism", "output": "context - option interaction", "neg_sample": ["subgraph - to - node message passing mechanism is used for OtherScientificTerm", "recent machine reading comprehension datasets such as reclor and logiqa require performing logical reasoning over text .", "conventional neural models are insufficient for logical reasoning , while symbolic reasoners can not directly apply to text ."], "relation": "used for", "id": "2022.acl-long.494", "year": 2022, "rel_sent": "We also implement a novel subgraph - to - node message passing mechanism to enhance context - option interaction for answering multiple - choice questions .", "forward": true, "src_ids": "2022.acl-long.494_322"}
+{"input": "crafty adversarial attacks is used for Method| context: social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale .", "entity": "crafty adversarial attacks", "output": "offensive language classifiers", "neg_sample": ["crafty adversarial attacks is used for Method", "social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale ."], "relation": "used for", "id": "2022.acl-long.513", "year": 2022, "rel_sent": "Our results on multiple datasets show that these crafty adversarial attacks can degrade the accuracy of offensive language classifiers by more than 50 % while also being able to preserve the readability and meaning of the modified text .", "forward": true, "src_ids": "2022.acl-long.513_323"}
+{"input": "offensive language classifiers is done by using Method| context: social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale . however , despite their real - world deployment , we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks . prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces .", "entity": "offensive language classifiers", "output": "crafty adversarial attacks", "neg_sample": ["offensive language classifiers is done by using Method", "social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale .", "however , despite their real - world deployment , we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks .", "prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces ."], "relation": "used for", "id": "2022.acl-long.513", "year": 2022, "rel_sent": "Our results on multiple datasets show that these crafty adversarial attacks can degrade the accuracy of offensive language classifiers by more than 50 % while also being able to preserve the readability and meaning of the modified text .", "forward": false, "src_ids": "2022.acl-long.513_324"}
+{"input": "word replacement is done by using OtherScientificTerm| context: social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale . however , despite their real - world deployment , we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks . prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces .", "entity": "word replacement", "output": "context - aware embeddings", "neg_sample": ["word replacement is done by using OtherScientificTerm", "social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale .", "however , despite their real - world deployment , we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks .", "prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces ."], "relation": "used for", "id": "2022.acl-long.513", "year": 2022, "rel_sent": "To address this gap , we systematically analyze the robustness of state - of - the - art offensive language classifiers against more crafty adversarial attacks that leverage greedy- and attention - based word selection and context - aware embeddings for word replacement .", "forward": false, "src_ids": "2022.acl-long.513_325"}
+{"input": "context - aware embeddings is used for Method| context: social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale . however , despite their real - world deployment , we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks . prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces .", "entity": "context - aware embeddings", "output": "word replacement", "neg_sample": ["context - aware embeddings is used for Method", "social media platforms are deploying machine learning based offensive language classification systems to combat hateful , racist , and other forms of offensive speech at scale .", "however , despite their real - world deployment , we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks .", "prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces ."], "relation": "used for", "id": "2022.acl-long.513", "year": 2022, "rel_sent": "To address this gap , we systematically analyze the robustness of state - of - the - art offensive language classifiers against more crafty adversarial attacks that leverage greedy- and attention - based word selection and context - aware embeddings for word replacement .", "forward": true, "src_ids": "2022.acl-long.513_326"}
+{"input": "semantic similarity is used for Task| context: word and sentence embeddings are useful feature representations in natural language processing . however , intrinsic evaluation for embeddings lags far behind , and there has been no significant update since the past decade . word and sentence similarity tasks have become the de facto evaluation method . it leads models to overfit to such evaluations , negatively impacting embedding models ' development .", "entity": "semantic similarity", "output": "word and sentence embedding evaluations", "neg_sample": ["semantic similarity is used for Task", "word and sentence embeddings are useful feature representations in natural language processing .", "however , intrinsic evaluation for embeddings lags far behind , and there has been no significant update since the past decade .", "word and sentence similarity tasks have become the de facto evaluation method .", "it leads models to overfit to such evaluations , negatively impacting embedding models ' development ."], "relation": "used for", "id": "2022.acl-long.419", "year": 2022, "rel_sent": "This paper first points out the problems using semantic similarity as the gold standard for word and sentence embedding evaluations .", "forward": true, "src_ids": "2022.acl-long.419_327"}
+{"input": "downstream tasks is done by using Method| context: word and sentence embeddings are useful feature representations in natural language processing . word and sentence similarity tasks have become the de facto evaluation method . it leads models to overfit to such evaluations , negatively impacting embedding models ' development .", "entity": "downstream tasks", "output": "intrinsic evaluation method", "neg_sample": ["downstream tasks is done by using Method", "word and sentence embeddings are useful feature representations in natural language processing .", "word and sentence similarity tasks have become the de facto evaluation method .", "it leads models to overfit to such evaluations , negatively impacting embedding models ' development ."], "relation": "used for", "id": "2022.acl-long.419", "year": 2022, "rel_sent": "Further , we propose a new intrinsic evaluation method called EvalRank , which shows a much stronger correlation with downstream tasks .", "forward": false, "src_ids": "2022.acl-long.419_328"}
+{"input": "intrinsic evaluation method is used for Task| context: word and sentence embeddings are useful feature representations in natural language processing . however , intrinsic evaluation for embeddings lags far behind , and there has been no significant update since the past decade . word and sentence similarity tasks have become the de facto evaluation method . it leads models to overfit to such evaluations , negatively impacting embedding models ' development .", "entity": "intrinsic evaluation method", "output": "downstream tasks", "neg_sample": ["intrinsic evaluation method is used for Task", "word and sentence embeddings are useful feature representations in natural language processing .", "however , intrinsic evaluation for embeddings lags far behind , and there has been no significant update since the past decade .", "word and sentence similarity tasks have become the de facto evaluation method .", "it leads models to overfit to such evaluations , negatively impacting embedding models ' development ."], "relation": "used for", "id": "2022.acl-long.419", "year": 2022, "rel_sent": "Further , we propose a new intrinsic evaluation method called EvalRank , which shows a much stronger correlation with downstream tasks .", "forward": true, "src_ids": "2022.acl-long.419_329"}
+{"input": "word and sentence embedding evaluations is done by using OtherScientificTerm| context: word and sentence embeddings are useful feature representations in natural language processing . however , intrinsic evaluation for embeddings lags far behind , and there has been no significant update since the past decade . word and sentence similarity tasks have become the de facto evaluation method . it leads models to overfit to such evaluations , negatively impacting embedding models ' development .", "entity": "word and sentence embedding evaluations", "output": "semantic similarity", "neg_sample": ["word and sentence embedding evaluations is done by using OtherScientificTerm", "word and sentence embeddings are useful feature representations in natural language processing .", "however , intrinsic evaluation for embeddings lags far behind , and there has been no significant update since the past decade .", "word and sentence similarity tasks have become the de facto evaluation method .", "it leads models to overfit to such evaluations , negatively impacting embedding models ' development ."], "relation": "used for", "id": "2022.acl-long.419", "year": 2022, "rel_sent": "This paper first points out the problems using semantic similarity as the gold standard for word and sentence embedding evaluations .", "forward": false, "src_ids": "2022.acl-long.419_330"}
+{"input": "dense retrieval is done by using Task| context: dense retrieval aims at searching for the most relevant documents to the given query by encoding texts in the embedding space , requiring a large amount of query - document pairs to train . since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever . however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance .", "entity": "dense retrieval", "output": "query generation", "neg_sample": ["dense retrieval is done by using Task", "dense retrieval aims at searching for the most relevant documents to the given query by encoding texts in the embedding space , requiring a large amount of query - document pairs to train .", "since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever .", "however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance ."], "relation": "used for", "id": "2022.deelio-1.3", "year": 2022, "rel_sent": "Query Generation with External Knowledge for Dense Retrieval.", "forward": false, "src_ids": "2022.deelio-1.3_331"}
+{"input": "query generation is used for Task| context: since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever . however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance .", "entity": "query generation", "output": "dense retrieval", "neg_sample": ["query generation is used for Task", "since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever .", "however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance ."], "relation": "used for", "id": "2022.deelio-1.3", "year": 2022, "rel_sent": "Query Generation with External Knowledge for Dense Retrieval.", "forward": true, "src_ids": "2022.deelio-1.3_332"}
+{"input": "triplet - based template form is used for OtherScientificTerm| context: dense retrieval aims at searching for the most relevant documents to the given query by encoding texts in the embedding space , requiring a large amount of query - document pairs to train . since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever . however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance .", "entity": "triplet - based template form", "output": "external information", "neg_sample": ["triplet - based template form is used for OtherScientificTerm", "dense retrieval aims at searching for the most relevant documents to the given query by encoding texts in the embedding space , requiring a large amount of query - document pairs to train .", "since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever .", "however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance ."], "relation": "used for", "id": "2022.deelio-1.3", "year": 2022, "rel_sent": "Specifically , we convert a query into a triplet - based template form to accommodate external information and transmit it to a pre - trained language model ( PLM ) .", "forward": true, "src_ids": "2022.deelio-1.3_333"}
+{"input": "external information is done by using OtherScientificTerm| context: dense retrieval aims at searching for the most relevant documents to the given query by encoding texts in the embedding space , requiring a large amount of query - document pairs to train . since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever . however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance .", "entity": "external information", "output": "triplet - based template form", "neg_sample": ["external information is done by using OtherScientificTerm", "dense retrieval aims at searching for the most relevant documents to the given query by encoding texts in the embedding space , requiring a large amount of query - document pairs to train .", "since manually constructing such training data is challenging , recent work has proposed to generate synthetic queries from documents and use them to train a dense retriever .", "however , compared to the manually composed queries , synthetic queries do not generally ask for implicit information , therefore leading to a degraded retrieval performance ."], "relation": "used for", "id": "2022.deelio-1.3", "year": 2022, "rel_sent": "Specifically , we convert a query into a triplet - based template form to accommodate external information and transmit it to a pre - trained language model ( PLM ) .", "forward": false, "src_ids": "2022.deelio-1.3_334"}
+{"input": "alignment is used for Material| context: simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy .", "entity": "alignment", "output": "streaming input", "neg_sample": ["alignment is used for Material", "simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy ."], "relation": "used for", "id": "2022.iwslt-1.3", "year": 2022, "rel_sent": "We use the alignment to segment a streaming input and fine - tune a translation model .", "forward": true, "src_ids": "2022.iwslt-1.3_335"}
+{"input": "alignment is used for Method| context: simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy .", "entity": "alignment", "output": "translation model", "neg_sample": ["alignment is used for Method", "simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy ."], "relation": "used for", "id": "2022.iwslt-1.3", "year": 2022, "rel_sent": "We use the alignment to segment a streaming input and fine - tune a translation model .", "forward": true, "src_ids": "2022.iwslt-1.3_336"}
+{"input": "streaming input is done by using Task| context: simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy .", "entity": "streaming input", "output": "alignment", "neg_sample": ["streaming input is done by using Task", "simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy ."], "relation": "used for", "id": "2022.iwslt-1.3", "year": 2022, "rel_sent": "We use the alignment to segment a streaming input and fine - tune a translation model .", "forward": false, "src_ids": "2022.iwslt-1.3_337"}
+{"input": "translation model is done by using Task| context: simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy .", "entity": "translation model", "output": "alignment", "neg_sample": ["translation model is done by using Task", "simultaneous translation is a task that requires starting translation before the speaker has finished speaking , so we face a trade - off between latency and accuracy ."], "relation": "used for", "id": "2022.iwslt-1.3", "year": 2022, "rel_sent": "We use the alignment to segment a streaming input and fine - tune a translation model .", "forward": false, "src_ids": "2022.iwslt-1.3_338"}
+{"input": "programming language is done by using Method| context: code completion , which aims to predict the following code token(s ) according to the code context , can improve the productivity of software development . recent work has proved that statistical language modeling with transformers can greatly improve the performance in the code completion task via learning from large - scale source code datasets . however , current approaches focus only on code context within the file or project , i.e. internal context .", "entity": "programming language", "output": "auto - regressive language model", "neg_sample": ["programming language is done by using Method", "code completion , which aims to predict the following code token(s ) according to the code context , can improve the productivity of software development .", "recent work has proved that statistical language modeling with transformers can greatly improve the performance in the code completion task via learning from large - scale source code datasets .", "however , current approaches focus only on code context within the file or project , i.e.", "internal context ."], "relation": "used for", "id": "2022.acl-long.431", "year": 2022, "rel_sent": "We adopt a stage - wise training approach that combines a source code retriever and an auto - regressive language model for programming language .", "forward": false, "src_ids": "2022.acl-long.431_339"}
+{"input": "auto - regressive language model is used for OtherScientificTerm| context: code completion , which aims to predict the following code token(s ) according to the code context , can improve the productivity of software development . recent work has proved that statistical language modeling with transformers can greatly improve the performance in the code completion task via learning from large - scale source code datasets . however , current approaches focus only on code context within the file or project , i.e. internal context .", "entity": "auto - regressive language model", "output": "programming language", "neg_sample": ["auto - regressive language model is used for OtherScientificTerm", "code completion , which aims to predict the following code token(s ) according to the code context , can improve the productivity of software development .", "recent work has proved that statistical language modeling with transformers can greatly improve the performance in the code completion task via learning from large - scale source code datasets .", "however , current approaches focus only on code context within the file or project , i.e.", "internal context ."], "relation": "used for", "id": "2022.acl-long.431", "year": 2022, "rel_sent": "We adopt a stage - wise training approach that combines a source code retriever and an auto - regressive language model for programming language .", "forward": true, "src_ids": "2022.acl-long.431_340"}
+{"input": "cross - lingual phrase retriever is used for Method| context: current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level .", "entity": "cross - lingual phrase retriever", "output": "phrase representations", "neg_sample": ["cross - lingual phrase retriever is used for Method", "current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level ."], "relation": "used for", "id": "2022.acl-long.288", "year": 2022, "rel_sent": "In this paper , we propose , a cross - lingual phrase retriever that extracts phrase representations from unlabeled example sentences .", "forward": true, "src_ids": "2022.acl-long.288_341"}
+{"input": "phrase representations is done by using Method| context: current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level . however , how to learn phrase representations for cross - lingual phrase retrieval is still an open problem .", "entity": "phrase representations", "output": "cross - lingual phrase retriever", "neg_sample": ["phrase representations is done by using Method", "current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level .", "however , how to learn phrase representations for cross - lingual phrase retrieval is still an open problem ."], "relation": "used for", "id": "2022.acl-long.288", "year": 2022, "rel_sent": "In this paper , we propose , a cross - lingual phrase retriever that extracts phrase representations from unlabeled example sentences .", "forward": false, "src_ids": "2022.acl-long.288_342"}
+{"input": "n - ary question answering is done by using Method| context: current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts . however , it neglects the n - ary facts , which contain more than two entities . in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e. , answering n - ary facts questions upon n - ary kgs . nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa .", "entity": "n - ary question answering", "output": "fact - tree reasoning framework", "neg_sample": ["n - ary question answering is done by using Method", "current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts .", "however , it neglects the n - ary facts , which contain more than two entities .", "in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e.", ", answering n - ary facts questions upon n - ary kgs .", "nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa ."], "relation": "used for", "id": "2022.findings-acl.66", "year": 2022, "rel_sent": "Fact - Tree Reasoning for N - ary Question Answering over Knowledge Graphs.", "forward": false, "src_ids": "2022.findings-acl.66_343"}
+{"input": "fact - tree reasoning framework is used for Task| context: current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts . however , it neglects the n - ary facts , which contain more than two entities . in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e. , answering n - ary facts questions upon n - ary kgs . nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa .", "entity": "fact - tree reasoning framework", "output": "n - ary question answering", "neg_sample": ["fact - tree reasoning framework is used for Task", "current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts .", "however , it neglects the n - ary facts , which contain more than two entities .", "in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e.", ", answering n - ary facts questions upon n - ary kgs .", "nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa ."], "relation": "used for", "id": "2022.findings-acl.66", "year": 2022, "rel_sent": "Fact - Tree Reasoning for N - ary Question Answering over Knowledge Graphs.", "forward": true, "src_ids": "2022.findings-acl.66_344"}
+{"input": "fact tree is done by using Method| context: current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts . however , it neglects the n - ary facts , which contain more than two entities . in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e. , answering n - ary facts questions upon n - ary kgs . nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa .", "entity": "fact tree", "output": "factree", "neg_sample": ["fact tree is done by using Method", "current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts .", "however , it neglects the n - ary facts , which contain more than two entities .", "in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e.", ", answering n - ary facts questions upon n - ary kgs .", "nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa ."], "relation": "used for", "id": "2022.findings-acl.66", "year": 2022, "rel_sent": "FacTree transforms the question into a fact tree and performs iterative fact reasoning on the fact tree to infer the correct answer .", "forward": false, "src_ids": "2022.findings-acl.66_345"}
+{"input": "factree is used for OtherScientificTerm| context: current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts . however , it neglects the n - ary facts , which contain more than two entities . in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e. , answering n - ary facts questions upon n - ary kgs . nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa .", "entity": "factree", "output": "fact tree", "neg_sample": ["factree is used for OtherScientificTerm", "current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts .", "however , it neglects the n - ary facts , which contain more than two entities .", "in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e.", ", answering n - ary facts questions upon n - ary kgs .", "nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa ."], "relation": "used for", "id": "2022.findings-acl.66", "year": 2022, "rel_sent": "FacTree transforms the question into a fact tree and performs iterative fact reasoning on the fact tree to infer the correct answer .", "forward": true, "src_ids": "2022.findings-acl.66_346"}
+{"input": "irish language learning is done by using Method| context: irish is a minority language which means that l2 learners have limited opportunities for exposure to the language , and additionally , there are also limited ( digital ) learning resources available .", "entity": "irish language learning", "output": "adaptive game", "neg_sample": ["irish language learning is done by using Method", "irish is a minority language which means that l2 learners have limited opportunities for exposure to the language , and additionally , there are also limited ( digital ) learning resources available ."], "relation": "used for", "id": "2022.computel-1.17", "year": 2022, "rel_sent": "Faoi Gheasa an adaptive game for Irish language learning.", "forward": false, "src_ids": "2022.computel-1.17_347"}
+{"input": "adaptive game is used for Task| context: irish is a minority language which means that l2 learners have limited opportunities for exposure to the language , and additionally , there are also limited ( digital ) learning resources available .", "entity": "adaptive game", "output": "irish language learning", "neg_sample": ["adaptive game is used for Task", "irish is a minority language which means that l2 learners have limited opportunities for exposure to the language , and additionally , there are also limited ( digital ) learning resources available ."], "relation": "used for", "id": "2022.computel-1.17", "year": 2022, "rel_sent": "Faoi Gheasa an adaptive game for Irish language learning.", "forward": true, "src_ids": "2022.computel-1.17_348"}
+{"input": "feed - forward networks is done by using OtherScientificTerm| context: recent work has shown that feed - forward networks ( ffns ) in pre - trained transformers are a key component , storing various linguistic and factual knowledge . however , the computational patterns of ffns are still unclear . in this work , we study the computational patterns of ffns and observe that most inputs only activate a tiny ratio of neurons of ffns . this phenomenon is similar to the sparsity of the human brain , which drives research on functional partitions of the human brain .", "entity": "feed - forward networks", "output": "fine - grained perspective", "neg_sample": ["feed - forward networks is done by using OtherScientificTerm", "recent work has shown that feed - forward networks ( ffns ) in pre - trained transformers are a key component , storing various linguistic and factual knowledge .", "however , the computational patterns of ffns are still unclear .", "in this work , we study the computational patterns of ffns and observe that most inputs only activate a tiny ratio of neurons of ffns .", "this phenomenon is similar to the sparsity of the human brain , which drives research on functional partitions of the human brain ."], "relation": "used for", "id": "2022.findings-acl.71", "year": 2022, "rel_sent": "Besides , MoEfication brings two advantages : ( 1 ) it significantly reduces the FLOPS of inference , i.e. , 2x speedup with 25 % of FFN parameters , and ( 2 ) it provides a fine - grained perspective to study the inner mechanism of FFNs .", "forward": false, "src_ids": "2022.findings-acl.71_349"}
+{"input": "fine - grained perspective is used for Method| context: this phenomenon is similar to the sparsity of the human brain , which drives research on functional partitions of the human brain .", "entity": "fine - grained perspective", "output": "feed - forward networks", "neg_sample": ["fine - grained perspective is used for Method", "this phenomenon is similar to the sparsity of the human brain , which drives research on functional partitions of the human brain ."], "relation": "used for", "id": "2022.findings-acl.71", "year": 2022, "rel_sent": "Besides , MoEfication brings two advantages : ( 1 ) it significantly reduces the FLOPS of inference , i.e. , 2x speedup with 25 % of FFN parameters , and ( 2 ) it provides a fine - grained perspective to study the inner mechanism of FFNs .", "forward": true, "src_ids": "2022.findings-acl.71_350"}
+{"input": "equality is done by using Task| context: hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals . hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways . hope speech offerssupport , reassurance , suggestions , inspirationand insight . hate speech is a prevalent practicethat society has to struggle with everyday . the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals .", "hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways .", "hope speech offerssupport , reassurance , suggestions , inspirationand insight .", "hate speech is a prevalent practicethat society has to struggle with everyday .", "the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred ."], "relation": "used for", "id": "2022.ltedi-1.32", "year": 2022, "rel_sent": "IIT Dhanbad @LT - EDI - ACL2022- Hope Speech Detection for Equality , Diversity , and Inclusion.", "forward": false, "src_ids": "2022.ltedi-1.32_351"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals . hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways . hope speech offerssupport , reassurance , suggestions , inspirationand insight . hate speech is a prevalent practicethat society has to struggle with everyday . the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals .", "hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways .", "hope speech offerssupport , reassurance , suggestions , inspirationand insight .", "hate speech is a prevalent practicethat society has to struggle with everyday .", "the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred ."], "relation": "used for", "id": "2022.ltedi-1.32", "year": 2022, "rel_sent": "IIT Dhanbad @LT - EDI - ACL2022- Hope Speech Detection for Equality , Diversity , and Inclusion.", "forward": true, "src_ids": "2022.ltedi-1.32_352"}
+{"input": "socialmedia comments is done by using Method| context: hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals . hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways . hope speech offerssupport , reassurance , suggestions , inspirationand insight . hate speech is a prevalent practicethat society has to struggle with everyday . the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred .", "entity": "socialmedia comments", "output": "machine learning models", "neg_sample": ["socialmedia comments is done by using Method", "hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals .", "hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways .", "hope speech offerssupport , reassurance , suggestions , inspirationand insight .", "hate speech is a prevalent practicethat society has to struggle with everyday .", "the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred ."], "relation": "used for", "id": "2022.ltedi-1.32", "year": 2022, "rel_sent": "We work withseveral machine learning models to classify socialmedia comments as hope speech or nonhopespeech in English .", "forward": false, "src_ids": "2022.ltedi-1.32_353"}
+{"input": "machine learning models is used for Material| context: hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals . hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways . hope speech offerssupport , reassurance , suggestions , inspirationand insight . hate speech is a prevalent practicethat society has to struggle with everyday . the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred .", "entity": "machine learning models", "output": "socialmedia comments", "neg_sample": ["machine learning models is used for Material", "hope is considered significant for the wellbeing , recuperation and restoration of humanlife by health professionals .", "hope speech reflectsthe belief that one can discover pathwaysto their desired objectives and become rousedto utilise those pathways .", "hope speech offerssupport , reassurance , suggestions , inspirationand insight .", "hate speech is a prevalent practicethat society has to struggle with everyday .", "the freedom of speech and ease of anonymitygranted by social media has also resulted inincitement to hatred ."], "relation": "used for", "id": "2022.ltedi-1.32", "year": 2022, "rel_sent": "We work withseveral machine learning models to classify socialmedia comments as hope speech or nonhopespeech in English .", "forward": true, "src_ids": "2022.ltedi-1.32_354"}
+{"input": "emotion analysis is done by using Method| context: using technology for analysis of human emotion is a relatively nascent research area . there are several types of data where emotion recognition can be employed , such as - text , images , audio and video . in this paper , the focus is on emotion recognition in text data . emotion recognition in text can be performed from both written comments and from conversations . while extensive research is being performed in this area , the language of the text plays a very important role . in this work , the focus is on the dravidian language of tamil .", "entity": "emotion analysis", "output": "cnn", "neg_sample": ["emotion analysis is done by using Method", "using technology for analysis of human emotion is a relatively nascent research area .", "there are several types of data where emotion recognition can be employed , such as - text , images , audio and video .", "in this paper , the focus is on emotion recognition in text data .", "emotion recognition in text can be performed from both written comments and from conversations .", "while extensive research is being performed in this area , the language of the text plays a very important role .", "in this work , the focus is on the dravidian language of tamil ."], "relation": "used for", "id": "2022.dravidianlangtech-1.9", "year": 2022, "rel_sent": "JudithJeyafreedaAndrew@TamilNLP - ACL2022 : CNN for Emotion Analysis in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.9_355"}
+{"input": "cnn is used for Task| context: using technology for analysis of human emotion is a relatively nascent research area . there are several types of data where emotion recognition can be employed , such as - text , images , audio and video . in this paper , the focus is on emotion recognition in text data . emotion recognition in text can be performed from both written comments and from conversations . while extensive research is being performed in this area , the language of the text plays a very important role . in this work , the focus is on the dravidian language of tamil .", "entity": "cnn", "output": "emotion analysis", "neg_sample": ["cnn is used for Task", "using technology for analysis of human emotion is a relatively nascent research area .", "there are several types of data where emotion recognition can be employed , such as - text , images , audio and video .", "in this paper , the focus is on emotion recognition in text data .", "emotion recognition in text can be performed from both written comments and from conversations .", "while extensive research is being performed in this area , the language of the text plays a very important role .", "in this work , the focus is on the dravidian language of tamil ."], "relation": "used for", "id": "2022.dravidianlangtech-1.9", "year": 2022, "rel_sent": "JudithJeyafreedaAndrew@TamilNLP - ACL2022 : CNN for Emotion Analysis in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.9_356"}
+{"input": "pre - processing methods is used for Material| context: using technology for analysis of human emotion is a relatively nascent research area . there are several types of data where emotion recognition can be employed , such as - text , images , audio and video . in this paper , the focus is on emotion recognition in text data . emotion recognition in text can be performed from both written comments and from conversations . while extensive research is being performed in this area , the language of the text plays a very important role .", "entity": "pre - processing methods", "output": "dravidian language of tamil", "neg_sample": ["pre - processing methods is used for Material", "using technology for analysis of human emotion is a relatively nascent research area .", "there are several types of data where emotion recognition can be employed , such as - text , images , audio and video .", "in this paper , the focus is on emotion recognition in text data .", "emotion recognition in text can be performed from both written comments and from conversations .", "while extensive research is being performed in this area , the language of the text plays a very important role ."], "relation": "used for", "id": "2022.dravidianlangtech-1.9", "year": 2022, "rel_sent": "The paper contributes to this by adapting various pre - processing methods to the Dravidian Language of Tamil .", "forward": true, "src_ids": "2022.dravidianlangtech-1.9_357"}
+{"input": "dravidian language of tamil is done by using Method| context: using technology for analysis of human emotion is a relatively nascent research area . there are several types of data where emotion recognition can be employed , such as - text , images , audio and video . in this paper , the focus is on emotion recognition in text data . emotion recognition in text can be performed from both written comments and from conversations . while extensive research is being performed in this area , the language of the text plays a very important role . in this work , the focus is on the dravidian language of tamil .", "entity": "dravidian language of tamil", "output": "pre - processing methods", "neg_sample": ["dravidian language of tamil is done by using Method", "using technology for analysis of human emotion is a relatively nascent research area .", "there are several types of data where emotion recognition can be employed , such as - text , images , audio and video .", "in this paper , the focus is on emotion recognition in text data .", "emotion recognition in text can be performed from both written comments and from conversations .", "while extensive research is being performed in this area , the language of the text plays a very important role .", "in this work , the focus is on the dravidian language of tamil ."], "relation": "used for", "id": "2022.dravidianlangtech-1.9", "year": 2022, "rel_sent": "The paper contributes to this by adapting various pre - processing methods to the Dravidian Language of Tamil .", "forward": false, "src_ids": "2022.dravidianlangtech-1.9_358"}
+{"input": "conversational bots is used for Task| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "conversational bots", "output": "psychotherapy", "neg_sample": ["conversational bots is used for Task", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "Conversational Bots for Psychotherapy : A Study of Generative Transformer Models Using Domain - specific Dialogues.", "forward": true, "src_ids": "2022.bionlp-1.27_359"}
+{"input": "deep trainable neural conversational model is done by using Method| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "deep trainable neural conversational model", "output": "deep neural generative language models", "neg_sample": ["deep trainable neural conversational model is done by using Method", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "Leveraging deep neural generative language models , we propose a deep trainable neural conversational model for therapy - oriented response generation .", "forward": false, "src_ids": "2022.bionlp-1.27_360"}
+{"input": "therapy - oriented response generation is done by using Method| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "therapy - oriented response generation", "output": "deep trainable neural conversational model", "neg_sample": ["therapy - oriented response generation is done by using Method", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "Leveraging deep neural generative language models , we propose a deep trainable neural conversational model for therapy - oriented response generation .", "forward": false, "src_ids": "2022.bionlp-1.27_361"}
+{"input": "deep neural generative language models is used for Method| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "deep neural generative language models", "output": "deep trainable neural conversational model", "neg_sample": ["deep neural generative language models is used for Method", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "Leveraging deep neural generative language models , we propose a deep trainable neural conversational model for therapy - oriented response generation .", "forward": true, "src_ids": "2022.bionlp-1.27_362"}
+{"input": "deep trainable neural conversational model is used for Task| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "deep trainable neural conversational model", "output": "therapy - oriented response generation", "neg_sample": ["deep trainable neural conversational model is used for Task", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "Leveraging deep neural generative language models , we propose a deep trainable neural conversational model for therapy - oriented response generation .", "forward": true, "src_ids": "2022.bionlp-1.27_363"}
+{"input": "therapy and counseling based data is done by using Method| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "therapy and counseling based data", "output": "transfer learning methods", "neg_sample": ["therapy and counseling based data is done by using Method", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "We leverage transfer learning methods during training on therapy and counseling based data from Reddit and AlexanderStreet .", "forward": false, "src_ids": "2022.bionlp-1.27_364"}
+{"input": "transfer learning methods is used for Material| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "transfer learning methods", "output": "therapy and counseling based data", "neg_sample": ["transfer learning methods is used for Material", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "We leverage transfer learning methods during training on therapy and counseling based data from Reddit and AlexanderStreet .", "forward": true, "src_ids": "2022.bionlp-1.27_365"}
+{"input": "automated dialog generation is done by using Method| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "automated dialog generation", "output": "generative models", "neg_sample": ["automated dialog generation is done by using Method", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "This was done to adapt existing generative models - GPT2 and DialoGPT - to the task of automated dialog generation .", "forward": false, "src_ids": "2022.bionlp-1.27_366"}
+{"input": "generative models is used for Task| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "generative models", "output": "automated dialog generation", "neg_sample": ["generative models is used for Task", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "This was done to adapt existing generative models - GPT2 and DialoGPT - to the task of automated dialog generation .", "forward": true, "src_ids": "2022.bionlp-1.27_367"}
+{"input": "dialogpt is used for Task| context: conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses .", "entity": "dialogpt", "output": "automated dialog generation", "neg_sample": ["dialogpt is used for Task", "conversational bots have become non - traditional methods for therapy among individuals suffering from psychological illnesses ."], "relation": "used for", "id": "2022.bionlp-1.27", "year": 2022, "rel_sent": "This was done to adapt existing generative models - GPT2 and DialoGPT - to the task of automated dialog generation .", "forward": true, "src_ids": "2022.bionlp-1.27_368"}
+{"input": "aspect sentiment triplet extraction is done by using Method| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "aspect sentiment triplet extraction", "output": "enhanced multi - channel graph convolutional network", "neg_sample": ["aspect sentiment triplet extraction is done by using Method", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Enhanced Multi - Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction.", "forward": false, "src_ids": "2022.acl-long.212_369"}
+{"input": "enhanced multi - channel graph convolutional network is used for Task| context: most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "enhanced multi - channel graph convolutional network", "output": "aspect sentiment triplet extraction", "neg_sample": ["enhanced multi - channel graph convolutional network is used for Task", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Enhanced Multi - Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction.", "forward": true, "src_ids": "2022.acl-long.212_370"}
+{"input": "word - pair representation refinement is done by using Method| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "word - pair representation refinement", "output": "emc - gcn", "neg_sample": ["word - pair representation refinement is done by using Method", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Finally , we design an effective refining strategy on EMC - GCN for word - pair representation refinement , which considers the implicit results of aspect and opinion extraction when determining whether word pairs match or not .", "forward": false, "src_ids": "2022.acl-long.212_371"}
+{"input": "refining strategy is used for Method| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "refining strategy", "output": "emc - gcn", "neg_sample": ["refining strategy is used for Method", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Finally , we design an effective refining strategy on EMC - GCN for word - pair representation refinement , which considers the implicit results of aspect and opinion extraction when determining whether word pairs match or not .", "forward": true, "src_ids": "2022.acl-long.212_372"}
+{"input": "emc - gcn model is done by using OtherScientificTerm| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "emc - gcn model", "output": "linguistic features", "neg_sample": ["emc - gcn model is done by using OtherScientificTerm", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Furthermore , we consider diverse linguistic features to enhance our EMC - GCN model .", "forward": false, "src_ids": "2022.acl-long.212_373"}
+{"input": "linguistic features is used for Method| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "linguistic features", "output": "emc - gcn model", "neg_sample": ["linguistic features is used for Method", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Furthermore , we consider diverse linguistic features to enhance our EMC - GCN model .", "forward": true, "src_ids": "2022.acl-long.212_374"}
+{"input": "emc - gcn is done by using Method| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "emc - gcn", "output": "refining strategy", "neg_sample": ["emc - gcn is done by using Method", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Finally , we design an effective refining strategy on EMC - GCN for word - pair representation refinement , which considers the implicit results of aspect and opinion extraction when determining whether word pairs match or not .", "forward": false, "src_ids": "2022.acl-long.212_375"}
+{"input": "refining strategy is used for Task| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "refining strategy", "output": "word - pair representation refinement", "neg_sample": ["refining strategy is used for Task", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Finally , we design an effective refining strategy on EMC - GCN for word - pair representation refinement , which considers the implicit results of aspect and opinion extraction when determining whether word pairs match or not .", "forward": true, "src_ids": "2022.acl-long.212_376"}
+{"input": "emc - gcn is used for Task| context: aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task . most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion . however , these methods ignore the relations between words for aste task .", "entity": "emc - gcn", "output": "word - pair representation refinement", "neg_sample": ["emc - gcn is used for Task", "aspect sentiment triplet extraction ( aste ) is an emerging sentiment analysis task .", "most of the existing studies focus on devising a new tagging scheme that enables the model to extract the sentiment triplets in an end - to - end fashion .", "however , these methods ignore the relations between words for aste task ."], "relation": "used for", "id": "2022.acl-long.212", "year": 2022, "rel_sent": "Finally , we design an effective refining strategy on EMC - GCN for word - pair representation refinement , which considers the implicit results of aspect and opinion extraction when determining whether word pairs match or not .", "forward": true, "src_ids": "2022.acl-long.212_377"}
+{"input": "publication year predictions is done by using OtherScientificTerm| context: in this paper , we describe a bert model trained on the eighteenth century collections online ( ecco ) dataset of digitized documents . the ecco dataset poses unique modelling challenges due to the presence of optical character recognition ( ocr ) artifacts .", "entity": "publication year predictions", "output": "features", "neg_sample": ["publication year predictions is done by using OtherScientificTerm", "in this paper , we describe a bert model trained on the eighteenth century collections online ( ecco ) dataset of digitized documents .", "the ecco dataset poses unique modelling challenges due to the presence of optical character recognition ( ocr ) artifacts ."], "relation": "used for", "id": "2022.lchange-1.7", "year": 2022, "rel_sent": "We also explore how language change over time affects the model by analyzing the features the model uses for publication year predictions as given by the Integrated Gradients model explanation method .", "forward": false, "src_ids": "2022.lchange-1.7_378"}
+{"input": "features is used for Task| context: in this paper , we describe a bert model trained on the eighteenth century collections online ( ecco ) dataset of digitized documents . the ecco dataset poses unique modelling challenges due to the presence of optical character recognition ( ocr ) artifacts .", "entity": "features", "output": "publication year predictions", "neg_sample": ["features is used for Task", "in this paper , we describe a bert model trained on the eighteenth century collections online ( ecco ) dataset of digitized documents .", "the ecco dataset poses unique modelling challenges due to the presence of optical character recognition ( ocr ) artifacts ."], "relation": "used for", "id": "2022.lchange-1.7", "year": 2022, "rel_sent": "We also explore how language change over time affects the model by analyzing the features the model uses for publication year predictions as given by the Integrated Gradients model explanation method .", "forward": true, "src_ids": "2022.lchange-1.7_379"}
+{"input": "end to end asr systems is done by using Task| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "entity": "end to end asr systems", "output": "domain adaptation", "neg_sample": ["end to end asr systems is done by using Task", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "A Simple Baseline for Domain Adaptation in End to End ASR Systems Using Synthetic Data.", "forward": false, "src_ids": "2022.ecnlp-1.28_380"}
+{"input": "baseline technique is used for Task| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "entity": "baseline technique", "output": "domain adaptation", "neg_sample": ["baseline technique is used for Task", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "In this work , we propose a simple baseline technique for domain adaptation in end - to - end speech recognition models .", "forward": true, "src_ids": "2022.ecnlp-1.28_381"}
+{"input": "domain adaptation is used for Task| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "domain adaptation", "output": "end to end asr systems", "neg_sample": ["domain adaptation is used for Task", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "A Simple Baseline for Domain Adaptation in End to End ASR Systems Using Synthetic Data.", "forward": true, "src_ids": "2022.ecnlp-1.28_382"}
+{"input": "domain adaptation is done by using Generic| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "domain adaptation", "output": "baseline technique", "neg_sample": ["domain adaptation is done by using Generic", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "In this work , we propose a simple baseline technique for domain adaptation in end - to - end speech recognition models .", "forward": false, "src_ids": "2022.ecnlp-1.28_383"}
+{"input": "dense layer of generic asr models is done by using Material| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "dense layer of generic asr models", "output": "parallel data", "neg_sample": ["dense layer of generic asr models is done by using Material", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "The parallel data in the target domain is then used tofine - tune the final dense layer of generic ASR models .", "forward": false, "src_ids": "2022.ecnlp-1.28_384"}
+{"input": "parallel data is used for Method| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "parallel data", "output": "dense layer of generic asr models", "neg_sample": ["parallel data is used for Method", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "The parallel data in the target domain is then used tofine - tune the final dense layer of generic ASR models .", "forward": true, "src_ids": "2022.ecnlp-1.28_385"}
+{"input": "attention - based models is done by using Material| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "attention - based models", "output": "text data", "neg_sample": ["attention - based models is done by using Material", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "We use text data from address and e - commerce search domains to show the effectiveness of our low - cost baseline approach on CTC and attention - based models .", "forward": false, "src_ids": "2022.ecnlp-1.28_386"}
+{"input": "ctc is done by using Material| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "ctc", "output": "text data", "neg_sample": ["ctc is done by using Material", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "We use text data from address and e - commerce search domains to show the effectiveness of our low - cost baseline approach on CTC and attention - based models .", "forward": false, "src_ids": "2022.ecnlp-1.28_387"}
+{"input": "text data is used for Method| context: automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models . these approaches require large amounts of labeled data in the form of audio - text pairs . moreover , these models are more susceptible to domain shift as compared to traditional models . it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets . we consider a more extreme case of domain adaptation where text - only corpus is available .", "entity": "text data", "output": "ctc", "neg_sample": ["text data is used for Method", "automatic speech recognition(asr ) has been dominated by deep learning - based end - to - end speech recognition models .", "these approaches require large amounts of labeled data in the form of audio - text pairs .", "moreover , these models are more susceptible to domain shift as compared to traditional models .", "it is common practice to train generic asr models and then adapt them to target domains using comparatively smaller data sets .", "we consider a more extreme case of domain adaptation where text - only corpus is available ."], "relation": "used for", "id": "2022.ecnlp-1.28", "year": 2022, "rel_sent": "We use text data from address and e - commerce search domains to show the effectiveness of our low - cost baseline approach on CTC and attention - based models .", "forward": true, "src_ids": "2022.ecnlp-1.28_388"}
+{"input": "inter - token distribution distance is done by using Method| context: named entity recognition ( ner ) in few - shot setting is imperative for entity tagging in low resource domains . existing approaches only learn class - specific semantic features and intermediate representations from source domains . this affects generalizability to unseen target domains , resulting in suboptimal performances .", "entity": "inter - token distribution distance", "output": "contrastive learning technique", "neg_sample": ["inter - token distribution distance is done by using Method", "named entity recognition ( ner ) in few - shot setting is imperative for entity tagging in low resource domains .", "existing approaches only learn class - specific semantic features and intermediate representations from source domains .", "this affects generalizability to unseen target domains , resulting in suboptimal performances ."], "relation": "used for", "id": "2022.acl-long.439", "year": 2022, "rel_sent": "To this end , we present CONTaiNER , a novel contrastive learning technique that optimizes the inter - token distribution distance for Few - Shot NER .", "forward": false, "src_ids": "2022.acl-long.439_389"}
+{"input": "contrastive learning technique is used for OtherScientificTerm| context: named entity recognition ( ner ) in few - shot setting is imperative for entity tagging in low resource domains . existing approaches only learn class - specific semantic features and intermediate representations from source domains . this affects generalizability to unseen target domains , resulting in suboptimal performances .", "entity": "contrastive learning technique", "output": "inter - token distribution distance", "neg_sample": ["contrastive learning technique is used for OtherScientificTerm", "named entity recognition ( ner ) in few - shot setting is imperative for entity tagging in low resource domains .", "existing approaches only learn class - specific semantic features and intermediate representations from source domains .", "this affects generalizability to unseen target domains , resulting in suboptimal performances ."], "relation": "used for", "id": "2022.acl-long.439", "year": 2022, "rel_sent": "To this end , we present CONTaiNER , a novel contrastive learning technique that optimizes the inter - token distribution distance for Few - Shot NER .", "forward": true, "src_ids": "2022.acl-long.439_390"}
+{"input": "distribution of candidate keywords is done by using Method| context: we propose a probabilistic approach to select a subset of a target domain representative keywords from a candidate set , contrasting with a context domain . such a task is crucial for many downstream tasks in natural language processing .", "entity": "distribution of candidate keywords", "output": "two - component mixture model concept", "neg_sample": ["distribution of candidate keywords is done by using Method", "we propose a probabilistic approach to select a subset of a target domain representative keywords from a candidate set , contrasting with a context domain .", "such a task is crucial for many downstream tasks in natural language processing ."], "relation": "used for", "id": "2022.findings-acl.56", "year": 2022, "rel_sent": "To contrast the target domain and the context domain , we adapt the two - component mixture model concept to generate a distribution of candidate keywords .", "forward": false, "src_ids": "2022.findings-acl.56_391"}
+{"input": "two - component mixture model concept is used for OtherScientificTerm| context: we propose a probabilistic approach to select a subset of a target domain representative keywords from a candidate set , contrasting with a context domain . such a task is crucial for many downstream tasks in natural language processing .", "entity": "two - component mixture model concept", "output": "distribution of candidate keywords", "neg_sample": ["two - component mixture model concept is used for OtherScientificTerm", "we propose a probabilistic approach to select a subset of a target domain representative keywords from a candidate set , contrasting with a context domain .", "such a task is crucial for many downstream tasks in natural language processing ."], "relation": "used for", "id": "2022.findings-acl.56", "year": 2022, "rel_sent": "To contrast the target domain and the context domain , we adapt the two - component mixture model concept to generate a distribution of candidate keywords .", "forward": true, "src_ids": "2022.findings-acl.56_392"}
+{"input": "target word surprisal is used for OtherScientificTerm| context: eye movement data are used in psycholinguistic studies to infer information regarding cognitive processes during reading .", "entity": "target word surprisal", "output": "regression features", "neg_sample": ["target word surprisal is used for OtherScientificTerm", "eye movement data are used in psycholinguistic studies to infer information regarding cognitive processes during reading ."], "relation": "used for", "id": "2022.cmcl-1.13", "year": 2022, "rel_sent": "We compared different regression models using features of the target word and its previous word , and target word surprisal as regression features .", "forward": true, "src_ids": "2022.cmcl-1.13_393"}
+{"input": "regression features is done by using OtherScientificTerm| context: eye movement data are used in psycholinguistic studies to infer information regarding cognitive processes during reading .", "entity": "regression features", "output": "target word surprisal", "neg_sample": ["regression features is done by using OtherScientificTerm", "eye movement data are used in psycholinguistic studies to infer information regarding cognitive processes during reading ."], "relation": "used for", "id": "2022.cmcl-1.13", "year": 2022, "rel_sent": "We compared different regression models using features of the target word and its previous word , and target word surprisal as regression features .", "forward": false, "src_ids": "2022.cmcl-1.13_394"}
+{"input": "parallel training data is done by using Method| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "parallel training data", "output": "pre - trained multilingual language model", "neg_sample": ["parallel training data is done by using Method", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "We show that introducing a pre - trained multilingual language model dramatically reduces the amount of parallel training data required to achieve good performance by 80 % .", "forward": false, "src_ids": "2022.acl-long.62_395"}
+{"input": "pre - trained multilingual language model is used for Material| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "pre - trained multilingual language model", "output": "parallel training data", "neg_sample": ["pre - trained multilingual language model is used for Material", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "We show that introducing a pre - trained multilingual language model dramatically reduces the amount of parallel training data required to achieve good performance by 80 % .", "forward": true, "src_ids": "2022.acl-long.62_396"}
+{"input": "nmt models is done by using Material| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "nmt models", "output": "parallel data", "neg_sample": ["nmt models is done by using Material", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en - zh and en - de .", "forward": false, "src_ids": "2022.acl-long.62_397"}
+{"input": "en - zh is done by using Method| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "en - zh", "output": "nmt models", "neg_sample": ["en - zh is done by using Method", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en - zh and en - de .", "forward": false, "src_ids": "2022.acl-long.62_398"}
+{"input": "en - de is done by using Method| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "en - de", "output": "nmt models", "neg_sample": ["en - de is done by using Method", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en - zh and en - de .", "forward": false, "src_ids": "2022.acl-long.62_399"}
+{"input": "parallel data is used for Method| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "parallel data", "output": "nmt models", "neg_sample": ["parallel data is used for Method", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en - zh and en - de .", "forward": true, "src_ids": "2022.acl-long.62_400"}
+{"input": "nmt models is used for Material| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "nmt models", "output": "en - zh", "neg_sample": ["nmt models is used for Material", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en - zh and en - de .", "forward": true, "src_ids": "2022.acl-long.62_401"}
+{"input": "nmt models is used for OtherScientificTerm| context: while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored .", "entity": "nmt models", "output": "en - de", "neg_sample": ["nmt models is used for OtherScientificTerm", "while bert is an effective method for learning monolingual sentence embeddings for semantic similarity and embedding based transfer learning bert based cross - lingual sentence embeddings have yet to be explored ."], "relation": "used for", "id": "2022.acl-long.62", "year": 2022, "rel_sent": "Parallel data mined from CommonCrawl using our best model is shown to train competitive NMT models for en - zh and en - de .", "forward": true, "src_ids": "2022.acl-long.62_402"}
+{"input": "lexically - constrained decoding is done by using Task| context: online alignment in machine translation refers to the task of aligning a target word to a source word when the target sequence has only been partially decoded . good online alignments facilitate important applications such as lexically constrained translation where user - defined dictionaries are used to inject lexical constraints into the translation model .", "entity": "lexically - constrained decoding", "output": "online posterior alignments", "neg_sample": ["lexically - constrained decoding is done by using Task", "online alignment in machine translation refers to the task of aligning a target word to a source word when the target sequence has only been partially decoded .", "good online alignments facilitate important applications such as lexically constrained translation where user - defined dictionaries are used to inject lexical constraints into the translation model ."], "relation": "used for", "id": "2022.acl-long.460", "year": 2022, "rel_sent": "Accurate Online Posterior Alignments for Principled Lexically - Constrained Decoding.", "forward": false, "src_ids": "2022.acl-long.460_403"}
+{"input": "online posterior alignments is used for Method| context: online alignment in machine translation refers to the task of aligning a target word to a source word when the target sequence has only been partially decoded . good online alignments facilitate important applications such as lexically constrained translation where user - defined dictionaries are used to inject lexical constraints into the translation model .", "entity": "online posterior alignments", "output": "lexically - constrained decoding", "neg_sample": ["online posterior alignments is used for Method", "online alignment in machine translation refers to the task of aligning a target word to a source word when the target sequence has only been partially decoded .", "good online alignments facilitate important applications such as lexically constrained translation where user - defined dictionaries are used to inject lexical constraints into the translation model ."], "relation": "used for", "id": "2022.acl-long.460", "year": 2022, "rel_sent": "Accurate Online Posterior Alignments for Principled Lexically - Constrained Decoding.", "forward": true, "src_ids": "2022.acl-long.460_404"}
+{"input": "gpt-3 is done by using OtherScientificTerm| context: gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities .", "entity": "gpt-3", "output": "in - context examples", "neg_sample": ["gpt-3 is done by using OtherScientificTerm", "gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities ."], "relation": "used for", "id": "2022.deelio-1.10", "year": 2022, "rel_sent": "What Makes Good In - Context Examples for GPT-3 ?.", "forward": false, "src_ids": "2022.deelio-1.10_405"}
+{"input": "neural networks is done by using Method| context: gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities . despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples .", "entity": "neural networks", "output": "retrieval module", "neg_sample": ["neural networks is done by using Method", "gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities .", "despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples ."], "relation": "used for", "id": "2022.deelio-1.10", "year": 2022, "rel_sent": "In this work , we investigate whether there are more effective strategies for judiciously selecting in - context examples ( relative to random sampling ) that better leverage GPT-3 's in - context learning capabilities . Inspired by the recent success of leveraging a retrieval module to augment neural networks , we propose to retrieve examples that are semantically - similar to a test query sample toformulate its corresponding prompt .", "forward": false, "src_ids": "2022.deelio-1.10_406"}
+{"input": "retrieval module is used for Method| context: gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities . despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples .", "entity": "retrieval module", "output": "neural networks", "neg_sample": ["retrieval module is used for Method", "gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities .", "despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples ."], "relation": "used for", "id": "2022.deelio-1.10", "year": 2022, "rel_sent": "In this work , we investigate whether there are more effective strategies for judiciously selecting in - context examples ( relative to random sampling ) that better leverage GPT-3 's in - context learning capabilities . Inspired by the recent success of leveraging a retrieval module to augment neural networks , we propose to retrieve examples that are semantically - similar to a test query sample toformulate its corresponding prompt .", "forward": true, "src_ids": "2022.deelio-1.10_407"}
+{"input": "retrieval is done by using Method| context: gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities . despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples .", "entity": "retrieval", "output": "sentence encoders", "neg_sample": ["retrieval is done by using Method", "gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities .", "despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples ."], "relation": "used for", "id": "2022.deelio-1.10", "year": 2022, "rel_sent": "Moreover , it is observed that the sentence encoders fine - tuned on task - related datasets yield even more helpful retrieval results .", "forward": false, "src_ids": "2022.deelio-1.10_408"}
+{"input": "sentence encoders is used for Task| context: gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities . despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples .", "entity": "sentence encoders", "output": "retrieval", "neg_sample": ["sentence encoders is used for Task", "gpt-3 has attracted lots of attention due to its superior performance across a wide range of nlp tasks , especially with its in - context learning abilities .", "despite its success , we found that the empirical results of gpt-3 depend heavily on the choice of in - context examples ."], "relation": "used for", "id": "2022.deelio-1.10", "year": 2022, "rel_sent": "Moreover , it is observed that the sentence encoders fine - tuned on task - related datasets yield even more helpful retrieval results .", "forward": true, "src_ids": "2022.deelio-1.10_409"}
+{"input": "natural language understanding is done by using Method| context: given the ubiquitous nature of numbers in text , reasoning with numbers to perform simple calculations is an important skill of ai systems . while many datasets and models have been developed to this end , state - of - the - art ai systems are brittle ; failing to perform the underlying mathematical reasoning when they appear in a slightly different scenario .", "entity": "natural language understanding", "output": "glue", "neg_sample": ["natural language understanding is done by using Method", "given the ubiquitous nature of numbers in text , reasoning with numbers to perform simple calculations is an important skill of ai systems .", "while many datasets and models have been developed to this end , state - of - the - art ai systems are brittle ; failing to perform the underlying mathematical reasoning when they appear in a slightly different scenario ."], "relation": "used for", "id": "2022.acl-long.246", "year": 2022, "rel_sent": "Drawing inspiration from GLUE that was proposed in the context of natural language understanding , we propose NumGLUE , a multi - task benchmark that evaluates the performance of AI systems on eight different tasks , that at their core require simple arithmetic understanding .", "forward": false, "src_ids": "2022.acl-long.246_410"}
+{"input": "glue is used for Task| context: given the ubiquitous nature of numbers in text , reasoning with numbers to perform simple calculations is an important skill of ai systems . while many datasets and models have been developed to this end , state - of - the - art ai systems are brittle ; failing to perform the underlying mathematical reasoning when they appear in a slightly different scenario .", "entity": "glue", "output": "natural language understanding", "neg_sample": ["glue is used for Task", "given the ubiquitous nature of numbers in text , reasoning with numbers to perform simple calculations is an important skill of ai systems .", "while many datasets and models have been developed to this end , state - of - the - art ai systems are brittle ; failing to perform the underlying mathematical reasoning when they appear in a slightly different scenario ."], "relation": "used for", "id": "2022.acl-long.246", "year": 2022, "rel_sent": "Drawing inspiration from GLUE that was proposed in the context of natural language understanding , we propose NumGLUE , a multi - task benchmark that evaluates the performance of AI systems on eight different tasks , that at their core require simple arithmetic understanding .", "forward": true, "src_ids": "2022.acl-long.246_411"}
+{"input": "language - agnostic model is used for Task| context: the task of identifying the emotion in a given text has many practical applications ranging from screening public health to business and management .", "entity": "language - agnostic model", "output": "emotion analysis", "neg_sample": ["language - agnostic model is used for Task", "the task of identifying the emotion in a given text has many practical applications ranging from screening public health to business and management ."], "relation": "used for", "id": "2022.dravidianlangtech-1.17", "year": 2022, "rel_sent": "In this paper , we propose a language - agnostic model that focuses on emotion analysis in Tamil text .", "forward": true, "src_ids": "2022.dravidianlangtech-1.17_412"}
+{"input": "emotion analysis is done by using Method| context: as the world around us continues to become increasingly digital , it has been acknowledged that there is a growing need for emotion analysis of social media content . the task of identifying the emotion in a given text has many practical applications ranging from screening public health to business and management .", "entity": "emotion analysis", "output": "language - agnostic model", "neg_sample": ["emotion analysis is done by using Method", "as the world around us continues to become increasingly digital , it has been acknowledged that there is a growing need for emotion analysis of social media content .", "the task of identifying the emotion in a given text has many practical applications ranging from screening public health to business and management ."], "relation": "used for", "id": "2022.dravidianlangtech-1.17", "year": 2022, "rel_sent": "In this paper , we propose a language - agnostic model that focuses on emotion analysis in Tamil text .", "forward": false, "src_ids": "2022.dravidianlangtech-1.17_413"}
+{"input": "labeled syntactic knowledge is used for Method| context: existing syntax - enriched neural machine translation ( nmt ) models work either with the single most - likely unlabeled parse or the set of n - best unlabeled parses coming out of an external parser . passing a single or n - best parses to the nmt model risks propagating parse errors . furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories .", "entity": "labeled syntactic knowledge", "output": "transformer", "neg_sample": ["labeled syntactic knowledge is used for Method", "existing syntax - enriched neural machine translation ( nmt ) models work either with the single most - likely unlabeled parse or the set of n - best unlabeled parses coming out of an external parser .", "passing a single or n - best parses to the nmt model risks propagating parse errors .", "furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories ."], "relation": "used for", "id": "2022.eamt-1.7", "year": 2022, "rel_sent": "In this paper we explore the question : Does passing both parser uncertainty and labeled syntactic knowledge to the Transformer improve its translation performance ?", "forward": true, "src_ids": "2022.eamt-1.7_414"}
+{"input": "neural machine translation is done by using OtherScientificTerm| context: existing syntax - enriched neural machine translation ( nmt ) models work either with the single most - likely unlabeled parse or the set of n - best unlabeled parses coming out of an external parser . passing a single or n - best parses to the nmt model risks propagating parse errors . furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories .", "entity": "neural machine translation", "output": "labeled dependency distributions", "neg_sample": ["neural machine translation is done by using OtherScientificTerm", "existing syntax - enriched neural machine translation ( nmt ) models work either with the single most - likely unlabeled parse or the set of n - best unlabeled parses coming out of an external parser .", "passing a single or n - best parses to the nmt model risks propagating parse errors .", "furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories ."], "relation": "used for", "id": "2022.eamt-1.7", "year": 2022, "rel_sent": "Passing Parser Uncertainty to the Transformer . Labeled Dependency Distributions for Neural Machine Translation ..", "forward": false, "src_ids": "2022.eamt-1.7_415"}
+{"input": "labeled dependency distributions is used for Task| context: passing a single or n - best parses to the nmt model risks propagating parse errors . furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories .", "entity": "labeled dependency distributions", "output": "neural machine translation", "neg_sample": ["labeled dependency distributions is used for Task", "passing a single or n - best parses to the nmt model risks propagating parse errors .", "furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories ."], "relation": "used for", "id": "2022.eamt-1.7", "year": 2022, "rel_sent": "Passing Parser Uncertainty to the Transformer . Labeled Dependency Distributions for Neural Machine Translation ..", "forward": true, "src_ids": "2022.eamt-1.7_416"}
+{"input": "transformer is done by using OtherScientificTerm| context: existing syntax - enriched neural machine translation ( nmt ) models work either with the single most - likely unlabeled parse or the set of n - best unlabeled parses coming out of an external parser . passing a single or n - best parses to the nmt model risks propagating parse errors . furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories .", "entity": "transformer", "output": "labeled syntactic knowledge", "neg_sample": ["transformer is done by using OtherScientificTerm", "existing syntax - enriched neural machine translation ( nmt ) models work either with the single most - likely unlabeled parse or the set of n - best unlabeled parses coming out of an external parser .", "passing a single or n - best parses to the nmt model risks propagating parse errors .", "furthermore , unlabeled parses represent only syntactic groupings without their linguistically relevant categories ."], "relation": "used for", "id": "2022.eamt-1.7", "year": 2022, "rel_sent": "In this paper we explore the question : Does passing both parser uncertainty and labeled syntactic knowledge to the Transformer improve its translation performance ?", "forward": false, "src_ids": "2022.eamt-1.7_417"}
+{"input": "compositional generalization is done by using Task| context: there is mounting evidence that existing neural network models , in particular the very popular sequence - to - sequence architecture , struggle to systematically generalize to unseen compositions of seen components . we demonstrate that one of the reasons hindering compositional generalization relates to representations being entangled .", "entity": "compositional generalization", "output": "disentangled sequence to sequence learning", "neg_sample": ["compositional generalization is done by using Task", "there is mounting evidence that existing neural network models , in particular the very popular sequence - to - sequence architecture , struggle to systematically generalize to unseen compositions of seen components .", "we demonstrate that one of the reasons hindering compositional generalization relates to representations being entangled ."], "relation": "used for", "id": "2022.acl-long.293", "year": 2022, "rel_sent": "Disentangled Sequence to Sequence Learning for Compositional Generalization.", "forward": false, "src_ids": "2022.acl-long.293_418"}
+{"input": "disentangled sequence to sequence learning is used for Task| context: there is mounting evidence that existing neural network models , in particular the very popular sequence - to - sequence architecture , struggle to systematically generalize to unseen compositions of seen components .", "entity": "disentangled sequence to sequence learning", "output": "compositional generalization", "neg_sample": ["disentangled sequence to sequence learning is used for Task", "there is mounting evidence that existing neural network models , in particular the very popular sequence - to - sequence architecture , struggle to systematically generalize to unseen compositions of seen components ."], "relation": "used for", "id": "2022.acl-long.293", "year": 2022, "rel_sent": "Disentangled Sequence to Sequence Learning for Compositional Generalization.", "forward": true, "src_ids": "2022.acl-long.293_419"}
+{"input": "task - adaptive pre - training is done by using Method| context: task - adaptive pre - training ( tapt ) alleviates the lack of labelled data and provides performance lift by adapting unlabelled data to downstream task . unfortunately , existing adaptations mainly involve deterministic rules that can not generalize well .", "entity": "task - adaptive pre - training", "output": "sequence - tagging based cloze answer extraction method", "neg_sample": ["task - adaptive pre - training is done by using Method", "task - adaptive pre - training ( tapt ) alleviates the lack of labelled data and provides performance lift by adapting unlabelled data to downstream task .", "unfortunately , existing adaptations mainly involve deterministic rules that can not generalize well ."], "relation": "used for", "id": "2022.repl4nlp-1.7", "year": 2022, "rel_sent": "Here , we propose Clozer , a sequence - tagging based cloze answer extraction method used in TAPT that is extendable for adaptation on any cloze - style machine reading comprehension ( MRC ) downstream tasks .", "forward": false, "src_ids": "2022.repl4nlp-1.7_420"}
+{"input": "sequence - tagging based cloze answer extraction method is used for Method| context: unfortunately , existing adaptations mainly involve deterministic rules that can not generalize well .", "entity": "sequence - tagging based cloze answer extraction method", "output": "task - adaptive pre - training", "neg_sample": ["sequence - tagging based cloze answer extraction method is used for Method", "unfortunately , existing adaptations mainly involve deterministic rules that can not generalize well ."], "relation": "used for", "id": "2022.repl4nlp-1.7", "year": 2022, "rel_sent": "Here , we propose Clozer , a sequence - tagging based cloze answer extraction method used in TAPT that is extendable for adaptation on any cloze - style machine reading comprehension ( MRC ) downstream tasks .", "forward": true, "src_ids": "2022.repl4nlp-1.7_421"}
+{"input": "conditional probability is done by using Method| context: we introduce a noisy channel approach for language model prompting in few - shot text classification .", "entity": "conditional probability", "output": "channel models", "neg_sample": ["conditional probability is done by using Method", "we introduce a noisy channel approach for language model prompting in few - shot text classification ."], "relation": "used for", "id": "2022.acl-long.365", "year": 2022, "rel_sent": "Instead of computing the likelihood of the label given the input ( referred as direct models ) , channel models compute the conditional probability of the input given the label , and are thereby required to explain every word in the input .", "forward": false, "src_ids": "2022.acl-long.365_422"}
+{"input": "few - shot learning methods is done by using Method| context: we introduce a noisy channel approach for language model prompting in few - shot text classification .", "entity": "few - shot learning methods", "output": "channel models", "neg_sample": ["few - shot learning methods is done by using Method", "we introduce a noisy channel approach for language model prompting in few - shot text classification ."], "relation": "used for", "id": "2022.acl-long.365", "year": 2022, "rel_sent": "We use channel models for recently proposed few - shot learning methods with no or very limited updates to the language model parameters , via either in - context demonstration or prompt tuning .", "forward": false, "src_ids": "2022.acl-long.365_423"}
+{"input": "channel models is used for OtherScientificTerm| context: we introduce a noisy channel approach for language model prompting in few - shot text classification .", "entity": "channel models", "output": "conditional probability", "neg_sample": ["channel models is used for OtherScientificTerm", "we introduce a noisy channel approach for language model prompting in few - shot text classification ."], "relation": "used for", "id": "2022.acl-long.365", "year": 2022, "rel_sent": "Instead of computing the likelihood of the label given the input ( referred as direct models ) , channel models compute the conditional probability of the input given the label , and are thereby required to explain every word in the input .", "forward": true, "src_ids": "2022.acl-long.365_424"}
+{"input": "channel models is used for Method| context: we introduce a noisy channel approach for language model prompting in few - shot text classification .", "entity": "channel models", "output": "few - shot learning methods", "neg_sample": ["channel models is used for Method", "we introduce a noisy channel approach for language model prompting in few - shot text classification ."], "relation": "used for", "id": "2022.acl-long.365", "year": 2022, "rel_sent": "We use channel models for recently proposed few - shot learning methods with no or very limited updates to the language model parameters , via either in - context demonstration or prompt tuning .", "forward": true, "src_ids": "2022.acl-long.365_425"}
+{"input": "dialogue disentanglement is done by using Method| context: tangled multi - party dialogue contexts lead to challenges for dialogue reading comprehension , where multiple dialogue threads flow simultaneously within a common dialogue record , increasing difficulties in understanding the dialogue history for both human and machine . previous studies mainly focus on utterance encoding methods with carefully designed features but pay inadequate attention to characteristic features of the structure of dialogues .", "entity": "dialogue disentanglement", "output": "structural characterization", "neg_sample": ["dialogue disentanglement is done by using Method", "tangled multi - party dialogue contexts lead to challenges for dialogue reading comprehension , where multiple dialogue threads flow simultaneously within a common dialogue record , increasing difficulties in understanding the dialogue history for both human and machine .", "previous studies mainly focus on utterance encoding methods with carefully designed features but pay inadequate attention to characteristic features of the structure of dialogues ."], "relation": "used for", "id": "2022.acl-long.23", "year": 2022, "rel_sent": "Structural Characterization for Dialogue Disentanglement.", "forward": false, "src_ids": "2022.acl-long.23_426"}
+{"input": "structural characterization is used for Task| context: tangled multi - party dialogue contexts lead to challenges for dialogue reading comprehension , where multiple dialogue threads flow simultaneously within a common dialogue record , increasing difficulties in understanding the dialogue history for both human and machine . previous studies mainly focus on utterance encoding methods with carefully designed features but pay inadequate attention to characteristic features of the structure of dialogues .", "entity": "structural characterization", "output": "dialogue disentanglement", "neg_sample": ["structural characterization is used for Task", "tangled multi - party dialogue contexts lead to challenges for dialogue reading comprehension , where multiple dialogue threads flow simultaneously within a common dialogue record , increasing difficulties in understanding the dialogue history for both human and machine .", "previous studies mainly focus on utterance encoding methods with carefully designed features but pay inadequate attention to characteristic features of the structure of dialogues ."], "relation": "used for", "id": "2022.acl-long.23", "year": 2022, "rel_sent": "Structural Characterization for Dialogue Disentanglement.", "forward": true, "src_ids": "2022.acl-long.23_427"}
+{"input": "mining event - centric opinions is done by using Task| context: events are considered as the fundamental building blocks of the world .", "entity": "mining event - centric opinions", "output": "mining event - centric opinions", "neg_sample": ["mining event - centric opinions is done by using Task", "events are considered as the fundamental building blocks of the world ."], "relation": "used for", "id": "2022.findings-acl.216", "year": 2022, "rel_sent": "In this paper , we propose and formulate the task of event - centric opinion mining based on event - argument structure and expression categorizing theory .", "forward": false, "src_ids": "2022.findings-acl.216_428"}
+{"input": "mining event - centric opinions is used for Task| context: events are considered as the fundamental building blocks of the world .", "entity": "mining event - centric opinions", "output": "mining event - centric opinions", "neg_sample": ["mining event - centric opinions is used for Task", "events are considered as the fundamental building blocks of the world ."], "relation": "used for", "id": "2022.findings-acl.216", "year": 2022, "rel_sent": "In this paper , we propose and formulate the task of event - centric opinion mining based on event - argument structure and expression categorizing theory .", "forward": true, "src_ids": "2022.findings-acl.216_429"}
+{"input": "dense passage retrieval is done by using Method| context: recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval . however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential .", "entity": "dense passage retrieval", "output": "unsupervised corpus aware language model pre - training", "neg_sample": ["dense passage retrieval is done by using Method", "recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval .", "however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential ."], "relation": "used for", "id": "2022.acl-long.203", "year": 2022, "rel_sent": "Unsupervised Corpus Aware Language Model Pre - training for Dense Passage Retrieval.", "forward": false, "src_ids": "2022.acl-long.203_430"}
+{"input": "unsupervised corpus aware language model pre - training is used for Task| context: recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval . however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential .", "entity": "unsupervised corpus aware language model pre - training", "output": "dense passage retrieval", "neg_sample": ["unsupervised corpus aware language model pre - training is used for Task", "recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval .", "however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential ."], "relation": "used for", "id": "2022.acl-long.203", "year": 2022, "rel_sent": "Unsupervised Corpus Aware Language Model Pre - training for Dense Passage Retrieval.", "forward": true, "src_ids": "2022.acl-long.203_431"}
+{"input": "passage embedding space is done by using OtherScientificTerm| context: recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval . however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential .", "entity": "passage embedding space", "output": "unsupervised corpus - level contrastive loss", "neg_sample": ["passage embedding space is done by using OtherScientificTerm", "recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval .", "however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential ."], "relation": "used for", "id": "2022.acl-long.203", "year": 2022, "rel_sent": "On top of it , we propose coCondenser , which adds an unsupervised corpus - level contrastive loss to warm up the passage embedding space .", "forward": false, "src_ids": "2022.acl-long.203_432"}
+{"input": "unsupervised corpus - level contrastive loss is used for OtherScientificTerm| context: recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval . however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential .", "entity": "unsupervised corpus - level contrastive loss", "output": "passage embedding space", "neg_sample": ["unsupervised corpus - level contrastive loss is used for OtherScientificTerm", "recent research demonstrates the effectiveness of using fine - tuned language models ( lm ) for dense retrieval .", "however , dense retrievers are hard to train , typically requiring heavily engineered fine - tuning pipelines to realize their full potential ."], "relation": "used for", "id": "2022.acl-long.203", "year": 2022, "rel_sent": "On top of it , we propose coCondenser , which adds an unsupervised corpus - level contrastive loss to warm up the passage embedding space .", "forward": true, "src_ids": "2022.acl-long.203_433"}
+{"input": "natural language interfaces is done by using OtherScientificTerm| context: a key challenge facing natural language interfaces is enabling users to understand the capabilities of the underlying system .", "entity": "natural language interfaces", "output": "counterfactual explanations", "neg_sample": ["natural language interfaces is done by using OtherScientificTerm", "a key challenge facing natural language interfaces is enabling users to understand the capabilities of the underlying system ."], "relation": "used for", "id": "2022.acl-short.14", "year": 2022, "rel_sent": "Counterfactual Explanations for Natural Language Interfaces.", "forward": false, "src_ids": "2022.acl-short.14_434"}
+{"input": "lyrics is done by using OtherScientificTerm| context: lyrics generation has been a very popular application of natural language generation . previous works mainly focused on generating lyrics based on a couple of attributes or keywords , rendering very limited control over the content of the lyrics .", "entity": "lyrics", "output": "passage - level text", "neg_sample": ["lyrics is done by using OtherScientificTerm", "lyrics generation has been a very popular application of natural language generation .", "previous works mainly focused on generating lyrics based on a couple of attributes or keywords , rendering very limited control over the content of the lyrics ."], "relation": "used for", "id": "2022.acl-demo.7", "year": 2022, "rel_sent": "By using the passage - level text as input , the content of generated lyrics is expected to reflect the nuances of users ' needs .", "forward": false, "src_ids": "2022.acl-demo.7_435"}
+{"input": "inference is done by using Method| context: we perform an in - depth error analysis of adversarial nli ( anli ) , a recently introduced large - scale human - and - model - in - the - loop natural language inference dataset collected over multiple rounds .", "entity": "inference", "output": "fine - grained annotation scheme", "neg_sample": ["inference is done by using Method", "we perform an in - depth error analysis of adversarial nli ( anli ) , a recently introduced large - scale human - and - model - in - the - loop natural language inference dataset collected over multiple rounds ."], "relation": "used for", "id": "2022.scil-1.3", "year": 2022, "rel_sent": "We propose a fine - grained annotation scheme of the different aspects of inference that are responsible for the gold classification labels , and use it to hand - code all three of the ANLI development sets .", "forward": false, "src_ids": "2022.scil-1.3_436"}
+{"input": "few shot named entity recognition is done by using OtherScientificTerm| context: we study the problem of few shot learning for named entity recognition . specifically , we leverage the semantic information in the names of the labels as a way of giving the model additional signal and enriched priors .", "entity": "few shot named entity recognition", "output": "label semantics", "neg_sample": ["few shot named entity recognition is done by using OtherScientificTerm", "we study the problem of few shot learning for named entity recognition .", "specifically , we leverage the semantic information in the names of the labels as a way of giving the model additional signal and enriched priors ."], "relation": "used for", "id": "2022.findings-acl.155", "year": 2022, "rel_sent": "Label Semantics for Few Shot Named Entity Recognition.", "forward": false, "src_ids": "2022.findings-acl.155_437"}
+{"input": "label semantics is used for Task| context: we study the problem of few shot learning for named entity recognition . specifically , we leverage the semantic information in the names of the labels as a way of giving the model additional signal and enriched priors .", "entity": "label semantics", "output": "few shot named entity recognition", "neg_sample": ["label semantics is used for Task", "we study the problem of few shot learning for named entity recognition .", "specifically , we leverage the semantic information in the names of the labels as a way of giving the model additional signal and enriched priors ."], "relation": "used for", "id": "2022.findings-acl.155", "year": 2022, "rel_sent": "Label Semantics for Few Shot Named Entity Recognition.", "forward": true, "src_ids": "2022.findings-acl.155_438"}
+{"input": "biomedical ie is done by using Method| context: information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text . the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction . sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine .", "entity": "biomedical ie", "output": "slot filling approach", "neg_sample": ["biomedical ie is done by using Method", "information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text .", "the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction .", "sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine ."], "relation": "used for", "id": "2022.bionlp-1.7", "year": 2022, "rel_sent": "In this work we present a slot filling approach to the task of biomedical IE , effectively replacing the need for entity and relation - specific training data , allowing us to deal with zero - shot settings .", "forward": false, "src_ids": "2022.bionlp-1.7_439"}
+{"input": "slot filling approach is used for Task| context: information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text . the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction . sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine .", "entity": "slot filling approach", "output": "biomedical ie", "neg_sample": ["slot filling approach is used for Task", "information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text .", "the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction .", "sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine ."], "relation": "used for", "id": "2022.bionlp-1.7", "year": 2022, "rel_sent": "In this work we present a slot filling approach to the task of biomedical IE , effectively replacing the need for entity and relation - specific training data , allowing us to deal with zero - shot settings .", "forward": true, "src_ids": "2022.bionlp-1.7_440"}
+{"input": "reading comprehension is done by using Material| context: information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text . the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction . sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine .", "entity": "reading comprehension", "output": "biomedical slot filling dataset", "neg_sample": ["reading comprehension is done by using Material", "information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text .", "the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction .", "sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine ."], "relation": "used for", "id": "2022.bionlp-1.7", "year": 2022, "rel_sent": "We assemble a biomedical slot filling dataset for both retrieval and reading comprehension and conduct a series of experiments demonstrating that our approach outperforms a number of simpler baselines .", "forward": false, "src_ids": "2022.bionlp-1.7_441"}
+{"input": "retrieval is done by using Material| context: information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text . the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction . sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine .", "entity": "retrieval", "output": "biomedical slot filling dataset", "neg_sample": ["retrieval is done by using Material", "information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text .", "the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction .", "sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine ."], "relation": "used for", "id": "2022.bionlp-1.7", "year": 2022, "rel_sent": "We assemble a biomedical slot filling dataset for both retrieval and reading comprehension and conduct a series of experiments demonstrating that our approach outperforms a number of simpler baselines .", "forward": false, "src_ids": "2022.bionlp-1.7_442"}
+{"input": "biomedical slot filling dataset is used for Task| context: information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text . the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction . sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine .", "entity": "biomedical slot filling dataset", "output": "retrieval", "neg_sample": ["biomedical slot filling dataset is used for Task", "information extraction ( ie ) from text refers to the task of extracting structured knowledge from unstructured text .", "the task typically consists of a series of sub - tasks such as named entity recognition and relation extraction .", "sourcing entity and relation type specific training data is a major bottleneck in domains with limited resources such as biomedicine ."], "relation": "used for", "id": "2022.bionlp-1.7", "year": 2022, "rel_sent": "We assemble a biomedical slot filling dataset for both retrieval and reading comprehension and conduct a series of experiments demonstrating that our approach outperforms a number of simpler baselines .", "forward": true, "src_ids": "2022.bionlp-1.7_443"}
+{"input": "linear transformation is used for Method| context: hyperbolic neural networks have shown great potential for modeling complex data . this hybrid method greatly limits the modeling ability of networks .", "entity": "linear transformation", "output": "hyperbolic networks", "neg_sample": ["linear transformation is used for Method", "hyperbolic neural networks have shown great potential for modeling complex data .", "this hybrid method greatly limits the modeling ability of networks ."], "relation": "used for", "id": "2022.acl-long.389", "year": 2022, "rel_sent": "Moreover , we also prove that linear transformation in tangent spaces used by existing hyperbolic networks is a relaxation of the Lorentz rotation and does not include the boost , implicitly limiting the capabilities of existing hyperbolic networks .", "forward": true, "src_ids": "2022.acl-long.389_444"}
+{"input": "hyperbolic networks is done by using OtherScientificTerm| context: hyperbolic neural networks have shown great potential for modeling complex data . however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model . this hybrid method greatly limits the modeling ability of networks .", "entity": "hyperbolic networks", "output": "linear transformation", "neg_sample": ["hyperbolic networks is done by using OtherScientificTerm", "hyperbolic neural networks have shown great potential for modeling complex data .", "however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model .", "this hybrid method greatly limits the modeling ability of networks ."], "relation": "used for", "id": "2022.acl-long.389", "year": 2022, "rel_sent": "Moreover , we also prove that linear transformation in tangent spaces used by existing hyperbolic networks is a relaxation of the Lorentz rotation and does not include the boost , implicitly limiting the capabilities of existing hyperbolic networks .", "forward": false, "src_ids": "2022.acl-long.389_445"}
+{"input": "fss - net is used for OtherScientificTerm| context: natural language processing for sign language video - including tasks like recognition , translation , and search - is crucial for making artificial intelligence technologies accessible to deaf individuals , and is gaining research interest in recent years .", "entity": "fss - net", "output": "fingerspelling", "neg_sample": ["fss - net is used for OtherScientificTerm", "natural language processing for sign language video - including tasks like recognition , translation , and search - is crucial for making artificial intelligence technologies accessible to deaf individuals , and is gaining research interest in recent years ."], "relation": "used for", "id": "2022.acl-long.119", "year": 2022, "rel_sent": "We propose an end - to - end model for this task , FSS - Net , that jointly detects fingerspelling and matches it to a text sequence .", "forward": true, "src_ids": "2022.acl-long.119_446"}
+{"input": "fingerspelling is done by using Method| context: natural language processing for sign language video - including tasks like recognition , translation , and search - is crucial for making artificial intelligence technologies accessible to deaf individuals , and is gaining research interest in recent years . this is an important task since significant content in sign language is often conveyed via fingerspelling , and to our knowledge the task has not been studied before .", "entity": "fingerspelling", "output": "fss - net", "neg_sample": ["fingerspelling is done by using Method", "natural language processing for sign language video - including tasks like recognition , translation , and search - is crucial for making artificial intelligence technologies accessible to deaf individuals , and is gaining research interest in recent years .", "this is an important task since significant content in sign language is often conveyed via fingerspelling , and to our knowledge the task has not been studied before ."], "relation": "used for", "id": "2022.acl-long.119", "year": 2022, "rel_sent": "We propose an end - to - end model for this task , FSS - Net , that jointly detects fingerspelling and matches it to a text sequence .", "forward": false, "src_ids": "2022.acl-long.119_447"}
+{"input": "finetuning lms is used for Method| context: in this work , we show that finetuning lms in the few - shot setting can considerably reduce the need for prompt engineering .", "entity": "finetuning lms", "output": "few - shot learning", "neg_sample": ["finetuning lms is used for Method", "in this work , we show that finetuning lms in the few - shot setting can considerably reduce the need for prompt engineering ."], "relation": "used for", "id": "2022.findings-acl.222", "year": 2022, "rel_sent": "All in all , we recommend finetuning LMs for few - shot learning as it is more accurate , robust to different prompts , and can be made nearly as efficient as using frozen LMs .", "forward": true, "src_ids": "2022.findings-acl.222_448"}
+{"input": "few - shot learning is done by using Method| context: prompting language models ( lms ) with training examples and task descriptions has been seen as critical to recent successes in few - shot learning .", "entity": "few - shot learning", "output": "finetuning lms", "neg_sample": ["few - shot learning is done by using Method", "prompting language models ( lms ) with training examples and task descriptions has been seen as critical to recent successes in few - shot learning ."], "relation": "used for", "id": "2022.findings-acl.222", "year": 2022, "rel_sent": "All in all , we recommend finetuning LMs for few - shot learning as it is more accurate , robust to different prompts , and can be made nearly as efficient as using frozen LMs .", "forward": false, "src_ids": "2022.findings-acl.222_449"}
+{"input": "annotations is done by using Method| context: docred is a widely used dataset for document - level relation extraction .", "entity": "annotations", "output": "recommend - revise", "neg_sample": ["annotations is done by using Method", "docred is a widely used dataset for document - level relation extraction ."], "relation": "used for", "id": "2022.acl-long.432", "year": 2022, "rel_sent": "Does Recommend - Revise Produce Reliable Annotations ? An Analysis on Missing Instances in DocRED.", "forward": false, "src_ids": "2022.acl-long.432_450"}
+{"input": "recommend - revise is used for OtherScientificTerm| context: docred is a widely used dataset for document - level relation extraction . in the large - scale annotation , a recommend - revise scheme is adopted to reduce the workload .", "entity": "recommend - revise", "output": "annotations", "neg_sample": ["recommend - revise is used for OtherScientificTerm", "docred is a widely used dataset for document - level relation extraction .", "in the large - scale annotation , a recommend - revise scheme is adopted to reduce the workload ."], "relation": "used for", "id": "2022.acl-long.432", "year": 2022, "rel_sent": "Does Recommend - Revise Produce Reliable Annotations ? An Analysis on Missing Instances in DocRED.", "forward": true, "src_ids": "2022.acl-long.432_451"}
+{"input": "referring image segmentation is done by using OtherScientificTerm| context: we investigate referring image segmentation ( ris ) , which outputs a segmentation map corresponding to the natural language description . addressing ris efficiently requires considering the interactions happening across visual and linguistic modalities and the interactions within each modality . existing methods are limited because they either compute different forms of interactions sequentially ( leading to error propagation ) or ignore intra - modal interactions .", "entity": "referring image segmentation", "output": "multi - modal interactions", "neg_sample": ["referring image segmentation is done by using OtherScientificTerm", "we investigate referring image segmentation ( ris ) , which outputs a segmentation map corresponding to the natural language description .", "addressing ris efficiently requires considering the interactions happening across visual and linguistic modalities and the interactions within each modality .", "existing methods are limited because they either compute different forms of interactions sequentially ( leading to error propagation ) or ignore intra - modal interactions ."], "relation": "used for", "id": "2022.findings-acl.270", "year": 2022, "rel_sent": "Comprehensive Multi - Modal Interactions for Referring Image Segmentation.", "forward": false, "src_ids": "2022.findings-acl.270_452"}
+{"input": "multi - modal interactions is used for Task| context: existing methods are limited because they either compute different forms of interactions sequentially ( leading to error propagation ) or ignore intra - modal interactions .", "entity": "multi - modal interactions", "output": "referring image segmentation", "neg_sample": ["multi - modal interactions is used for Task", "existing methods are limited because they either compute different forms of interactions sequentially ( leading to error propagation ) or ignore intra - modal interactions ."], "relation": "used for", "id": "2022.findings-acl.270", "year": 2022, "rel_sent": "Comprehensive Multi - Modal Interactions for Referring Image Segmentation.", "forward": true, "src_ids": "2022.findings-acl.270_453"}
+{"input": "structure - aware long document summarization is done by using Method| context: document structure is critical for efficient information consumption . however , it is challenging to encode it efficiently into the modern transformer architecture .", "entity": "structure - aware long document summarization", "output": "hibrids", "neg_sample": ["structure - aware long document summarization is done by using Method", "document structure is critical for efficient information consumption .", "however , it is challenging to encode it efficiently into the modern transformer architecture ."], "relation": "used for", "id": "2022.acl-long.58", "year": 2022, "rel_sent": "HIBRIDS : Attention with Hierarchical Biases for Structure - aware Long Document Summarization.", "forward": false, "src_ids": "2022.acl-long.58_454"}
+{"input": "hierarchical biases is done by using Method| context: document structure is critical for efficient information consumption . however , it is challenging to encode it efficiently into the modern transformer architecture .", "entity": "hierarchical biases", "output": "hibrids", "neg_sample": ["hierarchical biases is done by using Method", "document structure is critical for efficient information consumption .", "however , it is challenging to encode it efficiently into the modern transformer architecture ."], "relation": "used for", "id": "2022.acl-long.58", "year": 2022, "rel_sent": "In this work , we present HIBRIDS , which injects Hierarchical Biases foR Incorporating Document Structure into attention score calculation .", "forward": false, "src_ids": "2022.acl-long.58_455"}
+{"input": "hibrids is used for OtherScientificTerm| context: document structure is critical for efficient information consumption . however , it is challenging to encode it efficiently into the modern transformer architecture .", "entity": "hibrids", "output": "hierarchical biases", "neg_sample": ["hibrids is used for OtherScientificTerm", "document structure is critical for efficient information consumption .", "however , it is challenging to encode it efficiently into the modern transformer architecture ."], "relation": "used for", "id": "2022.acl-long.58", "year": 2022, "rel_sent": "In this work , we present HIBRIDS , which injects Hierarchical Biases foR Incorporating Document Structure into attention score calculation .", "forward": true, "src_ids": "2022.acl-long.58_456"}
+{"input": "hibrids is used for Task| context: document structure is critical for efficient information consumption . however , it is challenging to encode it efficiently into the modern transformer architecture .", "entity": "hibrids", "output": "structure - aware long document summarization", "neg_sample": ["hibrids is used for Task", "document structure is critical for efficient information consumption .", "however , it is challenging to encode it efficiently into the modern transformer architecture ."], "relation": "used for", "id": "2022.acl-long.58", "year": 2022, "rel_sent": "HIBRIDS : Attention with Hierarchical Biases for Structure - aware Long Document Summarization.", "forward": true, "src_ids": "2022.acl-long.58_457"}
+{"input": "nlg metric is done by using Method| context: fast and reliable evaluation metrics are key to r&d progress . while traditional natural language generation metrics are fast , they are not very reliable . conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources .", "entity": "nlg metric", "output": "frugalscore", "neg_sample": ["nlg metric is done by using Method", "fast and reliable evaluation metrics are key to r&d progress .", "while traditional natural language generation metrics are fast , they are not very reliable .", "conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources ."], "relation": "used for", "id": "2022.acl-long.93", "year": 2022, "rel_sent": "In this paper , we propose FrugalScore , an approach to learn a fixed , low cost version of any expensive NLG metric , while retaining most of its original performance .", "forward": false, "src_ids": "2022.acl-long.93_458"}
+{"input": "frugalscore is used for Metric| context: fast and reliable evaluation metrics are key to r&d progress . while traditional natural language generation metrics are fast , they are not very reliable . conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources .", "entity": "frugalscore", "output": "nlg metric", "neg_sample": ["frugalscore is used for Metric", "fast and reliable evaluation metrics are key to r&d progress .", "while traditional natural language generation metrics are fast , they are not very reliable .", "conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources ."], "relation": "used for", "id": "2022.acl-long.93", "year": 2022, "rel_sent": "In this paper , we propose FrugalScore , an approach to learn a fixed , low cost version of any expensive NLG metric , while retaining most of its original performance .", "forward": true, "src_ids": "2022.acl-long.93_459"}
+{"input": "summarization is done by using Metric| context: fast and reliable evaluation metrics are key to r&d progress . while traditional natural language generation metrics are fast , they are not very reliable . conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources .", "entity": "summarization", "output": "bertscore", "neg_sample": ["summarization is done by using Metric", "fast and reliable evaluation metrics are key to r&d progress .", "while traditional natural language generation metrics are fast , they are not very reliable .", "conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources ."], "relation": "used for", "id": "2022.acl-long.93", "year": 2022, "rel_sent": "Experiments with BERTScore and MoverScore on summarization and translation show that FrugalScore is on par with the original metrics ( and sometimes better ) , while having several orders of magnitude less parameters and running several times faster .", "forward": false, "src_ids": "2022.acl-long.93_460"}
+{"input": "summarization is done by using OtherScientificTerm| context: fast and reliable evaluation metrics are key to r&d progress . while traditional natural language generation metrics are fast , they are not very reliable . conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources .", "entity": "summarization", "output": "moverscore", "neg_sample": ["summarization is done by using OtherScientificTerm", "fast and reliable evaluation metrics are key to r&d progress .", "while traditional natural language generation metrics are fast , they are not very reliable .", "conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources ."], "relation": "used for", "id": "2022.acl-long.93", "year": 2022, "rel_sent": "Experiments with BERTScore and MoverScore on summarization and translation show that FrugalScore is on par with the original metrics ( and sometimes better ) , while having several orders of magnitude less parameters and running several times faster .", "forward": false, "src_ids": "2022.acl-long.93_461"}
+{"input": "bertscore is used for Task| context: fast and reliable evaluation metrics are key to r&d progress . while traditional natural language generation metrics are fast , they are not very reliable . conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources .", "entity": "bertscore", "output": "summarization", "neg_sample": ["bertscore is used for Task", "fast and reliable evaluation metrics are key to r&d progress .", "while traditional natural language generation metrics are fast , they are not very reliable .", "conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources ."], "relation": "used for", "id": "2022.acl-long.93", "year": 2022, "rel_sent": "Experiments with BERTScore and MoverScore on summarization and translation show that FrugalScore is on par with the original metrics ( and sometimes better ) , while having several orders of magnitude less parameters and running several times faster .", "forward": true, "src_ids": "2022.acl-long.93_462"}
+{"input": "moverscore is used for Task| context: fast and reliable evaluation metrics are key to r&d progress . while traditional natural language generation metrics are fast , they are not very reliable . conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources .", "entity": "moverscore", "output": "summarization", "neg_sample": ["moverscore is used for Task", "fast and reliable evaluation metrics are key to r&d progress .", "while traditional natural language generation metrics are fast , they are not very reliable .", "conversely , new metrics based on large pretrained language models are much more reliable , but require significant computational resources ."], "relation": "used for", "id": "2022.acl-long.93", "year": 2022, "rel_sent": "Experiments with BERTScore and MoverScore on summarization and translation show that FrugalScore is on par with the original metrics ( and sometimes better ) , while having several orders of magnitude less parameters and running several times faster .", "forward": true, "src_ids": "2022.acl-long.93_463"}
+{"input": "medical dialogue generation is done by using Method| context: medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses . the computational methods are still in their early stages and are not ready for widespread application despite their great potential . however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge .", "entity": "medical dialogue generation", "output": "transformer - based language models", "neg_sample": ["medical dialogue generation is done by using Method", "medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses .", "the computational methods are still in their early stages and are not ready for widespread application despite their great potential .", "however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge ."], "relation": "used for", "id": "2022.bionlp-1.10", "year": 2022, "rel_sent": "Incorporating Medical Knowledge to Transformer - based Language Models for Medical Dialogue Generation.", "forward": false, "src_ids": "2022.bionlp-1.10_464"}
+{"input": "transformer - based language models is used for Task| context: medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses . the computational methods are still in their early stages and are not ready for widespread application despite their great potential . existing transformer - based language models have shown promising results but lack domain - specific knowledge . however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge .", "entity": "transformer - based language models", "output": "medical dialogue generation", "neg_sample": ["transformer - based language models is used for Task", "medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses .", "the computational methods are still in their early stages and are not ready for widespread application despite their great potential .", "existing transformer - based language models have shown promising results but lack domain - specific knowledge .", "however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge ."], "relation": "used for", "id": "2022.bionlp-1.10", "year": 2022, "rel_sent": "Incorporating Medical Knowledge to Transformer - based Language Models for Medical Dialogue Generation.", "forward": true, "src_ids": "2022.bionlp-1.10_465"}
+{"input": "domain knowledge is done by using OtherScientificTerm| context: medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses . the computational methods are still in their early stages and are not ready for widespread application despite their great potential . existing transformer - based language models have shown promising results but lack domain - specific knowledge . however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge .", "entity": "domain knowledge", "output": "triples", "neg_sample": ["domain knowledge is done by using OtherScientificTerm", "medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses .", "the computational methods are still in their early stages and are not ready for widespread application despite their great potential .", "existing transformer - based language models have shown promising results but lack domain - specific knowledge .", "however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge ."], "relation": "used for", "id": "2022.bionlp-1.10", "year": 2022, "rel_sent": "We present a method that leverages an external medical knowledge graph and injects triples as domain knowledge into the utterances .", "forward": false, "src_ids": "2022.bionlp-1.10_466"}
+{"input": "triples is used for OtherScientificTerm| context: medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses . the computational methods are still in their early stages and are not ready for widespread application despite their great potential . existing transformer - based language models have shown promising results but lack domain - specific knowledge . however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge .", "entity": "triples", "output": "domain knowledge", "neg_sample": ["triples is used for OtherScientificTerm", "medical dialogue systems have the potential to assist doctors in expanding access to medical care , improving the quality of patient experiences , and lowering medical expenses .", "the computational methods are still in their early stages and are not ready for widespread application despite their great potential .", "existing transformer - based language models have shown promising results but lack domain - specific knowledge .", "however , to diagnose like doctors , an automatic medical diagnosis necessitates more stringent requirements for the rationality of the dialogue in the context of relevant knowledge ."], "relation": "used for", "id": "2022.bionlp-1.10", "year": 2022, "rel_sent": "We present a method that leverages an external medical knowledge graph and injects triples as domain knowledge into the utterances .", "forward": true, "src_ids": "2022.bionlp-1.10_467"}
+{"input": "distilled student network is used for Method| context: we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training .", "entity": "distilled student network", "output": "meta learning framework", "neg_sample": ["distilled student network is used for Method", "we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training ."], "relation": "used for", "id": "2022.acl-long.485", "year": 2022, "rel_sent": "We show the teacher network can learn to better transfer knowledge to the student network ( i.e. , learning to teach ) with the feedback from the performance of the distilled student network in a meta learning framework .", "forward": true, "src_ids": "2022.acl-long.485_468"}
+{"input": "meta learning algorithms is done by using Method| context: we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training .", "entity": "meta learning algorithms", "output": "pilot update mechanism", "neg_sample": ["meta learning algorithms is done by using Method", "we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training ."], "relation": "used for", "id": "2022.acl-long.485", "year": 2022, "rel_sent": "Moreover , we introduce a pilot update mechanism to improve the alignment between the inner - learner and meta - learner in meta learning algorithms that focus on an improved inner - learner .", "forward": false, "src_ids": "2022.acl-long.485_469"}
+{"input": "pilot update mechanism is used for Method| context: we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training .", "entity": "pilot update mechanism", "output": "meta learning algorithms", "neg_sample": ["pilot update mechanism is used for Method", "we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training ."], "relation": "used for", "id": "2022.acl-long.485", "year": 2022, "rel_sent": "Moreover , we introduce a pilot update mechanism to improve the alignment between the inner - learner and meta - learner in meta learning algorithms that focus on an improved inner - learner .", "forward": true, "src_ids": "2022.acl-long.485_470"}
+{"input": "meta learning framework is done by using Method| context: we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training .", "entity": "meta learning framework", "output": "distilled student network", "neg_sample": ["meta learning framework is done by using Method", "we present knowledge distillation with meta learning ( metadistil ) , a simple yet effective alternative to traditional knowledge distillation ( kd ) methods where the teacher model is fixed during training ."], "relation": "used for", "id": "2022.acl-long.485", "year": 2022, "rel_sent": "We show the teacher network can learn to better transfer knowledge to the student network ( i.e. , learning to teach ) with the feedback from the performance of the distilled student network in a meta learning framework .", "forward": false, "src_ids": "2022.acl-long.485_471"}
+{"input": "named entity recognition is done by using Method| context: previous work of class - incremental learning for named entity recognition ( ner ) relies on the assumption that there exists abundance of labeled data for the training of new classes .", "entity": "named entity recognition", "output": "few - shot class - incremental learning", "neg_sample": ["named entity recognition is done by using Method", "previous work of class - incremental learning for named entity recognition ( ner ) relies on the assumption that there exists abundance of labeled data for the training of new classes ."], "relation": "used for", "id": "2022.acl-long.43", "year": 2022, "rel_sent": "Few - Shot Class - Incremental Learning for Named Entity Recognition.", "forward": false, "src_ids": "2022.acl-long.43_472"}
+{"input": "unsupervised open - world classification is done by using Method| context: open - world classification in dialog systems require models to detect open intents , while ensuring the quality of in - domain ( id ) intent classification .", "entity": "unsupervised open - world classification", "output": "confidence - score based techniques", "neg_sample": ["unsupervised open - world classification is done by using Method", "open - world classification in dialog systems require models to detect open intents , while ensuring the quality of in - domain ( id ) intent classification ."], "relation": "used for", "id": "2022.insights-1.3", "year": 2022, "rel_sent": "Evaluating the Practical Utility of Confidence - score based Techniques for Unsupervised Open - world Classification.", "forward": false, "src_ids": "2022.insights-1.3_473"}
+{"input": "confidence - score based techniques is used for Task| context: open - world classification in dialog systems require models to detect open intents , while ensuring the quality of in - domain ( id ) intent classification .", "entity": "confidence - score based techniques", "output": "unsupervised open - world classification", "neg_sample": ["confidence - score based techniques is used for Task", "open - world classification in dialog systems require models to detect open intents , while ensuring the quality of in - domain ( id ) intent classification ."], "relation": "used for", "id": "2022.insights-1.3", "year": 2022, "rel_sent": "Evaluating the Practical Utility of Confidence - score based Techniques for Unsupervised Open - world Classification.", "forward": true, "src_ids": "2022.insights-1.3_474"}
+{"input": "synthetic dialogues is done by using OtherScientificTerm| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "synthetic dialogues", "output": "belief state annotations", "neg_sample": ["synthetic dialogues is done by using OtherScientificTerm", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "In this work , we introduce an augmentation framework that utilizes belief state annotations to match turns from various dialogues and form new synthetic dialogues in a bottom - up manner .", "forward": false, "src_ids": "2022.findings-acl.131_475"}
+{"input": "dialogue augmentation is done by using OtherScientificTerm| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "dialogue augmentation", "output": "belief state annotations", "neg_sample": ["dialogue augmentation is done by using OtherScientificTerm", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "Further analysis shows that our model performs better on seen values during training , and it is also more robust to unseen values . We conclude that exploiting belief state annotations enhances dialogue augmentation and results in improved models in n - shot training scenarios .", "forward": false, "src_ids": "2022.findings-acl.131_476"}
+{"input": "belief state annotations is used for OtherScientificTerm| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "belief state annotations", "output": "synthetic dialogues", "neg_sample": ["belief state annotations is used for OtherScientificTerm", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "In this work , we introduce an augmentation framework that utilizes belief state annotations to match turns from various dialogues and form new synthetic dialogues in a bottom - up manner .", "forward": true, "src_ids": "2022.findings-acl.131_477"}
+{"input": "new domain is done by using Method| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "new domain", "output": "dst model", "neg_sample": ["new domain is done by using Method", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "Our augmentation strategy yields significant improvements when both adapting a DST model to a new domain , and when adapting a language model to the DST task , on evaluations with TRADE and TOD - BERT models .", "forward": false, "src_ids": "2022.findings-acl.131_478"}
+{"input": "dst model is used for Material| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "dst model", "output": "new domain", "neg_sample": ["dst model is used for Material", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "Our augmentation strategy yields significant improvements when both adapting a DST model to a new domain , and when adapting a language model to the DST task , on evaluations with TRADE and TOD - BERT models .", "forward": true, "src_ids": "2022.findings-acl.131_479"}
+{"input": "dst task is done by using Method| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "dst task", "output": "language model", "neg_sample": ["dst task is done by using Method", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "Our augmentation strategy yields significant improvements when both adapting a DST model to a new domain , and when adapting a language model to the DST task , on evaluations with TRADE and TOD - BERT models .", "forward": false, "src_ids": "2022.findings-acl.131_480"}
+{"input": "language model is used for Task| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "language model", "output": "dst task", "neg_sample": ["language model is used for Task", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "Our augmentation strategy yields significant improvements when both adapting a DST model to a new domain , and when adapting a language model to the DST task , on evaluations with TRADE and TOD - BERT models .", "forward": true, "src_ids": "2022.findings-acl.131_481"}
+{"input": "belief state annotations is used for Task| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "belief state annotations", "output": "dialogue augmentation", "neg_sample": ["belief state annotations is used for Task", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.131", "year": 2022, "rel_sent": "Further analysis shows that our model performs better on seen values during training , and it is also more robust to unseen values . We conclude that exploiting belief state annotations enhances dialogue augmentation and results in improved models in n - shot training scenarios .", "forward": true, "src_ids": "2022.findings-acl.131_482"}
+{"input": "dialogue response is done by using Task| context: fact - checking is an essential tool to mitigate the spread of misinformation and disinformation .", "entity": "dialogue response", "output": "claim verification task", "neg_sample": ["dialogue response is done by using Task", "fact - checking is an essential tool to mitigate the spread of misinformation and disinformation ."], "relation": "used for", "id": "2022.acl-long.263", "year": 2022, "rel_sent": "There are three sub - tasks in DialFact : 1 ) Verifiable claim detection task distinguishes whether a response carries verifiable factual information ; 2 ) Evidence retrieval task retrieves the most relevant Wikipedia snippets as evidence ; 3 ) Claim verification task predicts a dialogue response to be supported , refuted , or not enough information .", "forward": false, "src_ids": "2022.acl-long.263_483"}
+{"input": "claim verification task is used for OtherScientificTerm| context: fact - checking is an essential tool to mitigate the spread of misinformation and disinformation .", "entity": "claim verification task", "output": "dialogue response", "neg_sample": ["claim verification task is used for OtherScientificTerm", "fact - checking is an essential tool to mitigate the spread of misinformation and disinformation ."], "relation": "used for", "id": "2022.acl-long.263", "year": 2022, "rel_sent": "There are three sub - tasks in DialFact : 1 ) Verifiable claim detection task distinguishes whether a response carries verifiable factual information ; 2 ) Evidence retrieval task retrieves the most relevant Wikipedia snippets as evidence ; 3 ) Claim verification task predicts a dialogue response to be supported , refuted , or not enough information .", "forward": true, "src_ids": "2022.acl-long.263_484"}
+{"input": "joint entity alignment ( ea ) is done by using Method| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "joint entity alignment ( ea )", "output": "unsupervised method", "neg_sample": ["joint entity alignment ( ea ) is done by using Method", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "An Accurate Unsupervised Method for Joint Entity Alignment and Dangling Entity Detection.", "forward": false, "src_ids": "2022.findings-acl.183_485"}
+{"input": "joint entity alignment ( ea ) is done by using Method| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "joint entity alignment ( ea )", "output": "unsupervised method", "neg_sample": ["joint entity alignment ( ea ) is done by using Method", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "In this paper , we propose a novel accurate Unsupervised method for joint Entity alignment ( EA ) and Dangling entity detection ( DED ) , called UED .", "forward": false, "src_ids": "2022.findings-acl.183_486"}
+{"input": "dangling entity detection ( ded ) is done by using Method| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "dangling entity detection ( ded )", "output": "unsupervised method", "neg_sample": ["dangling entity detection ( ded ) is done by using Method", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "In this paper , we propose a novel accurate Unsupervised method for joint Entity alignment ( EA ) and Dangling entity detection ( DED ) , called UED .", "forward": false, "src_ids": "2022.findings-acl.183_487"}
+{"input": "unsupervised method is used for Task| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "unsupervised method", "output": "joint entity alignment ( ea )", "neg_sample": ["unsupervised method is used for Task", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "An Accurate Unsupervised Method for Joint Entity Alignment and Dangling Entity Detection.", "forward": true, "src_ids": "2022.findings-acl.183_488"}
+{"input": "unsupervised method is used for Task| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "unsupervised method", "output": "joint entity alignment ( ea )", "neg_sample": ["unsupervised method is used for Task", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "In this paper , we propose a novel accurate Unsupervised method for joint Entity alignment ( EA ) and Dangling entity detection ( DED ) , called UED .", "forward": true, "src_ids": "2022.findings-acl.183_489"}
+{"input": "pseudo entity pairs is done by using Method| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "pseudo entity pairs", "output": "ued", "neg_sample": ["pseudo entity pairs is done by using Method", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": false, "src_ids": "2022.findings-acl.183_490"}
+{"input": "entity alignment is done by using Method| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "entity alignment", "output": "ued", "neg_sample": ["entity alignment is done by using Method", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": false, "src_ids": "2022.findings-acl.183_491"}
+{"input": "pseudo entity pairs is done by using OtherScientificTerm| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "pseudo entity pairs", "output": "literal semantic information", "neg_sample": ["pseudo entity pairs is done by using OtherScientificTerm", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": false, "src_ids": "2022.findings-acl.183_492"}
+{"input": "literal semantic information is used for OtherScientificTerm| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "literal semantic information", "output": "pseudo entity pairs", "neg_sample": ["literal semantic information is used for OtherScientificTerm", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": true, "src_ids": "2022.findings-acl.183_493"}
+{"input": "ued is used for OtherScientificTerm| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "ued", "output": "pseudo entity pairs", "neg_sample": ["ued is used for OtherScientificTerm", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": true, "src_ids": "2022.findings-acl.183_494"}
+{"input": "entity alignment is done by using OtherScientificTerm| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "entity alignment", "output": "globally guided alignment information", "neg_sample": ["entity alignment is done by using OtherScientificTerm", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": false, "src_ids": "2022.findings-acl.183_495"}
+{"input": "globally guided alignment information is used for Task| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "globally guided alignment information", "output": "entity alignment", "neg_sample": ["globally guided alignment information is used for Task", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": true, "src_ids": "2022.findings-acl.183_496"}
+{"input": "ued is used for Task| context: knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) . the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming .", "entity": "ued", "output": "entity alignment", "neg_sample": ["ued is used for Task", "knowledge graph integration typically suffers from the widely existing dangling entities that can not find alignment cross knowledge graphs ( kgs ) .", "the dangling entity set is unavailable in most real - world scenarios , and manually mining the entity pairs that consist of entities with the same meaning is labor - consuming ."], "relation": "used for", "id": "2022.findings-acl.183", "year": 2022, "rel_sent": "The UED mines the literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA and then utilizes the EA results to assist the DED .", "forward": true, "src_ids": "2022.findings-acl.183_497"}
+{"input": "human evaluation datasheet is used for Metric| context: originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al . ( 2019 ) , and gebru et al .", "entity": "human evaluation datasheet", "output": "meta - evaluation", "neg_sample": ["human evaluation datasheet is used for Metric", "originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al .", "( 2019 ) , and gebru et al ."], "relation": "used for", "id": "2022.humeval-1.6", "year": 2022, "rel_sent": "( 2020 ) , HEDS facilitates the recording of properties of human evaluations in sufficient detail , and with sufficient standardisation , to support comparability , meta - evaluation , and reproducibility assessments for human evaluations .", "forward": true, "src_ids": "2022.humeval-1.6_498"}
+{"input": "comparability is done by using Method| context: originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al . ( 2019 ) , and gebru et al .", "entity": "comparability", "output": "human evaluation datasheet", "neg_sample": ["comparability is done by using Method", "originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al .", "( 2019 ) , and gebru et al ."], "relation": "used for", "id": "2022.humeval-1.6", "year": 2022, "rel_sent": "( 2020 ) , HEDS facilitates the recording of properties of human evaluations in sufficient detail , and with sufficient standardisation , to support comparability , meta - evaluation , and reproducibility assessments for human evaluations .", "forward": false, "src_ids": "2022.humeval-1.6_499"}
+{"input": "meta - evaluation is done by using Method| context: originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al . ( 2019 ) , and gebru et al .", "entity": "meta - evaluation", "output": "human evaluation datasheet", "neg_sample": ["meta - evaluation is done by using Method", "originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al .", "( 2019 ) , and gebru et al ."], "relation": "used for", "id": "2022.humeval-1.6", "year": 2022, "rel_sent": "( 2020 ) , HEDS facilitates the recording of properties of human evaluations in sufficient detail , and with sufficient standardisation , to support comparability , meta - evaluation , and reproducibility assessments for human evaluations .", "forward": false, "src_ids": "2022.humeval-1.6_500"}
+{"input": "reproducibility assessments is done by using Method| context: originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al . ( 2019 ) , and gebru et al .", "entity": "reproducibility assessments", "output": "human evaluation datasheet", "neg_sample": ["reproducibility assessments is done by using Method", "originally taking inspiration from seminal papers by bender and friedman ( 2018 ) , mitchell et al .", "( 2019 ) , and gebru et al ."], "relation": "used for", "id": "2022.humeval-1.6", "year": 2022, "rel_sent": "( 2020 ) , HEDS facilitates the recording of properties of human evaluations in sufficient detail , and with sufficient standardisation , to support comparability , meta - evaluation , and reproducibility assessments for human evaluations .", "forward": false, "src_ids": "2022.humeval-1.6_501"}
+{"input": "inference scenario is done by using Material| context: the source discrepancy between training and inference hinders the translation performance of unmt models .", "entity": "inference scenario", "output": "pseudo parallel data", "neg_sample": ["inference scenario is done by using Material", "the source discrepancy between training and inference hinders the translation performance of unmt models ."], "relation": "used for", "id": "2022.acl-long.456", "year": 2022, "rel_sent": "To narrow the data gap , we propose an online self - training approach , which simultaneously uses the pseudo parallel data { natural source , translated target } to mimic the inference scenario .", "forward": false, "src_ids": "2022.acl-long.456_502"}
+{"input": "pseudo parallel data is used for Task| context: back - translation is a critical component of unsupervised neural machine translation ( unmt ) , which generates pseudo parallel data from target monolingual data . a unmt model is trained on the pseudo parallel data with translated source , and translates natural source sentences in inference . the source discrepancy between training and inference hinders the translation performance of unmt models .", "entity": "pseudo parallel data", "output": "inference scenario", "neg_sample": ["pseudo parallel data is used for Task", "back - translation is a critical component of unsupervised neural machine translation ( unmt ) , which generates pseudo parallel data from target monolingual data .", "a unmt model is trained on the pseudo parallel data with translated source , and translates natural source sentences in inference .", "the source discrepancy between training and inference hinders the translation performance of unmt models ."], "relation": "used for", "id": "2022.acl-long.456", "year": 2022, "rel_sent": "To narrow the data gap , we propose an online self - training approach , which simultaneously uses the pseudo parallel data { natural source , translated target } to mimic the inference scenario .", "forward": true, "src_ids": "2022.acl-long.456_503"}
+{"input": "data - linked prompts is done by using OtherScientificTerm| context: promptsource is a system for creating , sharing , and using natural language prompts . prompts are functions that map an example from a dataset to a natural language input and target output . using prompts to train and query language models is an emerging area in nlp that requires new tools that let users develop and refine these prompts collaboratively .", "entity": "data - linked prompts", "output": "templating language", "neg_sample": ["data - linked prompts is done by using OtherScientificTerm", "promptsource is a system for creating , sharing , and using natural language prompts .", "prompts are functions that map an example from a dataset to a natural language input and target output .", "using prompts to train and query language models is an emerging area in nlp that requires new tools that let users develop and refine these prompts collaboratively ."], "relation": "used for", "id": "2022.acl-demo.9", "year": 2022, "rel_sent": "PromptSource addresses the emergent challenges in this new setting with ( 1 ) a templating language for defining data - linked prompts , ( 2 ) an interface that lets users quickly iterate on prompt development by observing outputs of their prompts on many examples , and ( 3 ) a community - driven set of guidelines for contributing new prompts to a common pool .", "forward": false, "src_ids": "2022.acl-demo.9_504"}
+{"input": "templating language is used for OtherScientificTerm| context: promptsource is a system for creating , sharing , and using natural language prompts . prompts are functions that map an example from a dataset to a natural language input and target output . using prompts to train and query language models is an emerging area in nlp that requires new tools that let users develop and refine these prompts collaboratively .", "entity": "templating language", "output": "data - linked prompts", "neg_sample": ["templating language is used for OtherScientificTerm", "promptsource is a system for creating , sharing , and using natural language prompts .", "prompts are functions that map an example from a dataset to a natural language input and target output .", "using prompts to train and query language models is an emerging area in nlp that requires new tools that let users develop and refine these prompts collaboratively ."], "relation": "used for", "id": "2022.acl-demo.9", "year": 2022, "rel_sent": "PromptSource addresses the emergent challenges in this new setting with ( 1 ) a templating language for defining data - linked prompts , ( 2 ) an interface that lets users quickly iterate on prompt development by observing outputs of their prompts on many examples , and ( 3 ) a community - driven set of guidelines for contributing new prompts to a common pool .", "forward": true, "src_ids": "2022.acl-demo.9_505"}
+{"input": "human language modeling is done by using Method| context: natural language is generated by people , yet traditional language modeling views words or documents as if generated independently . here , we propose human language modeling ( hulm ) , a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents ( e.g.", "entity": "human language modeling", "output": "large - scale transformer model", "neg_sample": ["human language modeling is done by using Method", "natural language is generated by people , yet traditional language modeling views words or documents as if generated independently .", "here , we propose human language modeling ( hulm ) , a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents ( e.g."], "relation": "used for", "id": "2022.findings-acl.52", "year": 2022, "rel_sent": "We introduce , HaRT , a large - scale transformer model for solving HuLM , pre - trained on approximately 100,000 social media users , and demonstrate it 's effectiveness in terms of both language modeling ( perplexity ) for social media and fine - tuning for 4 downstream tasks spanning document- and user - levels .", "forward": false, "src_ids": "2022.findings-acl.52_506"}
+{"input": "large - scale transformer model is used for Task| context: natural language is generated by people , yet traditional language modeling views words or documents as if generated independently .", "entity": "large - scale transformer model", "output": "human language modeling", "neg_sample": ["large - scale transformer model is used for Task", "natural language is generated by people , yet traditional language modeling views words or documents as if generated independently ."], "relation": "used for", "id": "2022.findings-acl.52", "year": 2022, "rel_sent": "We introduce , HaRT , a large - scale transformer model for solving HuLM , pre - trained on approximately 100,000 social media users , and demonstrate it 's effectiveness in terms of both language modeling ( perplexity ) for social media and fine - tuning for 4 downstream tasks spanning document- and user - levels .", "forward": true, "src_ids": "2022.findings-acl.52_507"}
+{"input": "downstream tasks is done by using Method| context: natural language is generated by people , yet traditional language modeling views words or documents as if generated independently . here , we propose human language modeling ( hulm ) , a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents ( e.g.", "entity": "downstream tasks", "output": "fine - tuning", "neg_sample": ["downstream tasks is done by using Method", "natural language is generated by people , yet traditional language modeling views words or documents as if generated independently .", "here , we propose human language modeling ( hulm ) , a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents ( e.g."], "relation": "used for", "id": "2022.findings-acl.52", "year": 2022, "rel_sent": "We introduce , HaRT , a large - scale transformer model for solving HuLM , pre - trained on approximately 100,000 social media users , and demonstrate it 's effectiveness in terms of both language modeling ( perplexity ) for social media and fine - tuning for 4 downstream tasks spanning document- and user - levels .", "forward": false, "src_ids": "2022.findings-acl.52_508"}
+{"input": "fine - tuning is used for Task| context: natural language is generated by people , yet traditional language modeling views words or documents as if generated independently . here , we propose human language modeling ( hulm ) , a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents ( e.g.", "entity": "fine - tuning", "output": "downstream tasks", "neg_sample": ["fine - tuning is used for Task", "natural language is generated by people , yet traditional language modeling views words or documents as if generated independently .", "here , we propose human language modeling ( hulm ) , a hierarchical extension to the language modeling problem where by a human- level exists to connect sequences of documents ( e.g."], "relation": "used for", "id": "2022.findings-acl.52", "year": 2022, "rel_sent": "We introduce , HaRT , a large - scale transformer model for solving HuLM , pre - trained on approximately 100,000 social media users , and demonstrate it 's effectiveness in terms of both language modeling ( perplexity ) for social media and fine - tuning for 4 downstream tasks spanning document- and user - levels .", "forward": true, "src_ids": "2022.findings-acl.52_509"}
+{"input": "turkish is used for Task| context: having sufficient resources for language x lifts it from the under - resourced languages class , but not necessarily from the under - researched class .", "entity": "turkish", "output": "language modeling", "neg_sample": ["turkish is used for Task", "having sufficient resources for language x lifts it from the under - resourced languages class , but not necessarily from the under - researched class ."], "relation": "used for", "id": "2022.findings-acl.69", "year": 2022, "rel_sent": "Moreover , we present four new benchmarking datasets in Turkish for language modeling , sentence segmentation , and spell checking .", "forward": true, "src_ids": "2022.findings-acl.69_510"}
+{"input": "language modeling is done by using Material| context: having sufficient resources for language x lifts it from the under - resourced languages class , but not necessarily from the under - researched class .", "entity": "language modeling", "output": "turkish", "neg_sample": ["language modeling is done by using Material", "having sufficient resources for language x lifts it from the under - resourced languages class , but not necessarily from the under - researched class ."], "relation": "used for", "id": "2022.findings-acl.69", "year": 2022, "rel_sent": "Moreover , we present four new benchmarking datasets in Turkish for language modeling , sentence segmentation , and spell checking .", "forward": false, "src_ids": "2022.findings-acl.69_511"}
+{"input": "relation extraction task is done by using Method| context: biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery . pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction .", "entity": "relation extraction task", "output": "encoder - only and encoder - decoder transformers", "neg_sample": ["relation extraction task is done by using Method", "biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery .", "pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "Comparing Encoder - Only and Encoder - Decoder Transformers for Relation Extraction from Biomedical Texts : An Empirical Study on Ten Benchmark Datasets.", "forward": false, "src_ids": "2022.bionlp-1.37_512"}
+{"input": "encoder - only and encoder - decoder transformers is used for Task| context: biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery .", "entity": "encoder - only and encoder - decoder transformers", "output": "relation extraction task", "neg_sample": ["encoder - only and encoder - decoder transformers is used for Task", "biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "Comparing Encoder - Only and Encoder - Decoder Transformers for Relation Extraction from Biomedical Texts : An Empirical Study on Ten Benchmark Datasets.", "forward": true, "src_ids": "2022.bionlp-1.37_513"}
+{"input": "multi - task fine - tuning is used for Task| context: pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction .", "entity": "multi - task fine - tuning", "output": "biomedical relation extraction tasks", "neg_sample": ["multi - task fine - tuning is used for Task", "pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "We also explore the use of multi - task fine - tuning to investigate the correlation among major biomedical relation extraction tasks .", "forward": true, "src_ids": "2022.bionlp-1.37_514"}
+{"input": "multi - task learning is used for Task| context: pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction .", "entity": "multi - task learning", "output": "biomedical relation extraction tasks", "neg_sample": ["multi - task learning is used for Task", "pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "We report performance ( microf - score ) using T5 , BioBERT and PubMedBERT , demonstrating that T5 and multi - task learning can improve the performance of the biomedical relation extraction task .", "forward": true, "src_ids": "2022.bionlp-1.37_515"}
+{"input": "t5 is used for Task| context: pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction .", "entity": "t5", "output": "biomedical relation extraction tasks", "neg_sample": ["t5 is used for Task", "pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "We report performance ( microf - score ) using T5 , BioBERT and PubMedBERT , demonstrating that T5 and multi - task learning can improve the performance of the biomedical relation extraction task .", "forward": true, "src_ids": "2022.bionlp-1.37_516"}
+{"input": "biomedical relation extraction tasks is done by using Method| context: biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery . pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction .", "entity": "biomedical relation extraction tasks", "output": "t5", "neg_sample": ["biomedical relation extraction tasks is done by using Method", "biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery .", "pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "We report performance ( microf - score ) using T5 , BioBERT and PubMedBERT , demonstrating that T5 and multi - task learning can improve the performance of the biomedical relation extraction task .", "forward": false, "src_ids": "2022.bionlp-1.37_517"}
+{"input": "biomedical relation extraction tasks is done by using Method| context: biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery . pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction .", "entity": "biomedical relation extraction tasks", "output": "multi - task fine - tuning", "neg_sample": ["biomedical relation extraction tasks is done by using Method", "biomedical relation extraction , aiming to automatically discover high - quality and semantic relations between the entities from free text , is becoming a vital step for automated knowledge discovery .", "pretrained language models have achieved impressive performance on various natural language processing tasks , including relation extraction ."], "relation": "used for", "id": "2022.bionlp-1.37", "year": 2022, "rel_sent": "We also explore the use of multi - task fine - tuning to investigate the correlation among major biomedical relation extraction tasks .", "forward": false, "src_ids": "2022.bionlp-1.37_518"}
+{"input": "decision - making process is done by using Method| context: many recent deep learning - based solutions have adopted the attention mechanism in various tasks in the field of nlp . however , the inherent characteristics of deep learning models and the flexibility of the attention mechanism increase the models ' complexity , thus leading to challenges in model explainability .", "entity": "decision - making process", "output": "two - tier attention architecture", "neg_sample": ["decision - making process is done by using Method", "many recent deep learning - based solutions have adopted the attention mechanism in various tasks in the field of nlp .", "however , the inherent characteristics of deep learning models and the flexibility of the attention mechanism increase the models ' complexity , thus leading to challenges in model explainability ."], "relation": "used for", "id": "2022.findings-acl.178", "year": 2022, "rel_sent": "To address this challenge , we propose a novel practical framework by utilizing a two - tier attention architecture to decouple the complexity of explanation and the decision - making process .", "forward": false, "src_ids": "2022.findings-acl.178_519"}
+{"input": "two - tier attention architecture is used for Task| context: many recent deep learning - based solutions have adopted the attention mechanism in various tasks in the field of nlp . however , the inherent characteristics of deep learning models and the flexibility of the attention mechanism increase the models ' complexity , thus leading to challenges in model explainability .", "entity": "two - tier attention architecture", "output": "decision - making process", "neg_sample": ["two - tier attention architecture is used for Task", "many recent deep learning - based solutions have adopted the attention mechanism in various tasks in the field of nlp .", "however , the inherent characteristics of deep learning models and the flexibility of the attention mechanism increase the models ' complexity , thus leading to challenges in model explainability ."], "relation": "used for", "id": "2022.findings-acl.178", "year": 2022, "rel_sent": "To address this challenge , we propose a novel practical framework by utilizing a two - tier attention architecture to decouple the complexity of explanation and the decision - making process .", "forward": true, "src_ids": "2022.findings-acl.178_520"}
+{"input": "open - domain passage retrieval ( odpr ) is done by using Method| context: training dense passage representations via contrastive learning has been shown effective for open - domain passage retrieval ( odpr ) . existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining . however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity . specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal .", "entity": "open - domain passage retrieval ( odpr )", "output": "sentence - aware contrastive learning", "neg_sample": ["open - domain passage retrieval ( odpr ) is done by using Method", "training dense passage representations via contrastive learning has been shown effective for open - domain passage retrieval ( odpr ) .", "existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining .", "however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity .", "specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal ."], "relation": "used for", "id": "2022.acl-long.76", "year": 2022, "rel_sent": "Sentence - aware Contrastive Learning for Open - Domain Passage Retrieval.", "forward": false, "src_ids": "2022.acl-long.76_521"}
+{"input": "sentence - aware contrastive learning is used for Task| context: existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining . however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity . specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal .", "entity": "sentence - aware contrastive learning", "output": "open - domain passage retrieval ( odpr )", "neg_sample": ["sentence - aware contrastive learning is used for Task", "existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining .", "however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity .", "specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal ."], "relation": "used for", "id": "2022.acl-long.76", "year": 2022, "rel_sent": "Sentence - aware Contrastive Learning for Open - Domain Passage Retrieval.", "forward": true, "src_ids": "2022.acl-long.76_522"}
+{"input": "sentence representations is done by using Method| context: training dense passage representations via contrastive learning has been shown effective for open - domain passage retrieval ( odpr ) . existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining . however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity . specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal .", "entity": "sentence representations", "output": "in - passage negative sampling strategy", "neg_sample": ["sentence representations is done by using Method", "training dense passage representations via contrastive learning has been shown effective for open - domain passage retrieval ( odpr ) .", "existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining .", "however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity .", "specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal ."], "relation": "used for", "id": "2022.acl-long.76", "year": 2022, "rel_sent": "In detail , we introduce an in - passage negative sampling strategy to encourage a diverse generation of sentence representations within the same passage .", "forward": false, "src_ids": "2022.acl-long.76_523"}
+{"input": "in - passage negative sampling strategy is used for Method| context: training dense passage representations via contrastive learning has been shown effective for open - domain passage retrieval ( odpr ) . existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining . however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity . specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal .", "entity": "in - passage negative sampling strategy", "output": "sentence representations", "neg_sample": ["in - passage negative sampling strategy is used for Method", "training dense passage representations via contrastive learning has been shown effective for open - domain passage retrieval ( odpr ) .", "existing studies focus on further optimizing by improving negative sampling strategy or extra pretraining .", "however , these studies keep unknown in capturing passage with internal representation conflicts from improper modeling granularity .", "specifically , under our observation that a passage can be organized by multiple semantically different sentences , modeling such a passage as a unified dense vector is not optimal ."], "relation": "used for", "id": "2022.acl-long.76", "year": 2022, "rel_sent": "In detail , we introduce an in - passage negative sampling strategy to encourage a diverse generation of sentence representations within the same passage .", "forward": true, "src_ids": "2022.acl-long.76_524"}
+{"input": "nmt systems is done by using Method| context: neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows . consequently , human translators are increasingly working as post - editors for machine - translated content .", "entity": "nmt systems", "output": "error analysis", "neg_sample": ["nmt systems is done by using Method", "neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows .", "consequently , human translators are increasingly working as post - editors for machine - translated content ."], "relation": "used for", "id": "2022.eamt-1.46", "year": 2022, "rel_sent": "In the specific context of translation , we aim at developing new tools and methodologies to improve prediction attribution , error analysis , and controllable generation for NMT systems .", "forward": false, "src_ids": "2022.eamt-1.46_525"}
+{"input": "nmt systems is done by using Task| context: neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows . consequently , human translators are increasingly working as post - editors for machine - translated content .", "entity": "nmt systems", "output": "controllable generation", "neg_sample": ["nmt systems is done by using Task", "neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows .", "consequently , human translators are increasingly working as post - editors for machine - translated content ."], "relation": "used for", "id": "2022.eamt-1.46", "year": 2022, "rel_sent": "In the specific context of translation , we aim at developing new tools and methodologies to improve prediction attribution , error analysis , and controllable generation for NMT systems .", "forward": false, "src_ids": "2022.eamt-1.46_526"}
+{"input": "controllable generation is used for Method| context: neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows . consequently , human translators are increasingly working as post - editors for machine - translated content .", "entity": "controllable generation", "output": "nmt systems", "neg_sample": ["controllable generation is used for Method", "neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows .", "consequently , human translators are increasingly working as post - editors for machine - translated content ."], "relation": "used for", "id": "2022.eamt-1.46", "year": 2022, "rel_sent": "In the specific context of translation , we aim at developing new tools and methodologies to improve prediction attribution , error analysis , and controllable generation for NMT systems .", "forward": true, "src_ids": "2022.eamt-1.46_527"}
+{"input": "error analysis is used for Method| context: neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows . consequently , human translators are increasingly working as post - editors for machine - translated content .", "entity": "error analysis", "output": "nmt systems", "neg_sample": ["error analysis is used for Method", "neural machine translation ( nmt ) systems are nowadays essential components of professional translation workflows .", "consequently , human translators are increasingly working as post - editors for machine - translated content ."], "relation": "used for", "id": "2022.eamt-1.46", "year": 2022, "rel_sent": "In the specific context of translation , we aim at developing new tools and methodologies to improve prediction attribution , error analysis , and controllable generation for NMT systems .", "forward": true, "src_ids": "2022.eamt-1.46_528"}
+{"input": "personality prediction models is done by using OtherScientificTerm| context: research at the intersection of personality psychology , computer science , and linguistics has recently focused increasingly on modeling and predicting personality from language use .", "entity": "personality prediction models", "output": "psycholinguistic features", "neg_sample": ["personality prediction models is done by using OtherScientificTerm", "research at the intersection of personality psychology , computer science , and linguistics has recently focused increasingly on modeling and predicting personality from language use ."], "relation": "used for", "id": "2022.wassa-1.17", "year": 2022, "rel_sent": "In addition , we perform ablation experiments to quantify the impact of different categories of psycholinguistic features in the respective personality prediction models .", "forward": false, "src_ids": "2022.wassa-1.17_529"}
+{"input": "psycholinguistic features is used for Method| context: research at the intersection of personality psychology , computer science , and linguistics has recently focused increasingly on modeling and predicting personality from language use .", "entity": "psycholinguistic features", "output": "personality prediction models", "neg_sample": ["psycholinguistic features is used for Method", "research at the intersection of personality psychology , computer science , and linguistics has recently focused increasingly on modeling and predicting personality from language use ."], "relation": "used for", "id": "2022.wassa-1.17", "year": 2022, "rel_sent": "In addition , we perform ablation experiments to quantify the impact of different categories of psycholinguistic features in the respective personality prediction models .", "forward": true, "src_ids": "2022.wassa-1.17_530"}
+{"input": "nat student model is done by using Method| context: knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words .", "entity": "nat student model", "output": "monolingual kd", "neg_sample": ["nat student model is done by using Method", "knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words ."], "relation": "used for", "id": "2022.acl-long.172", "year": 2022, "rel_sent": "Monolingual KD is able to transfer both the knowledge of the original bilingual data ( implicitly encoded in the trained AT teacher model ) and that of the new monolingual data to the NAT student model .", "forward": false, "src_ids": "2022.acl-long.172_531"}
+{"input": "low - frequency word translation is done by using Method| context: knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words .", "entity": "low - frequency word translation", "output": "monolingual kd", "neg_sample": ["low - frequency word translation is done by using Method", "knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words ."], "relation": "used for", "id": "2022.acl-long.172", "year": 2022, "rel_sent": "Extensive experiments on eight WMT benchmarks over two advanced NAT models show that monolingual KD consistently outperforms the standard KD by improving low - frequency word translation , without introducing any computational cost .", "forward": false, "src_ids": "2022.acl-long.172_532"}
+{"input": "monolingual kd is used for Method| context: knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words .", "entity": "monolingual kd", "output": "nat student model", "neg_sample": ["monolingual kd is used for Method", "knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words ."], "relation": "used for", "id": "2022.acl-long.172", "year": 2022, "rel_sent": "Monolingual KD is able to transfer both the knowledge of the original bilingual data ( implicitly encoded in the trained AT teacher model ) and that of the new monolingual data to the NAT student model .", "forward": true, "src_ids": "2022.acl-long.172_533"}
+{"input": "monolingual data is used for Method| context: knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words .", "entity": "monolingual data", "output": "nat student model", "neg_sample": ["monolingual data is used for Method", "knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words ."], "relation": "used for", "id": "2022.acl-long.172", "year": 2022, "rel_sent": "Monolingual KD is able to transfer both the knowledge of the original bilingual data ( implicitly encoded in the trained AT teacher model ) and that of the new monolingual data to the NAT student model .", "forward": true, "src_ids": "2022.acl-long.172_534"}
+{"input": "monolingual kd is used for Task| context: knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words .", "entity": "monolingual kd", "output": "low - frequency word translation", "neg_sample": ["monolingual kd is used for Task", "knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words ."], "relation": "used for", "id": "2022.acl-long.172", "year": 2022, "rel_sent": "Extensive experiments on eight WMT benchmarks over two advanced NAT models show that monolingual KD consistently outperforms the standard KD by improving low - frequency word translation , without introducing any computational cost .", "forward": true, "src_ids": "2022.acl-long.172_535"}
+{"input": "nat student model is done by using Material| context: knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words .", "entity": "nat student model", "output": "monolingual data", "neg_sample": ["nat student model is done by using Material", "knowledge distillation ( kd ) is the preliminary step for training non - autoregressive translation ( nat ) models , which eases the training of nat models at the cost of losing important information for translating low - frequency words ."], "relation": "used for", "id": "2022.acl-long.172", "year": 2022, "rel_sent": "Monolingual KD is able to transfer both the knowledge of the original bilingual data ( implicitly encoded in the trained AT teacher model ) and that of the new monolingual data to the NAT student model .", "forward": false, "src_ids": "2022.acl-long.172_536"}
+{"input": "zero - shot cross - lingual transfer is done by using Task| context: recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) . however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting . towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data . the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages .", "entity": "zero - shot cross - lingual transfer", "output": "language clustering", "neg_sample": ["zero - shot cross - lingual transfer is done by using Task", "recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) .", "however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting .", "towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data .", "the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages ."], "relation": "used for", "id": "2022.findings-acl.24", "year": 2022, "rel_sent": "Meta - X_{NLG } : A Meta - Learning Approach Based on Language Clustering for Zero - Shot Cross - Lingual Transfer and Generation.", "forward": false, "src_ids": "2022.findings-acl.24_537"}
+{"input": "language clustering is used for Task| context: recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) . however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting . towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data . the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages .", "entity": "language clustering", "output": "zero - shot cross - lingual transfer", "neg_sample": ["language clustering is used for Task", "recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) .", "however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting .", "towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data .", "the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages ."], "relation": "used for", "id": "2022.findings-acl.24", "year": 2022, "rel_sent": "Meta - X_{NLG } : A Meta - Learning Approach Based on Language Clustering for Zero - Shot Cross - Lingual Transfer and Generation.", "forward": true, "src_ids": "2022.findings-acl.24_538"}
+{"input": "meta - learning framework is used for OtherScientificTerm| context: recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) . however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting . the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages .", "entity": "meta - learning framework", "output": "shareable structures", "neg_sample": ["meta - learning framework is used for OtherScientificTerm", "recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) .", "however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting .", "the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages ."], "relation": "used for", "id": "2022.findings-acl.24", "year": 2022, "rel_sent": "In this paper , we propose a novel meta - learning framework ( called Meta - X_{NLG } ) to learn shareable structures from typologically diverse languages based on meta - learning and language clustering .", "forward": true, "src_ids": "2022.findings-acl.24_539"}
+{"input": "shareable structures is done by using Method| context: recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) . however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting . towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data . the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages .", "entity": "shareable structures", "output": "meta - learning framework", "neg_sample": ["shareable structures is done by using Method", "recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) .", "however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting .", "towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data .", "the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages ."], "relation": "used for", "id": "2022.findings-acl.24", "year": 2022, "rel_sent": "In this paper , we propose a novel meta - learning framework ( called Meta - X_{NLG } ) to learn shareable structures from typologically diverse languages based on meta - learning and language clustering .", "forward": false, "src_ids": "2022.findings-acl.24_540"}
+{"input": "accidental translation problem is done by using OtherScientificTerm| context: recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) . however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting . towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data . the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages .", "entity": "accidental translation problem", "output": "end - to - end nlg setup", "neg_sample": ["accidental translation problem is done by using OtherScientificTerm", "recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) .", "however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting .", "towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data .", "the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages ."], "relation": "used for", "id": "2022.findings-acl.24", "year": 2022, "rel_sent": "The careful design of the model makes this end - to - end NLG setup less vulnerable to the accidental translation problem , which is a prominent concern in zero - shot cross - lingual NLG tasks .", "forward": false, "src_ids": "2022.findings-acl.24_541"}
+{"input": "end - to - end nlg setup is used for Task| context: recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) . however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting . towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data . the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages .", "entity": "end - to - end nlg setup", "output": "accidental translation problem", "neg_sample": ["end - to - end nlg setup is used for Task", "recently , the nlp community has witnessed a rapid advancement in multilingual and cross - lingual transfer research where the supervision is transferred from high - resource languages ( hrls ) to low - resource languages ( lrls ) .", "however , the cross - lingual transfer is not uniform across languages , particularly in the zero - shot setting .", "towards this goal , one promising research direction is to learn shareable structures across multiple tasks with limited annotated data .", "the downstream multilingual applications may benefit from such a learning setup as most of the languages across the globe are low - resource and share some structures with other languages ."], "relation": "used for", "id": "2022.findings-acl.24", "year": 2022, "rel_sent": "The careful design of the model makes this end - to - end NLG setup less vulnerable to the accidental translation problem , which is a prominent concern in zero - shot cross - lingual NLG tasks .", "forward": true, "src_ids": "2022.findings-acl.24_542"}
+{"input": "hybrid granularities is done by using Method| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "hybrid granularities", "output": "hierarchical contrastive learning framework", "neg_sample": ["hybrid granularities is done by using Method", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Keywords and Instances : A Hierarchical Contrastive Learning Framework Unifying Hybrid Granularities for Text Generation.", "forward": false, "src_ids": "2022.acl-long.304_543"}
+{"input": "hierarchical contrastive learning framework is used for OtherScientificTerm| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "hierarchical contrastive learning framework", "output": "hybrid granularities", "neg_sample": ["hierarchical contrastive learning framework is used for OtherScientificTerm", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Keywords and Instances : A Hierarchical Contrastive Learning Framework Unifying Hybrid Granularities for Text Generation.", "forward": true, "src_ids": "2022.acl-long.304_544"}
+{"input": "hybrid granularities semantic meaning is done by using Method| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "hybrid granularities semantic meaning", "output": "hierarchical contrastive learning mechanism", "neg_sample": ["hybrid granularities semantic meaning is done by using Method", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Hence , in this work , we propose a hierarchical contrastive learning mechanism , which can unify hybrid granularities semantic meaning in the input text .", "forward": false, "src_ids": "2022.acl-long.304_545"}
+{"input": "hierarchical contrastive learning mechanism is used for OtherScientificTerm| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "hierarchical contrastive learning mechanism", "output": "hybrid granularities semantic meaning", "neg_sample": ["hierarchical contrastive learning mechanism is used for OtherScientificTerm", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Hence , in this work , we propose a hierarchical contrastive learning mechanism , which can unify hybrid granularities semantic meaning in the input text .", "forward": true, "src_ids": "2022.acl-long.304_546"}
+{"input": "keyword representations is done by using OtherScientificTerm| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "keyword representations", "output": "keyword graph", "neg_sample": ["keyword representations is done by using OtherScientificTerm", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Concretely , we first propose a keyword graph via contrastive correlations of positive - negative pairs to iteratively polish the keyword representations .", "forward": false, "src_ids": "2022.acl-long.304_547"}
+{"input": "keyword graph is used for Method| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "keyword graph", "output": "keyword representations", "neg_sample": ["keyword graph is used for Method", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Concretely , we first propose a keyword graph via contrastive correlations of positive - negative pairs to iteratively polish the keyword representations .", "forward": true, "src_ids": "2022.acl-long.304_548"}
+{"input": "contrastive correlations of positive - negative pairs is used for Method| context: contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references . existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships .", "entity": "contrastive correlations of positive - negative pairs", "output": "keyword representations", "neg_sample": ["contrastive correlations of positive - negative pairs is used for Method", "contrastive learning has achieved impressive success in generation tasks to militate the ' exposure bias ' problem and discriminatively exploit the different quality of references .", "existing works mostly focus on contrastive learning on the instance - level without discriminating the contribution of each word , while keywords are the gist of the text and dominant the constrained mapping relationships ."], "relation": "used for", "id": "2022.acl-long.304", "year": 2022, "rel_sent": "Concretely , we first propose a keyword graph via contrastive correlations of positive - negative pairs to iteratively polish the keyword representations .", "forward": true, "src_ids": "2022.acl-long.304_549"}
+{"input": "continual learning strategy is used for Method| context: despite its importance , the time variable has been largely neglected in the nlp and language model literature .", "entity": "continual learning strategy", "output": "twitter - based language models", "neg_sample": ["continual learning strategy is used for Method", "despite its importance , the time variable has been largely neglected in the nlp and language model literature ."], "relation": "used for", "id": "2022.acl-demo.25", "year": 2022, "rel_sent": "We show that a continual learning strategy contributes to enhancing Twitter - based language models ' capacity to deal with future and out - of - distribution tweets , while making them competitive with standardized and more monolithic benchmarks .", "forward": true, "src_ids": "2022.acl-demo.25_550"}
+{"input": "twitter - based language models is done by using Method| context: despite its importance , the time variable has been largely neglected in the nlp and language model literature .", "entity": "twitter - based language models", "output": "continual learning strategy", "neg_sample": ["twitter - based language models is done by using Method", "despite its importance , the time variable has been largely neglected in the nlp and language model literature ."], "relation": "used for", "id": "2022.acl-demo.25", "year": 2022, "rel_sent": "We show that a continual learning strategy contributes to enhancing Twitter - based language models ' capacity to deal with future and out - of - distribution tweets , while making them competitive with standardized and more monolithic benchmarks .", "forward": false, "src_ids": "2022.acl-demo.25_551"}
+{"input": "unisectional and intersectional social biases is done by using Method| context: as natural language processing systems become more widespread , it is necessary to address fairness issues in their implementation and deployment to ensure that their negative impacts on society are understood and minimized . however , there is limited work that studies fairness using a multilingual and intersectional framework or on downstream tasks .", "entity": "unisectional and intersectional social biases", "output": "statistical framework", "neg_sample": ["unisectional and intersectional social biases is done by using Method", "as natural language processing systems become more widespread , it is necessary to address fairness issues in their implementation and deployment to ensure that their negative impacts on society are understood and minimized .", "however , there is limited work that studies fairness using a multilingual and intersectional framework or on downstream tasks ."], "relation": "used for", "id": "2022.ltedi-1.11", "year": 2022, "rel_sent": "In this paper , we introduce four multilingual Equity Evaluation Corpora , supplementary test sets designed to measure social biases , and a novel statistical framework for studying unisectional and intersectional social biases in natural language processing .", "forward": false, "src_ids": "2022.ltedi-1.11_552"}
+{"input": "statistical framework is used for OtherScientificTerm| context: as natural language processing systems become more widespread , it is necessary to address fairness issues in their implementation and deployment to ensure that their negative impacts on society are understood and minimized . however , there is limited work that studies fairness using a multilingual and intersectional framework or on downstream tasks .", "entity": "statistical framework", "output": "unisectional and intersectional social biases", "neg_sample": ["statistical framework is used for OtherScientificTerm", "as natural language processing systems become more widespread , it is necessary to address fairness issues in their implementation and deployment to ensure that their negative impacts on society are understood and minimized .", "however , there is limited work that studies fairness using a multilingual and intersectional framework or on downstream tasks ."], "relation": "used for", "id": "2022.ltedi-1.11", "year": 2022, "rel_sent": "In this paper , we introduce four multilingual Equity Evaluation Corpora , supplementary test sets designed to measure social biases , and a novel statistical framework for studying unisectional and intersectional social biases in natural language processing .", "forward": true, "src_ids": "2022.ltedi-1.11_553"}
+{"input": "million - scale concept events is done by using Method| context: events are fundamental building blocks of real - world happenings .", "entity": "million - scale concept events", "output": "induction strategy", "neg_sample": ["million - scale concept events is done by using Method", "events are fundamental building blocks of real - world happenings ."], "relation": "used for", "id": "2022.acl-demo.23", "year": 2022, "rel_sent": "We also develop an induction strategy to create million - scale concept events and a schema organizing all events and relations in MMEKG .", "forward": false, "src_ids": "2022.acl-demo.23_554"}
+{"input": "induction strategy is used for OtherScientificTerm| context: events are fundamental building blocks of real - world happenings .", "entity": "induction strategy", "output": "million - scale concept events", "neg_sample": ["induction strategy is used for OtherScientificTerm", "events are fundamental building blocks of real - world happenings ."], "relation": "used for", "id": "2022.acl-demo.23", "year": 2022, "rel_sent": "We also develop an induction strategy to create million - scale concept events and a schema organizing all events and relations in MMEKG .", "forward": true, "src_ids": "2022.acl-demo.23_555"}
+{"input": "prompts is used for Metric| context: neural machine translation ( nmt ) has obtained significant performance improvement over the recent years . moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific .", "entity": "prompts", "output": "flexibility", "neg_sample": ["prompts is used for Metric", "neural machine translation ( nmt ) has obtained significant performance improvement over the recent years .", "moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific ."], "relation": "used for", "id": "2022.findings-acl.203", "year": 2022, "rel_sent": "To this end , we propose prompt - driven neural machine translation to incorporate prompts for enhancing translation control and enriching flexibility .", "forward": true, "src_ids": "2022.findings-acl.203_556"}
+{"input": "translation control is done by using OtherScientificTerm| context: neural machine translation ( nmt ) has obtained significant performance improvement over the recent years . however , nmt models still face various challenges including fragility and lack of style flexibility . moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific .", "entity": "translation control", "output": "prompts", "neg_sample": ["translation control is done by using OtherScientificTerm", "neural machine translation ( nmt ) has obtained significant performance improvement over the recent years .", "however , nmt models still face various challenges including fragility and lack of style flexibility .", "moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific ."], "relation": "used for", "id": "2022.findings-acl.203", "year": 2022, "rel_sent": "To this end , we propose prompt - driven neural machine translation to incorporate prompts for enhancing translation control and enriching flexibility .", "forward": false, "src_ids": "2022.findings-acl.203_557"}
+{"input": "flexibility is done by using OtherScientificTerm| context: neural machine translation ( nmt ) has obtained significant performance improvement over the recent years . however , nmt models still face various challenges including fragility and lack of style flexibility . moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific .", "entity": "flexibility", "output": "prompts", "neg_sample": ["flexibility is done by using OtherScientificTerm", "neural machine translation ( nmt ) has obtained significant performance improvement over the recent years .", "however , nmt models still face various challenges including fragility and lack of style flexibility .", "moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific ."], "relation": "used for", "id": "2022.findings-acl.203", "year": 2022, "rel_sent": "To this end , we propose prompt - driven neural machine translation to incorporate prompts for enhancing translation control and enriching flexibility .", "forward": false, "src_ids": "2022.findings-acl.203_558"}
+{"input": "prompts is used for Task| context: neural machine translation ( nmt ) has obtained significant performance improvement over the recent years . however , nmt models still face various challenges including fragility and lack of style flexibility . moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific .", "entity": "prompts", "output": "translation control", "neg_sample": ["prompts is used for Task", "neural machine translation ( nmt ) has obtained significant performance improvement over the recent years .", "however , nmt models still face various challenges including fragility and lack of style flexibility .", "moreover , current methods for instance - level constraints are limited in that they are either constraint - specific or model - specific ."], "relation": "used for", "id": "2022.findings-acl.203", "year": 2022, "rel_sent": "To this end , we propose prompt - driven neural machine translation to incorporate prompts for enhancing translation control and enriching flexibility .", "forward": true, "src_ids": "2022.findings-acl.203_559"}
+{"input": "few - shot relation extraction is done by using Method| context: few - shot relation extraction aims at predicting the relation for a pair of entities in a sentence by training with a few labelled examples in each relation . some recent works have introduced relation information ( i.e. , relation labels or descriptions ) to assist model learning based on prototype network . however , most of them constrain the prototypes of each relation class implicitly with relation information , generally through designing complex network structures , like generating hybrid features , combining with contrastive learning or attention networks . we argue that relation information can be introduced more explicitly and effectively into the model .", "entity": "few - shot relation extraction", "output": "relation information guided approach", "neg_sample": ["few - shot relation extraction is done by using Method", "few - shot relation extraction aims at predicting the relation for a pair of entities in a sentence by training with a few labelled examples in each relation .", "some recent works have introduced relation information ( i.e.", ", relation labels or descriptions ) to assist model learning based on prototype network .", "however , most of them constrain the prototypes of each relation class implicitly with relation information , generally through designing complex network structures , like generating hybrid features , combining with contrastive learning or attention networks .", "we argue that relation information can be introduced more explicitly and effectively into the model ."], "relation": "used for", "id": "2022.findings-acl.62", "year": 2022, "rel_sent": "A Simple yet Effective Relation Information Guided Approach for Few - Shot Relation Extraction.", "forward": false, "src_ids": "2022.findings-acl.62_560"}
+{"input": "relation information guided approach is used for Task| context: some recent works have introduced relation information ( i.e. , relation labels or descriptions ) to assist model learning based on prototype network . however , most of them constrain the prototypes of each relation class implicitly with relation information , generally through designing complex network structures , like generating hybrid features , combining with contrastive learning or attention networks . we argue that relation information can be introduced more explicitly and effectively into the model .", "entity": "relation information guided approach", "output": "few - shot relation extraction", "neg_sample": ["relation information guided approach is used for Task", "some recent works have introduced relation information ( i.e.", ", relation labels or descriptions ) to assist model learning based on prototype network .", "however , most of them constrain the prototypes of each relation class implicitly with relation information , generally through designing complex network structures , like generating hybrid features , combining with contrastive learning or attention networks .", "we argue that relation information can be introduced more explicitly and effectively into the model ."], "relation": "used for", "id": "2022.findings-acl.62", "year": 2022, "rel_sent": "A Simple yet Effective Relation Information Guided Approach for Few - Shot Relation Extraction.", "forward": true, "src_ids": "2022.findings-acl.62_561"}
+{"input": "direct addition approach is used for OtherScientificTerm| context: few - shot relation extraction aims at predicting the relation for a pair of entities in a sentence by training with a few labelled examples in each relation .", "entity": "direct addition approach", "output": "relation information", "neg_sample": ["direct addition approach is used for OtherScientificTerm", "few - shot relation extraction aims at predicting the relation for a pair of entities in a sentence by training with a few labelled examples in each relation ."], "relation": "used for", "id": "2022.findings-acl.62", "year": 2022, "rel_sent": "Thus , this paper proposes a direct addition approach to introduce relation information .", "forward": true, "src_ids": "2022.findings-acl.62_562"}
+{"input": "relation information is done by using Method| context: few - shot relation extraction aims at predicting the relation for a pair of entities in a sentence by training with a few labelled examples in each relation . some recent works have introduced relation information ( i.e. , relation labels or descriptions ) to assist model learning based on prototype network . however , most of them constrain the prototypes of each relation class implicitly with relation information , generally through designing complex network structures , like generating hybrid features , combining with contrastive learning or attention networks . we argue that relation information can be introduced more explicitly and effectively into the model .", "entity": "relation information", "output": "direct addition approach", "neg_sample": ["relation information is done by using Method", "few - shot relation extraction aims at predicting the relation for a pair of entities in a sentence by training with a few labelled examples in each relation .", "some recent works have introduced relation information ( i.e.", ", relation labels or descriptions ) to assist model learning based on prototype network .", "however , most of them constrain the prototypes of each relation class implicitly with relation information , generally through designing complex network structures , like generating hybrid features , combining with contrastive learning or attention networks .", "we argue that relation information can be introduced more explicitly and effectively into the model ."], "relation": "used for", "id": "2022.findings-acl.62", "year": 2022, "rel_sent": "Thus , this paper proposes a direct addition approach to introduce relation information .", "forward": false, "src_ids": "2022.findings-acl.62_563"}
+{"input": "propositional structures is used for Task| context: we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains .", "entity": "propositional structures", "output": "lexical expansion", "neg_sample": ["propositional structures is used for Task", "we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains ."], "relation": "used for", "id": "2022.findings-acl.264", "year": 2022, "rel_sent": "Prudent ( automatic ) selection of terms from propositional structures for lexical expansion ( via semantic similarity ) produces new moral dimension lexicons at three levels of granularity beyond a strong baseline lexicon .", "forward": true, "src_ids": "2022.findings-acl.264_564"}
+{"input": "extraction of domain - dependent concern types is done by using Method| context: we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains .", "entity": "extraction of domain - dependent concern types", "output": "semi - automatic resource building", "neg_sample": ["extraction of domain - dependent concern types is done by using Method", "we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains ."], "relation": "used for", "id": "2022.findings-acl.264", "year": 2022, "rel_sent": "A key contribution is the combination of semi - automatic resource building for extraction of domain - dependent concern types ( with 2 - 4 hours of human labor per domain ) and an entirely automatic procedure for extraction of domain - independent moral dimensions and endorsement values .", "forward": false, "src_ids": "2022.findings-acl.264_565"}
+{"input": "semi - automatic resource building is used for Task| context: we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains .", "entity": "semi - automatic resource building", "output": "extraction of domain - dependent concern types", "neg_sample": ["semi - automatic resource building is used for Task", "we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains ."], "relation": "used for", "id": "2022.findings-acl.264", "year": 2022, "rel_sent": "A key contribution is the combination of semi - automatic resource building for extraction of domain - dependent concern types ( with 2 - 4 hours of human labor per domain ) and an entirely automatic procedure for extraction of domain - independent moral dimensions and endorsement values .", "forward": true, "src_ids": "2022.findings-acl.264_566"}
+{"input": "lexical expansion is done by using OtherScientificTerm| context: we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains .", "entity": "lexical expansion", "output": "propositional structures", "neg_sample": ["lexical expansion is done by using OtherScientificTerm", "we present a generalized paradigm for adaptation of propositional analysis ( predicate - argument pairs ) to new tasks and domains ."], "relation": "used for", "id": "2022.findings-acl.264", "year": 2022, "rel_sent": "Prudent ( automatic ) selection of terms from propositional structures for lexical expansion ( via semantic similarity ) produces new moral dimension lexicons at three levels of granularity beyond a strong baseline lexicon .", "forward": false, "src_ids": "2022.findings-acl.264_567"}
+{"input": "augmentative and alternative communication is done by using Method| context: many people with severely limited muscle control can only communicate through augmentative and alternative communication ( aac ) systems with a small number of buttons .", "entity": "augmentative and alternative communication", "output": "bayesian inference", "neg_sample": ["augmentative and alternative communication is done by using Method", "many people with severely limited muscle control can only communicate through augmentative and alternative communication ( aac ) systems with a small number of buttons ."], "relation": "used for", "id": "2022.slpat-1.2", "year": 2022, "rel_sent": "ColorCode : A Bayesian Approach to Augmentative and Alternative Communication with Two Buttons.", "forward": false, "src_ids": "2022.slpat-1.2_568"}
+{"input": "coherent and incoherent dialogues is done by using Method| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "coherent and incoherent dialogues", "output": "deam", "neg_sample": ["coherent and incoherent dialogues is done by using Method", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "We also show that DEAM can distinguish between coherent and incoherent dialogues generated by baseline manipulations , whereas those baseline models can not detect incoherent examples generated by DEAM .", "forward": false, "src_ids": "2022.acl-long.57_569"}
+{"input": "natural negative example generation is done by using Method| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "natural negative example generation", "output": "amr - based semantic manipulations", "neg_sample": ["natural negative example generation is done by using Method", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "Our results demonstrate the potential of AMR - based semantic manipulations for natural negative example generation .", "forward": false, "src_ids": "2022.acl-long.57_570"}
+{"input": "amr - based semantic manipulations is used for Task| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "amr - based semantic manipulations", "output": "natural negative example generation", "neg_sample": ["amr - based semantic manipulations is used for Task", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "Our results demonstrate the potential of AMR - based semantic manipulations for natural negative example generation .", "forward": true, "src_ids": "2022.acl-long.57_571"}
+{"input": "incoherent ( negative ) data generation is done by using Method| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "incoherent ( negative ) data generation", "output": "semantic - level manipulations", "neg_sample": ["incoherent ( negative ) data generation is done by using Method", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "To tackle this problem , we propose DEAM , a Dialogue coherence Evaluation metric that relies on Abstract Meaning Representation ( AMR ) to apply semantic - level Manipulations for incoherent ( negative ) data generation .", "forward": false, "src_ids": "2022.acl-long.57_572"}
+{"input": "semantic - level manipulations is used for Task| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "semantic - level manipulations", "output": "incoherent ( negative ) data generation", "neg_sample": ["semantic - level manipulations is used for Task", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "To tackle this problem , we propose DEAM , a Dialogue coherence Evaluation metric that relies on Abstract Meaning Representation ( AMR ) to apply semantic - level Manipulations for incoherent ( negative ) data generation .", "forward": true, "src_ids": "2022.acl-long.57_573"}
+{"input": "incoherence sources is done by using Method| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "incoherence sources", "output": "amrs", "neg_sample": ["incoherence sources is done by using Method", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "AMRs naturally facilitate the injection of various types of incoherence sources , such as coreference inconsistency , irrelevancy , contradictions , and decrease engagement , at the semantic level , thus resulting in more natural incoherent samples .", "forward": false, "src_ids": "2022.acl-long.57_574"}
+{"input": "amrs is used for OtherScientificTerm| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "amrs", "output": "incoherence sources", "neg_sample": ["amrs is used for OtherScientificTerm", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "AMRs naturally facilitate the injection of various types of incoherence sources , such as coreference inconsistency , irrelevancy , contradictions , and decrease engagement , at the semantic level , thus resulting in more natural incoherent samples .", "forward": true, "src_ids": "2022.acl-long.57_575"}
+{"input": "deam is used for OtherScientificTerm| context: automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models . although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data . prior works mainly resort to heuristic text - level manipulations ( e.g. utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) . such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans .", "entity": "deam", "output": "coherent and incoherent dialogues", "neg_sample": ["deam is used for OtherScientificTerm", "automatic evaluation metrics are essential for the rapid development of open - domain dialogue systems as they facilitate hyper - parameter tuning and comparison between models .", "although recently proposed trainable conversation - level metrics have shown encouraging results , the quality of the metrics is strongly dependent on the quality of training data .", "prior works mainly resort to heuristic text - level manipulations ( e.g.", "utterances shuffling ) to bootstrap incoherent conversations ( negative examples ) from coherent dialogues ( positive examples ) .", "such approaches are insufficient to appropriately reflect the incoherence that occurs in interactions between advanced dialogue models and humans ."], "relation": "used for", "id": "2022.acl-long.57", "year": 2022, "rel_sent": "We also show that DEAM can distinguish between coherent and incoherent dialogues generated by baseline manipulations , whereas those baseline models can not detect incoherent examples generated by DEAM .", "forward": true, "src_ids": "2022.acl-long.57_576"}
+{"input": "knowledge - enhanced language models is done by using Method| context: popular language models ( lms ) struggle to capture knowledge about rare tail facts and entities . since widely used systems such as search and personal - assistants must support the long tail of entities that users ask about , there has been significant effort towards enhancing these base lms with factual knowledge . we observe proposed methods typically start with a base lm and data that has been annotated with entity metadata , then change the model , by modifying the architecture or introducing auxiliary loss terms to better capture entity knowledge .", "entity": "knowledge - enhanced language models", "output": "metadata shaping", "neg_sample": ["knowledge - enhanced language models is done by using Method", "popular language models ( lms ) struggle to capture knowledge about rare tail facts and entities .", "since widely used systems such as search and personal - assistants must support the long tail of entities that users ask about , there has been significant effort towards enhancing these base lms with factual knowledge .", "we observe proposed methods typically start with a base lm and data that has been annotated with entity metadata , then change the model , by modifying the architecture or introducing auxiliary loss terms to better capture entity knowledge ."], "relation": "used for", "id": "2022.findings-acl.137", "year": 2022, "rel_sent": "Metadata Shaping : A Simple Approach for Knowledge - Enhanced Language Models.", "forward": false, "src_ids": "2022.findings-acl.137_577"}
+{"input": "metadata shaping is used for Task| context: popular language models ( lms ) struggle to capture knowledge about rare tail facts and entities . since widely used systems such as search and personal - assistants must support the long tail of entities that users ask about , there has been significant effort towards enhancing these base lms with factual knowledge . we observe proposed methods typically start with a base lm and data that has been annotated with entity metadata , then change the model , by modifying the architecture or introducing auxiliary loss terms to better capture entity knowledge .", "entity": "metadata shaping", "output": "knowledge - enhanced language models", "neg_sample": ["metadata shaping is used for Task", "popular language models ( lms ) struggle to capture knowledge about rare tail facts and entities .", "since widely used systems such as search and personal - assistants must support the long tail of entities that users ask about , there has been significant effort towards enhancing these base lms with factual knowledge .", "we observe proposed methods typically start with a base lm and data that has been annotated with entity metadata , then change the model , by modifying the architecture or introducing auxiliary loss terms to better capture entity knowledge ."], "relation": "used for", "id": "2022.findings-acl.137", "year": 2022, "rel_sent": "Metadata Shaping : A Simple Approach for Knowledge - Enhanced Language Models.", "forward": true, "src_ids": "2022.findings-acl.137_578"}
+{"input": "nmt models is used for OtherScientificTerm| context: transformer - based models are the modern work horses for neural machine translation ( nmt ) , reaching state of the art across several benchmarks . despite their impressive accuracy , we observe a systemic and rudimentary class of errors made by current state - of - the - art nmt models with regards to translating from a language that does n't mark gender on nouns into others that do . we find that even when the surrounding context provides unambiguous evidence of the appropriate grammatical gender marking , no tested model was able to accurately gender occupation nouns systematically .", "entity": "nmt models", "output": "gender morphology", "neg_sample": ["nmt models is used for OtherScientificTerm", "transformer - based models are the modern work horses for neural machine translation ( nmt ) , reaching state of the art across several benchmarks .", "despite their impressive accuracy , we observe a systemic and rudimentary class of errors made by current state - of - the - art nmt models with regards to translating from a language that does n't mark gender on nouns into others that do .", "we find that even when the surrounding context provides unambiguous evidence of the appropriate grammatical gender marking , no tested model was able to accurately gender occupation nouns systematically ."], "relation": "used for", "id": "2022.acl-long.243", "year": 2022, "rel_sent": "We release an evaluation scheme and dataset for measuring the ability of NMT models to translate gender morphology correctly in unambiguous contexts across syntactically diverse sentences .", "forward": true, "src_ids": "2022.acl-long.243_579"}
+{"input": "gender morphology is done by using Method| context: transformer - based models are the modern work horses for neural machine translation ( nmt ) , reaching state of the art across several benchmarks . we find that even when the surrounding context provides unambiguous evidence of the appropriate grammatical gender marking , no tested model was able to accurately gender occupation nouns systematically .", "entity": "gender morphology", "output": "nmt models", "neg_sample": ["gender morphology is done by using Method", "transformer - based models are the modern work horses for neural machine translation ( nmt ) , reaching state of the art across several benchmarks .", "we find that even when the surrounding context provides unambiguous evidence of the appropriate grammatical gender marking , no tested model was able to accurately gender occupation nouns systematically ."], "relation": "used for", "id": "2022.acl-long.243", "year": 2022, "rel_sent": "We release an evaluation scheme and dataset for measuring the ability of NMT models to translate gender morphology correctly in unambiguous contexts across syntactically diverse sentences .", "forward": false, "src_ids": "2022.acl-long.243_580"}
+{"input": "language understanding is done by using Method| context: transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks .", "entity": "language understanding", "output": "cross - modal systems", "neg_sample": ["language understanding is done by using Method", "transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks ."], "relation": "used for", "id": "2022.acl-short.52", "year": 2022, "rel_sent": "XDBERT : Distilling Visual Information to BERT from Cross - Modal Systems to Improve Language Understanding.", "forward": false, "src_ids": "2022.acl-short.52_581"}
+{"input": "visual - language tasks is done by using Method| context: transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks .", "entity": "visual - language tasks", "output": "cross - modal encoders", "neg_sample": ["visual - language tasks is done by using Method", "transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks ."], "relation": "used for", "id": "2022.acl-short.52", "year": 2022, "rel_sent": "Our framework is inspired by cross - modal encoders ' success in visual - language tasks while we alter the learning objective to cater to the language - heavy characteristics of NLU .", "forward": false, "src_ids": "2022.acl-short.52_582"}
+{"input": "pairwise sentence classification is done by using Method| context: identifying the relation between two sentences requires datasets with pairwise annotations . in many cases , these datasets contain instances that are annotated multiple times as part of different pairs . they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations .", "entity": "pairwise sentence classification", "output": "structure - aware batches", "neg_sample": ["pairwise sentence classification is done by using Method", "identifying the relation between two sentences requires datasets with pairwise annotations .", "in many cases , these datasets contain instances that are annotated multiple times as part of different pairs .", "they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations ."], "relation": "used for", "id": "2022.findings-acl.239", "year": 2022, "rel_sent": "Composing Structure - Aware Batches for Pairwise Sentence Classification.", "forward": false, "src_ids": "2022.findings-acl.239_583"}
+{"input": "structure - aware batches is used for Task| context: identifying the relation between two sentences requires datasets with pairwise annotations . in many cases , these datasets contain instances that are annotated multiple times as part of different pairs . they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations .", "entity": "structure - aware batches", "output": "pairwise sentence classification", "neg_sample": ["structure - aware batches is used for Task", "identifying the relation between two sentences requires datasets with pairwise annotations .", "in many cases , these datasets contain instances that are annotated multiple times as part of different pairs .", "they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations ."], "relation": "used for", "id": "2022.findings-acl.239", "year": 2022, "rel_sent": "Composing Structure - Aware Batches for Pairwise Sentence Classification.", "forward": true, "src_ids": "2022.findings-acl.239_584"}
+{"input": "training is done by using OtherScientificTerm| context: identifying the relation between two sentences requires datasets with pairwise annotations . in many cases , these datasets contain instances that are annotated multiple times as part of different pairs . they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations .", "entity": "training", "output": "structural dataset information", "neg_sample": ["training is done by using OtherScientificTerm", "identifying the relation between two sentences requires datasets with pairwise annotations .", "in many cases , these datasets contain instances that are annotated multiple times as part of different pairs .", "they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations ."], "relation": "used for", "id": "2022.findings-acl.239", "year": 2022, "rel_sent": "This paper investigates how this kind of structural dataset information can be exploited during training . We propose three batch composition strategies to incorporate such information and measure their performance over 14 heterogeneous pairwise sentence classification tasks .", "forward": false, "src_ids": "2022.findings-acl.239_585"}
+{"input": "structural dataset information is used for Task| context: identifying the relation between two sentences requires datasets with pairwise annotations . in many cases , these datasets contain instances that are annotated multiple times as part of different pairs . they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations .", "entity": "structural dataset information", "output": "training", "neg_sample": ["structural dataset information is used for Task", "identifying the relation between two sentences requires datasets with pairwise annotations .", "in many cases , these datasets contain instances that are annotated multiple times as part of different pairs .", "they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations ."], "relation": "used for", "id": "2022.findings-acl.239", "year": 2022, "rel_sent": "This paper investigates how this kind of structural dataset information can be exploited during training . We propose three batch composition strategies to incorporate such information and measure their performance over 14 heterogeneous pairwise sentence classification tasks .", "forward": true, "src_ids": "2022.findings-acl.239_586"}
+{"input": "glitter is used for OtherScientificTerm| context: data augmentation ( da ) is known to improve the generalizability of deep neural networks . most existing da techniques naively add a certain number of augmented samples without considering the quality and the added computational cost of these samples .", "entity": "glitter", "output": "worst - case samples", "neg_sample": ["glitter is used for OtherScientificTerm", "data augmentation ( da ) is known to improve the generalizability of deep neural networks .", "most existing da techniques naively add a certain number of augmented samples without considering the quality and the added computational cost of these samples ."], "relation": "used for", "id": "2022.findings-acl.84", "year": 2022, "rel_sent": "From a pre - generated pool of augmented samples , Glitter adaptively selects a subset of worst - case samples with maximal loss , analogous to adversarial DA .", "forward": true, "src_ids": "2022.findings-acl.84_587"}
+{"input": "worst - case samples is done by using Method| context: data augmentation ( da ) is known to improve the generalizability of deep neural networks . most existing da techniques naively add a certain number of augmented samples without considering the quality and the added computational cost of these samples .", "entity": "worst - case samples", "output": "glitter", "neg_sample": ["worst - case samples is done by using Method", "data augmentation ( da ) is known to improve the generalizability of deep neural networks .", "most existing da techniques naively add a certain number of augmented samples without considering the quality and the added computational cost of these samples ."], "relation": "used for", "id": "2022.findings-acl.84", "year": 2022, "rel_sent": "From a pre - generated pool of augmented samples , Glitter adaptively selects a subset of worst - case samples with maximal loss , analogous to adversarial DA .", "forward": false, "src_ids": "2022.findings-acl.84_588"}
+{"input": "medical domain is done by using Task| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "medical domain", "output": "incremental intent detection", "neg_sample": ["medical domain is done by using Task", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "Incremental Intent Detection for Medical Domain with Contrast Replay Networks.", "forward": false, "src_ids": "2022.findings-acl.280_589"}
+{"input": "incremental intent detection is used for Material| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "incremental intent detection", "output": "medical domain", "neg_sample": ["incremental intent detection is used for Material", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "Incremental Intent Detection for Medical Domain with Contrast Replay Networks.", "forward": true, "src_ids": "2022.findings-acl.280_590"}
+{"input": "incremental learning is used for Task| context: however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "incremental learning", "output": "medical intent detection", "neg_sample": ["incremental learning is used for Task", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "We first formulate incremental learning for medical intent detection .", "forward": true, "src_ids": "2022.findings-acl.280_591"}
+{"input": "medical intent detection is done by using Method| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "medical intent detection", "output": "incremental learning", "neg_sample": ["medical intent detection is done by using Method", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "We first formulate incremental learning for medical intent detection .", "forward": false, "src_ids": "2022.findings-acl.280_592"}
+{"input": "memory - based method is used for Method| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "memory - based method", "output": "incremental learning", "neg_sample": ["memory - based method is used for Method", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "Then , we employ a memory - based method to handle incremental learning .", "forward": true, "src_ids": "2022.findings-acl.280_593"}
+{"input": "incremental learning is done by using Method| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "incremental learning", "output": "memory - based method", "neg_sample": ["incremental learning is done by using Method", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "Then , we employ a memory - based method to handle incremental learning .", "forward": false, "src_ids": "2022.findings-acl.280_594"}
+{"input": "training data imbalance is done by using OtherScientificTerm| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "training data imbalance", "output": "multilevel distillation", "neg_sample": ["training data imbalance is done by using OtherScientificTerm", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "We further propose to enhance the method with contrast replay networks , which use multilevel distillation and contrast objective to address training data imbalance and medical rare words respectively .", "forward": false, "src_ids": "2022.findings-acl.280_595"}
+{"input": "medical rare words is done by using OtherScientificTerm| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "medical rare words", "output": "contrast objective", "neg_sample": ["medical rare words is done by using OtherScientificTerm", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "We further propose to enhance the method with contrast replay networks , which use multilevel distillation and contrast objective to address training data imbalance and medical rare words respectively .", "forward": false, "src_ids": "2022.findings-acl.280_596"}
+{"input": "contrast objective is used for OtherScientificTerm| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "contrast objective", "output": "training data imbalance", "neg_sample": ["contrast objective is used for OtherScientificTerm", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "We further propose to enhance the method with contrast replay networks , which use multilevel distillation and contrast objective to address training data imbalance and medical rare words respectively .", "forward": true, "src_ids": "2022.findings-acl.280_597"}
+{"input": "multilevel distillation is used for OtherScientificTerm| context: conventional approaches to medical intent detection require fixed pre - defined intent categories . however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical .", "entity": "multilevel distillation", "output": "training data imbalance", "neg_sample": ["multilevel distillation is used for OtherScientificTerm", "conventional approaches to medical intent detection require fixed pre - defined intent categories .", "however , due to the incessant emergence of new medical intents in the real world , such requirement is not practical ."], "relation": "used for", "id": "2022.findings-acl.280", "year": 2022, "rel_sent": "We further propose to enhance the method with contrast replay networks , which use multilevel distillation and contrast objective to address training data imbalance and medical rare words respectively .", "forward": true, "src_ids": "2022.findings-acl.280_598"}
+{"input": "labeled and directed dependency parse trees is done by using Method| context: probing has become an important tool for analyzing representations in natural language processing ( nlp ) .", "entity": "labeled and directed dependency parse trees", "output": "linear probes", "neg_sample": ["labeled and directed dependency parse trees is done by using Method", "probing has become an important tool for analyzing representations in natural language processing ( nlp ) ."], "relation": "used for", "id": "2022.acl-long.532", "year": 2022, "rel_sent": "This work introduces DepProbe , a linear probe which can extract labeled and directed dependency parse trees from embeddings while using fewer parameters and compute than prior methods .", "forward": false, "src_ids": "2022.acl-long.532_599"}
+{"input": "linear probes is used for OtherScientificTerm| context: probing has become an important tool for analyzing representations in natural language processing ( nlp ) . for graphical nlp tasks such as dependency parsing , linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task .", "entity": "linear probes", "output": "labeled and directed dependency parse trees", "neg_sample": ["linear probes is used for OtherScientificTerm", "probing has become an important tool for analyzing representations in natural language processing ( nlp ) .", "for graphical nlp tasks such as dependency parsing , linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task ."], "relation": "used for", "id": "2022.acl-long.532", "year": 2022, "rel_sent": "This work introduces DepProbe , a linear probe which can extract labeled and directed dependency parse trees from embeddings while using fewer parameters and compute than prior methods .", "forward": true, "src_ids": "2022.acl-long.532_600"}
+{"input": "full biaffine attention parser is done by using Material| context: probing has become an important tool for analyzing representations in natural language processing ( nlp ) . for graphical nlp tasks such as dependency parsing , linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task .", "entity": "full biaffine attention parser", "output": "transfer language", "neg_sample": ["full biaffine attention parser is done by using Material", "probing has become an important tool for analyzing representations in natural language processing ( nlp ) .", "for graphical nlp tasks such as dependency parsing , linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task ."], "relation": "used for", "id": "2022.acl-long.532", "year": 2022, "rel_sent": "Leveraging its full task coverage and lightweight parametrization , we investigate its predictive power for selecting the best transfer language for training a full biaffine attention parser .", "forward": false, "src_ids": "2022.acl-long.532_601"}
+{"input": "transfer language is used for Method| context: probing has become an important tool for analyzing representations in natural language processing ( nlp ) . for graphical nlp tasks such as dependency parsing , linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task .", "entity": "transfer language", "output": "full biaffine attention parser", "neg_sample": ["transfer language is used for Method", "probing has become an important tool for analyzing representations in natural language processing ( nlp ) .", "for graphical nlp tasks such as dependency parsing , linear probes are currently limited to extracting undirected or unlabeled parse trees which do not capture the full task ."], "relation": "used for", "id": "2022.acl-long.532", "year": 2022, "rel_sent": "Leveraging its full task coverage and lightweight parametrization , we investigate its predictive power for selecting the best transfer language for training a full biaffine attention parser .", "forward": true, "src_ids": "2022.acl-long.532_602"}
+{"input": "dialogue state tracking is done by using Method| context: the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) . however , substantial noise has been discovered in its state annotations . such noise brings about huge challenges for training dst models robustly . although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set . besides , it is costly to rectify all the problematic annotations .", "entity": "dialogue state tracking", "output": "assist", "neg_sample": ["dialogue state tracking is done by using Method", "the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) .", "however , substantial noise has been discovered in its state annotations .", "such noise brings about huge challenges for training dst models robustly .", "although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set .", "besides , it is costly to rectify all the problematic annotations ."], "relation": "used for", "id": "2022.findings-acl.214", "year": 2022, "rel_sent": "Experimental results also demonstrate that ASSIST improves the joint goal accuracy of DST by up to 28.16 % on MultiWOZ 2.0 and 8.41 % on MultiWOZ 2.4 , compared to using only the vanilla noisy labels .", "forward": false, "src_ids": "2022.findings-acl.214_603"}
+{"input": "primary model is done by using OtherScientificTerm| context: the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) . however , substantial noise has been discovered in its state annotations . such noise brings about huge challenges for training dst models robustly . although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set . besides , it is costly to rectify all the problematic annotations .", "entity": "primary model", "output": "pseudo labels", "neg_sample": ["primary model is done by using OtherScientificTerm", "the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) .", "however , substantial noise has been discovered in its state annotations .", "such noise brings about huge challenges for training dst models robustly .", "although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set .", "besides , it is costly to rectify all the problematic annotations ."], "relation": "used for", "id": "2022.findings-acl.214", "year": 2022, "rel_sent": "ASSIST first generates pseudo labels for each sample in the training set by using an auxiliary model trained on a small clean dataset , then puts the generated pseudo labels and vanilla noisy labels together to train the primary model .", "forward": false, "src_ids": "2022.findings-acl.214_604"}
+{"input": "vanilla noisy labels is used for Generic| context: the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) . however , substantial noise has been discovered in its state annotations . such noise brings about huge challenges for training dst models robustly . although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set . besides , it is costly to rectify all the problematic annotations .", "entity": "vanilla noisy labels", "output": "primary model", "neg_sample": ["vanilla noisy labels is used for Generic", "the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) .", "however , substantial noise has been discovered in its state annotations .", "such noise brings about huge challenges for training dst models robustly .", "although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set .", "besides , it is costly to rectify all the problematic annotations ."], "relation": "used for", "id": "2022.findings-acl.214", "year": 2022, "rel_sent": "ASSIST first generates pseudo labels for each sample in the training set by using an auxiliary model trained on a small clean dataset , then puts the generated pseudo labels and vanilla noisy labels together to train the primary model .", "forward": true, "src_ids": "2022.findings-acl.214_605"}
+{"input": "pseudo labels is used for Generic| context: the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) . however , substantial noise has been discovered in its state annotations . such noise brings about huge challenges for training dst models robustly . although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set . besides , it is costly to rectify all the problematic annotations .", "entity": "pseudo labels", "output": "primary model", "neg_sample": ["pseudo labels is used for Generic", "the multiwoz 2.0 dataset has greatly boosted the research on dialogue state tracking ( dst ) .", "however , substantial noise has been discovered in its state annotations .", "such noise brings about huge challenges for training dst models robustly .", "although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set .", "besides , it is costly to rectify all the problematic annotations ."], "relation": "used for", "id": "2022.findings-acl.214", "year": 2022, "rel_sent": "ASSIST first generates pseudo labels for each sample in the training set by using an auxiliary model trained on a small clean dataset , then puts the generated pseudo labels and vanilla noisy labels together to train the primary model .", "forward": true, "src_ids": "2022.findings-acl.214_606"}
+{"input": "assist is used for Task| context: however , substantial noise has been discovered in its state annotations . such noise brings about huge challenges for training dst models robustly . although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set . besides , it is costly to rectify all the problematic annotations .", "entity": "assist", "output": "dialogue state tracking", "neg_sample": ["assist is used for Task", "however , substantial noise has been discovered in its state annotations .", "such noise brings about huge challenges for training dst models robustly .", "although several refined versions , including multiwoz 2.1 - 2.4 , have been published recently , there are still lots of noisy labels , especially in the training set .", "besides , it is costly to rectify all the problematic annotations ."], "relation": "used for", "id": "2022.findings-acl.214", "year": 2022, "rel_sent": "Experimental results also demonstrate that ASSIST improves the joint goal accuracy of DST by up to 28.16 % on MultiWOZ 2.0 and 8.41 % on MultiWOZ 2.4 , compared to using only the vanilla noisy labels .", "forward": true, "src_ids": "2022.findings-acl.214_607"}
+{"input": "hierarchy - aware text representation is done by using Method| context: hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy . existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "hierarchy - aware text representation", "output": "text encoder", "neg_sample": ["hierarchy - aware text representation is done by using Method", "hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy .", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "By pulling together the input text and its positive sample , the text encoder can learn to generate the hierarchy - aware text representation independently .", "forward": false, "src_ids": "2022.acl-long.491_608"}
+{"input": "hierarchy - guided contrastive learning ( hgclr ) is used for Method| context: hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy . existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "hierarchy - guided contrastive learning ( hgclr )", "output": "text encoder", "neg_sample": ["hierarchy - guided contrastive learning ( hgclr ) is used for Method", "hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy .", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "Instead of modeling them separately , in this work , we propose Hierarchy - guided Contrastive Learning ( HGCLR ) to directly embed the hierarchy into a text encoder .", "forward": true, "src_ids": "2022.acl-long.491_609"}
+{"input": "hierarchical text classification is done by using Method| context: hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy . existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "hierarchical text classification", "output": "contrastive learning approach", "neg_sample": ["hierarchical text classification is done by using Method", "hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy .", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "Incorporating Hierarchy into Text Encoder : a Contrastive Learning Approach for Hierarchical Text Classification.", "forward": false, "src_ids": "2022.acl-long.491_610"}
+{"input": "contrastive learning approach is used for Task| context: existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "contrastive learning approach", "output": "hierarchical text classification", "neg_sample": ["contrastive learning approach is used for Task", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "Incorporating Hierarchy into Text Encoder : a Contrastive Learning Approach for Hierarchical Text Classification.", "forward": true, "src_ids": "2022.acl-long.491_611"}
+{"input": "text encoder is done by using Method| context: hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy . existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "text encoder", "output": "hierarchy - guided contrastive learning ( hgclr )", "neg_sample": ["text encoder is done by using Method", "hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy .", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "Instead of modeling them separately , in this work , we propose Hierarchy - guided Contrastive Learning ( HGCLR ) to directly embed the hierarchy into a text encoder .", "forward": false, "src_ids": "2022.acl-long.491_612"}
+{"input": "hierarchy is done by using Method| context: hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy . existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "hierarchy", "output": "hierarchy - guided contrastive learning ( hgclr )", "neg_sample": ["hierarchy is done by using Method", "hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy .", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "Instead of modeling them separately , in this work , we propose Hierarchy - guided Contrastive Learning ( HGCLR ) to directly embed the hierarchy into a text encoder .", "forward": false, "src_ids": "2022.acl-long.491_613"}
+{"input": "text encoder is used for Method| context: hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy . existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text .", "entity": "text encoder", "output": "hierarchy - aware text representation", "neg_sample": ["text encoder is used for Method", "hierarchical text classification is a challenging subtask of multi - label classification due to its complex label hierarchy .", "existing methods encode text and label hierarchy separately and mix their representations for classification , where the hierarchy remains unchanged for all input text ."], "relation": "used for", "id": "2022.acl-long.491", "year": 2022, "rel_sent": "By pulling together the input text and its positive sample , the text encoder can learn to generate the hierarchy - aware text representation independently .", "forward": true, "src_ids": "2022.acl-long.491_614"}
+{"input": "number agreement task is done by using Task| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "entity": "number agreement task", "output": "encoding", "neg_sample": ["number agreement task is done by using Task", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "As a case study , we focus on how BERT encodes grammatical number , and on how it uses this encoding to solve the number agreement task .", "forward": false, "src_ids": "2022.acl-long.603_615"}
+{"input": "grammatical number is done by using Method| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "grammatical number", "output": "bert", "neg_sample": ["grammatical number is done by using Method", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "As a case study , we focus on how BERT encodes grammatical number , and on how it uses this encoding to solve the number agreement task .", "forward": false, "src_ids": "2022.acl-long.603_616"}
+{"input": "bert is used for OtherScientificTerm| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "bert", "output": "grammatical number", "neg_sample": ["bert is used for OtherScientificTerm", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "As a case study , we focus on how BERT encodes grammatical number , and on how it uses this encoding to solve the number agreement task .", "forward": true, "src_ids": "2022.acl-long.603_617"}
+{"input": "encoding is used for Task| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "encoding", "output": "number agreement task", "neg_sample": ["encoding is used for Task", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "As a case study , we focus on how BERT encodes grammatical number , and on how it uses this encoding to solve the number agreement task .", "forward": true, "src_ids": "2022.acl-long.603_618"}
+{"input": "behavioral output is done by using Method| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "behavioral output", "output": "linear encoding of grammatical number", "neg_sample": ["behavioral output is done by using Method", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "Experimentally , we find that BERT relies on a linear encoding of grammatical number to produce the correct behavioral output .", "forward": false, "src_ids": "2022.acl-long.603_619"}
+{"input": "linear encoding of grammatical number is used for OtherScientificTerm| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "linear encoding of grammatical number", "output": "behavioral output", "neg_sample": ["linear encoding of grammatical number is used for OtherScientificTerm", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "Experimentally , we find that BERT relies on a linear encoding of grammatical number to produce the correct behavioral output .", "forward": true, "src_ids": "2022.acl-long.603_620"}
+{"input": "nouns is done by using OtherScientificTerm| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "nouns", "output": "encoding of grammatical number", "neg_sample": ["nouns is done by using OtherScientificTerm", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "We alsofind that BERT uses a separate encoding of grammatical number for nouns and verbs .", "forward": false, "src_ids": "2022.acl-long.603_621"}
+{"input": "verbs is done by using OtherScientificTerm| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "verbs", "output": "encoding of grammatical number", "neg_sample": ["verbs is done by using OtherScientificTerm", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "We alsofind that BERT uses a separate encoding of grammatical number for nouns and verbs .", "forward": false, "src_ids": "2022.acl-long.603_622"}
+{"input": "encoding of grammatical number is used for OtherScientificTerm| context: a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations . an encoding , however , might be spurious - i.e. , the model might not rely on it when making predictions .", "entity": "encoding of grammatical number", "output": "nouns", "neg_sample": ["encoding of grammatical number is used for OtherScientificTerm", "a central quest of probing is to uncover how pre - trained models encode a linguistic property within their representations .", "an encoding , however , might be spurious - i.e.", ", the model might not rely on it when making predictions ."], "relation": "used for", "id": "2022.acl-long.603", "year": 2022, "rel_sent": "We alsofind that BERT uses a separate encoding of grammatical number for nouns and verbs .", "forward": true, "src_ids": "2022.acl-long.603_623"}
+{"input": "document - level neural machine translation is done by using Task| context: document - level neural machine translation ( docnmt ) achieves coherent translations by incorporating cross - sentence context . however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available .", "entity": "document - level neural machine translation", "output": "multilingual transfer", "neg_sample": ["document - level neural machine translation is done by using Task", "document - level neural machine translation ( docnmt ) achieves coherent translations by incorporating cross - sentence context .", "however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available ."], "relation": "used for", "id": "2022.acl-long.287", "year": 2022, "rel_sent": "Our experiments on Europarl-7 and IWSLT-10 show the feasibility of multilingual transfer for DocNMT , particularly on document - specific metrics .", "forward": false, "src_ids": "2022.acl-long.287_624"}
+{"input": "document - level neural machine translation is done by using Method| context: document - level neural machine translation ( docnmt ) achieves coherent translations by incorporating cross - sentence context . however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available .", "entity": "document - level neural machine translation", "output": "contextual modeling", "neg_sample": ["document - level neural machine translation is done by using Method", "document - level neural machine translation ( docnmt ) achieves coherent translations by incorporating cross - sentence context .", "however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available ."], "relation": "used for", "id": "2022.acl-long.287", "year": 2022, "rel_sent": "In this paper , we study whether and how contextual modeling in DocNMT is transferable via multilingual modeling .", "forward": false, "src_ids": "2022.acl-long.287_625"}
+{"input": "contextual modeling is used for Task| context: however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available .", "entity": "contextual modeling", "output": "document - level neural machine translation", "neg_sample": ["contextual modeling is used for Task", "however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available ."], "relation": "used for", "id": "2022.acl-long.287", "year": 2022, "rel_sent": "In this paper , we study whether and how contextual modeling in DocNMT is transferable via multilingual modeling .", "forward": true, "src_ids": "2022.acl-long.287_626"}
+{"input": "multilingual transfer is used for Task| context: however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available .", "entity": "multilingual transfer", "output": "document - level neural machine translation", "neg_sample": ["multilingual transfer is used for Task", "however , for most language pairs there 's a shortage of parallel documents , although parallel sentences are readily available ."], "relation": "used for", "id": "2022.acl-long.287", "year": 2022, "rel_sent": "Our experiments on Europarl-7 and IWSLT-10 show the feasibility of multilingual transfer for DocNMT , particularly on document - specific metrics .", "forward": true, "src_ids": "2022.acl-long.287_627"}
+{"input": "minimalist grammars is done by using Method| context: one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles .", "entity": "minimalist grammars", "output": "parser", "neg_sample": ["minimalist grammars is done by using Method", "one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles ."], "relation": "used for", "id": "2022.scil-1.5", "year": 2022, "rel_sent": "Here , we evaluate this hypothesis by testing the predictions of a parser for Minimalist grammars for PR and RC structures in Italian .", "forward": false, "src_ids": "2022.scil-1.5_628"}
+{"input": "pr and rc structures is done by using Method| context: grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference . one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles .", "entity": "pr and rc structures", "output": "minimalist grammars", "neg_sample": ["pr and rc structures is done by using Method", "grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference .", "one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles ."], "relation": "used for", "id": "2022.scil-1.5", "year": 2022, "rel_sent": "Here , we evaluate this hypothesis by testing the predictions of a parser for Minimalist grammars for PR and RC structures in Italian .", "forward": false, "src_ids": "2022.scil-1.5_629"}
+{"input": "parser is used for Method| context: grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference . one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles .", "entity": "parser", "output": "minimalist grammars", "neg_sample": ["parser is used for Method", "grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference .", "one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles ."], "relation": "used for", "id": "2022.scil-1.5", "year": 2022, "rel_sent": "Here , we evaluate this hypothesis by testing the predictions of a parser for Minimalist grammars for PR and RC structures in Italian .", "forward": true, "src_ids": "2022.scil-1.5_630"}
+{"input": "minimalist grammars is used for OtherScientificTerm| context: grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference . one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles .", "entity": "minimalist grammars", "output": "pr and rc structures", "neg_sample": ["minimalist grammars is used for OtherScientificTerm", "grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference .", "one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles ."], "relation": "used for", "id": "2022.scil-1.5", "year": 2022, "rel_sent": "Here , we evaluate this hypothesis by testing the predictions of a parser for Minimalist grammars for PR and RC structures in Italian .", "forward": true, "src_ids": "2022.scil-1.5_631"}
+{"input": "economy considerations is done by using Method| context: grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference . one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles .", "entity": "economy considerations", "output": "computational models", "neg_sample": ["economy considerations is done by using Method", "grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference .", "one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles ."], "relation": "used for", "id": "2022.scil-1.5", "year": 2022, "rel_sent": "We discuss the relevance of our results for PR - first explanations of the cross - linguistic variability of RC attachment biases , and highlight the role that computational models can play in evaluating the cognitive plausibility of economy considerations tied to fine - grained structural analyses .", "forward": false, "src_ids": "2022.scil-1.5_632"}
+{"input": "computational models is used for Task| context: grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference . one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles .", "entity": "computational models", "output": "economy considerations", "neg_sample": ["computational models is used for Task", "grillo and costa ( 2014 ) argue for a pseudo - relative ( pr ) first account of relative clause attachment preferences ( rc ) such that , when faced with a sentence ambiguous between a pr and a rc interpretation , the parser prefers committing to a pr structure first , thus giving rise to what looks like a high - attachment preference .", "one possible explanation for this parsing choice is in terms of simplicity of the pr structure , and overall economy principles ."], "relation": "used for", "id": "2022.scil-1.5", "year": 2022, "rel_sent": "We discuss the relevance of our results for PR - first explanations of the cross - linguistic variability of RC attachment biases , and highlight the role that computational models can play in evaluating the cognitive plausibility of economy considerations tied to fine - grained structural analyses .", "forward": true, "src_ids": "2022.scil-1.5_633"}
+{"input": "multilingual sememe kb is done by using Task| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "multilingual sememe kb", "output": "sememe prediction", "neg_sample": ["multilingual sememe kb is done by using Task", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "To address this issue , the task of sememe prediction for BabelNet synsets ( SPBS ) is presented , aiming to build a multilingual sememe KB based on BabelNet , a multilingual encyclopedia dictionary .", "forward": false, "src_ids": "2022.findings-acl.15_634"}
+{"input": "babelnet synsets ( spbs ) is done by using Task| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "babelnet synsets ( spbs )", "output": "sememe prediction", "neg_sample": ["babelnet synsets ( spbs ) is done by using Task", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "To address this issue , the task of sememe prediction for BabelNet synsets ( SPBS ) is presented , aiming to build a multilingual sememe KB based on BabelNet , a multilingual encyclopedia dictionary .", "forward": false, "src_ids": "2022.findings-acl.15_635"}
+{"input": "multimodal information fusion model is used for Task| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "multimodal information fusion model", "output": "sememe prediction", "neg_sample": ["multimodal information fusion model is used for Task", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "We design a multimodal information fusion model to encode and combine this information for sememe prediction .", "forward": true, "src_ids": "2022.findings-acl.15_636"}
+{"input": "sememe prediction is used for Material| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "sememe prediction", "output": "babelnet synsets ( spbs )", "neg_sample": ["sememe prediction is used for Material", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "To address this issue , the task of sememe prediction for BabelNet synsets ( SPBS ) is presented , aiming to build a multilingual sememe KB based on BabelNet , a multilingual encyclopedia dictionary .", "forward": true, "src_ids": "2022.findings-acl.15_637"}
+{"input": "sememes is used for Material| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "sememes", "output": "babelnet synsets ( spbs )", "neg_sample": ["sememes is used for Material", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "By automatically predicting sememes for a BabelNet synset , the words in many languages in the synset would obtain sememe annotations simultaneously .", "forward": true, "src_ids": "2022.findings-acl.15_638"}
+{"input": "babelnet synsets ( spbs ) is done by using OtherScientificTerm| context: however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "babelnet synsets ( spbs )", "output": "sememes", "neg_sample": ["babelnet synsets ( spbs ) is done by using OtherScientificTerm", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "By automatically predicting sememes for a BabelNet synset , the words in many languages in the synset would obtain sememe annotations simultaneously .", "forward": false, "src_ids": "2022.findings-acl.15_639"}
+{"input": "sememe prediction is used for Method| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "sememe prediction", "output": "multilingual sememe kb", "neg_sample": ["sememe prediction is used for Method", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "To address this issue , the task of sememe prediction for BabelNet synsets ( SPBS ) is presented , aiming to build a multilingual sememe KB based on BabelNet , a multilingual encyclopedia dictionary .", "forward": true, "src_ids": "2022.findings-acl.15_640"}
+{"input": "sememe prediction for babelnet synsets is done by using OtherScientificTerm| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "sememe prediction for babelnet synsets", "output": "multilingual glosses", "neg_sample": ["sememe prediction for babelnet synsets is done by using OtherScientificTerm", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "In this paper , we utilize the multilingual synonyms , multilingual glosses and images in BabelNet for SPBS .", "forward": false, "src_ids": "2022.findings-acl.15_641"}
+{"input": "sememe prediction for babelnet synsets is done by using Material| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "sememe prediction for babelnet synsets", "output": "images", "neg_sample": ["sememe prediction for babelnet synsets is done by using Material", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "In this paper , we utilize the multilingual synonyms , multilingual glosses and images in BabelNet for SPBS .", "forward": false, "src_ids": "2022.findings-acl.15_642"}
+{"input": "images is used for Task| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "images", "output": "sememe prediction for babelnet synsets", "neg_sample": ["images is used for Task", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "In this paper , we utilize the multilingual synonyms , multilingual glosses and images in BabelNet for SPBS .", "forward": true, "src_ids": "2022.findings-acl.15_643"}
+{"input": "multilingual glosses is used for Task| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "multilingual glosses", "output": "sememe prediction for babelnet synsets", "neg_sample": ["multilingual glosses is used for Task", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "In this paper , we utilize the multilingual synonyms , multilingual glosses and images in BabelNet for SPBS .", "forward": true, "src_ids": "2022.findings-acl.15_644"}
+{"input": "sememe prediction is done by using Method| context: in linguistics , a sememe is defined as the minimum semantic unit of languages . sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks . however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes . however , previous spbs methods have not taken full advantage of the abundant information in babelnet .", "entity": "sememe prediction", "output": "multimodal information fusion model", "neg_sample": ["sememe prediction is done by using Method", "in linguistics , a sememe is defined as the minimum semantic unit of languages .", "sememe knowledge bases ( kbs ) , which are built by manually annotating words with sememes , have been successfully applied to various nlp tasks .", "however , existing sememe kbs only cover a few languages , which hinders the wide utilization of sememes .", "however , previous spbs methods have not taken full advantage of the abundant information in babelnet ."], "relation": "used for", "id": "2022.findings-acl.15", "year": 2022, "rel_sent": "We design a multimodal information fusion model to encode and combine this information for sememe prediction .", "forward": false, "src_ids": "2022.findings-acl.15_645"}
+{"input": "question - answering is done by using Method| context: accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g. , bert ) . these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g. , squad ) . however , qa datasets on such a scale are not available for most of the other languages . multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages . since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages . however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data .", "entity": "question - answering", "output": "multilingual contrastive training", "neg_sample": ["question - answering is done by using Method", "accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g.", ", bert ) .", "these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g.", ", squad ) .", "however , qa datasets on such a scale are not available for most of the other languages .", "multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages .", "since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages .", "however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.3", "year": 2022, "rel_sent": "MuCoT : Multilingual Contrastive Training for Question - Answering in Low - resource Languages.", "forward": false, "src_ids": "2022.dravidianlangtech-1.3_646"}
+{"input": "multilingual contrastive training is used for Task| context: accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g. , bert ) . these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g. , squad ) . however , qa datasets on such a scale are not available for most of the other languages . multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages . since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages . however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data .", "entity": "multilingual contrastive training", "output": "question - answering", "neg_sample": ["multilingual contrastive training is used for Task", "accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g.", ", bert ) .", "these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g.", ", squad ) .", "however , qa datasets on such a scale are not available for most of the other languages .", "multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages .", "since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages .", "however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.3", "year": 2022, "rel_sent": "MuCoT : Multilingual Contrastive Training for Question - Answering in Low - resource Languages.", "forward": true, "src_ids": "2022.dravidianlangtech-1.3_647"}
+{"input": "fine - tuning process is done by using OtherScientificTerm| context: accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g. , bert ) . these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g. , squad ) . however , qa datasets on such a scale are not available for most of the other languages . multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages . since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages . however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data .", "entity": "fine - tuning process", "output": "contrastive loss", "neg_sample": ["fine - tuning process is done by using OtherScientificTerm", "accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g.", ", bert ) .", "these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g.", ", squad ) .", "however , qa datasets on such a scale are not available for most of the other languages .", "multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages .", "since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages .", "however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.3", "year": 2022, "rel_sent": "We further show that introducing a contrastive loss between the translated question - context feature pairs during the fine - tuning process , prevents such degradation with cross - lingual family translations and leads to marginal improvement .", "forward": false, "src_ids": "2022.dravidianlangtech-1.3_648"}
+{"input": "translated question - context feature pairs is used for Method| context: accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g. , bert ) . these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g. , squad ) . however , qa datasets on such a scale are not available for most of the other languages . multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages . since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages . however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data .", "entity": "translated question - context feature pairs", "output": "fine - tuning process", "neg_sample": ["translated question - context feature pairs is used for Method", "accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g.", ", bert ) .", "these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g.", ", squad ) .", "however , qa datasets on such a scale are not available for most of the other languages .", "multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages .", "since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages .", "however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.3", "year": 2022, "rel_sent": "We further show that introducing a contrastive loss between the translated question - context feature pairs during the fine - tuning process , prevents such degradation with cross - lingual family translations and leads to marginal improvement .", "forward": true, "src_ids": "2022.dravidianlangtech-1.3_649"}
+{"input": "contrastive loss is used for Method| context: accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g. , bert ) . these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g. , squad ) . however , qa datasets on such a scale are not available for most of the other languages . multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages . since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages . however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data .", "entity": "contrastive loss", "output": "fine - tuning process", "neg_sample": ["contrastive loss is used for Method", "accuracy of english - language question answering ( qa ) systems has improved significantly in recent years with the advent of transformer - based models ( e.g.", ", bert ) .", "these models are pre - trained in a self - supervised fashion with a large english text corpus and further fine - tuned with a massive english qa dataset ( e.g.", ", squad ) .", "however , qa datasets on such a scale are not available for most of the other languages .", "multi - lingual bert - based models ( mbert ) are often used to transfer knowledge from high - resource languages to low - resource languages .", "since these models are pre - trained with huge text corpora containing multiple languages , they typically learn language - agnostic embeddings for tokens from different languages .", "however , directly training an mbert - based qa system for low - resource languages is challenging due to the paucity of training data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.3", "year": 2022, "rel_sent": "We further show that introducing a contrastive loss between the translated question - context feature pairs during the fine - tuning process , prevents such degradation with cross - lingual family translations and leads to marginal improvement .", "forward": true, "src_ids": "2022.dravidianlangtech-1.3_650"}
+{"input": "pomak is done by using Material| context: as a case study we use greek and pomak , the latter being an endangered oral slavic language of the balkans ( including thrace / greece ) .", "entity": "pomak", "output": "morphologically annotated text corpora", "neg_sample": ["pomak is done by using Material", "as a case study we use greek and pomak , the latter being an endangered oral slavic language of the balkans ( including thrace / greece ) ."], "relation": "used for", "id": "2022.computel-1.22", "year": 2022, "rel_sent": "We describe our experience in the development of a Latin - based orthography and morphologically annotated text corpora of Pomak with state - of - the - art NLP technology .", "forward": false, "src_ids": "2022.computel-1.22_651"}
+{"input": "pomak is done by using Material| context: as a case study we use greek and pomak , the latter being an endangered oral slavic language of the balkans ( including thrace / greece ) .", "entity": "pomak", "output": "gold annotated corpora", "neg_sample": ["pomak is done by using Material", "as a case study we use greek and pomak , the latter being an endangered oral slavic language of the balkans ( including thrace / greece ) ."], "relation": "used for", "id": "2022.computel-1.22", "year": 2022, "rel_sent": "These resources will be made openly available on the XXXX site and the gold annotated corpora of Pomak will be made available on the Universal Dependencies treebank repository .", "forward": false, "src_ids": "2022.computel-1.22_652"}
+{"input": "pomak is done by using Task| context: as a case study we use greek and pomak , the latter being an endangered oral slavic language of the balkans ( including thrace / greece ) .", "entity": "pomak", "output": "linguistic documentation", "neg_sample": ["pomak is done by using Task", "as a case study we use greek and pomak , the latter being an endangered oral slavic language of the balkans ( including thrace / greece ) ."], "relation": "used for", "id": "2022.computel-1.22", "year": 2022, "rel_sent": "The linguistic documentation of Pomak is an ongoing work by an interdisciplinary team in close cooperation with the Pomak community of Greece .", "forward": false, "src_ids": "2022.computel-1.22_653"}
+{"input": "realistic adversarial attack is done by using OtherScientificTerm| context: unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts .", "entity": "realistic adversarial attack", "output": "human - written text perturbations", "neg_sample": ["realistic adversarial attack is done by using OtherScientificTerm", "unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts ."], "relation": "used for", "id": "2022.findings-acl.232", "year": 2022, "rel_sent": "Perturbations in the Wild : Leveraging Human - Written Text Perturbations for Realistic Adversarial Attack and Defense.", "forward": false, "src_ids": "2022.findings-acl.232_654"}
+{"input": "human - written text perturbations is used for Task| context: unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts .", "entity": "human - written text perturbations", "output": "realistic adversarial attack", "neg_sample": ["human - written text perturbations is used for Task", "unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts ."], "relation": "used for", "id": "2022.findings-acl.232", "year": 2022, "rel_sent": "Perturbations in the Wild : Leveraging Human - Written Text Perturbations for Realistic Adversarial Attack and Defense.", "forward": true, "src_ids": "2022.findings-acl.232_655"}
+{"input": "bert classifier is done by using Method| context: unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts .", "entity": "bert classifier", "output": "anthro", "neg_sample": ["bert classifier is done by using Method", "unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts ."], "relation": "used for", "id": "2022.findings-acl.232", "year": 2022, "rel_sent": "ANTHRO can further enhance a BERT classifier 's performance in understanding different variations of human - written toxic texts via adversarial training when compared to the Perspective API .", "forward": false, "src_ids": "2022.findings-acl.232_656"}
+{"input": "human - written toxic texts is done by using Method| context: unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts .", "entity": "human - written toxic texts", "output": "bert classifier", "neg_sample": ["human - written toxic texts is done by using Method", "unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts ."], "relation": "used for", "id": "2022.findings-acl.232", "year": 2022, "rel_sent": "ANTHRO can further enhance a BERT classifier 's performance in understanding different variations of human - written toxic texts via adversarial training when compared to the Perspective API .", "forward": false, "src_ids": "2022.findings-acl.232_657"}
+{"input": "anthro is used for Method| context: unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts .", "entity": "anthro", "output": "bert classifier", "neg_sample": ["anthro is used for Method", "unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts ."], "relation": "used for", "id": "2022.findings-acl.232", "year": 2022, "rel_sent": "ANTHRO can further enhance a BERT classifier 's performance in understanding different variations of human - written toxic texts via adversarial training when compared to the Perspective API .", "forward": true, "src_ids": "2022.findings-acl.232_658"}
+{"input": "bert classifier is used for Material| context: unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts .", "entity": "bert classifier", "output": "human - written toxic texts", "neg_sample": ["bert classifier is used for Material", "unlike existing character - based attacks which often deductively hypothesize a set of manipulation strategies , our work is grounded on actual observations from real - world texts ."], "relation": "used for", "id": "2022.findings-acl.232", "year": 2022, "rel_sent": "ANTHRO can further enhance a BERT classifier 's performance in understanding different variations of human - written toxic texts via adversarial training when compared to the Perspective API .", "forward": true, "src_ids": "2022.findings-acl.232_659"}
+{"input": "multi - document news summarization is done by using Method| context: a common method for extractive multi - document news summarization is to re - formulate it as a single - document summarization problem by concatenating all documents as a single meta - document . however , this method neglects the relative importance of documents .", "entity": "multi - document news summarization", "output": "document reordering approach", "neg_sample": ["multi - document news summarization is done by using Method", "a common method for extractive multi - document news summarization is to re - formulate it as a single - document summarization problem by concatenating all documents as a single meta - document .", "however , this method neglects the relative importance of documents ."], "relation": "used for", "id": "2022.findings-acl.51", "year": 2022, "rel_sent": "Read Top News First : A Document Reordering Approach for Multi - Document News Summarization.", "forward": false, "src_ids": "2022.findings-acl.51_660"}
+{"input": "document reordering approach is used for Task| context: however , this method neglects the relative importance of documents .", "entity": "document reordering approach", "output": "multi - document news summarization", "neg_sample": ["document reordering approach is used for Task", "however , this method neglects the relative importance of documents ."], "relation": "used for", "id": "2022.findings-acl.51", "year": 2022, "rel_sent": "Read Top News First : A Document Reordering Approach for Multi - Document News Summarization.", "forward": true, "src_ids": "2022.findings-acl.51_661"}
+{"input": "calibration method is used for OtherScientificTerm| context: early stopping , which is widely used to prevent overfitting , is generally based on a separate validation set . however , in low resource settings , validation - based stopping can be risky because a small validation set may not be sufficiently representative , and the reduction in the number of samples by validation split may result in insufficient samples for training .", "entity": "calibration method", "output": "unlabeled samples", "neg_sample": ["calibration method is used for OtherScientificTerm", "early stopping , which is widely used to prevent overfitting , is generally based on a separate validation set .", "however , in low resource settings , validation - based stopping can be risky because a small validation set may not be sufficiently representative , and the reduction in the number of samples by validation split may result in insufficient samples for training ."], "relation": "used for", "id": "2022.acl-long.52", "year": 2022, "rel_sent": "Tofurther improve the performance , we present a calibration method to better estimate the class distribution of the unlabeled samples .", "forward": true, "src_ids": "2022.acl-long.52_662"}
+{"input": "class distribution is done by using Method| context: early stopping , which is widely used to prevent overfitting , is generally based on a separate validation set . however , in low resource settings , validation - based stopping can be risky because a small validation set may not be sufficiently representative , and the reduction in the number of samples by validation split may result in insufficient samples for training .", "entity": "class distribution", "output": "calibration method", "neg_sample": ["class distribution is done by using Method", "early stopping , which is widely used to prevent overfitting , is generally based on a separate validation set .", "however , in low resource settings , validation - based stopping can be risky because a small validation set may not be sufficiently representative , and the reduction in the number of samples by validation split may result in insufficient samples for training ."], "relation": "used for", "id": "2022.acl-long.52", "year": 2022, "rel_sent": "Tofurther improve the performance , we present a calibration method to better estimate the class distribution of the unlabeled samples .", "forward": false, "src_ids": "2022.acl-long.52_663"}
+{"input": "unlabeled samples is done by using Method| context: early stopping , which is widely used to prevent overfitting , is generally based on a separate validation set . however , in low resource settings , validation - based stopping can be risky because a small validation set may not be sufficiently representative , and the reduction in the number of samples by validation split may result in insufficient samples for training .", "entity": "unlabeled samples", "output": "calibration method", "neg_sample": ["unlabeled samples is done by using Method", "early stopping , which is widely used to prevent overfitting , is generally based on a separate validation set .", "however , in low resource settings , validation - based stopping can be risky because a small validation set may not be sufficiently representative , and the reduction in the number of samples by validation split may result in insufficient samples for training ."], "relation": "used for", "id": "2022.acl-long.52", "year": 2022, "rel_sent": "Tofurther improve the performance , we present a calibration method to better estimate the class distribution of the unlabeled samples .", "forward": false, "src_ids": "2022.acl-long.52_664"}
+{"input": "named entity recognition is done by using Method| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "named entity recognition", "output": "semantic - based data augmentation method", "neg_sample": ["named entity recognition is done by using Method", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "Simple Semantic - based Data Augmentation for Named Entity Recognition in Biomedical Texts.", "forward": false, "src_ids": "2022.bionlp-1.12_665"}
+{"input": "biomedical ner is done by using Method| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "biomedical ner", "output": "semantic - based data augmentation method", "neg_sample": ["biomedical ner is done by using Method", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "To that end , in this paper we propose a simple semantic - based data augmentation method for biomedical NER .", "forward": false, "src_ids": "2022.bionlp-1.12_666"}
+{"input": "semantic - based data augmentation method is used for Task| context: data augmentation is important in addressing data sparsity and low resources in nlp .", "entity": "semantic - based data augmentation method", "output": "named entity recognition", "neg_sample": ["semantic - based data augmentation method is used for Task", "data augmentation is important in addressing data sparsity and low resources in nlp ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "Simple Semantic - based Data Augmentation for Named Entity Recognition in Biomedical Texts.", "forward": true, "src_ids": "2022.bionlp-1.12_667"}
+{"input": "semantic - based data augmentation method is used for Task| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "semantic - based data augmentation method", "output": "biomedical ner", "neg_sample": ["semantic - based data augmentation method is used for Task", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "To that end , in this paper we propose a simple semantic - based data augmentation method for biomedical NER .", "forward": true, "src_ids": "2022.bionlp-1.12_668"}
+{"input": "entity - level and sentence - level is done by using OtherScientificTerm| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "entity - level and sentence - level", "output": "semantic information", "neg_sample": ["entity - level and sentence - level is done by using OtherScientificTerm", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "Our method leverages semantic information from pre - trained language models for both entity - level and sentence - level .", "forward": false, "src_ids": "2022.bionlp-1.12_669"}
+{"input": "entity - level and sentence - level is done by using Method| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "entity - level and sentence - level", "output": "pre - trained language models", "neg_sample": ["entity - level and sentence - level is done by using Method", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "Our method leverages semantic information from pre - trained language models for both entity - level and sentence - level .", "forward": false, "src_ids": "2022.bionlp-1.12_670"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "pre - trained language models", "output": "entity - level and sentence - level", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "Our method leverages semantic information from pre - trained language models for both entity - level and sentence - level .", "forward": true, "src_ids": "2022.bionlp-1.12_671"}
+{"input": "semantic information is used for OtherScientificTerm| context: data augmentation is important in addressing data sparsity and low resources in nlp . unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities .", "entity": "semantic information", "output": "entity - level and sentence - level", "neg_sample": ["semantic information is used for OtherScientificTerm", "data augmentation is important in addressing data sparsity and low resources in nlp .", "unlike data augmentation for other tasks such as sentence - level and sentence - pair ones , data augmentation for named entity recognition ( ner ) requires preserving the semantic of entities ."], "relation": "used for", "id": "2022.bionlp-1.12", "year": 2022, "rel_sent": "Our method leverages semantic information from pre - trained language models for both entity - level and sentence - level .", "forward": true, "src_ids": "2022.bionlp-1.12_672"}
+{"input": "zero - shot setting is done by using Method| context: large multilingual pretrained language models such as mbert and xlm - roberta have been found to be surprisingly effective for cross - lingual transfer of syntactic parsing models wu and dredze ( 2019 ) , but only between related languages . however , source and training languages are rarely related , when parsing truly low - resource languages .", "entity": "zero - shot setting", "output": "uniform and size - proportional sampling", "neg_sample": ["zero - shot setting is done by using Method", "large multilingual pretrained language models such as mbert and xlm - roberta have been found to be surprisingly effective for cross - lingual transfer of syntactic parsing models wu and dredze ( 2019 ) , but only between related languages .", "however , source and training languages are rarely related , when parsing truly low - resource languages ."], "relation": "used for", "id": "2022.acl-short.64", "year": 2022, "rel_sent": "We show that this approach is significantly better than uniform and size - proportional sampling in the zero - shot setting .", "forward": false, "src_ids": "2022.acl-short.64_673"}
+{"input": "uniform and size - proportional sampling is used for OtherScientificTerm| context: large multilingual pretrained language models such as mbert and xlm - roberta have been found to be surprisingly effective for cross - lingual transfer of syntactic parsing models wu and dredze ( 2019 ) , but only between related languages . however , source and training languages are rarely related , when parsing truly low - resource languages .", "entity": "uniform and size - proportional sampling", "output": "zero - shot setting", "neg_sample": ["uniform and size - proportional sampling is used for OtherScientificTerm", "large multilingual pretrained language models such as mbert and xlm - roberta have been found to be surprisingly effective for cross - lingual transfer of syntactic parsing models wu and dredze ( 2019 ) , but only between related languages .", "however , source and training languages are rarely related , when parsing truly low - resource languages ."], "relation": "used for", "id": "2022.acl-short.64", "year": 2022, "rel_sent": "We show that this approach is significantly better than uniform and size - proportional sampling in the zero - shot setting .", "forward": true, "src_ids": "2022.acl-short.64_674"}
+{"input": "multiple baseline systems is used for Task| context: multimodal machine translation and textual chat translation have received considerable attention in recent years . although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations .", "entity": "multiple baseline systems", "output": "multimodal chat translation ( mct )", "neg_sample": ["multiple baseline systems is used for Task", "multimodal machine translation and textual chat translation have received considerable attention in recent years .", "although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations ."], "relation": "used for", "id": "2022.acl-long.186", "year": 2022, "rel_sent": "Then , we benchmark the task by establishing multiple baseline systems that incorporate multimodal and sentiment features for MCT .", "forward": true, "src_ids": "2022.acl-long.186_675"}
+{"input": "multimodal and sentiment features is used for Task| context: multimodal machine translation and textual chat translation have received considerable attention in recent years . although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations .", "entity": "multimodal and sentiment features", "output": "multimodal chat translation ( mct )", "neg_sample": ["multimodal and sentiment features is used for Task", "multimodal machine translation and textual chat translation have received considerable attention in recent years .", "although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations ."], "relation": "used for", "id": "2022.acl-long.186", "year": 2022, "rel_sent": "Then , we benchmark the task by establishing multiple baseline systems that incorporate multimodal and sentiment features for MCT .", "forward": true, "src_ids": "2022.acl-long.186_676"}
+{"input": "conversational scene is done by using OtherScientificTerm| context: multimodal machine translation and textual chat translation have received considerable attention in recent years . although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations .", "entity": "conversational scene", "output": "visual context", "neg_sample": ["conversational scene is done by using OtherScientificTerm", "multimodal machine translation and textual chat translation have received considerable attention in recent years .", "although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations ."], "relation": "used for", "id": "2022.acl-long.186", "year": 2022, "rel_sent": "Each utterance pair , corresponding to the visual context that reflects the current conversational scene , is annotated with a sentiment label .", "forward": false, "src_ids": "2022.acl-long.186_677"}
+{"input": "visual context is used for OtherScientificTerm| context: multimodal machine translation and textual chat translation have received considerable attention in recent years . although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations .", "entity": "visual context", "output": "conversational scene", "neg_sample": ["visual context is used for OtherScientificTerm", "multimodal machine translation and textual chat translation have received considerable attention in recent years .", "although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations ."], "relation": "used for", "id": "2022.acl-long.186", "year": 2022, "rel_sent": "Each utterance pair , corresponding to the visual context that reflects the current conversational scene , is annotated with a sentiment label .", "forward": true, "src_ids": "2022.acl-long.186_678"}
+{"input": "multimodal chat translation ( mct ) is done by using Method| context: multimodal machine translation and textual chat translation have received considerable attention in recent years . although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations .", "entity": "multimodal chat translation ( mct )", "output": "multiple baseline systems", "neg_sample": ["multimodal chat translation ( mct ) is done by using Method", "multimodal machine translation and textual chat translation have received considerable attention in recent years .", "although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations ."], "relation": "used for", "id": "2022.acl-long.186", "year": 2022, "rel_sent": "Then , we benchmark the task by establishing multiple baseline systems that incorporate multimodal and sentiment features for MCT .", "forward": false, "src_ids": "2022.acl-long.186_679"}
+{"input": "multimodal chat translation ( mct ) is done by using OtherScientificTerm| context: multimodal machine translation and textual chat translation have received considerable attention in recent years . although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations .", "entity": "multimodal chat translation ( mct )", "output": "multimodal and sentiment features", "neg_sample": ["multimodal chat translation ( mct ) is done by using OtherScientificTerm", "multimodal machine translation and textual chat translation have received considerable attention in recent years .", "although the conversation in its natural form is usually multimodal , there still lacks work on multimodal machine translation in conversations ."], "relation": "used for", "id": "2022.acl-long.186", "year": 2022, "rel_sent": "Then , we benchmark the task by establishing multiple baseline systems that incorporate multimodal and sentiment features for MCT .", "forward": false, "src_ids": "2022.acl-long.186_680"}
+{"input": "entity disambiguation ( ed ) is done by using Method| context: local models for entity disambiguation ( ed ) have today become extremely powerful , in most part thanks to the advent of large pre - trained language models . however , despite their significant performance achievements , most of these approaches frame ed through classification formulations that have intrinsic limitations , both computationally and from a modeling perspective .", "entity": "entity disambiguation ( ed )", "output": "local formulation", "neg_sample": ["entity disambiguation ( ed ) is done by using Method", "local models for entity disambiguation ( ed ) have today become extremely powerful , in most part thanks to the advent of large pre - trained language models .", "however , despite their significant performance achievements , most of these approaches frame ed through classification formulations that have intrinsic limitations , both computationally and from a modeling perspective ."], "relation": "used for", "id": "2022.acl-long.177", "year": 2022, "rel_sent": "In contrast with this trend , here we propose ExtEnD , a novel local formulation for ED where we frame this task as a text extraction problem , and present two Transformer - based architectures that implement it .", "forward": false, "src_ids": "2022.acl-long.177_681"}
+{"input": "object - level concepts is done by using Method| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "object - level concepts", "output": "visitron", "neg_sample": ["object - level concepts is done by using Method", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "VISITRON is trained to : i ) identify and associate object - level concepts and semantics between the environment and dialogue history , ii ) identify when to interact vs. navigate via imitation learning of a binary classification head .", "forward": false, "src_ids": "2022.findings-acl.157_682"}
+{"input": "cooperative vision - and - dialog navigation is done by using Method| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "cooperative vision - and - dialog navigation", "output": "visitron", "neg_sample": ["cooperative vision - and - dialog navigation is done by using Method", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "We perform extensive pre - training and fine - tuning ablations with VISITRON to gain empirical insights and improve performance on CVDN .", "forward": false, "src_ids": "2022.findings-acl.157_683"}
+{"input": "interactive regime is done by using Method| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "interactive regime", "output": "multi - modal transformer - based navigator", "neg_sample": ["interactive regime is done by using Method", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "In this paper , we present VISITRON , a multi - modal Transformer - based navigator better suited to the interactive regime inherent to Cooperative Vision - and - Dialog Navigation ( CVDN ) .", "forward": false, "src_ids": "2022.findings-acl.157_684"}
+{"input": "multi - modal transformer - based navigator is used for OtherScientificTerm| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "multi - modal transformer - based navigator", "output": "interactive regime", "neg_sample": ["multi - modal transformer - based navigator is used for OtherScientificTerm", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "In this paper , we present VISITRON , a multi - modal Transformer - based navigator better suited to the interactive regime inherent to Cooperative Vision - and - Dialog Navigation ( CVDN ) .", "forward": true, "src_ids": "2022.findings-acl.157_685"}
+{"input": "visitron is used for OtherScientificTerm| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "visitron", "output": "object - level concepts", "neg_sample": ["visitron is used for OtherScientificTerm", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "VISITRON is trained to : i ) identify and associate object - level concepts and semantics between the environment and dialogue history , ii ) identify when to interact vs. navigate via imitation learning of a binary classification head .", "forward": true, "src_ids": "2022.findings-acl.157_686"}
+{"input": "pre - training and fine - tuning ablations is used for Method| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "pre - training and fine - tuning ablations", "output": "cooperative vision - and - dialog navigation", "neg_sample": ["pre - training and fine - tuning ablations is used for Method", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "We perform extensive pre - training and fine - tuning ablations with VISITRON to gain empirical insights and improve performance on CVDN .", "forward": true, "src_ids": "2022.findings-acl.157_687"}
+{"input": "visitron is used for Method| context: interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) .", "entity": "visitron", "output": "cooperative vision - and - dialog navigation", "neg_sample": ["visitron is used for Method", "interactive robots navigating photo - realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision - and - language navigation ( vln ) ."], "relation": "used for", "id": "2022.findings-acl.157", "year": 2022, "rel_sent": "We perform extensive pre - training and fine - tuning ablations with VISITRON to gain empirical insights and improve performance on CVDN .", "forward": true, "src_ids": "2022.findings-acl.157_688"}
+{"input": "neural machine translation is done by using Method| context: machine translation typically adopts an encoder - to - decoder framework , in which the decoder generates the target sentence word - by - word in an auto - regressive manner . however , the auto - regressive decoder faces a deep - rooted one - pass issue whereby each generated word is considered as one element of the final output regardless of whether it is correct or not . these generated wrong words further constitute the target historical context to affect the generation of subsequent target words .", "entity": "neural machine translation", "output": "synchronous refinement method", "neg_sample": ["neural machine translation is done by using Method", "machine translation typically adopts an encoder - to - decoder framework , in which the decoder generates the target sentence word - by - word in an auto - regressive manner .", "however , the auto - regressive decoder faces a deep - rooted one - pass issue whereby each generated word is considered as one element of the final output regardless of whether it is correct or not .", "these generated wrong words further constitute the target historical context to affect the generation of subsequent target words ."], "relation": "used for", "id": "2022.findings-acl.235", "year": 2022, "rel_sent": "Synchronous Refinement for Neural Machine Translation.", "forward": false, "src_ids": "2022.findings-acl.235_689"}
+{"input": "synchronous refinement method is used for Task| context: machine translation typically adopts an encoder - to - decoder framework , in which the decoder generates the target sentence word - by - word in an auto - regressive manner . however , the auto - regressive decoder faces a deep - rooted one - pass issue whereby each generated word is considered as one element of the final output regardless of whether it is correct or not . these generated wrong words further constitute the target historical context to affect the generation of subsequent target words .", "entity": "synchronous refinement method", "output": "neural machine translation", "neg_sample": ["synchronous refinement method is used for Task", "machine translation typically adopts an encoder - to - decoder framework , in which the decoder generates the target sentence word - by - word in an auto - regressive manner .", "however , the auto - regressive decoder faces a deep - rooted one - pass issue whereby each generated word is considered as one element of the final output regardless of whether it is correct or not .", "these generated wrong words further constitute the target historical context to affect the generation of subsequent target words ."], "relation": "used for", "id": "2022.findings-acl.235", "year": 2022, "rel_sent": "Synchronous Refinement for Neural Machine Translation.", "forward": true, "src_ids": "2022.findings-acl.235_690"}
+{"input": "anti - asian hate sentiment is done by using Material| context: covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact . however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope .", "entity": "anti - asian hate sentiment", "output": "twitter", "neg_sample": ["anti - asian hate sentiment is done by using Material", "covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact .", "however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope ."], "relation": "used for", "id": "2022.wassa-1.2", "year": 2022, "rel_sent": "We introduce new datasets from Twitter related to anti - Asian hate sentiment before and during the pandemic .", "forward": false, "src_ids": "2022.wassa-1.2_691"}
+{"input": "twitter is used for OtherScientificTerm| context: covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact . however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope .", "entity": "twitter", "output": "anti - asian hate sentiment", "neg_sample": ["twitter is used for OtherScientificTerm", "covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact .", "however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope ."], "relation": "used for", "id": "2022.wassa-1.2", "year": 2022, "rel_sent": "We introduce new datasets from Twitter related to anti - Asian hate sentiment before and during the pandemic .", "forward": true, "src_ids": "2022.wassa-1.2_692"}
+{"input": "hateful and counter - hate tweets is done by using Material| context: covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact . however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope .", "entity": "hateful and counter - hate tweets", "output": "twitter 's academic api", "neg_sample": ["hateful and counter - hate tweets is done by using Material", "covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact .", "however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope ."], "relation": "used for", "id": "2022.wassa-1.2", "year": 2022, "rel_sent": "Relying on Twitter 's academic API , we retrieve hateful and counter - hate tweets from the Twitter Historical Database .", "forward": false, "src_ids": "2022.wassa-1.2_693"}
+{"input": "twitter 's academic api is used for OtherScientificTerm| context: covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact . however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope .", "entity": "twitter 's academic api", "output": "hateful and counter - hate tweets", "neg_sample": ["twitter 's academic api is used for OtherScientificTerm", "covid-19 has disproportionately threatened minority communities in the u.s , not only in health but also in societal impact .", "however , social scientists and policymakers lack critical data to capture the dynamics of the anti - asian hate trend and to evaluate its scale and scope ."], "relation": "used for", "id": "2022.wassa-1.2", "year": 2022, "rel_sent": "Relying on Twitter 's academic API , we retrieve hateful and counter - hate tweets from the Twitter Historical Database .", "forward": true, "src_ids": "2022.wassa-1.2_694"}
+{"input": "traits is done by using Method| context: item response theory ( irt ) has been extensively used to numerically characterize question difficulty and discrimination for human subjects in domains including cognitive psychology and education ( primi et al . , 2014 ; downing , 2003 ) . more recently , irt has been used to similarly characterize item difficulty and discrimination for natural language models across various datasets ( lalor et al . , 2019 ; vania et al . , 2021 ; rodriguez et al . , 2021 ) .", "entity": "traits", "output": "predictive models", "neg_sample": ["traits is done by using Method", "item response theory ( irt ) has been extensively used to numerically characterize question difficulty and discrimination for human subjects in domains including cognitive psychology and education ( primi et al .", ", 2014 ; downing , 2003 ) .", "more recently , irt has been used to similarly characterize item difficulty and discrimination for natural language models across various datasets ( lalor et al .", ", 2019 ; vania et al .", ", 2021 ; rodriguez et al .", ", 2021 ) ."], "relation": "used for", "id": "2022.acl-short.15", "year": 2022, "rel_sent": "In this work , we explore predictive models for directly estimating and explaining these traits for natural language questions in a question - answering context .", "forward": false, "src_ids": "2022.acl-short.15_695"}
+{"input": "predictive models is used for OtherScientificTerm| context: item response theory ( irt ) has been extensively used to numerically characterize question difficulty and discrimination for human subjects in domains including cognitive psychology and education ( primi et al . , 2014 ; downing , 2003 ) . more recently , irt has been used to similarly characterize item difficulty and discrimination for natural language models across various datasets ( lalor et al . , 2019 ; vania et al . , 2021 ; rodriguez et al . , 2021 ) .", "entity": "predictive models", "output": "traits", "neg_sample": ["predictive models is used for OtherScientificTerm", "item response theory ( irt ) has been extensively used to numerically characterize question difficulty and discrimination for human subjects in domains including cognitive psychology and education ( primi et al .", ", 2014 ; downing , 2003 ) .", "more recently , irt has been used to similarly characterize item difficulty and discrimination for natural language models across various datasets ( lalor et al .", ", 2019 ; vania et al .", ", 2021 ; rodriguez et al .", ", 2021 ) ."], "relation": "used for", "id": "2022.acl-short.15", "year": 2022, "rel_sent": "In this work , we explore predictive models for directly estimating and explaining these traits for natural language questions in a question - answering context .", "forward": true, "src_ids": "2022.acl-short.15_696"}
+{"input": "open - domain qa is done by using Task| context: the retriever - reader pipeline has shown promising performance in open - domain qa but suffers from a very slow inference speed . recently proposed question retrieval models tackle this problem by indexing question - answer pairs and searching for similar questions . these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models .", "entity": "open - domain qa", "output": "two - step question retrieval", "neg_sample": ["open - domain qa is done by using Task", "the retriever - reader pipeline has shown promising performance in open - domain qa but suffers from a very slow inference speed .", "recently proposed question retrieval models tackle this problem by indexing question - answer pairs and searching for similar questions .", "these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models ."], "relation": "used for", "id": "2022.findings-acl.117", "year": 2022, "rel_sent": "Two - Step Question Retrieval for Open - Domain QA.", "forward": false, "src_ids": "2022.findings-acl.117_697"}
+{"input": "two - step question retrieval is used for Task| context: recently proposed question retrieval models tackle this problem by indexing question - answer pairs and searching for similar questions . these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models .", "entity": "two - step question retrieval", "output": "open - domain qa", "neg_sample": ["two - step question retrieval is used for Task", "recently proposed question retrieval models tackle this problem by indexing question - answer pairs and searching for similar questions .", "these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models ."], "relation": "used for", "id": "2022.findings-acl.117", "year": 2022, "rel_sent": "Two - Step Question Retrieval for Open - Domain QA.", "forward": true, "src_ids": "2022.findings-acl.117_698"}
+{"input": "question retrieval is done by using Method| context: the retriever - reader pipeline has shown promising performance in open - domain qa but suffers from a very slow inference speed . recently proposed question retrieval models tackle this problem by indexing question - answer pairs and searching for similar questions . these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models .", "entity": "question retrieval", "output": "bi - encoders", "neg_sample": ["question retrieval is done by using Method", "the retriever - reader pipeline has shown promising performance in open - domain qa but suffers from a very slow inference speed .", "recently proposed question retrieval models tackle this problem by indexing question - answer pairs and searching for similar questions .", "these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models ."], "relation": "used for", "id": "2022.findings-acl.117", "year": 2022, "rel_sent": "SQuID uses two bi - encoders for question retrieval .", "forward": false, "src_ids": "2022.findings-acl.117_699"}
+{"input": "bi - encoders is used for Task| context: the retriever - reader pipeline has shown promising performance in open - domain qa but suffers from a very slow inference speed . these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models .", "entity": "bi - encoders", "output": "question retrieval", "neg_sample": ["bi - encoders is used for Task", "the retriever - reader pipeline has shown promising performance in open - domain qa but suffers from a very slow inference speed .", "these models have shown a significant increase in inference speed , but at the cost of lower qa performance compared to the retriever - reader models ."], "relation": "used for", "id": "2022.findings-acl.117", "year": 2022, "rel_sent": "SQuID uses two bi - encoders for question retrieval .", "forward": true, "src_ids": "2022.findings-acl.117_700"}
+{"input": "premise paragraphs is done by using Material| context: to create models that are robust across a wide range of test inputs , training datasets should include diverse examples that span numerous phenomena . dynamic adversarial data collection ( dadc ) , where annotators craft examples that challenge continually improving models , holds promise as an approach for generating such diverse training sets . prior work has shown that running dadc over 1 - 3 rounds can help models fix some error types , but it does not necessarily lead to better generalization beyond adversarial test data .", "entity": "premise paragraphs", "output": "nli examples", "neg_sample": ["premise paragraphs is done by using Material", "to create models that are robust across a wide range of test inputs , training datasets should include diverse examples that span numerous phenomena .", "dynamic adversarial data collection ( dadc ) , where annotators craft examples that challenge continually improving models , holds promise as an approach for generating such diverse training sets .", "prior work has shown that running dadc over 1 - 3 rounds can help models fix some error types , but it does not necessarily lead to better generalization beyond adversarial test data ."], "relation": "used for", "id": "2022.findings-acl.18", "year": 2022, "rel_sent": "We present the first study of longer - term DADC , where we collect 20 rounds of NLI examples for a small set of premise paragraphs , with both adversarial and non - adversarial approaches .", "forward": false, "src_ids": "2022.findings-acl.18_701"}
+{"input": "nli examples is used for OtherScientificTerm| context: to create models that are robust across a wide range of test inputs , training datasets should include diverse examples that span numerous phenomena . dynamic adversarial data collection ( dadc ) , where annotators craft examples that challenge continually improving models , holds promise as an approach for generating such diverse training sets . prior work has shown that running dadc over 1 - 3 rounds can help models fix some error types , but it does not necessarily lead to better generalization beyond adversarial test data .", "entity": "nli examples", "output": "premise paragraphs", "neg_sample": ["nli examples is used for OtherScientificTerm", "to create models that are robust across a wide range of test inputs , training datasets should include diverse examples that span numerous phenomena .", "dynamic adversarial data collection ( dadc ) , where annotators craft examples that challenge continually improving models , holds promise as an approach for generating such diverse training sets .", "prior work has shown that running dadc over 1 - 3 rounds can help models fix some error types , but it does not necessarily lead to better generalization beyond adversarial test data ."], "relation": "used for", "id": "2022.findings-acl.18", "year": 2022, "rel_sent": "We present the first study of longer - term DADC , where we collect 20 rounds of NLI examples for a small set of premise paragraphs , with both adversarial and non - adversarial approaches .", "forward": true, "src_ids": "2022.findings-acl.18_702"}
+{"input": "contrastive summaries is done by using Task| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? '", "entity": "contrastive summaries", "output": "comparative opinion summarization task", "neg_sample": ["contrastive summaries is done by using Task", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? '"], "relation": "used for", "id": "2022.findings-acl.261", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews . We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries . Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": false, "src_ids": "2022.findings-acl.261_703"}
+{"input": "comparative opinion summarization task is used for OtherScientificTerm| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? '", "entity": "comparative opinion summarization task", "output": "contrastive summaries", "neg_sample": ["comparative opinion summarization task is used for OtherScientificTerm", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? '"], "relation": "used for", "id": "2022.findings-acl.261", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews . We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries . Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": true, "src_ids": "2022.findings-acl.261_704"}
+{"input": "contrastive and common summaries is done by using Method| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? '", "entity": "contrastive and common summaries", "output": "base summarization models", "neg_sample": ["contrastive and common summaries is done by using Method", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? '"], "relation": "used for", "id": "2022.findings-acl.261", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews . We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries . Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": false, "src_ids": "2022.findings-acl.261_705"}
+{"input": "base summarization models is used for OtherScientificTerm| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? '", "entity": "base summarization models", "output": "contrastive and common summaries", "neg_sample": ["base summarization models is used for OtherScientificTerm", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? '"], "relation": "used for", "id": "2022.findings-acl.261", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews . We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries . Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": true, "src_ids": "2022.findings-acl.261_706"}
+{"input": "cocosum is used for OtherScientificTerm| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? '", "entity": "cocosum", "output": "contrastive and common summaries", "neg_sample": ["cocosum is used for OtherScientificTerm", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? '"], "relation": "used for", "id": "2022.findings-acl.261", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews . We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries . Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": true, "src_ids": "2022.findings-acl.261_707"}
+{"input": "consistency loss is used for OtherScientificTerm| context: cross - lingual natural language inference ( xnli ) is a fundamental task in cross - lingual natural language understanding . recently this task is commonly addressed by pre - trained cross - lingual language models . existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora . these additional data , however , are rare in practice , especially for low - resource languages .", "entity": "consistency loss", "output": "answer probability distribution", "neg_sample": ["consistency loss is used for OtherScientificTerm", "cross - lingual natural language inference ( xnli ) is a fundamental task in cross - lingual natural language understanding .", "recently this task is commonly addressed by pre - trained cross - lingual language models .", "existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora .", "these additional data , however , are rare in practice , especially for low - resource languages ."], "relation": "used for", "id": "2022.acl-long.134", "year": 2022, "rel_sent": "To enforce correspondence between different languages , the framework augments a new question for every question using a sampled template in another language and then introduces a consistency loss to make the answer probability distribution obtained from the new question as similar as possible with the corresponding distribution obtained from the original question .", "forward": true, "src_ids": "2022.acl-long.134_708"}
+{"input": "xnli is done by using Method| context: cross - lingual natural language inference ( xnli ) is a fundamental task in cross - lingual natural language understanding . recently this task is commonly addressed by pre - trained cross - lingual language models . existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora . these additional data , however , are rare in practice , especially for low - resource languages .", "entity": "xnli", "output": "prompt - learning based framework", "neg_sample": ["xnli is done by using Method", "cross - lingual natural language inference ( xnli ) is a fundamental task in cross - lingual natural language understanding .", "recently this task is commonly addressed by pre - trained cross - lingual language models .", "existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora .", "these additional data , however , are rare in practice , especially for low - resource languages ."], "relation": "used for", "id": "2022.acl-long.134", "year": 2022, "rel_sent": "Inspired by recent promising results achieved by prompt - learning , this paper proposes a novel prompt - learning based framework for enhancing XNLI .", "forward": false, "src_ids": "2022.acl-long.134_709"}
+{"input": "prompt - learning based framework is used for Method| context: recently this task is commonly addressed by pre - trained cross - lingual language models . existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora . these additional data , however , are rare in practice , especially for low - resource languages .", "entity": "prompt - learning based framework", "output": "xnli", "neg_sample": ["prompt - learning based framework is used for Method", "recently this task is commonly addressed by pre - trained cross - lingual language models .", "existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora .", "these additional data , however , are rare in practice , especially for low - resource languages ."], "relation": "used for", "id": "2022.acl-long.134", "year": 2022, "rel_sent": "Inspired by recent promising results achieved by prompt - learning , this paper proposes a novel prompt - learning based framework for enhancing XNLI .", "forward": true, "src_ids": "2022.acl-long.134_710"}
+{"input": "answer probability distribution is done by using OtherScientificTerm| context: cross - lingual natural language inference ( xnli ) is a fundamental task in cross - lingual natural language understanding . recently this task is commonly addressed by pre - trained cross - lingual language models . existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora . these additional data , however , are rare in practice , especially for low - resource languages .", "entity": "answer probability distribution", "output": "consistency loss", "neg_sample": ["answer probability distribution is done by using OtherScientificTerm", "cross - lingual natural language inference ( xnli ) is a fundamental task in cross - lingual natural language understanding .", "recently this task is commonly addressed by pre - trained cross - lingual language models .", "existing methods usually enhance pre - trained language models with additional data , such as annotated parallel corpora .", "these additional data , however , are rare in practice , especially for low - resource languages ."], "relation": "used for", "id": "2022.acl-long.134", "year": 2022, "rel_sent": "To enforce correspondence between different languages , the framework augments a new question for every question using a sampled template in another language and then introduces a consistency loss to make the answer probability distribution obtained from the new question as similar as possible with the corresponding distribution obtained from the original question .", "forward": false, "src_ids": "2022.acl-long.134_711"}
+{"input": "logical reasoning is done by using Method| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "logical reasoning", "output": "meta - path guided contrastive learning method", "neg_sample": ["logical reasoning is done by using Method", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "MERIt : Meta - Path Guided Contrastive Learning for Logical Reasoning.", "forward": false, "src_ids": "2022.findings-acl.276_712"}
+{"input": "logical reasoning of text is done by using Method| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "logical reasoning of text", "output": "meta - path guided contrastive learning method", "neg_sample": ["logical reasoning of text is done by using Method", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "To address these two problems , in this paper , we propose MERIt , a MEta - path guided contrastive learning method for logical ReasonIng of text , to perform self - supervised pre - training on abundant unlabeled text data .", "forward": false, "src_ids": "2022.findings-acl.276_713"}
+{"input": "self - supervised pre - training is done by using Method| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "self - supervised pre - training", "output": "meta - path guided contrastive learning method", "neg_sample": ["self - supervised pre - training is done by using Method", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "To address these two problems , in this paper , we propose MERIt , a MEta - path guided contrastive learning method for logical ReasonIng of text , to perform self - supervised pre - training on abundant unlabeled text data .", "forward": false, "src_ids": "2022.findings-acl.276_714"}
+{"input": "meta - path guided contrastive learning method is used for Method| context: previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "meta - path guided contrastive learning method", "output": "logical reasoning", "neg_sample": ["meta - path guided contrastive learning method is used for Method", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "MERIt : Meta - Path Guided Contrastive Learning for Logical Reasoning.", "forward": true, "src_ids": "2022.findings-acl.276_715"}
+{"input": "meta - path guided contrastive learning method is used for Task| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "meta - path guided contrastive learning method", "output": "logical reasoning of text", "neg_sample": ["meta - path guided contrastive learning method is used for Task", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "To address these two problems , in this paper , we propose MERIt , a MEta - path guided contrastive learning method for logical ReasonIng of text , to perform self - supervised pre - training on abundant unlabeled text data .", "forward": true, "src_ids": "2022.findings-acl.276_716"}
+{"input": "merit is used for Method| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "merit", "output": "self - supervised pre - training", "neg_sample": ["merit is used for Method", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "To address these two problems , in this paper , we propose MERIt , a MEta - path guided contrastive learning method for logical ReasonIng of text , to perform self - supervised pre - training on abundant unlabeled text data .", "forward": true, "src_ids": "2022.findings-acl.276_717"}
+{"input": "meta - path guided contrastive learning method is used for Method| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "meta - path guided contrastive learning method", "output": "self - supervised pre - training", "neg_sample": ["meta - path guided contrastive learning method is used for Method", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "To address these two problems , in this paper , we propose MERIt , a MEta - path guided contrastive learning method for logical ReasonIng of text , to perform self - supervised pre - training on abundant unlabeled text data .", "forward": true, "src_ids": "2022.findings-acl.276_718"}
+{"input": "logical structure is done by using OtherScientificTerm| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "logical structure", "output": "meta - path", "neg_sample": ["logical structure is done by using OtherScientificTerm", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "In particular , a strategy based on meta - path is devised to discover the logical structure in natural texts , followed by a counterfactual data augmentation strategy to eliminate the information shortcut induced by pre - training .", "forward": false, "src_ids": "2022.findings-acl.276_719"}
+{"input": "meta - path is used for OtherScientificTerm| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "meta - path", "output": "logical structure", "neg_sample": ["meta - path is used for OtherScientificTerm", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "In particular , a strategy based on meta - path is devised to discover the logical structure in natural texts , followed by a counterfactual data augmentation strategy to eliminate the information shortcut induced by pre - training .", "forward": true, "src_ids": "2022.findings-acl.276_720"}
+{"input": "information shortcut is done by using Method| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "information shortcut", "output": "counterfactual data augmentation strategy", "neg_sample": ["information shortcut is done by using Method", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "In particular , a strategy based on meta - path is devised to discover the logical structure in natural texts , followed by a counterfactual data augmentation strategy to eliminate the information shortcut induced by pre - training .", "forward": false, "src_ids": "2022.findings-acl.276_721"}
+{"input": "counterfactual data augmentation strategy is used for OtherScientificTerm| context: logical reasoning is of vital importance to natural language understanding . previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation . these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity .", "entity": "counterfactual data augmentation strategy", "output": "information shortcut", "neg_sample": ["counterfactual data augmentation strategy is used for OtherScientificTerm", "logical reasoning is of vital importance to natural language understanding .", "previous studies either employ graph - based models to incorporate prior knowledge about logical relations , or introduce symbolic logic into neural models through data augmentation .", "these methods , however , heavily depend on annotated training data , and thus suffer from over - fitting and poor generalization problems due to the dataset sparsity ."], "relation": "used for", "id": "2022.findings-acl.276", "year": 2022, "rel_sent": "In particular , a strategy based on meta - path is devised to discover the logical structure in natural texts , followed by a counterfactual data augmentation strategy to eliminate the information shortcut induced by pre - training .", "forward": true, "src_ids": "2022.findings-acl.276_722"}
+{"input": "divergence of patterns is done by using Method| context: math word problem ( mwp ) solving needs to discover the quantitative relationships over natural language narratives . recent work shows that existing models memorize procedures from context and rely on shallow heuristics to solve mwps . in this paper , we look at this issue and argue that the cause is a lack of overall understanding of mwp patterns .", "entity": "divergence of patterns", "output": "neural network", "neg_sample": ["divergence of patterns is done by using Method", "math word problem ( mwp ) solving needs to discover the quantitative relationships over natural language narratives .", "recent work shows that existing models memorize procedures from context and rely on shallow heuristics to solve mwps .", "in this paper , we look at this issue and argue that the cause is a lack of overall understanding of mwp patterns ."], "relation": "used for", "id": "2022.findings-acl.195", "year": 2022, "rel_sent": "Inspired by it , we propose a contrastive learning approach , where the neural network perceives the divergence of patterns .", "forward": false, "src_ids": "2022.findings-acl.195_723"}
+{"input": "neural network is used for OtherScientificTerm| context: math word problem ( mwp ) solving needs to discover the quantitative relationships over natural language narratives . recent work shows that existing models memorize procedures from context and rely on shallow heuristics to solve mwps . in this paper , we look at this issue and argue that the cause is a lack of overall understanding of mwp patterns .", "entity": "neural network", "output": "divergence of patterns", "neg_sample": ["neural network is used for OtherScientificTerm", "math word problem ( mwp ) solving needs to discover the quantitative relationships over natural language narratives .", "recent work shows that existing models memorize procedures from context and rely on shallow heuristics to solve mwps .", "in this paper , we look at this issue and argue that the cause is a lack of overall understanding of mwp patterns ."], "relation": "used for", "id": "2022.findings-acl.195", "year": 2022, "rel_sent": "Inspired by it , we propose a contrastive learning approach , where the neural network perceives the divergence of patterns .", "forward": true, "src_ids": "2022.findings-acl.195_724"}
+{"input": "text - only ic model is done by using Material| context: item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce . as a special text classification task , fine - tuning pre - trained models , e.g. , bert , has become a mainstream solution . although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems .", "entity": "text - only ic model", "output": "product images", "neg_sample": ["text - only ic model is done by using Material", "item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce .", "as a special text classification task , fine - tuning pre - trained models , e.g.", ", bert , has become a mainstream solution .", "although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems ."], "relation": "used for", "id": "2022.ecnlp-1.25", "year": 2022, "rel_sent": "In this paper , we proposed a new way of using product images to improve text - only IC model : leveraging cross - modal signals between products ' titles and associated images to adapt BERT models in a self - supervised learning ( SSL ) way .", "forward": false, "src_ids": "2022.ecnlp-1.25_725"}
+{"input": "product images is used for Method| context: item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce . as a special text classification task , fine - tuning pre - trained models , e.g. , bert , has become a mainstream solution . to improve ic performance further , other product metadata , e.g. , product images , have been used . although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems .", "entity": "product images", "output": "text - only ic model", "neg_sample": ["product images is used for Method", "item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce .", "as a special text classification task , fine - tuning pre - trained models , e.g.", ", bert , has become a mainstream solution .", "to improve ic performance further , other product metadata , e.g.", ", product images , have been used .", "although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems ."], "relation": "used for", "id": "2022.ecnlp-1.25", "year": 2022, "rel_sent": "In this paper , we proposed a new way of using product images to improve text - only IC model : leveraging cross - modal signals between products ' titles and associated images to adapt BERT models in a self - supervised learning ( SSL ) way .", "forward": true, "src_ids": "2022.ecnlp-1.25_726"}
+{"input": "bert models is done by using OtherScientificTerm| context: item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce . as a special text classification task , fine - tuning pre - trained models , e.g. , bert , has become a mainstream solution . to improve ic performance further , other product metadata , e.g. , product images , have been used . although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems .", "entity": "bert models", "output": "cross - modal signals", "neg_sample": ["bert models is done by using OtherScientificTerm", "item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce .", "as a special text classification task , fine - tuning pre - trained models , e.g.", ", bert , has become a mainstream solution .", "to improve ic performance further , other product metadata , e.g.", ", product images , have been used .", "although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems ."], "relation": "used for", "id": "2022.ecnlp-1.25", "year": 2022, "rel_sent": "In this paper , we proposed a new way of using product images to improve text - only IC model : leveraging cross - modal signals between products ' titles and associated images to adapt BERT models in a self - supervised learning ( SSL ) way .", "forward": false, "src_ids": "2022.ecnlp-1.25_727"}
+{"input": "cross - modal signals is used for Method| context: item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce . as a special text classification task , fine - tuning pre - trained models , e.g. , bert , has become a mainstream solution . to improve ic performance further , other product metadata , e.g. , product images , have been used . although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems .", "entity": "cross - modal signals", "output": "bert models", "neg_sample": ["cross - modal signals is used for Method", "item categorization ( ic ) is a core natural language processing ( nlp ) task in e - commerce .", "as a special text classification task , fine - tuning pre - trained models , e.g.", ", bert , has become a mainstream solution .", "to improve ic performance further , other product metadata , e.g.", ", product images , have been used .", "although multimodal ic ( mic ) systems show higher performance , expanding from processing text to more resource - demanding images brings large engineering impacts and hinders the deployment of such dual - input mic systems ."], "relation": "used for", "id": "2022.ecnlp-1.25", "year": 2022, "rel_sent": "In this paper , we proposed a new way of using product images to improve text - only IC model : leveraging cross - modal signals between products ' titles and associated images to adapt BERT models in a self - supervised learning ( SSL ) way .", "forward": true, "src_ids": "2022.ecnlp-1.25_728"}
+{"input": "generative extraction is done by using OtherScientificTerm| context: named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence . depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner . among the existing approaches , only the generative model can be uniformly adapted to these three subtasks . however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases .", "entity": "generative extraction", "output": "de - bias", "neg_sample": ["generative extraction is done by using OtherScientificTerm", "named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence .", "depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner .", "among the existing approaches , only the generative model can be uniformly adapted to these three subtasks .", "however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases ."], "relation": "used for", "id": "2022.acl-long.59", "year": 2022, "rel_sent": "De - Bias for Generative Extraction in Unified NER Task.", "forward": false, "src_ids": "2022.acl-long.59_729"}
+{"input": "unified ner task is done by using Task| context: named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence . depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner . among the existing approaches , only the generative model can be uniformly adapted to these three subtasks . however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases .", "entity": "unified ner task", "output": "generative extraction", "neg_sample": ["unified ner task is done by using Task", "named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence .", "depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner .", "among the existing approaches , only the generative model can be uniformly adapted to these three subtasks .", "however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases ."], "relation": "used for", "id": "2022.acl-long.59", "year": 2022, "rel_sent": "De - Bias for Generative Extraction in Unified NER Task.", "forward": false, "src_ids": "2022.acl-long.59_730"}
+{"input": "de - bias is used for Task| context: named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence . depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner . among the existing approaches , only the generative model can be uniformly adapted to these three subtasks . however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases .", "entity": "de - bias", "output": "generative extraction", "neg_sample": ["de - bias is used for Task", "named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence .", "depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner .", "among the existing approaches , only the generative model can be uniformly adapted to these three subtasks .", "however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases ."], "relation": "used for", "id": "2022.acl-long.59", "year": 2022, "rel_sent": "De - Bias for Generative Extraction in Unified NER Task.", "forward": true, "src_ids": "2022.acl-long.59_731"}
+{"input": "generative extraction is used for Task| context: named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence . depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner . among the existing approaches , only the generative model can be uniformly adapted to these three subtasks . however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases .", "entity": "generative extraction", "output": "unified ner task", "neg_sample": ["generative extraction is used for Task", "named entity recognition ( ner ) is a fundamental task to recognize specific types of entities from a given sentence .", "depending on how the entities appear in the sentence , it can be divided into three subtasks , namely , flat ner , nested ner , and discontinuous ner .", "among the existing approaches , only the generative model can be uniformly adapted to these three subtasks .", "however , when the generative model is applied to ner , its optimization objective is not consistent with the task , which makes the model vulnerable to the incorrect biases ."], "relation": "used for", "id": "2022.acl-long.59", "year": 2022, "rel_sent": "De - Bias for Generative Extraction in Unified NER Task.", "forward": true, "src_ids": "2022.acl-long.59_732"}
+{"input": "nlp is done by using Method| context: the mainstream machine learning paradigms for nlp often work with two underlying presumptions . first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively . second , the supervision of a task mainly comes from a set of labeled examples .", "entity": "nlp", "output": "continual learning from task instructions", "neg_sample": ["nlp is done by using Method", "the mainstream machine learning paradigms for nlp often work with two underlying presumptions .", "first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively .", "second , the supervision of a task mainly comes from a set of labeled examples ."], "relation": "used for", "id": "2022.acl-long.218", "year": 2022, "rel_sent": "To our knowledge , this is the first time to study ConTinTin in NLP .", "forward": false, "src_ids": "2022.acl-long.218_733"}
+{"input": "continual learning from task instructions is used for Task| context: first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively . second , the supervision of a task mainly comes from a set of labeled examples .", "entity": "continual learning from task instructions", "output": "nlp", "neg_sample": ["continual learning from task instructions is used for Task", "first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively .", "second , the supervision of a task mainly comes from a set of labeled examples ."], "relation": "used for", "id": "2022.acl-long.218", "year": 2022, "rel_sent": "To our knowledge , this is the first time to study ConTinTin in NLP .", "forward": true, "src_ids": "2022.acl-long.218_734"}
+{"input": "task instructions is used for OtherScientificTerm| context: the mainstream machine learning paradigms for nlp often work with two underlying presumptions . first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively . second , the supervision of a task mainly comes from a set of labeled examples .", "entity": "task instructions", "output": "forward - transfer", "neg_sample": ["task instructions is used for OtherScientificTerm", "the mainstream machine learning paradigms for nlp often work with two underlying presumptions .", "first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively .", "second , the supervision of a task mainly comes from a set of labeled examples ."], "relation": "used for", "id": "2022.acl-long.218", "year": 2022, "rel_sent": "Technically , our method InstructionSpeak contains two strategies that make full use of task instructions to improve forward - transfer and backward - transfer : one is to learn from negative outputs , the other is to re - visit instructions of previous tasks .", "forward": true, "src_ids": "2022.acl-long.218_735"}
+{"input": "forward - transfer is done by using OtherScientificTerm| context: the mainstream machine learning paradigms for nlp often work with two underlying presumptions . first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively . second , the supervision of a task mainly comes from a set of labeled examples .", "entity": "forward - transfer", "output": "task instructions", "neg_sample": ["forward - transfer is done by using OtherScientificTerm", "the mainstream machine learning paradigms for nlp often work with two underlying presumptions .", "first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively .", "second , the supervision of a task mainly comes from a set of labeled examples ."], "relation": "used for", "id": "2022.acl-long.218", "year": 2022, "rel_sent": "Technically , our method InstructionSpeak contains two strategies that make full use of task instructions to improve forward - transfer and backward - transfer : one is to learn from negative outputs , the other is to re - visit instructions of previous tasks .", "forward": false, "src_ids": "2022.acl-long.218_736"}
+{"input": "backward - transfer is done by using OtherScientificTerm| context: the mainstream machine learning paradigms for nlp often work with two underlying presumptions . first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively . second , the supervision of a task mainly comes from a set of labeled examples .", "entity": "backward - transfer", "output": "task instructions", "neg_sample": ["backward - transfer is done by using OtherScientificTerm", "the mainstream machine learning paradigms for nlp often work with two underlying presumptions .", "first , the target task is predefined and static ; a system merely needs to learn to solve it exclusively .", "second , the supervision of a task mainly comes from a set of labeled examples ."], "relation": "used for", "id": "2022.acl-long.218", "year": 2022, "rel_sent": "Technically , our method InstructionSpeak contains two strategies that make full use of task instructions to improve forward - transfer and backward - transfer : one is to learn from negative outputs , the other is to re - visit instructions of previous tasks .", "forward": false, "src_ids": "2022.acl-long.218_737"}
+{"input": "theory - driven framework is used for OtherScientificTerm| context: previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human - like interactions .", "entity": "theory - driven framework", "output": "sarcastic responses", "neg_sample": ["theory - driven framework is used for OtherScientificTerm", "previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human - like interactions ."], "relation": "used for", "id": "2022.acl-long.530", "year": 2022, "rel_sent": "Next , we use a theory - driven framework for generating sarcastic responses , which allows us to control the linguistic devices included during generation .", "forward": true, "src_ids": "2022.acl-long.530_738"}
+{"input": "sarcastic responses is done by using Method| context: previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human - like interactions .", "entity": "sarcastic responses", "output": "theory - driven framework", "neg_sample": ["sarcastic responses is done by using Method", "previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human - like interactions ."], "relation": "used for", "id": "2022.acl-long.530", "year": 2022, "rel_sent": "Next , we use a theory - driven framework for generating sarcastic responses , which allows us to control the linguistic devices included during generation .", "forward": false, "src_ids": "2022.acl-long.530_739"}
+{"input": "generation is done by using OtherScientificTerm| context: previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human - like interactions .", "entity": "generation", "output": "linguistic devices", "neg_sample": ["generation is done by using OtherScientificTerm", "previous sarcasm generation research has focused on how to generate text that people perceive as sarcastic to create more human - like interactions ."], "relation": "used for", "id": "2022.acl-long.530", "year": 2022, "rel_sent": "Next , we use a theory - driven framework for generating sarcastic responses , which allows us to control the linguistic devices included during generation .", "forward": false, "src_ids": "2022.acl-long.530_740"}
+{"input": "image - based model vgg16 is used for Task| context: social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities . one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it . the situation is more complex for multilingual(e.g . , tamil ) memes due to the lack of benchmark datasets and models .", "entity": "image - based model vgg16", "output": "troll - meme classification", "neg_sample": ["image - based model vgg16 is used for Task", "social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities .", "one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it .", "the situation is more complex for multilingual(e.g .", ", tamil ) memes due to the lack of benchmark datasets and models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.8", "year": 2022, "rel_sent": "We observe while the text - based model MURIL performs better for Non - troll meme classification , the image - based model VGG16 performs better for Troll - meme classification .", "forward": true, "src_ids": "2022.dravidianlangtech-1.8_741"}
+{"input": "non - troll meme classification is done by using Method| context: social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities . one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it . the situation is more complex for multilingual(e.g . , tamil ) memes due to the lack of benchmark datasets and models .", "entity": "non - troll meme classification", "output": "text - based model", "neg_sample": ["non - troll meme classification is done by using Method", "social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities .", "one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it .", "the situation is more complex for multilingual(e.g .", ", tamil ) memes due to the lack of benchmark datasets and models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.8", "year": 2022, "rel_sent": "We observe while the text - based model MURIL performs better for Non - troll meme classification , the image - based model VGG16 performs better for Troll - meme classification .", "forward": false, "src_ids": "2022.dravidianlangtech-1.8_742"}
+{"input": "text - based model is used for Task| context: social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities . one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it . the situation is more complex for multilingual(e.g . , tamil ) memes due to the lack of benchmark datasets and models .", "entity": "text - based model", "output": "non - troll meme classification", "neg_sample": ["text - based model is used for Task", "social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities .", "one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it .", "the situation is more complex for multilingual(e.g .", ", tamil ) memes due to the lack of benchmark datasets and models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.8", "year": 2022, "rel_sent": "We observe while the text - based model MURIL performs better for Non - troll meme classification , the image - based model VGG16 performs better for Troll - meme classification .", "forward": true, "src_ids": "2022.dravidianlangtech-1.8_743"}
+{"input": "troll - meme classification is done by using Method| context: social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities . one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it . the situation is more complex for multilingual(e.g . , tamil ) memes due to the lack of benchmark datasets and models .", "entity": "troll - meme classification", "output": "image - based model vgg16", "neg_sample": ["troll - meme classification is done by using Method", "social media platforms often act as breeding grounds for various forms of trolling or malicious content targeting users or communities .", "one way of trolling users is by creating memes , which in most cases unites an image with a short piece of text embedded on top of it .", "the situation is more complex for multilingual(e.g .", ", tamil ) memes due to the lack of benchmark datasets and models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.8", "year": 2022, "rel_sent": "We observe while the text - based model MURIL performs better for Non - troll meme classification , the image - based model VGG16 performs better for Troll - meme classification .", "forward": false, "src_ids": "2022.dravidianlangtech-1.8_744"}
+{"input": "crosslinguistic low - resource morphological segmentation is done by using Task| context: common designs of model evaluation typically focus on monolingual settings , where different models are compared according to their performance on a single data set that is assumed to be representative of all possible data for the task at hand . while this may be reasonable for a large data set , this assumption is difficult to maintain in low - resource scenarios , where artifacts of the data collection can yield data sets that are outliers , potentially making conclusions about model performance coincidental .", "entity": "crosslinguistic low - resource morphological segmentation", "output": "data - driven model generalizability", "neg_sample": ["crosslinguistic low - resource morphological segmentation is done by using Task", "common designs of model evaluation typically focus on monolingual settings , where different models are compared according to their performance on a single data set that is assumed to be representative of all possible data for the task at hand .", "while this may be reasonable for a large data set , this assumption is difficult to maintain in low - resource scenarios , where artifacts of the data collection can yield data sets that are outliers , potentially making conclusions about model performance coincidental ."], "relation": "used for", "id": "2022.tacl-1.23", "year": 2022, "rel_sent": "Data - driven Model Generalizability in Crosslinguistic Low - resource Morphological Segmentation.", "forward": false, "src_ids": "2022.tacl-1.23_745"}
+{"input": "data - driven model generalizability is used for Task| context: common designs of model evaluation typically focus on monolingual settings , where different models are compared according to their performance on a single data set that is assumed to be representative of all possible data for the task at hand . while this may be reasonable for a large data set , this assumption is difficult to maintain in low - resource scenarios , where artifacts of the data collection can yield data sets that are outliers , potentially making conclusions about model performance coincidental .", "entity": "data - driven model generalizability", "output": "crosslinguistic low - resource morphological segmentation", "neg_sample": ["data - driven model generalizability is used for Task", "common designs of model evaluation typically focus on monolingual settings , where different models are compared according to their performance on a single data set that is assumed to be representative of all possible data for the task at hand .", "while this may be reasonable for a large data set , this assumption is difficult to maintain in low - resource scenarios , where artifacts of the data collection can yield data sets that are outliers , potentially making conclusions about model performance coincidental ."], "relation": "used for", "id": "2022.tacl-1.23", "year": 2022, "rel_sent": "Data - driven Model Generalizability in Crosslinguistic Low - resource Morphological Segmentation.", "forward": true, "src_ids": "2022.tacl-1.23_746"}
+{"input": "automatic metrics is used for Task| context: visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain .", "entity": "automatic metrics", "output": "text generation", "neg_sample": ["automatic metrics is used for Task", "visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain ."], "relation": "used for", "id": "2022.acl-long.441", "year": 2022, "rel_sent": "We first show that the results from commonly adopted automatic metrics for text generation have little correlation with those obtained from human evaluation , which motivates us to directly utilize human evaluation results to learn the automatic evaluation model .", "forward": true, "src_ids": "2022.acl-long.441_747"}
+{"input": "story evaluation is done by using Metric| context: visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain . however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist .", "entity": "story evaluation", "output": "reference - free vist metric", "neg_sample": ["story evaluation is done by using Metric", "visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain .", "however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist ."], "relation": "used for", "id": "2022.acl-long.441", "year": 2022, "rel_sent": "In this paper , we present the VHED ( VIST Human Evaluation Data ) dataset , which first re - purposes human evaluation results for automatic evaluation ; hence we develop Vrank ( VIST Ranker ) , a novel reference - free VIST metric for story evaluation .", "forward": false, "src_ids": "2022.acl-long.441_748"}
+{"input": "reference - free vist metric is used for Task| context: visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain . however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist .", "entity": "reference - free vist metric", "output": "story evaluation", "neg_sample": ["reference - free vist metric is used for Task", "visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain .", "however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist ."], "relation": "used for", "id": "2022.acl-long.441", "year": 2022, "rel_sent": "In this paper , we present the VHED ( VIST Human Evaluation Data ) dataset , which first re - purposes human evaluation results for automatic evaluation ; hence we develop Vrank ( VIST Ranker ) , a novel reference - free VIST metric for story evaluation .", "forward": true, "src_ids": "2022.acl-long.441_749"}
+{"input": "text generation is done by using Metric| context: visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain . however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist .", "entity": "text generation", "output": "automatic metrics", "neg_sample": ["text generation is done by using Metric", "visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain .", "however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist ."], "relation": "used for", "id": "2022.acl-long.441", "year": 2022, "rel_sent": "We first show that the results from commonly adopted automatic metrics for text generation have little correlation with those obtained from human evaluation , which motivates us to directly utilize human evaluation results to learn the automatic evaluation model .", "forward": false, "src_ids": "2022.acl-long.441_750"}
+{"input": "automatic evaluation model is done by using OtherScientificTerm| context: visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain . however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist .", "entity": "automatic evaluation model", "output": "human evaluation results", "neg_sample": ["automatic evaluation model is done by using OtherScientificTerm", "visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain .", "however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist ."], "relation": "used for", "id": "2022.acl-long.441", "year": 2022, "rel_sent": "We first show that the results from commonly adopted automatic metrics for text generation have little correlation with those obtained from human evaluation , which motivates us to directly utilize human evaluation results to learn the automatic evaluation model .", "forward": false, "src_ids": "2022.acl-long.441_751"}
+{"input": "human evaluation results is used for Method| context: visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain . however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist .", "entity": "human evaluation results", "output": "automatic evaluation model", "neg_sample": ["human evaluation results is used for Method", "visual storytelling ( vist ) is a typical vision and language task that has seen extensive development in the natural language generation research domain .", "however , it remains unclear whether conventional automatic evaluation metrics for text generation are applicable on vist ."], "relation": "used for", "id": "2022.acl-long.441", "year": 2022, "rel_sent": "We first show that the results from commonly adopted automatic metrics for text generation have little correlation with those obtained from human evaluation , which motivates us to directly utilize human evaluation results to learn the automatic evaluation model .", "forward": true, "src_ids": "2022.acl-long.441_752"}
+{"input": "text spans is done by using Method| context: higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level .", "entity": "text spans", "output": "text span proposal module", "neg_sample": ["text spans is done by using Method", "higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level ."], "relation": "used for", "id": "2022.acl-long.173", "year": 2022, "rel_sent": "It consists of two modules : the text span proposal module which proposes candidate text spans , each of which represents a subtree in the dependency tree denoted by ( root , start , end ) ; and the span linking module , which constructs links between proposed spans .", "forward": false, "src_ids": "2022.acl-long.173_753"}
+{"input": "text span proposal module is used for OtherScientificTerm| context: higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level .", "entity": "text span proposal module", "output": "text spans", "neg_sample": ["text span proposal module is used for OtherScientificTerm", "higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level ."], "relation": "used for", "id": "2022.acl-long.173", "year": 2022, "rel_sent": "It consists of two modules : the text span proposal module which proposes candidate text spans , each of which represents a subtree in the dependency tree denoted by ( root , start , end ) ; and the span linking module , which constructs links between proposed spans .", "forward": true, "src_ids": "2022.acl-long.173_754"}
+{"input": "missing spans is done by using Method| context: higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level .", "entity": "missing spans", "output": "mrc framework", "neg_sample": ["missing spans is done by using Method", "higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level ."], "relation": "used for", "id": "2022.acl-long.173", "year": 2022, "rel_sent": "The proposed method has the following merits : ( 1 ) it addresses the fundamental problem that edges in a dependency tree should be constructed between subtrees ; ( 2 ) the MRC framework allows the method to retrieve missing spans in the span proposal stage , which leads to higher recall for eligible spans .", "forward": false, "src_ids": "2022.acl-long.173_755"}
+{"input": "mrc framework is used for OtherScientificTerm| context: higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level .", "entity": "mrc framework", "output": "missing spans", "neg_sample": ["mrc framework is used for OtherScientificTerm", "higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level ."], "relation": "used for", "id": "2022.acl-long.173", "year": 2022, "rel_sent": "The proposed method has the following merits : ( 1 ) it addresses the fundamental problem that edges in a dependency tree should be constructed between subtrees ; ( 2 ) the MRC framework allows the method to retrieve missing spans in the span proposal stage , which leads to higher recall for eligible spans .", "forward": true, "src_ids": "2022.acl-long.173_756"}
+{"input": "fine - grained representations is used for Task| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept .", "entity": "fine - grained representations", "output": "automatic biomedical term clustering", "neg_sample": ["fine - grained representations is used for Task", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": true, "src_ids": "2022.bionlp-1.8_757"}
+{"input": "dynamic hard positive and negative samples is used for Method| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "dynamic hard positive and negative samples", "output": "contrastive learning", "neg_sample": ["dynamic hard positive and negative samples is used for Method", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": true, "src_ids": "2022.bionlp-1.8_758"}
+{"input": "pretraining term embeddings is done by using Method| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "pretraining term embeddings", "output": "sampling strategy", "neg_sample": ["pretraining term embeddings is done by using Method", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": false, "src_ids": "2022.bionlp-1.8_759"}
+{"input": "fine - grained representations is done by using Method| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "fine - grained representations", "output": "sampling strategy", "neg_sample": ["fine - grained representations is done by using Method", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": false, "src_ids": "2022.bionlp-1.8_760"}
+{"input": "sampling strategy is used for Method| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "sampling strategy", "output": "pretraining term embeddings", "neg_sample": ["sampling strategy is used for Method", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": true, "src_ids": "2022.bionlp-1.8_761"}
+{"input": "fine - grained representations is done by using OtherScientificTerm| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "fine - grained representations", "output": "dynamic hard positive and negative samples", "neg_sample": ["fine - grained representations is done by using OtherScientificTerm", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": false, "src_ids": "2022.bionlp-1.8_762"}
+{"input": "contrastive learning is done by using OtherScientificTerm| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "contrastive learning", "output": "dynamic hard positive and negative samples", "neg_sample": ["contrastive learning is done by using OtherScientificTerm", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": false, "src_ids": "2022.bionlp-1.8_763"}
+{"input": "automatic biomedical term clustering is done by using Method| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "automatic biomedical term clustering", "output": "fine - grained representations", "neg_sample": ["automatic biomedical term clustering is done by using Method", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": false, "src_ids": "2022.bionlp-1.8_764"}
+{"input": "dynamic hard positive and negative samples is used for Method| context: term clustering is important in biomedical knowledge graph construction . using similarities between terms embedding is helpful for term clustering . state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning . these embeddings provide close embeddings for terms belonging to the same concept . however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering .", "entity": "dynamic hard positive and negative samples", "output": "fine - grained representations", "neg_sample": ["dynamic hard positive and negative samples is used for Method", "term clustering is important in biomedical knowledge graph construction .", "using similarities between terms embedding is helpful for term clustering .", "state - of - the - art term embeddings leverage pretrained language models to encode terms , and use synonyms and relation knowledge from knowledge graphs to guide contrastive learning .", "these embeddings provide close embeddings for terms belonging to the same concept .", "however , from our probing experiments , these embeddings are not sensitive to minor textual differences which leads tofailure for biomedical term clustering ."], "relation": "used for", "id": "2022.bionlp-1.8", "year": 2022, "rel_sent": "To alleviate this problem , we adjust the sampling strategy in pretraining term embeddings by providing dynamic hard positive and negative samples during contrastive learning to learn fine - grained representations which result in better biomedical term clustering .", "forward": true, "src_ids": "2022.bionlp-1.8_765"}
+{"input": "comparable corpora is used for Task| context: the past few years have seen the multiplication of studies on post - editese , following the massive adoption of post - editing in professional translation workflows . these studies mainly rely on the comparison of post - edited machine translation and human translation on artificial parallel corpora . by contrast , we investigate here post - editese on comparable corpora of authentic translation jobs for the language direction english intofrench .", "entity": "comparable corpora", "output": "post - editese research", "neg_sample": ["comparable corpora is used for Task", "the past few years have seen the multiplication of studies on post - editese , following the massive adoption of post - editing in professional translation workflows .", "these studies mainly rely on the comparison of post - edited machine translation and human translation on artificial parallel corpora .", "by contrast , we investigate here post - editese on comparable corpora of authentic translation jobs for the language direction english intofrench ."], "relation": "used for", "id": "2022.eamt-1.10", "year": 2022, "rel_sent": "Finally , our study highlights some of the challenges of working with comparable corpora in post - editese research .", "forward": true, "src_ids": "2022.eamt-1.10_766"}
+{"input": "post - editese research is done by using Material| context: the past few years have seen the multiplication of studies on post - editese , following the massive adoption of post - editing in professional translation workflows . these studies mainly rely on the comparison of post - edited machine translation and human translation on artificial parallel corpora .", "entity": "post - editese research", "output": "comparable corpora", "neg_sample": ["post - editese research is done by using Material", "the past few years have seen the multiplication of studies on post - editese , following the massive adoption of post - editing in professional translation workflows .", "these studies mainly rely on the comparison of post - edited machine translation and human translation on artificial parallel corpora ."], "relation": "used for", "id": "2022.eamt-1.10", "year": 2022, "rel_sent": "Finally , our study highlights some of the challenges of working with comparable corpora in post - editese research .", "forward": false, "src_ids": "2022.eamt-1.10_767"}
+{"input": "parallel data is used for Task| context: quality assessment has been an ongoing activity of the series of paracrawl efforts to crawl massive amounts of parallel data from multilingual websites for 29 languages .", "entity": "parallel data", "output": "machine translation", "neg_sample": ["parallel data is used for Task", "quality assessment has been an ongoing activity of the series of paracrawl efforts to crawl massive amounts of parallel data from multilingual websites for 29 languages ."], "relation": "used for", "id": "2022.humeval-1.4", "year": 2022, "rel_sent": "The goal of ParaCrawl is to get parallel data that is good for machine translation .", "forward": true, "src_ids": "2022.humeval-1.4_768"}
+{"input": "helpfulness prediction is done by using OtherScientificTerm| context: over the years , the review helpfulness prediction task has been the subject of several works , but remains being a challenging issue in natural language processing , as results vary a lot depending on the domain , on the adopted features and on the chosen classification strategy .", "entity": "helpfulness prediction", "output": "content features", "neg_sample": ["helpfulness prediction is done by using OtherScientificTerm", "over the years , the review helpfulness prediction task has been the subject of several works , but remains being a challenging issue in natural language processing , as results vary a lot depending on the domain , on the adopted features and on the chosen classification strategy ."], "relation": "used for", "id": "2022.wassa-1.19", "year": 2022, "rel_sent": "Evaluating Content Features and Classification Methods for Helpfulness Prediction of Online Reviews : Establishing a Benchmark for Portuguese.", "forward": false, "src_ids": "2022.wassa-1.19_769"}
+{"input": "helpfulness prediction is done by using Method| context: over the years , the review helpfulness prediction task has been the subject of several works , but remains being a challenging issue in natural language processing , as results vary a lot depending on the domain , on the adopted features and on the chosen classification strategy .", "entity": "helpfulness prediction", "output": "classification methods", "neg_sample": ["helpfulness prediction is done by using Method", "over the years , the review helpfulness prediction task has been the subject of several works , but remains being a challenging issue in natural language processing , as results vary a lot depending on the domain , on the adopted features and on the chosen classification strategy ."], "relation": "used for", "id": "2022.wassa-1.19", "year": 2022, "rel_sent": "Evaluating Content Features and Classification Methods for Helpfulness Prediction of Online Reviews : Establishing a Benchmark for Portuguese.", "forward": false, "src_ids": "2022.wassa-1.19_770"}
+{"input": "helpfulness prediction is done by using Method| context: over the years , the review helpfulness prediction task has been the subject of several works , but remains being a challenging issue in natural language processing , as results vary a lot depending on the domain , on the adopted features and on the chosen classification strategy .", "entity": "helpfulness prediction", "output": "classification methods", "neg_sample": ["helpfulness prediction is done by using Method", "over the years , the review helpfulness prediction task has been the subject of several works , but remains being a challenging issue in natural language processing , as results vary a lot depending on the domain , on the adopted features and on the chosen classification strategy ."], "relation": "used for", "id": "2022.wassa-1.19", "year": 2022, "rel_sent": "We show that simple features and classical classification methods are powerful for the task of helpfulness prediction , but are largely outperformed by a convolutional neural network - based solution .", "forward": false, "src_ids": "2022.wassa-1.19_771"}
+{"input": "knowledgedriven discussions is done by using Task| context: in this work we study giving access to this information to conversational agents . large language models , even though they store an impressive amount of knowledge within their weights , are known to hallucinate facts when generating dialogue ( shuster et al . , 2021 ) ; moreover , those facts are frozen in time at the point of model training .", "entity": "knowledgedriven discussions", "output": "internet search", "neg_sample": ["knowledgedriven discussions is done by using Task", "in this work we study giving access to this information to conversational agents .", "large language models , even though they store an impressive amount of knowledge within their weights , are known to hallucinate facts when generating dialogue ( shuster et al .", ", 2021 ) ; moreover , those facts are frozen in time at the point of model training ."], "relation": "used for", "id": "2022.acl-long.579", "year": 2022, "rel_sent": "We train and evaluate such models on a newly collected dataset of human - human conversations whereby one of the speakers is given access to internet search during knowledgedriven discussions in order to ground their responses .", "forward": false, "src_ids": "2022.acl-long.579_772"}
+{"input": "internet search is used for Task| context: the largest store of continually updating knowledge on our planet can be accessed via internet search . in this work we study giving access to this information to conversational agents . large language models , even though they store an impressive amount of knowledge within their weights , are known to hallucinate facts when generating dialogue ( shuster et al . , 2021 ) ; moreover , those facts are frozen in time at the point of model training .", "entity": "internet search", "output": "knowledgedriven discussions", "neg_sample": ["internet search is used for Task", "the largest store of continually updating knowledge on our planet can be accessed via internet search .", "in this work we study giving access to this information to conversational agents .", "large language models , even though they store an impressive amount of knowledge within their weights , are known to hallucinate facts when generating dialogue ( shuster et al .", ", 2021 ) ; moreover , those facts are frozen in time at the point of model training ."], "relation": "used for", "id": "2022.acl-long.579", "year": 2022, "rel_sent": "We train and evaluate such models on a newly collected dataset of human - human conversations whereby one of the speakers is given access to internet search during knowledgedriven discussions in order to ground their responses .", "forward": true, "src_ids": "2022.acl-long.579_773"}
+{"input": "ood intent discovery is done by using Method| context: discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system . the key challenge is how to transfer prior ind knowledge to ood clustering .", "entity": "ood intent discovery", "output": "disentangled knowledge transfer method", "neg_sample": ["ood intent discovery is done by using Method", "discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system .", "the key challenge is how to transfer prior ind knowledge to ood clustering ."], "relation": "used for", "id": "2022.acl-short.6", "year": 2022, "rel_sent": "Disentangled Knowledge Transfer for OOD Intent Discovery with Unified Contrastive Learning.", "forward": false, "src_ids": "2022.acl-short.6_774"}
+{"input": "shared intent representation is used for Method| context: discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system . the key challenge is how to transfer prior ind knowledge to ood clustering .", "entity": "shared intent representation", "output": "disentangled knowledge transfer method", "neg_sample": ["shared intent representation is used for Method", "discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system .", "the key challenge is how to transfer prior ind knowledge to ood clustering ."], "relation": "used for", "id": "2022.acl-short.6", "year": 2022, "rel_sent": "Different from existing work based on shared intent representation , we propose a novel disentangled knowledge transfer method via a unified multi - head contrastive learning framework .", "forward": true, "src_ids": "2022.acl-short.6_775"}
+{"input": "disentangled knowledge transfer method is used for Task| context: discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system . the key challenge is how to transfer prior ind knowledge to ood clustering .", "entity": "disentangled knowledge transfer method", "output": "ood intent discovery", "neg_sample": ["disentangled knowledge transfer method is used for Task", "discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system .", "the key challenge is how to transfer prior ind knowledge to ood clustering ."], "relation": "used for", "id": "2022.acl-short.6", "year": 2022, "rel_sent": "Disentangled Knowledge Transfer for OOD Intent Discovery with Unified Contrastive Learning.", "forward": true, "src_ids": "2022.acl-short.6_776"}
+{"input": "disentangled knowledge transfer method is done by using Method| context: discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system . the key challenge is how to transfer prior ind knowledge to ood clustering .", "entity": "disentangled knowledge transfer method", "output": "shared intent representation", "neg_sample": ["disentangled knowledge transfer method is done by using Method", "discovering out - of - domain(ood ) intents is essential for developing new skills in a task - oriented dialogue system .", "the key challenge is how to transfer prior ind knowledge to ood clustering ."], "relation": "used for", "id": "2022.acl-short.6", "year": 2022, "rel_sent": "Different from existing work based on shared intent representation , we propose a novel disentangled knowledge transfer method via a unified multi - head contrastive learning framework .", "forward": false, "src_ids": "2022.acl-short.6_777"}
+{"input": "spoken language tasks is done by using Method| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "spoken language tasks", "output": "mtl - slt", "neg_sample": ["spoken language tasks is done by using Method", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "Tofacilitate the development of spoken language research , we introduce MTL - SLT , a multi - task learning framework for spoken language tasks .", "forward": false, "src_ids": "2022.nlp4convai-1.11_778"}
+{"input": "spoken language understanding is done by using Method| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "spoken language understanding", "output": "mtl - slt", "neg_sample": ["spoken language understanding is done by using Method", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "MTL - SLT takes speech as input , and outputs transcription , intent , named entities , summaries , and answers to text queries , supporting the tasks of spoken language understanding , spoken summarization and spoken question answering respectively .", "forward": false, "src_ids": "2022.nlp4convai-1.11_779"}
+{"input": "spoken question answering is done by using Method| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "spoken question answering", "output": "mtl - slt", "neg_sample": ["spoken question answering is done by using Method", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "MTL - SLT takes speech as input , and outputs transcription , intent , named entities , summaries , and answers to text queries , supporting the tasks of spoken language understanding , spoken summarization and spoken question answering respectively .", "forward": false, "src_ids": "2022.nlp4convai-1.11_780"}
+{"input": "spoken summarization is done by using Method| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "spoken summarization", "output": "mtl - slt", "neg_sample": ["spoken summarization is done by using Method", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "MTL - SLT takes speech as input , and outputs transcription , intent , named entities , summaries , and answers to text queries , supporting the tasks of spoken language understanding , spoken summarization and spoken question answering respectively .", "forward": false, "src_ids": "2022.nlp4convai-1.11_781"}
+{"input": "multi - task learning framework is used for Task| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "multi - task learning framework", "output": "spoken language tasks", "neg_sample": ["multi - task learning framework is used for Task", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "Tofacilitate the development of spoken language research , we introduce MTL - SLT , a multi - task learning framework for spoken language tasks .", "forward": true, "src_ids": "2022.nlp4convai-1.11_782"}
+{"input": "mtl - slt is used for Task| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "mtl - slt", "output": "spoken language tasks", "neg_sample": ["mtl - slt is used for Task", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "Tofacilitate the development of spoken language research , we introduce MTL - SLT , a multi - task learning framework for spoken language tasks .", "forward": true, "src_ids": "2022.nlp4convai-1.11_783"}
+{"input": "mtl - slt is used for Task| context: language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications . prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task .", "entity": "mtl - slt", "output": "spoken language understanding", "neg_sample": ["mtl - slt is used for Task", "language understanding in speech - based systems has attracted extensive interest from both academic and industrial communities in recent years with the growing demand for voice - based applications .", "prior works focus on independent research by the automatic speech recognition ( asr ) and natural language processing ( nlp ) communities , or on jointly modeling the speech and nlp problems focusing on a single dataset or single nlp task ."], "relation": "used for", "id": "2022.nlp4convai-1.11", "year": 2022, "rel_sent": "MTL - SLT takes speech as input , and outputs transcription , intent , named entities , summaries , and answers to text queries , supporting the tasks of spoken language understanding , spoken summarization and spoken question answering respectively .", "forward": true, "src_ids": "2022.nlp4convai-1.11_784"}
+{"input": "sentence compression models is done by using Method| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality . recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time .", "entity": "sentence compression models", "output": "reinforcement learning", "neg_sample": ["sentence compression models is done by using Method", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "In this work , we explore the use of reinforcement learning to train effective sentence compression models that are alsofast when generating predictions .", "forward": false, "src_ids": "2022.acl-long.90_785"}
+{"input": "sentence compression is done by using Method| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality . recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time .", "entity": "sentence compression", "output": "unsupervised objective driven methods", "neg_sample": ["sentence compression is done by using Method", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "Unsupervised objective driven methods for sentence compression can be used to create customized models without the need for ground - truth training data , while allowing flexibility in the objective function(s ) that are used for learning and inference .", "forward": false, "src_ids": "2022.acl-long.90_786"}
+{"input": "inference is done by using OtherScientificTerm| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality . recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time .", "entity": "inference", "output": "objective function(s )", "neg_sample": ["inference is done by using OtherScientificTerm", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "Unsupervised objective driven methods for sentence compression can be used to create customized models without the need for ground - truth training data , while allowing flexibility in the objective function(s ) that are used for learning and inference .", "forward": false, "src_ids": "2022.acl-long.90_787"}
+{"input": "learning is done by using OtherScientificTerm| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality . recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time .", "entity": "learning", "output": "objective function(s )", "neg_sample": ["learning is done by using OtherScientificTerm", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "Unsupervised objective driven methods for sentence compression can be used to create customized models without the need for ground - truth training data , while allowing flexibility in the objective function(s ) that are used for learning and inference .", "forward": false, "src_ids": "2022.acl-long.90_788"}
+{"input": "objective function(s ) is used for Task| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality . recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time .", "entity": "objective function(s )", "output": "learning", "neg_sample": ["objective function(s ) is used for Task", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "Unsupervised objective driven methods for sentence compression can be used to create customized models without the need for ground - truth training data , while allowing flexibility in the objective function(s ) that are used for learning and inference .", "forward": true, "src_ids": "2022.acl-long.90_789"}
+{"input": "objective function(s ) is used for Task| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "entity": "objective function(s )", "output": "inference", "neg_sample": ["objective function(s ) is used for Task", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "Unsupervised objective driven methods for sentence compression can be used to create customized models without the need for ground - truth training data , while allowing flexibility in the objective function(s ) that are used for learning and inference .", "forward": true, "src_ids": "2022.acl-long.90_790"}
+{"input": "reinforcement learning is used for Method| context: sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality . recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time .", "entity": "reinforcement learning", "output": "sentence compression models", "neg_sample": ["reinforcement learning is used for Method", "sentence compression reduces the length of text by removing non - essential content while preserving important facts and grammaticality .", "recent unsupervised sentence compression approaches use custom objectives to guide discrete search ; however , guided search is expensive at inference time ."], "relation": "used for", "id": "2022.acl-long.90", "year": 2022, "rel_sent": "In this work , we explore the use of reinforcement learning to train effective sentence compression models that are alsofast when generating predictions .", "forward": true, "src_ids": "2022.acl-long.90_791"}
+{"input": "learning is done by using Method| context: hedges have an important role in the management of rapport .", "entity": "learning", "output": "tutoring agent", "neg_sample": ["learning is done by using Method", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "In peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback . Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.153_792"}
+{"input": "rapport is done by using Method| context: hedges have an important role in the management of rapport .", "entity": "rapport", "output": "tutoring agent", "neg_sample": ["rapport is done by using Method", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "In peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback . Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.153_793"}
+{"input": "tutoring agent is used for Task| context: hedges have an important role in the management of rapport .", "entity": "tutoring agent", "output": "learning", "neg_sample": ["tutoring agent is used for Task", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "In peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback . Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": true, "src_ids": "2022.acl-long.153_794"}
+{"input": "computational framework is done by using Material| context: hedges have an important role in the management of rapport .", "entity": "computational framework", "output": "multimodal peer - tutoring dataset", "neg_sample": ["computational framework is done by using Material", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "In peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback . Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.153_795"}
+{"input": "hedges is done by using Method| context: hedges have an important role in the management of rapport .", "entity": "hedges", "output": "computational framework", "neg_sample": ["hedges is done by using Method", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "In peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback . Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.153_796"}
+{"input": "multimodal peer - tutoring dataset is used for Method| context: hedges have an important role in the management of rapport .", "entity": "multimodal peer - tutoring dataset", "output": "computational framework", "neg_sample": ["multimodal peer - tutoring dataset is used for Method", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "In peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback . Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": true, "src_ids": "2022.acl-long.153_797"}
+{"input": "features is done by using Method| context: hedges have an important role in the management of rapport .", "entity": "features", "output": "model explainability tool", "neg_sample": ["features is done by using Method", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "We employ a model explainability tool to explore the features that characterize hedges in peer - tutoring conversations , and we identify some novel features , and the benefits of a such a hybrid model approach .", "forward": false, "src_ids": "2022.acl-long.153_798"}
+{"input": "model explainability tool is used for OtherScientificTerm| context: hedges have an important role in the management of rapport .", "entity": "model explainability tool", "output": "features", "neg_sample": ["model explainability tool is used for OtherScientificTerm", "hedges have an important role in the management of rapport ."], "relation": "used for", "id": "2022.acl-long.153", "year": 2022, "rel_sent": "We employ a model explainability tool to explore the features that characterize hedges in peer - tutoring conversations , and we identify some novel features , and the benefits of a such a hybrid model approach .", "forward": true, "src_ids": "2022.acl-long.153_799"}
+{"input": "coherent counterfactuals is done by using Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "coherent counterfactuals", "output": "docogen", "neg_sample": ["coherent counterfactuals is done by using Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We show that DoCoGen can generate coherent counterfactuals consisting of multiple sentences .", "forward": false, "src_ids": "2022.acl-long.533_800"}
+{"input": "sentiment classifier is done by using Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "sentiment classifier", "output": "docogen", "neg_sample": ["sentiment classifier is done by using Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": false, "src_ids": "2022.acl-long.533_801"}
+{"input": "multi - label intent classifier is done by using Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "multi - label intent classifier", "output": "docogen", "neg_sample": ["multi - label intent classifier is done by using Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": false, "src_ids": "2022.acl-long.533_802"}
+{"input": "low resource domain adaptation is done by using Task| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "low resource domain adaptation", "output": "domain counterfactual generation", "neg_sample": ["low resource domain adaptation is done by using Task", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "DoCoGen : Domain Counterfactual Generation for Low Resource Domain Adaptation.", "forward": false, "src_ids": "2022.acl-long.533_803"}
+{"input": "domain counterfactual generation is used for Task| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "domain counterfactual generation", "output": "low resource domain adaptation", "neg_sample": ["domain counterfactual generation is used for Task", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "DoCoGen : Domain Counterfactual Generation for Low Resource Domain Adaptation.", "forward": true, "src_ids": "2022.acl-long.533_804"}
+{"input": "domain adaptation ( da ) challenge is done by using Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "domain adaptation ( da ) challenge", "output": "controllable generation approach", "neg_sample": ["domain adaptation ( da ) challenge is done by using Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "In this paper we propose a controllable generation approach in order to deal with this domain adaptation ( DA ) challenge .", "forward": false, "src_ids": "2022.acl-long.533_805"}
+{"input": "controllable generation approach is used for Task| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "controllable generation approach", "output": "domain adaptation ( da ) challenge", "neg_sample": ["controllable generation approach is used for Task", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "In this paper we propose a controllable generation approach in order to deal with this domain adaptation ( DA ) challenge .", "forward": true, "src_ids": "2022.acl-long.533_806"}
+{"input": "docogen is used for OtherScientificTerm| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "docogen", "output": "coherent counterfactuals", "neg_sample": ["docogen is used for OtherScientificTerm", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We show that DoCoGen can generate coherent counterfactuals consisting of multiple sentences .", "forward": true, "src_ids": "2022.acl-long.533_807"}
+{"input": "sentiment classifier is done by using OtherScientificTerm| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "sentiment classifier", "output": "d - cons", "neg_sample": ["sentiment classifier is done by using OtherScientificTerm", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": false, "src_ids": "2022.acl-long.533_808"}
+{"input": "multi - label intent classifier is done by using OtherScientificTerm| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "multi - label intent classifier", "output": "d - cons", "neg_sample": ["multi - label intent classifier is done by using OtherScientificTerm", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": false, "src_ids": "2022.acl-long.533_809"}
+{"input": "d - cons is used for Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "d - cons", "output": "sentiment classifier", "neg_sample": ["d - cons is used for Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": true, "src_ids": "2022.acl-long.533_810"}
+{"input": "docogen is used for Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "docogen", "output": "sentiment classifier", "neg_sample": ["docogen is used for Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.533", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": true, "src_ids": "2022.acl-long.533_811"}
+{"input": "language models is used for Task| context: language models are increasingly becoming popular in ai - powered scientific ir systems .", "entity": "language models", "output": "scholarly retrieval", "neg_sample": ["language models is used for Task", "language models are increasingly becoming popular in ai - powered scientific ir systems ."], "relation": "used for", "id": "2022.findings-acl.249", "year": 2022, "rel_sent": "The Inefficiency of Language Models in Scholarly Retrieval : An Experimental Walk - through.", "forward": true, "src_ids": "2022.findings-acl.249_812"}
+{"input": "textual neighbors is done by using Method| context: language models are increasingly becoming popular in ai - powered scientific ir systems .", "entity": "textual neighbors", "output": "scientific language models", "neg_sample": ["textual neighbors is done by using Method", "language models are increasingly becoming popular in ai - powered scientific ir systems ."], "relation": "used for", "id": "2022.findings-acl.249", "year": 2022, "rel_sent": "This paper evaluates popular scientific language models in handling ( i ) short - query texts and ( ii ) textual neighbors .", "forward": false, "src_ids": "2022.findings-acl.249_813"}
+{"input": "short - query texts is done by using Method| context: language models are increasingly becoming popular in ai - powered scientific ir systems .", "entity": "short - query texts", "output": "scientific language models", "neg_sample": ["short - query texts is done by using Method", "language models are increasingly becoming popular in ai - powered scientific ir systems ."], "relation": "used for", "id": "2022.findings-acl.249", "year": 2022, "rel_sent": "This paper evaluates popular scientific language models in handling ( i ) short - query texts and ( ii ) textual neighbors .", "forward": false, "src_ids": "2022.findings-acl.249_814"}
+{"input": "scientific language models is used for Material| context: language models are increasingly becoming popular in ai - powered scientific ir systems .", "entity": "scientific language models", "output": "short - query texts", "neg_sample": ["scientific language models is used for Material", "language models are increasingly becoming popular in ai - powered scientific ir systems ."], "relation": "used for", "id": "2022.findings-acl.249", "year": 2022, "rel_sent": "This paper evaluates popular scientific language models in handling ( i ) short - query texts and ( ii ) textual neighbors .", "forward": true, "src_ids": "2022.findings-acl.249_815"}
+{"input": "scientific language models is used for OtherScientificTerm| context: language models are increasingly becoming popular in ai - powered scientific ir systems .", "entity": "scientific language models", "output": "textual neighbors", "neg_sample": ["scientific language models is used for OtherScientificTerm", "language models are increasingly becoming popular in ai - powered scientific ir systems ."], "relation": "used for", "id": "2022.findings-acl.249", "year": 2022, "rel_sent": "This paper evaluates popular scientific language models in handling ( i ) short - query texts and ( ii ) textual neighbors .", "forward": true, "src_ids": "2022.findings-acl.249_816"}
+{"input": "emotion analysis is done by using Method| context: emotion analysis is the process of identifying and analyzing the underlying emotions expressed in textual data . identifying emotions from a textual conversation is a challenging task due to the absence of gestures , vocal intonation , and facial expressions . once the chatbots and messengers detect and report the emotions of the user , a comfortable conversation can be carried out with no misunderstandings .", "entity": "emotion analysis", "output": "transformer based approach", "neg_sample": ["emotion analysis is done by using Method", "emotion analysis is the process of identifying and analyzing the underlying emotions expressed in textual data .", "identifying emotions from a textual conversation is a challenging task due to the absence of gestures , vocal intonation , and facial expressions .", "once the chatbots and messengers detect and report the emotions of the user , a comfortable conversation can be carried out with no misunderstandings ."], "relation": "used for", "id": "2022.dravidianlangtech-1.20", "year": 2022, "rel_sent": "SSNCSE_NLP@TamilNLP - ACL2022 : Transformer based approach for Emotion analysis in Tamil language.", "forward": false, "src_ids": "2022.dravidianlangtech-1.20_817"}
+{"input": "transformer based approach is used for Task| context: identifying emotions from a textual conversation is a challenging task due to the absence of gestures , vocal intonation , and facial expressions . once the chatbots and messengers detect and report the emotions of the user , a comfortable conversation can be carried out with no misunderstandings .", "entity": "transformer based approach", "output": "emotion analysis", "neg_sample": ["transformer based approach is used for Task", "identifying emotions from a textual conversation is a challenging task due to the absence of gestures , vocal intonation , and facial expressions .", "once the chatbots and messengers detect and report the emotions of the user , a comfortable conversation can be carried out with no misunderstandings ."], "relation": "used for", "id": "2022.dravidianlangtech-1.20", "year": 2022, "rel_sent": "SSNCSE_NLP@TamilNLP - ACL2022 : Transformer based approach for Emotion analysis in Tamil language.", "forward": true, "src_ids": "2022.dravidianlangtech-1.20_818"}
+{"input": "emotions is done by using Method| context: emotion analysis is the process of identifying and analyzing the underlying emotions expressed in textual data . identifying emotions from a textual conversation is a challenging task due to the absence of gestures , vocal intonation , and facial expressions . once the chatbots and messengers detect and report the emotions of the user , a comfortable conversation can be carried out with no misunderstandings .", "entity": "emotions", "output": "transformer model approach", "neg_sample": ["emotions is done by using Method", "emotion analysis is the process of identifying and analyzing the underlying emotions expressed in textual data .", "identifying emotions from a textual conversation is a challenging task due to the absence of gestures , vocal intonation , and facial expressions .", "once the chatbots and messengers detect and report the emotions of the user , a comfortable conversation can be carried out with no misunderstandings ."], "relation": "used for", "id": "2022.dravidianlangtech-1.20", "year": 2022, "rel_sent": "We have adopted the transformer model approach to identify the emotions present in the text sequence .", "forward": false, "src_ids": "2022.dravidianlangtech-1.20_819"}
+{"input": "probing fine - tuned models is done by using OtherScientificTerm| context: several studies have investigated the reasons behind the effectiveness of fine - tuning , usually through the lens of probing . however , these studies often neglect the role of the size of the dataset on which the model is fine - tuned .", "entity": "probing fine - tuned models", "output": "data size", "neg_sample": ["probing fine - tuned models is done by using OtherScientificTerm", "several studies have investigated the reasons behind the effectiveness of fine - tuning , usually through the lens of probing .", "however , these studies often neglect the role of the size of the dataset on which the model is fine - tuned ."], "relation": "used for", "id": "2022.findings-acl.20", "year": 2022, "rel_sent": "On the Importance of Data Size in Probing Fine - tuned Models.", "forward": false, "src_ids": "2022.findings-acl.20_820"}
+{"input": "data size is used for Task| context: several studies have investigated the reasons behind the effectiveness of fine - tuning , usually through the lens of probing . however , these studies often neglect the role of the size of the dataset on which the model is fine - tuned .", "entity": "data size", "output": "probing fine - tuned models", "neg_sample": ["data size is used for Task", "several studies have investigated the reasons behind the effectiveness of fine - tuning , usually through the lens of probing .", "however , these studies often neglect the role of the size of the dataset on which the model is fine - tuned ."], "relation": "used for", "id": "2022.findings-acl.20", "year": 2022, "rel_sent": "On the Importance of Data Size in Probing Fine - tuned Models.", "forward": true, "src_ids": "2022.findings-acl.20_821"}
+{"input": "massive weakly - labeled dataset is done by using Method| context: pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature . manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive . a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems . the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community .", "entity": "massive weakly - labeled dataset", "output": "distant - cto", "neg_sample": ["massive weakly - labeled dataset is done by using Method", "pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature .", "manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive .", "a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems .", "the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community ."], "relation": "used for", "id": "2022.bionlp-1.34", "year": 2022, "rel_sent": "To break through the bottleneck , we propose DISTANT - CTO , a novel distantly supervised PICO entity extraction approach using the clinical trials literature , to generate a massive weakly - labeled dataset with more than a million ' Intervention ' and ' Comparator ' entity annotations .", "forward": false, "src_ids": "2022.bionlp-1.34_822"}
+{"input": "low - resource entity extraction is done by using Method| context: pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature . manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive . a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems . the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community .", "entity": "low - resource entity extraction", "output": "distantly supervised approach", "neg_sample": ["low - resource entity extraction is done by using Method", "pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature .", "manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive .", "a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems .", "the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community ."], "relation": "used for", "id": "2022.bionlp-1.34", "year": 2022, "rel_sent": "DISTANT - CTO : A Zero Cost , Distantly Supervised Approach to Improve Low - Resource Entity Extraction Using Clinical Trials Literature.", "forward": false, "src_ids": "2022.bionlp-1.34_823"}
+{"input": "distantly supervised approach is used for Task| context: pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature . manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive . a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems . the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community .", "entity": "distantly supervised approach", "output": "low - resource entity extraction", "neg_sample": ["distantly supervised approach is used for Task", "pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature .", "manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive .", "a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems .", "the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community ."], "relation": "used for", "id": "2022.bionlp-1.34", "year": 2022, "rel_sent": "DISTANT - CTO : A Zero Cost , Distantly Supervised Approach to Improve Low - Resource Entity Extraction Using Clinical Trials Literature.", "forward": true, "src_ids": "2022.bionlp-1.34_824"}
+{"input": "distant - cto is used for Material| context: pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature . manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive . a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems . the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community .", "entity": "distant - cto", "output": "massive weakly - labeled dataset", "neg_sample": ["distant - cto is used for Material", "pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature .", "manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive .", "a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems .", "the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community ."], "relation": "used for", "id": "2022.bionlp-1.34", "year": 2022, "rel_sent": "To break through the bottleneck , we propose DISTANT - CTO , a novel distantly supervised PICO entity extraction approach using the clinical trials literature , to generate a massive weakly - labeled dataset with more than a million ' Intervention ' and ' Comparator ' entity annotations .", "forward": true, "src_ids": "2022.bionlp-1.34_825"}
+{"input": "distantly supervised pico entity extraction approach is used for Material| context: pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature . manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive . a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems . the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community .", "entity": "distantly supervised pico entity extraction approach", "output": "massive weakly - labeled dataset", "neg_sample": ["distantly supervised pico entity extraction approach is used for Material", "pico recognition is an information extraction task for identifying participant , intervention , comparator , and outcome information from clinical literature .", "manually identifying pico information is the most time - consuming step for conducting systematic reviews ( sr ) , which is already labor - intensive .", "a lack of diversified and large , annotated corpora restricts innovation and adoption of automated pico recognition systems .", "the largest - available pico entity / span corpus is manually annotated which is too expensive for a majority of the scientific community ."], "relation": "used for", "id": "2022.bionlp-1.34", "year": 2022, "rel_sent": "To break through the bottleneck , we propose DISTANT - CTO , a novel distantly supervised PICO entity extraction approach using the clinical trials literature , to generate a massive weakly - labeled dataset with more than a million ' Intervention ' and ' Comparator ' entity annotations .", "forward": true, "src_ids": "2022.bionlp-1.34_826"}
+{"input": "bert 's priors is done by using OtherScientificTerm| context: sampling is a promising bottom - up method for exposing what generative models have learned about language , but it remains unclear how to generate representative samples from popular masked language models ( mlms ) like bert . the mlm objective yields a dependency network with no guarantee of consistent conditional distributions , posing a problem for naive approaches .", "entity": "bert 's priors", "output": "serial reproduction chains", "neg_sample": ["bert 's priors is done by using OtherScientificTerm", "sampling is a promising bottom - up method for exposing what generative models have learned about language , but it remains unclear how to generate representative samples from popular masked language models ( mlms ) like bert .", "the mlm objective yields a dependency network with no guarantee of consistent conditional distributions , posing a problem for naive approaches ."], "relation": "used for", "id": "2022.findings-acl.314", "year": 2022, "rel_sent": "Drawing from theories of iterated learning in cognitive science , we explore the use of serial reproduction chains to sample from BERT 's priors .", "forward": false, "src_ids": "2022.findings-acl.314_827"}
+{"input": "serial reproduction chains is used for OtherScientificTerm| context: sampling is a promising bottom - up method for exposing what generative models have learned about language , but it remains unclear how to generate representative samples from popular masked language models ( mlms ) like bert . the mlm objective yields a dependency network with no guarantee of consistent conditional distributions , posing a problem for naive approaches .", "entity": "serial reproduction chains", "output": "bert 's priors", "neg_sample": ["serial reproduction chains is used for OtherScientificTerm", "sampling is a promising bottom - up method for exposing what generative models have learned about language , but it remains unclear how to generate representative samples from popular masked language models ( mlms ) like bert .", "the mlm objective yields a dependency network with no guarantee of consistent conditional distributions , posing a problem for naive approaches ."], "relation": "used for", "id": "2022.findings-acl.314", "year": 2022, "rel_sent": "Drawing from theories of iterated learning in cognitive science , we explore the use of serial reproduction chains to sample from BERT 's priors .", "forward": true, "src_ids": "2022.findings-acl.314_828"}
+{"input": "teacher model is done by using Method| context: while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated .", "entity": "teacher model", "output": "student model", "neg_sample": ["teacher model is done by using Method", "while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated ."], "relation": "used for", "id": "2022.tacl-1.21", "year": 2022, "rel_sent": "In this work , we introduce a framework to quantify the value of explanations via the accuracy gains that they confer on a student model trained to simulate a teacher model .", "forward": false, "src_ids": "2022.tacl-1.21_829"}
+{"input": "student model is used for Method| context: while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated .", "entity": "student model", "output": "teacher model", "neg_sample": ["student model is used for Method", "while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated ."], "relation": "used for", "id": "2022.tacl-1.21", "year": 2022, "rel_sent": "In this work , we introduce a framework to quantify the value of explanations via the accuracy gains that they confer on a student model trained to simulate a teacher model .", "forward": true, "src_ids": "2022.tacl-1.21_830"}
+{"input": "text classification is done by using Method| context: while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated .", "entity": "text classification", "output": "attribution methods", "neg_sample": ["text classification is done by using Method", "while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated ."], "relation": "used for", "id": "2022.tacl-1.21", "year": 2022, "rel_sent": "Using our framework , we compare numerous attribution methods for text classification and question answering , and observe quantitative differences that are consistent ( to a moderate to high degree ) across different student model architectures and learning strategies.1", "forward": false, "src_ids": "2022.tacl-1.21_831"}
+{"input": "question answering is done by using Method| context: while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated .", "entity": "question answering", "output": "attribution methods", "neg_sample": ["question answering is done by using Method", "while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated ."], "relation": "used for", "id": "2022.tacl-1.21", "year": 2022, "rel_sent": "Using our framework , we compare numerous attribution methods for text classification and question answering , and observe quantitative differences that are consistent ( to a moderate to high degree ) across different student model architectures and learning strategies.1", "forward": false, "src_ids": "2022.tacl-1.21_832"}
+{"input": "attribution methods is used for Task| context: while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated .", "entity": "attribution methods", "output": "text classification", "neg_sample": ["attribution methods is used for Task", "while many methods purport to explain predictions by highlighting salient features , what aims these explanations serve and how they ought to be evaluated often go unstated ."], "relation": "used for", "id": "2022.tacl-1.21", "year": 2022, "rel_sent": "Using our framework , we compare numerous attribution methods for text classification and question answering , and observe quantitative differences that are consistent ( to a moderate to high degree ) across different student model architectures and learning strategies.1", "forward": true, "src_ids": "2022.tacl-1.21_833"}
+{"input": "contrastive learning of sentence representations ( vascl ) is done by using Method| context: despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge . this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language .", "entity": "contrastive learning of sentence representations ( vascl )", "output": "virtual augmentation", "neg_sample": ["contrastive learning of sentence representations ( vascl ) is done by using Method", "despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge .", "this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language ."], "relation": "used for", "id": "2022.findings-acl.70", "year": 2022, "rel_sent": "Virtual Augmentation Supported Contrastive Learning of Sentence Representations.", "forward": false, "src_ids": "2022.findings-acl.70_834"}
+{"input": "contrastive learning of sentence representations ( vascl ) is done by using Method| context: despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge . this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language .", "entity": "contrastive learning of sentence representations ( vascl )", "output": "virtual augmentation", "neg_sample": ["contrastive learning of sentence representations ( vascl ) is done by using Method", "despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge .", "this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language ."], "relation": "used for", "id": "2022.findings-acl.70", "year": 2022, "rel_sent": "We tackle this challenge by presenting a Virtual augmentation Supported Contrastive Learning of sentence representations ( VaSCL ) .", "forward": false, "src_ids": "2022.findings-acl.70_835"}
+{"input": "virtual augmentation is used for Method| context: despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge . this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language .", "entity": "virtual augmentation", "output": "contrastive learning of sentence representations ( vascl )", "neg_sample": ["virtual augmentation is used for Method", "despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge .", "this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language ."], "relation": "used for", "id": "2022.findings-acl.70", "year": 2022, "rel_sent": "Virtual Augmentation Supported Contrastive Learning of Sentence Representations.", "forward": true, "src_ids": "2022.findings-acl.70_836"}
+{"input": "virtual augmentation is used for Method| context: despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge . this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language .", "entity": "virtual augmentation", "output": "contrastive learning of sentence representations ( vascl )", "neg_sample": ["virtual augmentation is used for Method", "despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge .", "this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language ."], "relation": "used for", "id": "2022.findings-acl.70", "year": 2022, "rel_sent": "We tackle this challenge by presenting a Virtual augmentation Supported Contrastive Learning of sentence representations ( VaSCL ) .", "forward": true, "src_ids": "2022.findings-acl.70_837"}
+{"input": "data augmentations is done by using OtherScientificTerm| context: despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge . this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language .", "entity": "data augmentations", "output": "neighborhood", "neg_sample": ["data augmentations is done by using OtherScientificTerm", "despite profound successes , contrastive representation learning relies on carefully designed data augmentations using domain - specific knowledge .", "this challenge is magnified in natural language processing , where no general rules exist for data augmentation due to the discrete nature of natural language ."], "relation": "used for", "id": "2022.findings-acl.70", "year": 2022, "rel_sent": "Originating from the interpretation that data augmentation essentially constructs the neighborhoods of each training instance , we , in turn , utilize the neighborhood to generate effective data augmentations .", "forward": false, "src_ids": "2022.findings-acl.70_838"}
+{"input": "machine translation is done by using Method| context: this paper considers some ethical implications of machine translation for low - resourced languages . i use armenian as a case study and investigate specific needs for and concerns arising from the creation and deployment of improved machine translation between english and armenian .", "entity": "machine translation", "output": "language technology", "neg_sample": ["machine translation is done by using Method", "this paper considers some ethical implications of machine translation for low - resourced languages .", "i use armenian as a case study and investigate specific needs for and concerns arising from the creation and deployment of improved machine translation between english and armenian ."], "relation": "used for", "id": "2022.acl-srw.5", "year": 2022, "rel_sent": "Based on these scenarios , I recommend 1 ) collaborating with stakeholders in order to create more useful and reliable machine translation tools , and 2 ) determining which other forms of language technology should be developed alongside efforts to improve machine translation in order to mitigate harms rendered to vulnerable language communities .", "forward": false, "src_ids": "2022.acl-srw.5_839"}
+{"input": "product defect triage is done by using Method| context: defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce . inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams .", "entity": "product defect triage", "output": "few - shot label fused contextual representation learning", "neg_sample": ["product defect triage is done by using Method", "defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce .", "inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams ."], "relation": "used for", "id": "2022.ecnlp-1.1", "year": 2022, "rel_sent": "DEFTri : A Few - Shot Label Fused Contextual Representation Learning For Product Defect Triage in e - Commerce.", "forward": false, "src_ids": "2022.ecnlp-1.1_840"}
+{"input": "few - shot label fused contextual representation learning is used for Task| context: defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce . inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams .", "entity": "few - shot label fused contextual representation learning", "output": "product defect triage", "neg_sample": ["few - shot label fused contextual representation learning is used for Task", "defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce .", "inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams ."], "relation": "used for", "id": "2022.ecnlp-1.1", "year": 2022, "rel_sent": "DEFTri : A Few - Shot Label Fused Contextual Representation Learning For Product Defect Triage in e - Commerce.", "forward": true, "src_ids": "2022.ecnlp-1.1_841"}
+{"input": "contextual representations is done by using OtherScientificTerm| context: defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce . inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams .", "entity": "contextual representations", "output": "labels fused text embeddings", "neg_sample": ["contextual representations is done by using OtherScientificTerm", "defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce .", "inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams ."], "relation": "used for", "id": "2022.ecnlp-1.1", "year": 2022, "rel_sent": "This work proposes a novel framework for automated defect triage ( DEFTri ) using fine - tuned state - of - the - art pre - trained BERT on labels fused text embeddings to improve contextual representations from human - generated product defects .", "forward": false, "src_ids": "2022.ecnlp-1.1_842"}
+{"input": "labels fused text embeddings is used for Method| context: defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce . inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams .", "entity": "labels fused text embeddings", "output": "contextual representations", "neg_sample": ["labels fused text embeddings is used for Method", "defect triage is a time - sensitive and critical process in a large - scale agile software development lifecycle for e - commerce .", "inefficiencies arising from human and process dependencies in this domain have motivated research in automated approaches using machine learning to accurately assign defects to qualified teams ."], "relation": "used for", "id": "2022.ecnlp-1.1", "year": 2022, "rel_sent": "This work proposes a novel framework for automated defect triage ( DEFTri ) using fine - tuned state - of - the - art pre - trained BERT on labels fused text embeddings to improve contextual representations from human - generated product defects .", "forward": true, "src_ids": "2022.ecnlp-1.1_843"}
+{"input": "covid-19 misinformation is done by using Task| context: during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially . unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths . the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation . however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually . thus , it becomes imperative to verify claims automatically without human interventions .", "entity": "covid-19 misinformation", "output": "document retrieval", "neg_sample": ["covid-19 misinformation is done by using Task", "during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially .", "unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths .", "the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation .", "however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually .", "thus , it becomes imperative to verify claims automatically without human interventions ."], "relation": "used for", "id": "2022.constraint-1.8", "year": 2022, "rel_sent": "Document Retrieval and Claim Verification to Mitigate COVID-19 Misinformation.", "forward": false, "src_ids": "2022.constraint-1.8_844"}
+{"input": "claim verification is used for OtherScientificTerm| context: during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially . unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths . the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation . however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually . thus , it becomes imperative to verify claims automatically without human interventions .", "entity": "claim verification", "output": "covid-19 misinformation", "neg_sample": ["claim verification is used for OtherScientificTerm", "during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially .", "unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths .", "the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation .", "however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually .", "thus , it becomes imperative to verify claims automatically without human interventions ."], "relation": "used for", "id": "2022.constraint-1.8", "year": 2022, "rel_sent": "Document Retrieval and Claim Verification to Mitigate COVID-19 Misinformation.", "forward": true, "src_ids": "2022.constraint-1.8_845"}
+{"input": "document retrieval is used for OtherScientificTerm| context: during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially . unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths . the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation . however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually . thus , it becomes imperative to verify claims automatically without human interventions .", "entity": "document retrieval", "output": "covid-19 misinformation", "neg_sample": ["document retrieval is used for OtherScientificTerm", "during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially .", "unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths .", "the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation .", "however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually .", "thus , it becomes imperative to verify claims automatically without human interventions ."], "relation": "used for", "id": "2022.constraint-1.8", "year": 2022, "rel_sent": "Document Retrieval and Claim Verification to Mitigate COVID-19 Misinformation.", "forward": true, "src_ids": "2022.constraint-1.8_846"}
+{"input": "re - ranking is done by using Method| context: during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially . unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths . the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation . however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually . thus , it becomes imperative to verify claims automatically without human interventions .", "entity": "re - ranking", "output": "transformer - based models", "neg_sample": ["re - ranking is done by using Method", "during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially .", "unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths .", "the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation .", "however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually .", "thus , it becomes imperative to verify claims automatically without human interventions ."], "relation": "used for", "id": "2022.constraint-1.8", "year": 2022, "rel_sent": "For the retrieval module , we employ a hybrid search - based system with BM25 as a base retriever and experiment with recent state - of - the - art transformer - based models for re - ranking .", "forward": false, "src_ids": "2022.constraint-1.8_847"}
+{"input": "transformer - based models is used for Task| context: during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially . unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths . the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation . however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually . thus , it becomes imperative to verify claims automatically without human interventions .", "entity": "transformer - based models", "output": "re - ranking", "neg_sample": ["transformer - based models is used for Task", "during the covid-19 pandemic , the spread of misinformation on online social media has grown exponentially .", "unverified bogus claims on these platforms regularly mislead people , leading them to believe in half - baked truths .", "the current vogue is to employ manual fact - checkers to verify claims to combat this avalanche of misinformation .", "however , establishing such claims ' veracity is becoming increasingly challenging , partly due to the plethora of information available , which is difficult to process manually .", "thus , it becomes imperative to verify claims automatically without human interventions ."], "relation": "used for", "id": "2022.constraint-1.8", "year": 2022, "rel_sent": "For the retrieval module , we employ a hybrid search - based system with BM25 as a base retriever and experiment with recent state - of - the - art transformer - based models for re - ranking .", "forward": true, "src_ids": "2022.constraint-1.8_848"}
+{"input": "fine - grained morphological and syntactic signals is done by using Method| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "fine - grained morphological and syntactic signals", "output": "language models", "neg_sample": ["fine - grained morphological and syntactic signals is done by using Method", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "This indicates that language models do not fully cover the fine - grained morphological and syntactic signals that are explicitly represented in grammatical profiles .", "forward": false, "src_ids": "2022.lchange-1.6_849"}
+{"input": "grammatical profiles is used for Method| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "grammatical profiles", "output": "language models", "neg_sample": ["grammatical profiles is used for Method", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "Do Not Fire the Linguist : Grammatical Profiles Help Language Models Detect Semantic Change.", "forward": true, "src_ids": "2022.lchange-1.6_850"}
+{"input": "lexical semantic change detection is done by using Method| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "lexical semantic change detection", "output": "contextualised language models", "neg_sample": ["lexical semantic change detection is done by using Method", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "In this work , we explore whether large pre - trained contextualised language models , a common tool for lexical semantic change detection , are sensitive to such morphosyntactic changes .", "forward": false, "src_ids": "2022.lchange-1.6_851"}
+{"input": "morphosyntactic changes is done by using Method| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "morphosyntactic changes", "output": "contextualised language models", "neg_sample": ["morphosyntactic changes is done by using Method", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "In this work , we explore whether large pre - trained contextualised language models , a common tool for lexical semantic change detection , are sensitive to such morphosyntactic changes .", "forward": false, "src_ids": "2022.lchange-1.6_852"}
+{"input": "contextualised language models is used for Task| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "contextualised language models", "output": "lexical semantic change detection", "neg_sample": ["contextualised language models is used for Task", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "In this work , we explore whether large pre - trained contextualised language models , a common tool for lexical semantic change detection , are sensitive to such morphosyntactic changes .", "forward": true, "src_ids": "2022.lchange-1.6_853"}
+{"input": "contextualised language models is used for OtherScientificTerm| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "contextualised language models", "output": "morphosyntactic changes", "neg_sample": ["contextualised language models is used for OtherScientificTerm", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "In this work , we explore whether large pre - trained contextualised language models , a common tool for lexical semantic change detection , are sensitive to such morphosyntactic changes .", "forward": true, "src_ids": "2022.lchange-1.6_854"}
+{"input": "language models is used for OtherScientificTerm| context: morphological and syntactic changes in word usage - as captured , e.g. , by grammatical profiles - have been shown to be good predictors of a word 's meaning change .", "entity": "language models", "output": "fine - grained morphological and syntactic signals", "neg_sample": ["language models is used for OtherScientificTerm", "morphological and syntactic changes in word usage - as captured , e.g.", ", by grammatical profiles - have been shown to be good predictors of a word 's meaning change ."], "relation": "used for", "id": "2022.lchange-1.6", "year": 2022, "rel_sent": "This indicates that language models do not fully cover the fine - grained morphological and syntactic signals that are explicitly represented in grammatical profiles .", "forward": true, "src_ids": "2022.lchange-1.6_855"}
+{"input": "severity of depression is done by using Method| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "severity of depression", "output": "bag of words and document embeddings based framework", "neg_sample": ["severity of depression is done by using Method", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "IISERB@LT - EDI - ACL2022 : A Bag of Words and Document Embeddings Based Framework to Identify Severity of Depression Over Social Media.", "forward": false, "src_ids": "2022.ltedi-1.33_856"}
+{"input": "bag of words and document embeddings based framework is used for OtherScientificTerm| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "bag of words and document embeddings based framework", "output": "severity of depression", "neg_sample": ["bag of words and document embeddings based framework is used for OtherScientificTerm", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "IISERB@LT - EDI - ACL2022 : A Bag of Words and Document Embeddings Based Framework to Identify Severity of Depression Over Social Media.", "forward": true, "src_ids": "2022.ltedi-1.33_857"}
+{"input": "text classification is done by using OtherScientificTerm| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "text classification", "output": "features", "neg_sample": ["text classification is done by using OtherScientificTerm", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "The objective of this work is to identify relevant features from the given social media texts for effective text classification .", "forward": false, "src_ids": "2022.ltedi-1.33_858"}
+{"input": "features is used for Task| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "features", "output": "text classification", "neg_sample": ["features is used for Task", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "The objective of this work is to identify relevant features from the given social media texts for effective text classification .", "forward": true, "src_ids": "2022.ltedi-1.33_859"}
+{"input": "scale of depression is done by using Method| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "scale of depression", "output": "adaptive boosting", "neg_sample": ["scale of depression is done by using Method", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "Subsequently , adaptive boosting , logistic regression , random forest and support vector machine ( SVM ) classifiers were used to identify the scale of depression from the given texts .", "forward": false, "src_ids": "2022.ltedi-1.33_860"}
+{"input": "support vector machine ( svm ) classifiers is used for OtherScientificTerm| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "support vector machine ( svm ) classifiers", "output": "scale of depression", "neg_sample": ["support vector machine ( svm ) classifiers is used for OtherScientificTerm", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "Subsequently , adaptive boosting , logistic regression , random forest and support vector machine ( SVM ) classifiers were used to identify the scale of depression from the given texts .", "forward": true, "src_ids": "2022.ltedi-1.33_861"}
+{"input": "adaptive boosting is used for OtherScientificTerm| context: the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models .", "entity": "adaptive boosting", "output": "scale of depression", "neg_sample": ["adaptive boosting is used for OtherScientificTerm", "the bionlp group at department of data science and engineering in indian institute of science education and research bhopal ( iiserb ) has participated in this challenge and submitted three runs based on three different text mining models ."], "relation": "used for", "id": "2022.ltedi-1.33", "year": 2022, "rel_sent": "Subsequently , adaptive boosting , logistic regression , random forest and support vector machine ( SVM ) classifiers were used to identify the scale of depression from the given texts .", "forward": true, "src_ids": "2022.ltedi-1.33_862"}
+{"input": "neural cross - lingual summarization is done by using Method| context: the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls . existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective . however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize .", "entity": "neural cross - lingual summarization", "output": "variational hierarchical model", "neg_sample": ["neural cross - lingual summarization is done by using Method", "the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls .", "existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective .", "however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize ."], "relation": "used for", "id": "2022.acl-long.148", "year": 2022, "rel_sent": "A Variational Hierarchical Model for Neural Cross - Lingual Summarization.", "forward": false, "src_ids": "2022.acl-long.148_863"}
+{"input": "variational hierarchical model is used for Task| context: the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls . existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective . however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize .", "entity": "variational hierarchical model", "output": "neural cross - lingual summarization", "neg_sample": ["variational hierarchical model is used for Task", "the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls .", "existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective .", "however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize ."], "relation": "used for", "id": "2022.acl-long.148", "year": 2022, "rel_sent": "A Variational Hierarchical Model for Neural Cross - Lingual Summarization.", "forward": true, "src_ids": "2022.acl-long.148_864"}
+{"input": "hierarchical model is used for Task| context: existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective . however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize .", "entity": "hierarchical model", "output": "cls task", "neg_sample": ["hierarchical model is used for Task", "existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective .", "however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize ."], "relation": "used for", "id": "2022.acl-long.148", "year": 2022, "rel_sent": "To address this issue , we propose a hierarchical model for the CLS task , based on the conditional variational auto - encoder .", "forward": true, "src_ids": "2022.acl-long.148_865"}
+{"input": "cls task is done by using Method| context: the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls . existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective . however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize .", "entity": "cls task", "output": "hierarchical model", "neg_sample": ["cls task is done by using Method", "the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls .", "existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective .", "however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize ."], "relation": "used for", "id": "2022.acl-long.148", "year": 2022, "rel_sent": "To address this issue , we propose a hierarchical model for the CLS task , based on the conditional variational auto - encoder .", "forward": false, "src_ids": "2022.acl-long.148_866"}
+{"input": "cross - lingual summarization ( cls ) is done by using OtherScientificTerm| context: the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls . existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective . however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize .", "entity": "cross - lingual summarization ( cls )", "output": "latent variables", "neg_sample": ["cross - lingual summarization ( cls ) is done by using OtherScientificTerm", "the cls task is essentially the combination of machine translation ( mt ) and monolingual summarization ( ms ) , and thus there exists the hierarchical relationship between mt&ms and cls .", "existing studies on cls mainly focus on utilizing pipeline methods or jointly training an end - to - end model through an auxiliary mt or ms objective .", "however , it is very challenging for the model to directly conduct cls as it requires both the abilities to translate and summarize ."], "relation": "used for", "id": "2022.acl-long.148", "year": 2022, "rel_sent": "As for the global level , there is another latent variable for cross - lingual summarization conditioned on the two local - level variables .", "forward": false, "src_ids": "2022.acl-long.148_867"}
+{"input": "syntactic information is used for Task| context: probing is popular to analyze whether linguistic information can be captured by a well - trained deep neural model , but it is hard to answer how the change of the encoded linguistic information will affect task performance .", "entity": "syntactic information", "output": "nlp tasks", "neg_sample": ["syntactic information is used for Task", "probing is popular to analyze whether linguistic information can be captured by a well - trained deep neural model , but it is hard to answer how the change of the encoded linguistic information will affect task performance ."], "relation": "used for", "id": "2022.findings-acl.35", "year": 2022, "rel_sent": "Our empirical findings suggest that some syntactic information is helpful for NLP tasks whereas encoding more syntactic information does not necessarily lead to better performance , because the model architecture is also an important factor .", "forward": true, "src_ids": "2022.findings-acl.35_868"}
+{"input": "nlp tasks is done by using OtherScientificTerm| context: probing is popular to analyze whether linguistic information can be captured by a well - trained deep neural model , but it is hard to answer how the change of the encoded linguistic information will affect task performance .", "entity": "nlp tasks", "output": "syntactic information", "neg_sample": ["nlp tasks is done by using OtherScientificTerm", "probing is popular to analyze whether linguistic information can be captured by a well - trained deep neural model , but it is hard to answer how the change of the encoded linguistic information will affect task performance ."], "relation": "used for", "id": "2022.findings-acl.35", "year": 2022, "rel_sent": "Our empirical findings suggest that some syntactic information is helpful for NLP tasks whereas encoding more syntactic information does not necessarily lead to better performance , because the model architecture is also an important factor .", "forward": false, "src_ids": "2022.findings-acl.35_869"}
+{"input": "indic - transformer based classifier is done by using Method| context: social media platforms have grown their reach worldwide . as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions . tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages . abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models .", "entity": "indic - transformer based classifier", "output": "data augmentation", "neg_sample": ["indic - transformer based classifier is done by using Method", "social media platforms have grown their reach worldwide .", "as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions .", "tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages .", "abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.22", "year": 2022, "rel_sent": "BpHigh@TamilNLP - ACL2022 : Effects of Data Augmentation on Indic - Transformer based classifier for Abusive Comments Detection in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.22_870"}
+{"input": "abusive comments detection is done by using Method| context: social media platforms have grown their reach worldwide . as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions . tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages . abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models .", "entity": "abusive comments detection", "output": "indic - transformer based classifier", "neg_sample": ["abusive comments detection is done by using Method", "social media platforms have grown their reach worldwide .", "as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions .", "tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages .", "abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.22", "year": 2022, "rel_sent": "BpHigh@TamilNLP - ACL2022 : Effects of Data Augmentation on Indic - Transformer based classifier for Abusive Comments Detection in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.22_871"}
+{"input": "data augmentation is used for Method| context: social media platforms have grown their reach worldwide . as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions . tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages . abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models .", "entity": "data augmentation", "output": "indic - transformer based classifier", "neg_sample": ["data augmentation is used for Method", "social media platforms have grown their reach worldwide .", "as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions .", "tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages .", "abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.22", "year": 2022, "rel_sent": "BpHigh@TamilNLP - ACL2022 : Effects of Data Augmentation on Indic - Transformer based classifier for Abusive Comments Detection in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.22_872"}
+{"input": "indic - transformer based classifier is used for Task| context: social media platforms have grown their reach worldwide . as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions . tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages . abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models .", "entity": "indic - transformer based classifier", "output": "abusive comments detection", "neg_sample": ["indic - transformer based classifier is used for Task", "social media platforms have grown their reach worldwide .", "as an effect of this growth , many vernacular social media platforms have also emerged , focusing more on the diverse languages in the specific regions .", "tamil has also emerged as a popular language for use on social media platforms due to the increasing penetration of vernacular media like sharechat and moj , which focus more on local indian languages than english and encourage their users to converse in indic languages .", "abusive language remains a significant challenge in the social media framework and more so when we consider languages like tamil , which are low - resource languages and have poor performance on multilingual models and lack language - specific models ."], "relation": "used for", "id": "2022.dravidianlangtech-1.22", "year": 2022, "rel_sent": "BpHigh@TamilNLP - ACL2022 : Effects of Data Augmentation on Indic - Transformer based classifier for Abusive Comments Detection in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.22_873"}
+{"input": "structured knowledge is used for Task| context: while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "structured knowledge", "output": "open - domain question answering", "neg_sample": ["structured knowledge is used for Task", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "In this work , we bridge this gap and use the data - to - text method as a means for encoding structured knowledge for open - domain question answering .", "forward": true, "src_ids": "2022.acl-long.113_874"}
+{"input": "verbalizer - retriever - reader framework is used for Task| context: while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "verbalizer - retriever - reader framework", "output": "open - domain question answering", "neg_sample": ["verbalizer - retriever - reader framework is used for Task", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "Specifically , we propose a verbalizer - retriever - reader framework for ODQA over data and text where verbalized tables from Wikipedia and graphs from Wikidata are used as augmented knowledge sources .", "forward": true, "src_ids": "2022.acl-long.113_875"}
+{"input": "open - domain question answering is done by using OtherScientificTerm| context: while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "open - domain question answering", "output": "structured knowledge", "neg_sample": ["open - domain question answering is done by using OtherScientificTerm", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "In this work , we bridge this gap and use the data - to - text method as a means for encoding structured knowledge for open - domain question answering .", "forward": false, "src_ids": "2022.acl-long.113_876"}
+{"input": "data - to - text method is used for OtherScientificTerm| context: while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "data - to - text method", "output": "structured knowledge", "neg_sample": ["data - to - text method is used for OtherScientificTerm", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "In this work , we bridge this gap and use the data - to - text method as a means for encoding structured knowledge for open - domain question answering .", "forward": true, "src_ids": "2022.acl-long.113_877"}
+{"input": "structured knowledge is done by using Method| context: the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge . although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question . while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "structured knowledge", "output": "data - to - text method", "neg_sample": ["structured knowledge is done by using Method", "the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge .", "although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question .", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "In this work , we bridge this gap and use the data - to - text method as a means for encoding structured knowledge for open - domain question answering .", "forward": false, "src_ids": "2022.acl-long.113_878"}
+{"input": "open - domain question answering is done by using Method| context: the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge . although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question . while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "open - domain question answering", "output": "verbalizer - retriever - reader framework", "neg_sample": ["open - domain question answering is done by using Method", "the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge .", "although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question .", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "Specifically , we propose a verbalizer - retriever - reader framework for ODQA over data and text where verbalized tables from Wikipedia and graphs from Wikidata are used as augmented knowledge sources .", "forward": false, "src_ids": "2022.acl-long.113_879"}
+{"input": "augmented knowledge sources is done by using OtherScientificTerm| context: the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge . although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question . while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "augmented knowledge sources", "output": "verbalized tables", "neg_sample": ["augmented knowledge sources is done by using OtherScientificTerm", "the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge .", "although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question .", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "Specifically , we propose a verbalizer - retriever - reader framework for ODQA over data and text where verbalized tables from Wikipedia and graphs from Wikidata are used as augmented knowledge sources .", "forward": false, "src_ids": "2022.acl-long.113_880"}
+{"input": "verbalized tables is used for OtherScientificTerm| context: the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge . although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question . while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "verbalized tables", "output": "augmented knowledge sources", "neg_sample": ["verbalized tables is used for OtherScientificTerm", "the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge .", "although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question .", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "Specifically , we propose a verbalizer - retriever - reader framework for ODQA over data and text where verbalized tables from Wikipedia and graphs from Wikidata are used as augmented knowledge sources .", "forward": true, "src_ids": "2022.acl-long.113_881"}
+{"input": "answer reasoning is done by using OtherScientificTerm| context: the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge . although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question . while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "answer reasoning", "output": "verbalized knowledge", "neg_sample": ["answer reasoning is done by using OtherScientificTerm", "the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge .", "although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question .", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "Furthermore , our analyses indicate that verbalized knowledge is preferred for answer reasoning for both adapted and hot - swap settings .", "forward": false, "src_ids": "2022.acl-long.113_882"}
+{"input": "verbalized knowledge is used for Method| context: the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge . although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question . while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown .", "entity": "verbalized knowledge", "output": "answer reasoning", "neg_sample": ["verbalized knowledge is used for Method", "the retriever - reader framework is popular for open - domain question answering ( odqa ) due to its ability to use explicit knowledge .", "although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text , accessing heterogeneous knowledge sources through a unified interface remains an open question .", "while data - to - text generation has the potential to serve as a universal interface for data and text , its feasibility for downstream tasks remains largely unknown ."], "relation": "used for", "id": "2022.acl-long.113", "year": 2022, "rel_sent": "Furthermore , our analyses indicate that verbalized knowledge is preferred for answer reasoning for both adapted and hot - swap settings .", "forward": true, "src_ids": "2022.acl-long.113_883"}
+{"input": "pseudo ambiguous question generation is used for OtherScientificTerm| context: question answering ( qa ) with disambiguation questions is essential for practical qa systems because user questions often do not contain information enough tofind their answers . we call this task clarifying question answering , a task tofind answers to ambiguous user questions by disambiguating their intents through interactions . there are two major problems in building a clarifying question answering system : data preparation of possible ambiguous questions and the generation of clarifying questions .", "entity": "pseudo ambiguous question generation", "output": "ambiguity", "neg_sample": ["pseudo ambiguous question generation is used for OtherScientificTerm", "question answering ( qa ) with disambiguation questions is essential for practical qa systems because user questions often do not contain information enough tofind their answers .", "we call this task clarifying question answering , a task tofind answers to ambiguous user questions by disambiguating their intents through interactions .", "there are two major problems in building a clarifying question answering system : data preparation of possible ambiguous questions and the generation of clarifying questions ."], "relation": "used for", "id": "2022.dialdoc-1.4", "year": 2022, "rel_sent": "Our experimental results verify that our pseudo ambiguous question generation successfully adds ambiguity to questions .", "forward": true, "src_ids": "2022.dialdoc-1.4_884"}
+{"input": "ambiguity is done by using Method| context: question answering ( qa ) with disambiguation questions is essential for practical qa systems because user questions often do not contain information enough tofind their answers . we call this task clarifying question answering , a task tofind answers to ambiguous user questions by disambiguating their intents through interactions . there are two major problems in building a clarifying question answering system : data preparation of possible ambiguous questions and the generation of clarifying questions .", "entity": "ambiguity", "output": "pseudo ambiguous question generation", "neg_sample": ["ambiguity is done by using Method", "question answering ( qa ) with disambiguation questions is essential for practical qa systems because user questions often do not contain information enough tofind their answers .", "we call this task clarifying question answering , a task tofind answers to ambiguous user questions by disambiguating their intents through interactions .", "there are two major problems in building a clarifying question answering system : data preparation of possible ambiguous questions and the generation of clarifying questions ."], "relation": "used for", "id": "2022.dialdoc-1.4", "year": 2022, "rel_sent": "Our experimental results verify that our pseudo ambiguous question generation successfully adds ambiguity to questions .", "forward": false, "src_ids": "2022.dialdoc-1.4_885"}
+{"input": "mel models is done by using Method| context: multimodal entity linking ( mel ) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base ( e.g. , wikipedia ) , is an essential task for many multimodal applications . although much attention has been paid to mel , the shortcomings of existing mel datasets including limited contextual topics and entity types , simplified mention ambiguity , and restricted availability , have caused great obstacles to the research and application of mel .", "entity": "mel models", "output": "wikidiverse", "neg_sample": ["mel models is done by using Method", "multimodal entity linking ( mel ) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base ( e.g.", ", wikipedia ) , is an essential task for many multimodal applications .", "although much attention has been paid to mel , the shortcomings of existing mel datasets including limited contextual topics and entity types , simplified mention ambiguity , and restricted availability , have caused great obstacles to the research and application of mel ."], "relation": "used for", "id": "2022.acl-long.328", "year": 2022, "rel_sent": "Based on WikiDiverse , a sequence of well - designed MEL models with intra - modality and inter - modality attentions are implemented , which utilize the visual information of images more adequately than existing MEL models do .", "forward": false, "src_ids": "2022.acl-long.328_886"}
+{"input": "wikidiverse is used for Method| context: multimodal entity linking ( mel ) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base ( e.g. , wikipedia ) , is an essential task for many multimodal applications . although much attention has been paid to mel , the shortcomings of existing mel datasets including limited contextual topics and entity types , simplified mention ambiguity , and restricted availability , have caused great obstacles to the research and application of mel .", "entity": "wikidiverse", "output": "mel models", "neg_sample": ["wikidiverse is used for Method", "multimodal entity linking ( mel ) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base ( e.g.", ", wikipedia ) , is an essential task for many multimodal applications .", "although much attention has been paid to mel , the shortcomings of existing mel datasets including limited contextual topics and entity types , simplified mention ambiguity , and restricted availability , have caused great obstacles to the research and application of mel ."], "relation": "used for", "id": "2022.acl-long.328", "year": 2022, "rel_sent": "Based on WikiDiverse , a sequence of well - designed MEL models with intra - modality and inter - modality attentions are implemented , which utilize the visual information of images more adequately than existing MEL models do .", "forward": true, "src_ids": "2022.acl-long.328_887"}
+{"input": "multimodal entity linking is done by using OtherScientificTerm| context: multimodal entity linking ( mel ) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base ( e.g. , wikipedia ) , is an essential task for many multimodal applications . although much attention has been paid to mel , the shortcomings of existing mel datasets including limited contextual topics and entity types , simplified mention ambiguity , and restricted availability , have caused great obstacles to the research and application of mel .", "entity": "multimodal entity linking", "output": "modalities", "neg_sample": ["multimodal entity linking is done by using OtherScientificTerm", "multimodal entity linking ( mel ) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base ( e.g.", ", wikipedia ) , is an essential task for many multimodal applications .", "although much attention has been paid to mel , the shortcomings of existing mel datasets including limited contextual topics and entity types , simplified mention ambiguity , and restricted availability , have caused great obstacles to the research and application of mel ."], "relation": "used for", "id": "2022.acl-long.328", "year": 2022, "rel_sent": "Extensive experimental analyses are conducted to investigate the contributions of different modalities in terms of MEL , facilitating the future research on this task .", "forward": false, "src_ids": "2022.acl-long.328_888"}
+{"input": "controllable gpt2 generation is done by using Method| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "controllable gpt2 generation", "output": "lightweight framework", "neg_sample": ["controllable gpt2 generation is done by using Method", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "In this work , we propose a novel lightweight framework for controllable GPT2 generation , which utilizes a set of small attribute - specific vectors , called prefixes ( Li and Liang , 2021 ) , to steer natural language generation .", "forward": false, "src_ids": "2022.findings-acl.229_889"}
+{"input": "lightweight framework is used for Task| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "lightweight framework", "output": "controllable gpt2 generation", "neg_sample": ["lightweight framework is used for Task", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "In this work , we propose a novel lightweight framework for controllable GPT2 generation , which utilizes a set of small attribute - specific vectors , called prefixes ( Li and Liang , 2021 ) , to steer natural language generation .", "forward": true, "src_ids": "2022.findings-acl.229_890"}
+{"input": "natural language generation is done by using OtherScientificTerm| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "natural language generation", "output": "small attribute - specific vectors", "neg_sample": ["natural language generation is done by using OtherScientificTerm", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "In this work , we propose a novel lightweight framework for controllable GPT2 generation , which utilizes a set of small attribute - specific vectors , called prefixes ( Li and Liang , 2021 ) , to steer natural language generation .", "forward": false, "src_ids": "2022.findings-acl.229_891"}
+{"input": "unsupervised method is used for OtherScientificTerm| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "unsupervised method", "output": "prefixes", "neg_sample": ["unsupervised method is used for OtherScientificTerm", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "We propose a novel supervised method and also an unsupervised method to train the prefixes for single - aspect control while the combination of these two methods can achieve multi - aspect control .", "forward": true, "src_ids": "2022.findings-acl.229_892"}
+{"input": "small attribute - specific vectors is used for Task| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "small attribute - specific vectors", "output": "natural language generation", "neg_sample": ["small attribute - specific vectors is used for Task", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "In this work , we propose a novel lightweight framework for controllable GPT2 generation , which utilizes a set of small attribute - specific vectors , called prefixes ( Li and Liang , 2021 ) , to steer natural language generation .", "forward": true, "src_ids": "2022.findings-acl.229_893"}
+{"input": "prefixes is done by using Method| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "prefixes", "output": "unsupervised method", "neg_sample": ["prefixes is done by using Method", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "We propose a novel supervised method and also an unsupervised method to train the prefixes for single - aspect control while the combination of these two methods can achieve multi - aspect control .", "forward": false, "src_ids": "2022.findings-acl.229_894"}
+{"input": "single - aspect control is done by using Method| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "single - aspect control", "output": "unsupervised method", "neg_sample": ["single - aspect control is done by using Method", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "We propose a novel supervised method and also an unsupervised method to train the prefixes for single - aspect control while the combination of these two methods can achieve multi - aspect control .", "forward": false, "src_ids": "2022.findings-acl.229_895"}
+{"input": "unsupervised method is used for Task| context: to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator .", "entity": "unsupervised method", "output": "single - aspect control", "neg_sample": ["unsupervised method is used for Task", "to guide the generation of large pretrained language models ( lm ) , previous work has focused on directly fine - tuning the language model or utilizing an attribute discriminator ."], "relation": "used for", "id": "2022.findings-acl.229", "year": 2022, "rel_sent": "We propose a novel supervised method and also an unsupervised method to train the prefixes for single - aspect control while the combination of these two methods can achieve multi - aspect control .", "forward": true, "src_ids": "2022.findings-acl.229_896"}
+{"input": "tokenizer is used for Method| context: subword tokenization is a commonly used input pre - processing step in most recent nlp models . its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations . additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains .", "entity": "tokenizer", "output": "end - to - end task learning", "neg_sample": ["tokenizer is used for Method", "subword tokenization is a commonly used input pre - processing step in most recent nlp models .", "its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations .", "additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains ."], "relation": "used for", "id": "2022.repl4nlp-1.10", "year": 2022, "rel_sent": "Unlike the predefined and fixed vocabularies in subword methods , our tokenizer allows end - to - end task learning , resulting in optimal task - specific tokenization .", "forward": true, "src_ids": "2022.repl4nlp-1.10_897"}
+{"input": "subword methods is done by using OtherScientificTerm| context: subword tokenization is a commonly used input pre - processing step in most recent nlp models . however , it limits the models ' ability to leverage end - to - end task learning . its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations . additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains .", "entity": "subword methods", "output": "predefined and fixed vocabularies", "neg_sample": ["subword methods is done by using OtherScientificTerm", "subword tokenization is a commonly used input pre - processing step in most recent nlp models .", "however , it limits the models ' ability to leverage end - to - end task learning .", "its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations .", "additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains ."], "relation": "used for", "id": "2022.repl4nlp-1.10", "year": 2022, "rel_sent": "Unlike the predefined and fixed vocabularies in subword methods , our tokenizer allows end - to - end task learning , resulting in optimal task - specific tokenization .", "forward": false, "src_ids": "2022.repl4nlp-1.10_898"}
+{"input": "predefined and fixed vocabularies is used for Method| context: subword tokenization is a commonly used input pre - processing step in most recent nlp models . however , it limits the models ' ability to leverage end - to - end task learning . its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations . additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains .", "entity": "predefined and fixed vocabularies", "output": "subword methods", "neg_sample": ["predefined and fixed vocabularies is used for Method", "subword tokenization is a commonly used input pre - processing step in most recent nlp models .", "however , it limits the models ' ability to leverage end - to - end task learning .", "its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations .", "additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains ."], "relation": "used for", "id": "2022.repl4nlp-1.10", "year": 2022, "rel_sent": "Unlike the predefined and fixed vocabularies in subword methods , our tokenizer allows end - to - end task learning , resulting in optimal task - specific tokenization .", "forward": true, "src_ids": "2022.repl4nlp-1.10_899"}
+{"input": "optimal task - specific tokenization is done by using Method| context: subword tokenization is a commonly used input pre - processing step in most recent nlp models . however , it limits the models ' ability to leverage end - to - end task learning . its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations . additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains .", "entity": "optimal task - specific tokenization", "output": "tokenizer", "neg_sample": ["optimal task - specific tokenization is done by using Method", "subword tokenization is a commonly used input pre - processing step in most recent nlp models .", "however , it limits the models ' ability to leverage end - to - end task learning .", "its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations .", "additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains ."], "relation": "used for", "id": "2022.repl4nlp-1.10", "year": 2022, "rel_sent": "Unlike the predefined and fixed vocabularies in subword methods , our tokenizer allows end - to - end task learning , resulting in optimal task - specific tokenization .", "forward": false, "src_ids": "2022.repl4nlp-1.10_900"}
+{"input": "end - to - end task learning is done by using Method| context: subword tokenization is a commonly used input pre - processing step in most recent nlp models . however , it limits the models ' ability to leverage end - to - end task learning . its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations . additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains .", "entity": "end - to - end task learning", "output": "tokenizer", "neg_sample": ["end - to - end task learning is done by using Method", "subword tokenization is a commonly used input pre - processing step in most recent nlp models .", "however , it limits the models ' ability to leverage end - to - end task learning .", "its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations .", "additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains ."], "relation": "used for", "id": "2022.repl4nlp-1.10", "year": 2022, "rel_sent": "Unlike the predefined and fixed vocabularies in subword methods , our tokenizer allows end - to - end task learning , resulting in optimal task - specific tokenization .", "forward": false, "src_ids": "2022.repl4nlp-1.10_901"}
+{"input": "tokenizer is used for Task| context: subword tokenization is a commonly used input pre - processing step in most recent nlp models . however , it limits the models ' ability to leverage end - to - end task learning . its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations . additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains .", "entity": "tokenizer", "output": "optimal task - specific tokenization", "neg_sample": ["tokenizer is used for Task", "subword tokenization is a commonly used input pre - processing step in most recent nlp models .", "however , it limits the models ' ability to leverage end - to - end task learning .", "its frequency - based vocabulary creation compromises tokenization in low - resource languages , leading models to produce suboptimal representations .", "additionally , the dependency on a fixed vocabulary limits the subword models ' adaptability across languages and domains ."], "relation": "used for", "id": "2022.repl4nlp-1.10", "year": 2022, "rel_sent": "Unlike the predefined and fixed vocabularies in subword methods , our tokenizer allows end - to - end task learning , resulting in optimal task - specific tokenization .", "forward": true, "src_ids": "2022.repl4nlp-1.10_902"}
+{"input": "neural constituency parsing is done by using OtherScientificTerm| context: thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features .", "entity": "neural constituency parsing", "output": "non - local features", "neg_sample": ["neural constituency parsing is done by using OtherScientificTerm", "thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features ."], "relation": "used for", "id": "2022.acl-long.146", "year": 2022, "rel_sent": "Investigating Non - local Features for Neural Constituency Parsing.", "forward": false, "src_ids": "2022.acl-long.146_903"}
+{"input": "local span - based parser is done by using OtherScientificTerm| context: thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features .", "entity": "local span - based parser", "output": "non - local features", "neg_sample": ["local span - based parser is done by using OtherScientificTerm", "thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features ."], "relation": "used for", "id": "2022.acl-long.146", "year": 2022, "rel_sent": "In this paper , we investigate injecting non - local features into the training process of a local span - based parser , by predicting constituent n - gram non - local patterns and ensuring consistency between non - local patterns and local constituents .", "forward": false, "src_ids": "2022.acl-long.146_904"}
+{"input": "non - local features is used for Task| context: thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features . recently , it has been shown that non - local features in crf structures lead to improvements .", "entity": "non - local features", "output": "neural constituency parsing", "neg_sample": ["non - local features is used for Task", "thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features .", "recently , it has been shown that non - local features in crf structures lead to improvements ."], "relation": "used for", "id": "2022.acl-long.146", "year": 2022, "rel_sent": "Investigating Non - local Features for Neural Constituency Parsing.", "forward": true, "src_ids": "2022.acl-long.146_905"}
+{"input": "non - local features is used for Method| context: thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features . recently , it has been shown that non - local features in crf structures lead to improvements .", "entity": "non - local features", "output": "local span - based parser", "neg_sample": ["non - local features is used for Method", "thanks to the strong representation power of neural encoders , neural chart - based parsers have achieved highly competitive performance by using local features .", "recently , it has been shown that non - local features in crf structures lead to improvements ."], "relation": "used for", "id": "2022.acl-long.146", "year": 2022, "rel_sent": "In this paper , we investigate injecting non - local features into the training process of a local span - based parser , by predicting constituent n - gram non - local patterns and ensuring consistency between non - local patterns and local constituents .", "forward": true, "src_ids": "2022.acl-long.146_906"}
+{"input": "snp is done by using Method| context: self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech . recent developments in genomics has also adopted these pre - training methods for genome understanding . however , they focus only on understanding haploid sequences , which hinders their applicability towards understanding genetic variations , also known as single nucleotide polymorphisms ( snps ) , which is crucial for genome - wide association study .", "entity": "snp", "output": "snp2vec", "neg_sample": ["snp is done by using Method", "self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech .", "recent developments in genomics has also adopted these pre - training methods for genome understanding .", "however , they focus only on understanding haploid sequences , which hinders their applicability towards understanding genetic variations , also known as single nucleotide polymorphisms ( snps ) , which is crucial for genome - wide association study ."], "relation": "used for", "id": "2022.bionlp-1.14", "year": 2022, "rel_sent": "In this paper , we introduce SNP2Vec , a scalable self - supervised pre - training approach for understanding SNP .", "forward": false, "src_ids": "2022.bionlp-1.14_907"}
+{"input": "long - sequence genomics modeling is done by using Method| context: self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech . recent developments in genomics has also adopted these pre - training methods for genome understanding . however , they focus only on understanding haploid sequences , which hinders their applicability towards understanding genetic variations , also known as single nucleotide polymorphisms ( snps ) , which is crucial for genome - wide association study .", "entity": "long - sequence genomics modeling", "output": "snp2vec", "neg_sample": ["long - sequence genomics modeling is done by using Method", "self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech .", "recent developments in genomics has also adopted these pre - training methods for genome understanding .", "however , they focus only on understanding haploid sequences , which hinders their applicability towards understanding genetic variations , also known as single nucleotide polymorphisms ( snps ) , which is crucial for genome - wide association study ."], "relation": "used for", "id": "2022.bionlp-1.14", "year": 2022, "rel_sent": "We apply SNP2Vec to perform long - sequence genomics modeling , and we evaluate the effectiveness of our approach on predicting Alzheimer 's disease risk in a Chinese cohort .", "forward": false, "src_ids": "2022.bionlp-1.14_908"}
+{"input": "scalable self - supervised pre - training approach is used for OtherScientificTerm| context: self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech . recent developments in genomics has also adopted these pre - training methods for genome understanding .", "entity": "scalable self - supervised pre - training approach", "output": "snp", "neg_sample": ["scalable self - supervised pre - training approach is used for OtherScientificTerm", "self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech .", "recent developments in genomics has also adopted these pre - training methods for genome understanding ."], "relation": "used for", "id": "2022.bionlp-1.14", "year": 2022, "rel_sent": "In this paper , we introduce SNP2Vec , a scalable self - supervised pre - training approach for understanding SNP .", "forward": true, "src_ids": "2022.bionlp-1.14_909"}
+{"input": "snp2vec is used for OtherScientificTerm| context: self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech . recent developments in genomics has also adopted these pre - training methods for genome understanding .", "entity": "snp2vec", "output": "snp", "neg_sample": ["snp2vec is used for OtherScientificTerm", "self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech .", "recent developments in genomics has also adopted these pre - training methods for genome understanding ."], "relation": "used for", "id": "2022.bionlp-1.14", "year": 2022, "rel_sent": "In this paper , we introduce SNP2Vec , a scalable self - supervised pre - training approach for understanding SNP .", "forward": true, "src_ids": "2022.bionlp-1.14_910"}
+{"input": "snp2vec is used for Task| context: self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech . recent developments in genomics has also adopted these pre - training methods for genome understanding . however , they focus only on understanding haploid sequences , which hinders their applicability towards understanding genetic variations , also known as single nucleotide polymorphisms ( snps ) , which is crucial for genome - wide association study .", "entity": "snp2vec", "output": "long - sequence genomics modeling", "neg_sample": ["snp2vec is used for Task", "self - supervised pre - training methods have brought remarkable breakthroughs in the understanding of text , image , and speech .", "recent developments in genomics has also adopted these pre - training methods for genome understanding .", "however , they focus only on understanding haploid sequences , which hinders their applicability towards understanding genetic variations , also known as single nucleotide polymorphisms ( snps ) , which is crucial for genome - wide association study ."], "relation": "used for", "id": "2022.bionlp-1.14", "year": 2022, "rel_sent": "We apply SNP2Vec to perform long - sequence genomics modeling , and we evaluate the effectiveness of our approach on predicting Alzheimer 's disease risk in a Chinese cohort .", "forward": true, "src_ids": "2022.bionlp-1.14_911"}
+{"input": "translating out - of - domain sentences is done by using Method| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "translating out - of - domain sentences", "output": "mt systems", "neg_sample": ["translating out - of - domain sentences is done by using Method", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "It aims to alleviate the performance degradation of advanced MT systems in translating out - of - domain sentences by coordinating with an additional token - level feature - based retrieval module constructed from in - domain data .", "forward": false, "src_ids": "2022.acl-long.154_912"}
+{"input": "mt systems is used for Task| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "mt systems", "output": "translating out - of - domain sentences", "neg_sample": ["mt systems is used for Task", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "It aims to alleviate the performance degradation of advanced MT systems in translating out - of - domain sentences by coordinating with an additional token - level feature - based retrieval module constructed from in - domain data .", "forward": true, "src_ids": "2022.acl-long.154_913"}
+{"input": "retrieval efficiency is done by using Task| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "retrieval efficiency", "output": "clustering", "neg_sample": ["retrieval efficiency is done by using Task", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "To make it practical , in this paper , we explore a more efficient kNN - MT and propose to use clustering to improve the retrieval efficiency .", "forward": false, "src_ids": "2022.acl-long.154_914"}
+{"input": "clustering is used for Metric| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "clustering", "output": "retrieval efficiency", "neg_sample": ["clustering is used for Metric", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "To make it practical , in this paper , we explore a more efficient kNN - MT and propose to use clustering to improve the retrieval efficiency .", "forward": true, "src_ids": "2022.acl-long.154_915"}
+{"input": "feature reduction is done by using Method| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "feature reduction", "output": "cluster - based compact network", "neg_sample": ["feature reduction is done by using Method", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "Concretely , we first propose a cluster - based Compact Network for feature reduction in a contrastive learning manner to compress context features into 90+% lower dimensional vectors .", "forward": false, "src_ids": "2022.acl-long.154_916"}
+{"input": "context features is done by using Method| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "context features", "output": "cluster - based compact network", "neg_sample": ["context features is done by using Method", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "Concretely , we first propose a cluster - based Compact Network for feature reduction in a contrastive learning manner to compress context features into 90+% lower dimensional vectors .", "forward": false, "src_ids": "2022.acl-long.154_917"}
+{"input": "cluster - based compact network is used for Task| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "cluster - based compact network", "output": "feature reduction", "neg_sample": ["cluster - based compact network is used for Task", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "Concretely , we first propose a cluster - based Compact Network for feature reduction in a contrastive learning manner to compress context features into 90+% lower dimensional vectors .", "forward": true, "src_ids": "2022.acl-long.154_918"}
+{"input": "cluster - based compact network is used for OtherScientificTerm| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "cluster - based compact network", "output": "context features", "neg_sample": ["cluster - based compact network is used for OtherScientificTerm", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "Concretely , we first propose a cluster - based Compact Network for feature reduction in a contrastive learning manner to compress context features into 90+% lower dimensional vectors .", "forward": true, "src_ids": "2022.acl-long.154_919"}
+{"input": "redundant nodes is done by using Method| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "redundant nodes", "output": "cluster - based pruning solution", "neg_sample": ["redundant nodes is done by using Method", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "We then suggest a cluster - based pruning solution tofilter out 10 % 40 % redundant nodes in large datastores while retaining translation quality .", "forward": false, "src_ids": "2022.acl-long.154_920"}
+{"input": "cluster - based pruning solution is used for OtherScientificTerm| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "cluster - based pruning solution", "output": "redundant nodes", "neg_sample": ["cluster - based pruning solution is used for OtherScientificTerm", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.154", "year": 2022, "rel_sent": "We then suggest a cluster - based pruning solution tofilter out 10 % 40 % redundant nodes in large datastores while retaining translation quality .", "forward": true, "src_ids": "2022.acl-long.154_921"}
+{"input": "friendly adversarial data is done by using Method| context: recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness . however , we observe that a too large number of search steps can hurt accuracy .", "entity": "friendly adversarial data", "output": "friendly adversarial data augmentation ( fada )", "neg_sample": ["friendly adversarial data is done by using Method", "recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness .", "however , we observe that a too large number of search steps can hurt accuracy ."], "relation": "used for", "id": "2022.findings-acl.246", "year": 2022, "rel_sent": "Inspired by this , we propose friendly adversarial data augmentation ( FADA ) to generate friendly adversarial data .", "forward": false, "src_ids": "2022.findings-acl.246_922"}
+{"input": "friendly adversarial data augmentation ( fada ) is used for Material| context: recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness . however , we observe that a too large number of search steps can hurt accuracy .", "entity": "friendly adversarial data augmentation ( fada )", "output": "friendly adversarial data", "neg_sample": ["friendly adversarial data augmentation ( fada ) is used for Material", "recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness .", "however , we observe that a too large number of search steps can hurt accuracy ."], "relation": "used for", "id": "2022.findings-acl.246", "year": 2022, "rel_sent": "Inspired by this , we propose friendly adversarial data augmentation ( FADA ) to generate friendly adversarial data .", "forward": true, "src_ids": "2022.findings-acl.246_923"}
+{"input": "geometry - aware adversarial training ( gat ) is used for Material| context: recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness . however , we observe that a too large number of search steps can hurt accuracy .", "entity": "geometry - aware adversarial training ( gat )", "output": "friendly adversarial data", "neg_sample": ["geometry - aware adversarial training ( gat ) is used for Material", "recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness .", "however , we observe that a too large number of search steps can hurt accuracy ."], "relation": "used for", "id": "2022.findings-acl.246", "year": 2022, "rel_sent": "On top of FADA , we propose geometry - aware adversarial training ( GAT ) to perform adversarial training on friendly adversarial data so that we can save a large number of search steps .", "forward": true, "src_ids": "2022.findings-acl.246_924"}
+{"input": "adversarial training is done by using Method| context: recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness . however , we observe that a too large number of search steps can hurt accuracy .", "entity": "adversarial training", "output": "geometry - aware adversarial training ( gat )", "neg_sample": ["adversarial training is done by using Method", "recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness .", "however , we observe that a too large number of search steps can hurt accuracy ."], "relation": "used for", "id": "2022.findings-acl.246", "year": 2022, "rel_sent": "On top of FADA , we propose geometry - aware adversarial training ( GAT ) to perform adversarial training on friendly adversarial data so that we can save a large number of search steps .", "forward": false, "src_ids": "2022.findings-acl.246_925"}
+{"input": "friendly adversarial data is done by using Method| context: recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness . however , we observe that a too large number of search steps can hurt accuracy .", "entity": "friendly adversarial data", "output": "geometry - aware adversarial training ( gat )", "neg_sample": ["friendly adversarial data is done by using Method", "recent studies have found that removing the norm - bounded projection and increasing search steps in adversarial training can significantly improve robustness .", "however , we observe that a too large number of search steps can hurt accuracy ."], "relation": "used for", "id": "2022.findings-acl.246", "year": 2022, "rel_sent": "On top of FADA , we propose geometry - aware adversarial training ( GAT ) to perform adversarial training on friendly adversarial data so that we can save a large number of search steps .", "forward": false, "src_ids": "2022.findings-acl.246_926"}
+{"input": "geometry - aware adversarial training ( gat ) is used for Method| context: however , we observe that a too large number of search steps can hurt accuracy .", "entity": "geometry - aware adversarial training ( gat )", "output": "adversarial training", "neg_sample": ["geometry - aware adversarial training ( gat ) is used for Method", "however , we observe that a too large number of search steps can hurt accuracy ."], "relation": "used for", "id": "2022.findings-acl.246", "year": 2022, "rel_sent": "On top of FADA , we propose geometry - aware adversarial training ( GAT ) to perform adversarial training on friendly adversarial data so that we can save a large number of search steps .", "forward": true, "src_ids": "2022.findings-acl.246_927"}
+{"input": "computational argumentation is done by using Method| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "computational argumentation", "output": "fair and argumentative language modeling", "neg_sample": ["computational argumentation is done by using Method", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "Fair and Argumentative Language Modeling for Computational Argumentation.", "forward": false, "src_ids": "2022.acl-long.541_928"}
+{"input": "bias measurement is done by using Method| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "bias measurement", "output": "abba", "neg_sample": ["bias measurement is done by using Method", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "To this end , we introduce ABBA , a novel resource for bias measurement specifically tailored to argumentation .", "forward": false, "src_ids": "2022.acl-long.541_929"}
+{"input": "argumentation is done by using Task| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "argumentation", "output": "bias measurement", "neg_sample": ["argumentation is done by using Task", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "To this end , we introduce ABBA , a novel resource for bias measurement specifically tailored to argumentation .", "forward": false, "src_ids": "2022.acl-long.541_930"}
+{"input": "abba is used for Task| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "abba", "output": "bias measurement", "neg_sample": ["abba is used for Task", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "To this end , we introduce ABBA , a novel resource for bias measurement specifically tailored to argumentation .", "forward": true, "src_ids": "2022.acl-long.541_931"}
+{"input": "intrinsic bias is done by using Method| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "intrinsic bias", "output": "argumentative fine - tuning", "neg_sample": ["intrinsic bias is done by using Method", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "We employ our resource to assess the effect of argumentative fine - tuning and debiasing on the intrinsic bias found in transformer - based language models using a lightweight adapter - based approach that is more sustainable and parameter - efficient than full fine - tuning .", "forward": false, "src_ids": "2022.acl-long.541_932"}
+{"input": "intrinsic bias is done by using Task| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "intrinsic bias", "output": "debiasing", "neg_sample": ["intrinsic bias is done by using Task", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "We employ our resource to assess the effect of argumentative fine - tuning and debiasing on the intrinsic bias found in transformer - based language models using a lightweight adapter - based approach that is more sustainable and parameter - efficient than full fine - tuning .", "forward": false, "src_ids": "2022.acl-long.541_933"}
+{"input": "argumentative fine - tuning is used for OtherScientificTerm| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "argumentative fine - tuning", "output": "intrinsic bias", "neg_sample": ["argumentative fine - tuning is used for OtherScientificTerm", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "We employ our resource to assess the effect of argumentative fine - tuning and debiasing on the intrinsic bias found in transformer - based language models using a lightweight adapter - based approach that is more sustainable and parameter - efficient than full fine - tuning .", "forward": true, "src_ids": "2022.acl-long.541_934"}
+{"input": "debiasing is used for OtherScientificTerm| context: although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy .", "entity": "debiasing", "output": "intrinsic bias", "neg_sample": ["debiasing is used for OtherScientificTerm", "although much work in nlp has focused on measuring and mitigating stereotypical bias in semantic spaces , research addressing bias in computational argumentation is still in its infancy ."], "relation": "used for", "id": "2022.acl-long.541", "year": 2022, "rel_sent": "We employ our resource to assess the effect of argumentative fine - tuning and debiasing on the intrinsic bias found in transformer - based language models using a lightweight adapter - based approach that is more sustainable and parameter - efficient than full fine - tuning .", "forward": true, "src_ids": "2022.acl-long.541_935"}
+{"input": "social media comments is done by using Material| context: this paper presents the overview of the shared task on emotional analysis in tamil . the result of the shared task is presented at the workshop .", "entity": "social media comments", "output": "emotions annotated data", "neg_sample": ["social media comments is done by using Material", "this paper presents the overview of the shared task on emotional analysis in tamil .", "the result of the shared task is presented at the workshop ."], "relation": "used for", "id": "2022.dravidianlangtech-1.42", "year": 2022, "rel_sent": "Task A is carried with 11 emotions annotated data for social media comments in Tamil and Task B is organized with 31 fine - grained emotion annotated data for social media comments in Tamil .", "forward": false, "src_ids": "2022.dravidianlangtech-1.42_936"}
+{"input": "emotions annotated data is used for Material| context: this paper presents the overview of the shared task on emotional analysis in tamil . the result of the shared task is presented at the workshop .", "entity": "emotions annotated data", "output": "social media comments", "neg_sample": ["emotions annotated data is used for Material", "this paper presents the overview of the shared task on emotional analysis in tamil .", "the result of the shared task is presented at the workshop ."], "relation": "used for", "id": "2022.dravidianlangtech-1.42", "year": 2022, "rel_sent": "Task A is carried with 11 emotions annotated data for social media comments in Tamil and Task B is organized with 31 fine - grained emotion annotated data for social media comments in Tamil .", "forward": true, "src_ids": "2022.dravidianlangtech-1.42_937"}
+{"input": "fine - grained emotion annotated data is used for Material| context: this paper presents the overview of the shared task on emotional analysis in tamil . the result of the shared task is presented at the workshop .", "entity": "fine - grained emotion annotated data", "output": "social media comments", "neg_sample": ["fine - grained emotion annotated data is used for Material", "this paper presents the overview of the shared task on emotional analysis in tamil .", "the result of the shared task is presented at the workshop ."], "relation": "used for", "id": "2022.dravidianlangtech-1.42", "year": 2022, "rel_sent": "Task A is carried with 11 emotions annotated data for social media comments in Tamil and Task B is organized with 31 fine - grained emotion annotated data for social media comments in Tamil .", "forward": true, "src_ids": "2022.dravidianlangtech-1.42_938"}
+{"input": "classifiers is done by using Method| context: we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence . there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span .", "entity": "classifiers", "output": "seed bootstrapping technique", "neg_sample": ["classifiers is done by using Method", "we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence .", "there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span ."], "relation": "used for", "id": "2022.findings-acl.101", "year": 2022, "rel_sent": "A seed bootstrapping technique prepares the data to train these classifiers .", "forward": false, "src_ids": "2022.findings-acl.101_939"}
+{"input": "inductive bias is done by using Method| context: we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence . there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span .", "entity": "inductive bias", "output": "minimal heuristics", "neg_sample": ["inductive bias is done by using Method", "we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence .", "there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span ."], "relation": "used for", "id": "2022.findings-acl.101", "year": 2022, "rel_sent": "Our analyses further validate that such an approach in conjunction with weak supervision using prior branching knowledge of a known language ( left / right - branching ) and minimal heuristics injects strong inductive bias into the parser , achieving 63.1 F1 on the English ( PTB ) test set .", "forward": false, "src_ids": "2022.findings-acl.101_940"}
+{"input": "minimal heuristics is used for OtherScientificTerm| context: we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence . there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span .", "entity": "minimal heuristics", "output": "inductive bias", "neg_sample": ["minimal heuristics is used for OtherScientificTerm", "we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence .", "there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span ."], "relation": "used for", "id": "2022.findings-acl.101", "year": 2022, "rel_sent": "Our analyses further validate that such an approach in conjunction with weak supervision using prior branching knowledge of a known language ( left / right - branching ) and minimal heuristics injects strong inductive bias into the parser , achieving 63.1 F1 on the English ( PTB ) test set .", "forward": true, "src_ids": "2022.findings-acl.101_941"}
+{"input": "japanese ( ktb ) is done by using Material| context: we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence . there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span .", "entity": "japanese ( ktb )", "output": "treebanks", "neg_sample": ["japanese ( ktb ) is done by using Material", "we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence .", "there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span ."], "relation": "used for", "id": "2022.findings-acl.101", "year": 2022, "rel_sent": "In addition , we show the effectiveness of our architecture by evaluating on treebanks for Chinese ( CTB ) and Japanese ( KTB ) and achieve new state - of - the - art results .", "forward": false, "src_ids": "2022.findings-acl.101_942"}
+{"input": "treebanks is used for Material| context: we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence . there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span .", "entity": "treebanks", "output": "japanese ( ktb )", "neg_sample": ["treebanks is used for Material", "we introduce a method for unsupervised parsing that relies on bootstrapping classifiers to identify if a node dominates a specific span in a sentence .", "there are two types of classifiers , an inside classifier that acts on a span , and an outside classifier that acts on everything outside of a given span ."], "relation": "used for", "id": "2022.findings-acl.101", "year": 2022, "rel_sent": "In addition , we show the effectiveness of our architecture by evaluating on treebanks for Chinese ( CTB ) and Japanese ( KTB ) and achieve new state - of - the - art results .", "forward": true, "src_ids": "2022.findings-acl.101_943"}
+{"input": "understanding of rare words is done by using Method| context: since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus . therefore , the embeddings of rare words on the tail are usually poorly optimized .", "entity": "understanding of rare words", "output": "dict - bert", "neg_sample": ["understanding of rare words is done by using Method", "since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus .", "therefore , the embeddings of rare words on the tail are usually poorly optimized ."], "relation": "used for", "id": "2022.findings-acl.150", "year": 2022, "rel_sent": "Extensive experiments demonstrate that Dict - BERT can significantly improve the understanding of rare words and boost model performance on various NLP downstream tasks .", "forward": false, "src_ids": "2022.findings-acl.150_944"}
+{"input": "language modeling representation is done by using Task| context: since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus . therefore , the embeddings of rare words on the tail are usually poorly optimized .", "entity": "language modeling representation", "output": "self - supervised pre - training tasks", "neg_sample": ["language modeling representation is done by using Task", "since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus .", "therefore , the embeddings of rare words on the tail are usually poorly optimized ."], "relation": "used for", "id": "2022.findings-acl.150", "year": 2022, "rel_sent": "In addition to training with the masked language modeling objective , we propose two novel self - supervised pre - training tasks on word and sentence - level alignment between input text sequence and rare word definitions to enhance language modeling representation with dictionary .", "forward": false, "src_ids": "2022.findings-acl.150_945"}
+{"input": "self - supervised pre - training tasks is used for Method| context: since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus . therefore , the embeddings of rare words on the tail are usually poorly optimized .", "entity": "self - supervised pre - training tasks", "output": "language modeling representation", "neg_sample": ["self - supervised pre - training tasks is used for Method", "since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus .", "therefore , the embeddings of rare words on the tail are usually poorly optimized ."], "relation": "used for", "id": "2022.findings-acl.150", "year": 2022, "rel_sent": "In addition to training with the masked language modeling objective , we propose two novel self - supervised pre - training tasks on word and sentence - level alignment between input text sequence and rare word definitions to enhance language modeling representation with dictionary .", "forward": true, "src_ids": "2022.findings-acl.150_946"}
+{"input": "dict - bert is used for Task| context: since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus . therefore , the embeddings of rare words on the tail are usually poorly optimized .", "entity": "dict - bert", "output": "understanding of rare words", "neg_sample": ["dict - bert is used for Task", "since plms capture word semantics in different contexts , the quality of word representations highly depends on word frequency , which usually follows a heavy - tailed distributions in the pre - training corpus .", "therefore , the embeddings of rare words on the tail are usually poorly optimized ."], "relation": "used for", "id": "2022.findings-acl.150", "year": 2022, "rel_sent": "Extensive experiments demonstrate that Dict - BERT can significantly improve the understanding of rare words and boost model performance on various NLP downstream tasks .", "forward": true, "src_ids": "2022.findings-acl.150_947"}
+{"input": "identifying abusive comments is done by using Method| context: social media platforms along with many other public forums on the internet have shown a significant rise in the cases of abusive behavior such as misogynism , misandry , homophobia , and cyberbullying .", "entity": "identifying abusive comments", "output": "transformer based approach", "neg_sample": ["identifying abusive comments is done by using Method", "social media platforms along with many other public forums on the internet have shown a significant rise in the cases of abusive behavior such as misogynism , misandry , homophobia , and cyberbullying ."], "relation": "used for", "id": "2022.dravidianlangtech-1.25", "year": 2022, "rel_sent": "SSNCSE NLP@TamilNLP - ACL2022 : Transformer based approach for detection of abusive comment for Tamil language.", "forward": false, "src_ids": "2022.dravidianlangtech-1.25_948"}
+{"input": "transformer based approach is used for Task| context: social media platforms along with many other public forums on the internet have shown a significant rise in the cases of abusive behavior such as misogynism , misandry , homophobia , and cyberbullying .", "entity": "transformer based approach", "output": "identifying abusive comments", "neg_sample": ["transformer based approach is used for Task", "social media platforms along with many other public forums on the internet have shown a significant rise in the cases of abusive behavior such as misogynism , misandry , homophobia , and cyberbullying ."], "relation": "used for", "id": "2022.dravidianlangtech-1.25", "year": 2022, "rel_sent": "SSNCSE NLP@TamilNLP - ACL2022 : Transformer based approach for detection of abusive comment for Tamil language.", "forward": true, "src_ids": "2022.dravidianlangtech-1.25_949"}
+{"input": "clinical nlp applications is done by using Method| context: this work presents the first large - scale biomedical spanish language models trained from scratch , using large biomedical corpora consisting of a total of 1.1b tokens and an ehr corpus of 95 m tokens .", "entity": "clinical nlp applications", "output": "pretrained biomedical language models", "neg_sample": ["clinical nlp applications is done by using Method", "this work presents the first large - scale biomedical spanish language models trained from scratch , using large biomedical corpora consisting of a total of 1.1b tokens and an ehr corpus of 95 m tokens ."], "relation": "used for", "id": "2022.bionlp-1.19", "year": 2022, "rel_sent": "Pretrained Biomedical Language Models for Clinical NLP in Spanish.", "forward": false, "src_ids": "2022.bionlp-1.19_950"}
+{"input": "pretrained biomedical language models is used for Task| context: this work presents the first large - scale biomedical spanish language models trained from scratch , using large biomedical corpora consisting of a total of 1.1b tokens and an ehr corpus of 95 m tokens .", "entity": "pretrained biomedical language models", "output": "clinical nlp applications", "neg_sample": ["pretrained biomedical language models is used for Task", "this work presents the first large - scale biomedical spanish language models trained from scratch , using large biomedical corpora consisting of a total of 1.1b tokens and an ehr corpus of 95 m tokens ."], "relation": "used for", "id": "2022.bionlp-1.19", "year": 2022, "rel_sent": "Pretrained Biomedical Language Models for Clinical NLP in Spanish.", "forward": true, "src_ids": "2022.bionlp-1.19_951"}
+{"input": "spanish is done by using Method| context: this work presents the first large - scale biomedical spanish language models trained from scratch , using large biomedical corpora consisting of a total of 1.1b tokens and an ehr corpus of 95 m tokens .", "entity": "spanish", "output": "domain - specific models", "neg_sample": ["spanish is done by using Method", "this work presents the first large - scale biomedical spanish language models trained from scratch , using large biomedical corpora consisting of a total of 1.1b tokens and an ehr corpus of 95 m tokens ."], "relation": "used for", "id": "2022.bionlp-1.19", "year": 2022, "rel_sent": "We compared them against general - domain and other domain - specific models for Spanish on three clinical NER tasks .", "forward": false, "src_ids": "2022.bionlp-1.19_952"}
+{"input": "multimodal troll memes is done by using Task| context: multimodal sentiment analysis deals with the identification of sentiment from video . in addition to video data , the task requires the analysis of corresponding text and audiofeatures for the classification of movie reviews intofive classes .", "entity": "multimodal troll memes", "output": "troll meme classification task", "neg_sample": ["multimodal troll memes is done by using Task", "multimodal sentiment analysis deals with the identification of sentiment from video .", "in addition to video data , the task requires the analysis of corresponding text and audiofeatures for the classification of movie reviews intofive classes ."], "relation": "used for", "id": "2022.dravidianlangtech-1.39", "year": 2022, "rel_sent": "The Troll meme classification task aims to classify multimodal Troll memes into two categories .", "forward": false, "src_ids": "2022.dravidianlangtech-1.39_953"}
+{"input": "troll meme classification task is used for Material| context: multimodal sentiment analysis deals with the identification of sentiment from video . in addition to video data , the task requires the analysis of corresponding text and audiofeatures for the classification of movie reviews intofive classes .", "entity": "troll meme classification task", "output": "multimodal troll memes", "neg_sample": ["troll meme classification task is used for Material", "multimodal sentiment analysis deals with the identification of sentiment from video .", "in addition to video data , the task requires the analysis of corresponding text and audiofeatures for the classification of movie reviews intofive classes ."], "relation": "used for", "id": "2022.dravidianlangtech-1.39", "year": 2022, "rel_sent": "The Troll meme classification task aims to classify multimodal Troll memes into two categories .", "forward": true, "src_ids": "2022.dravidianlangtech-1.39_954"}
+{"input": "later - layer representations of words is done by using OtherScientificTerm| context: for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . '", "entity": "later - layer representations of words", "output": "word order", "neg_sample": ["later - layer representations of words is done by using OtherScientificTerm", "for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . '"], "relation": "used for", "id": "2022.acl-short.71", "year": 2022, "rel_sent": "We find that , while early layer embeddings are largely lexical , word order is in fact crucial in defining the later - layer representations of words in semantically non - prototypical positions .", "forward": false, "src_ids": "2022.acl-short.71_955"}
+{"input": "contextualization process is done by using OtherScientificTerm| context: for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . '", "entity": "contextualization process", "output": "word order", "neg_sample": ["contextualization process is done by using OtherScientificTerm", "for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . '"], "relation": "used for", "id": "2022.acl-short.71", "year": 2022, "rel_sent": "Our experiments isolate the effect of word order on the contextualization process , and highlight how models use context in the uncommon , but critical , instances where it matters .", "forward": false, "src_ids": "2022.acl-short.71_956"}
+{"input": "word order is used for Method| context: because meaning can often be inferred from lexical semantics alone , word order is often a redundant cue in natural language . for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . ' recent work has shown large language models to be surprisingly word order invariant , but crucially has largely considered natural prototypical inputs , where compositional meaning mostly matches lexical expectations . to overcome this confound , we probe grammatical role representation in english bert and gpt-2 , on instances where lexical expectations are not sufficient , and word order knowledge is necessary for correct classification .", "entity": "word order", "output": "later - layer representations of words", "neg_sample": ["word order is used for Method", "because meaning can often be inferred from lexical semantics alone , word order is often a redundant cue in natural language .", "for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . '", "recent work has shown large language models to be surprisingly word order invariant , but crucially has largely considered natural prototypical inputs , where compositional meaning mostly matches lexical expectations .", "to overcome this confound , we probe grammatical role representation in english bert and gpt-2 , on instances where lexical expectations are not sufficient , and word order knowledge is necessary for correct classification ."], "relation": "used for", "id": "2022.acl-short.71", "year": 2022, "rel_sent": "We find that , while early layer embeddings are largely lexical , word order is in fact crucial in defining the later - layer representations of words in semantically non - prototypical positions .", "forward": true, "src_ids": "2022.acl-short.71_957"}
+{"input": "word order is used for Task| context: because meaning can often be inferred from lexical semantics alone , word order is often a redundant cue in natural language . for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . ' recent work has shown large language models to be surprisingly word order invariant , but crucially has largely considered natural prototypical inputs , where compositional meaning mostly matches lexical expectations . to overcome this confound , we probe grammatical role representation in english bert and gpt-2 , on instances where lexical expectations are not sufficient , and word order knowledge is necessary for correct classification .", "entity": "word order", "output": "contextualization process", "neg_sample": ["word order is used for Task", "because meaning can often be inferred from lexical semantics alone , word order is often a redundant cue in natural language .", "for example , the words chopped , chef , and onion are more likely used to convey ' the chef chopped the onion , ' not ' the onion chopped the chef . '", "recent work has shown large language models to be surprisingly word order invariant , but crucially has largely considered natural prototypical inputs , where compositional meaning mostly matches lexical expectations .", "to overcome this confound , we probe grammatical role representation in english bert and gpt-2 , on instances where lexical expectations are not sufficient , and word order knowledge is necessary for correct classification ."], "relation": "used for", "id": "2022.acl-short.71", "year": 2022, "rel_sent": "Our experiments isolate the effect of word order on the contextualization process , and highlight how models use context in the uncommon , but critical , instances where it matters .", "forward": true, "src_ids": "2022.acl-short.71_958"}
+{"input": "phonetic spelling errors is done by using OtherScientificTerm| context: in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries . however , resolving phonetic errors is a critical but much overlooked area . the query with phonetic spelling errors tends to appear correct based on pronunciation but is nonetheless inaccurate in spelling ( e.g. , ' bluetooth sound system ' vs. ' blutut sant sistam ' ) with numerous noisy forms and sparse occurrences .", "entity": "phonetic spelling errors", "output": "phonetics", "neg_sample": ["phonetic spelling errors is done by using OtherScientificTerm", "in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries .", "however , resolving phonetic errors is a critical but much overlooked area .", "the query with phonetic spelling errors tends to appear correct based on pronunciation but is nonetheless inaccurate in spelling ( e.g.", ", ' bluetooth sound system ' vs. ' blutut sant sistam ' ) with numerous noisy forms and sparse occurrences ."], "relation": "used for", "id": "2022.ecnlp-1.9", "year": 2022, "rel_sent": "In this work , we propose a generalized spelling correction system integrating phonetics to address phonetic errors in E - commerce search without additional latency cost .", "forward": false, "src_ids": "2022.ecnlp-1.9_959"}
+{"input": "phonetic spelling errors is done by using Method| context: in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries . however , resolving phonetic errors is a critical but much overlooked area . the query with phonetic spelling errors tends to appear correct based on pronunciation but is nonetheless inaccurate in spelling ( e.g. , ' bluetooth sound system ' vs. ' blutut sant sistam ' ) with numerous noisy forms and sparse occurrences .", "entity": "phonetic spelling errors", "output": "generalized spelling correction system", "neg_sample": ["phonetic spelling errors is done by using Method", "in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries .", "however , resolving phonetic errors is a critical but much overlooked area .", "the query with phonetic spelling errors tends to appear correct based on pronunciation but is nonetheless inaccurate in spelling ( e.g.", ", ' bluetooth sound system ' vs. ' blutut sant sistam ' ) with numerous noisy forms and sparse occurrences ."], "relation": "used for", "id": "2022.ecnlp-1.9", "year": 2022, "rel_sent": "In this work , we propose a generalized spelling correction system integrating phonetics to address phonetic errors in E - commerce search without additional latency cost .", "forward": false, "src_ids": "2022.ecnlp-1.9_960"}
+{"input": "generalized spelling correction system is used for OtherScientificTerm| context: in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries .", "entity": "generalized spelling correction system", "output": "phonetic spelling errors", "neg_sample": ["generalized spelling correction system is used for OtherScientificTerm", "in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries ."], "relation": "used for", "id": "2022.ecnlp-1.9", "year": 2022, "rel_sent": "In this work , we propose a generalized spelling correction system integrating phonetics to address phonetic errors in E - commerce search without additional latency cost .", "forward": true, "src_ids": "2022.ecnlp-1.9_961"}
+{"input": "phonetics is used for OtherScientificTerm| context: in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries .", "entity": "phonetics", "output": "phonetic spelling errors", "neg_sample": ["phonetics is used for OtherScientificTerm", "in e - commerce search , spelling correction plays an important role tofind desired products for customers in processing user - typed search queries ."], "relation": "used for", "id": "2022.ecnlp-1.9", "year": 2022, "rel_sent": "In this work , we propose a generalized spelling correction system integrating phonetics to address phonetic errors in E - commerce search without additional latency cost .", "forward": true, "src_ids": "2022.ecnlp-1.9_962"}
+{"input": "inconsistency detection is done by using Method| context: in the summarization domain , a key requirement for summaries is to be factually consistent with the input document . previous work has found that natural language inference ( nli ) models do not perform competitively when applied to inconsistency detection . in this work , we revisit the use of nli for inconsistency detection , finding that past work suffered from a mismatch in input granularity between nli datasets ( sentence - level ) , and inconsistency detection ( document level ) .", "entity": "inconsistency detection", "output": "re - visiting nli - based models", "neg_sample": ["inconsistency detection is done by using Method", "in the summarization domain , a key requirement for summaries is to be factually consistent with the input document .", "previous work has found that natural language inference ( nli ) models do not perform competitively when applied to inconsistency detection .", "in this work , we revisit the use of nli for inconsistency detection , finding that past work suffered from a mismatch in input granularity between nli datasets ( sentence - level ) , and inconsistency detection ( document level ) ."], "relation": "used for", "id": "2022.tacl-1.10", "year": 2022, "rel_sent": "SummaC : Re - Visiting NLI - based Models for Inconsistency Detection in Summarization.", "forward": false, "src_ids": "2022.tacl-1.10_963"}
+{"input": "re - visiting nli - based models is used for Task| context: in the summarization domain , a key requirement for summaries is to be factually consistent with the input document .", "entity": "re - visiting nli - based models", "output": "inconsistency detection", "neg_sample": ["re - visiting nli - based models is used for Task", "in the summarization domain , a key requirement for summaries is to be factually consistent with the input document ."], "relation": "used for", "id": "2022.tacl-1.10", "year": 2022, "rel_sent": "SummaC : Re - Visiting NLI - based Models for Inconsistency Detection in Summarization.", "forward": true, "src_ids": "2022.tacl-1.10_964"}
+{"input": "nli models is done by using Method| context: in the summarization domain , a key requirement for summaries is to be factually consistent with the input document . previous work has found that natural language inference ( nli ) models do not perform competitively when applied to inconsistency detection . in this work , we revisit the use of nli for inconsistency detection , finding that past work suffered from a mismatch in input granularity between nli datasets ( sentence - level ) , and inconsistency detection ( document level ) .", "entity": "nli models", "output": "light - weight method", "neg_sample": ["nli models is done by using Method", "in the summarization domain , a key requirement for summaries is to be factually consistent with the input document .", "previous work has found that natural language inference ( nli ) models do not perform competitively when applied to inconsistency detection .", "in this work , we revisit the use of nli for inconsistency detection , finding that past work suffered from a mismatch in input granularity between nli datasets ( sentence - level ) , and inconsistency detection ( document level ) ."], "relation": "used for", "id": "2022.tacl-1.10", "year": 2022, "rel_sent": "We provide a highly effective and light - weight method called SummaCConv that enables NLI models to be successfully used for this task by segmenting documents into sentence units and aggregating scores between pairs of sentences .", "forward": false, "src_ids": "2022.tacl-1.10_965"}
+{"input": "light - weight method is used for Method| context: in the summarization domain , a key requirement for summaries is to be factually consistent with the input document . previous work has found that natural language inference ( nli ) models do not perform competitively when applied to inconsistency detection . in this work , we revisit the use of nli for inconsistency detection , finding that past work suffered from a mismatch in input granularity between nli datasets ( sentence - level ) , and inconsistency detection ( document level ) .", "entity": "light - weight method", "output": "nli models", "neg_sample": ["light - weight method is used for Method", "in the summarization domain , a key requirement for summaries is to be factually consistent with the input document .", "previous work has found that natural language inference ( nli ) models do not perform competitively when applied to inconsistency detection .", "in this work , we revisit the use of nli for inconsistency detection , finding that past work suffered from a mismatch in input granularity between nli datasets ( sentence - level ) , and inconsistency detection ( document level ) ."], "relation": "used for", "id": "2022.tacl-1.10", "year": 2022, "rel_sent": "We provide a highly effective and light - weight method called SummaCConv that enables NLI models to be successfully used for this task by segmenting documents into sentence units and aggregating scores between pairs of sentences .", "forward": true, "src_ids": "2022.tacl-1.10_966"}
+{"input": "cross - lingual deception classification is done by using Method| context: with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language .", "entity": "cross - lingual deception classification", "output": "multilingual classification models", "neg_sample": ["cross - lingual deception classification is done by using Method", "with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language ."], "relation": "used for", "id": "2022.acl-srw.30", "year": 2022, "rel_sent": "We propose to study the efficacy of multilingual classification models vs translation for cross - lingual deception classification .", "forward": false, "src_ids": "2022.acl-srw.30_967"}
+{"input": "cross - lingual deception classification is done by using Task| context: with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language .", "entity": "cross - lingual deception classification", "output": "translation", "neg_sample": ["cross - lingual deception classification is done by using Task", "with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language ."], "relation": "used for", "id": "2022.acl-srw.30", "year": 2022, "rel_sent": "We propose to study the efficacy of multilingual classification models vs translation for cross - lingual deception classification .", "forward": false, "src_ids": "2022.acl-srw.30_968"}
+{"input": "translation is used for Task| context: with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language .", "entity": "translation", "output": "cross - lingual deception classification", "neg_sample": ["translation is used for Task", "with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language ."], "relation": "used for", "id": "2022.acl-srw.30", "year": 2022, "rel_sent": "We propose to study the efficacy of multilingual classification models vs translation for cross - lingual deception classification .", "forward": true, "src_ids": "2022.acl-srw.30_969"}
+{"input": "multilingual classification models is used for Task| context: with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language .", "entity": "multilingual classification models", "output": "cross - lingual deception classification", "neg_sample": ["multilingual classification models is used for Task", "with the increase of deception and misinformation especially in social media , it has become crucial to be able to develop machine learning methods to automatically identify deceptive language ."], "relation": "used for", "id": "2022.acl-srw.30", "year": 2022, "rel_sent": "We propose to study the efficacy of multilingual classification models vs translation for cross - lingual deception classification .", "forward": true, "src_ids": "2022.acl-srw.30_970"}
+{"input": "self - attention is done by using OtherScientificTerm| context: transformer - based pre - trained models , such as bert , have shown extraordinary success in achieving state - of - the - art results in many natural language processing applications . however , deploying these models can be prohibitively costly , as the standard self - attention mechanism of the transformer suffers from quadratic computational cost in the input sequence length .", "entity": "self - attention", "output": "computational sequence length", "neg_sample": ["self - attention is done by using OtherScientificTerm", "transformer - based pre - trained models , such as bert , have shown extraordinary success in achieving state - of - the - art results in many natural language processing applications .", "however , deploying these models can be prohibitively costly , as the standard self - attention mechanism of the transformer suffers from quadratic computational cost in the input sequence length ."], "relation": "used for", "id": "2022.acl-long.330", "year": 2022, "rel_sent": "To confront this , we propose FCA , a fine- and coarse - granularity hybrid self - attention that reduces the computation cost through progressively shortening the computational sequence length in self - attention .", "forward": false, "src_ids": "2022.acl-long.330_971"}
+{"input": "computational sequence length is used for OtherScientificTerm| context: transformer - based pre - trained models , such as bert , have shown extraordinary success in achieving state - of - the - art results in many natural language processing applications .", "entity": "computational sequence length", "output": "self - attention", "neg_sample": ["computational sequence length is used for OtherScientificTerm", "transformer - based pre - trained models , such as bert , have shown extraordinary success in achieving state - of - the - art results in many natural language processing applications ."], "relation": "used for", "id": "2022.acl-long.330", "year": 2022, "rel_sent": "To confront this , we propose FCA , a fine- and coarse - granularity hybrid self - attention that reduces the computation cost through progressively shortening the computational sequence length in self - attention .", "forward": true, "src_ids": "2022.acl-long.330_972"}
+{"input": "post - editing is done by using Method| context: with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al . , 2020 ) .", "entity": "post - editing", "output": "survey - based study", "neg_sample": ["post - editing is done by using Method", "with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al .", ", 2020 ) ."], "relation": "used for", "id": "2022.eamt-1.30", "year": 2022, "rel_sent": "To shed light on current professional practices and provide new pedagogical perspectives , we set up a survey - based study to investigate how PE and revision are carried out in professional settings .", "forward": false, "src_ids": "2022.eamt-1.30_973"}
+{"input": "revision is done by using Method| context: with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al . , 2020 ) .", "entity": "revision", "output": "survey - based study", "neg_sample": ["revision is done by using Method", "with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al .", ", 2020 ) ."], "relation": "used for", "id": "2022.eamt-1.30", "year": 2022, "rel_sent": "To shed light on current professional practices and provide new pedagogical perspectives , we set up a survey - based study to investigate how PE and revision are carried out in professional settings .", "forward": false, "src_ids": "2022.eamt-1.30_974"}
+{"input": "human - translated and machine - translated texts is done by using OtherScientificTerm| context: with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al . , 2020 ) .", "entity": "human - translated and machine - translated texts", "output": "reading strategies", "neg_sample": ["human - translated and machine - translated texts is done by using OtherScientificTerm", "with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al .", ", 2020 ) ."], "relation": "used for", "id": "2022.eamt-1.30", "year": 2022, "rel_sent": "Although the differences between the two activities seem to be clear for in - house linguists , our findings show that they tend to use the same reading strategies when working with human - translated and machine - translated texts .", "forward": false, "src_ids": "2022.eamt-1.30_975"}
+{"input": "reading strategies is used for Material| context: with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al . , 2020 ) .", "entity": "reading strategies", "output": "human - translated and machine - translated texts", "neg_sample": ["reading strategies is used for Material", "with the arrival of neural machine translation , the boundaries between revision and post - editing ( pe ) have started to blur ( koponen et al .", ", 2020 ) ."], "relation": "used for", "id": "2022.eamt-1.30", "year": 2022, "rel_sent": "Although the differences between the two activities seem to be clear for in - house linguists , our findings show that they tend to use the same reading strategies when working with human - translated and machine - translated texts .", "forward": true, "src_ids": "2022.eamt-1.30_976"}
+{"input": "troll meme classification is done by using Generic| context: trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media .", "entity": "troll meme classification", "output": "classical and pre - trained models", "neg_sample": ["troll meme classification is done by using Generic", "trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media ."], "relation": "used for", "id": "2022.dravidianlangtech-1.24", "year": 2022, "rel_sent": "BPHC@DravidianLangTech - ACL2022 - A comparative analysis of classical and pre - trained models for troll meme classification in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.24_977"}
+{"input": "classical and pre - trained models is used for Task| context: trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media .", "entity": "classical and pre - trained models", "output": "troll meme classification", "neg_sample": ["classical and pre - trained models is used for Task", "trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media ."], "relation": "used for", "id": "2022.dravidianlangtech-1.24", "year": 2022, "rel_sent": "BPHC@DravidianLangTech - ACL2022 - A comparative analysis of classical and pre - trained models for troll meme classification in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.24_978"}
+{"input": "raw code - mixed text is done by using OtherScientificTerm| context: trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media .", "entity": "raw code - mixed text", "output": "embeddings", "neg_sample": ["raw code - mixed text is done by using OtherScientificTerm", "trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media ."], "relation": "used for", "id": "2022.dravidianlangtech-1.24", "year": 2022, "rel_sent": "Embeddings are obtained for raw code - mixed text and the translated and transliterated version of the text and their relative performances are compared .", "forward": false, "src_ids": "2022.dravidianlangtech-1.24_979"}
+{"input": "translated and transliterated version of the text is done by using OtherScientificTerm| context: trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media .", "entity": "translated and transliterated version of the text", "output": "embeddings", "neg_sample": ["translated and transliterated version of the text is done by using OtherScientificTerm", "trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media ."], "relation": "used for", "id": "2022.dravidianlangtech-1.24", "year": 2022, "rel_sent": "Embeddings are obtained for raw code - mixed text and the translated and transliterated version of the text and their relative performances are compared .", "forward": false, "src_ids": "2022.dravidianlangtech-1.24_980"}
+{"input": "embeddings is used for Material| context: trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media .", "entity": "embeddings", "output": "raw code - mixed text", "neg_sample": ["embeddings is used for Material", "trolling refers to any user behaviour on the internet to intentionally provoke or instigate conflict predominantly in social media ."], "relation": "used for", "id": "2022.dravidianlangtech-1.24", "year": 2022, "rel_sent": "Embeddings are obtained for raw code - mixed text and the translated and transliterated version of the text and their relative performances are compared .", "forward": true, "src_ids": "2022.dravidianlangtech-1.24_981"}
+{"input": "utterance encoder is done by using Method| context: task - oriented dialogue ( tod ) systems allow users to accomplish tasks by giving directions to the system using natural language utterances . with the widespread adoption of conversational agents and chat platforms , tod has become mainstream in nlp research today . however , developing tod systems require massive amounts of data , and there has been limited work done for tod in low - resource languages like tamil .", "entity": "utterance encoder", "output": "xlm - roberta", "neg_sample": ["utterance encoder is done by using Method", "task - oriented dialogue ( tod ) systems allow users to accomplish tasks by giving directions to the system using natural language utterances .", "with the widespread adoption of conversational agents and chat platforms , tod has become mainstream in nlp research today .", "however , developing tod systems require massive amounts of data , and there has been limited work done for tod in low - resource languages like tamil ."], "relation": "used for", "id": "2022.dravidianlangtech-1.4", "year": 2022, "rel_sent": "The joint BERT model with XLM - Roberta as utterance encoder achieved the highest score with an intent accuracy of 96.26 % and slot F1 of 94.01 % .", "forward": false, "src_ids": "2022.dravidianlangtech-1.4_982"}
+{"input": "xlm - roberta is used for Method| context: task - oriented dialogue ( tod ) systems allow users to accomplish tasks by giving directions to the system using natural language utterances . with the widespread adoption of conversational agents and chat platforms , tod has become mainstream in nlp research today . however , developing tod systems require massive amounts of data , and there has been limited work done for tod in low - resource languages like tamil .", "entity": "xlm - roberta", "output": "utterance encoder", "neg_sample": ["xlm - roberta is used for Method", "task - oriented dialogue ( tod ) systems allow users to accomplish tasks by giving directions to the system using natural language utterances .", "with the widespread adoption of conversational agents and chat platforms , tod has become mainstream in nlp research today .", "however , developing tod systems require massive amounts of data , and there has been limited work done for tod in low - resource languages like tamil ."], "relation": "used for", "id": "2022.dravidianlangtech-1.4", "year": 2022, "rel_sent": "The joint BERT model with XLM - Roberta as utterance encoder achieved the highest score with an intent accuracy of 96.26 % and slot F1 of 94.01 % .", "forward": true, "src_ids": "2022.dravidianlangtech-1.4_983"}
+{"input": "language models is used for Task| context: large - scale language models are rapidly improving , performing well on a variety of tasks with little to no customization .", "entity": "language models", "output": "science writing", "neg_sample": ["language models is used for Task", "large - scale language models are rapidly improving , performing well on a variety of tasks with little to no customization ."], "relation": "used for", "id": "2022.in2writing-1.12", "year": 2022, "rel_sent": "In this work we investigate how language models can support science writing , a challenging writing task that is both open - ended and highly constrained .", "forward": true, "src_ids": "2022.in2writing-1.12_984"}
+{"input": "event prediction is done by using Method| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "event prediction", "output": "graph enhanced bert model", "neg_sample": ["event prediction is done by using Method", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.206", "year": 2022, "rel_sent": "A Graph Enhanced BERT Model for Event Prediction.", "forward": false, "src_ids": "2022.findings-acl.206_985"}
+{"input": "graph enhanced bert model is used for Task| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "graph enhanced bert model", "output": "event prediction", "neg_sample": ["graph enhanced bert model is used for Task", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.206", "year": 2022, "rel_sent": "A Graph Enhanced BERT Model for Event Prediction.", "forward": true, "src_ids": "2022.findings-acl.206_986"}
+{"input": "event connections is done by using OtherScientificTerm| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "event connections", "output": "structured variable", "neg_sample": ["event connections is done by using OtherScientificTerm", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.206", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process . Hence , in the test process , the connection relationship for unseen events can be predicted by the structured variable . Results on two event prediction tasks : script event prediction and story ending prediction , show that our approach can outperform state - of - the - art baseline methods .", "forward": false, "src_ids": "2022.findings-acl.206_987"}
+{"input": "event connections is done by using Method| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "event connections", "output": "bert", "neg_sample": ["event connections is done by using Method", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.206", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process . Hence , in the test process , the connection relationship for unseen events can be predicted by the structured variable . Results on two event prediction tasks : script event prediction and story ending prediction , show that our approach can outperform state - of - the - art baseline methods .", "forward": false, "src_ids": "2022.findings-acl.206_988"}
+{"input": "structured variable is used for OtherScientificTerm| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "structured variable", "output": "event connections", "neg_sample": ["structured variable is used for OtherScientificTerm", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.206", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process . Hence , in the test process , the connection relationship for unseen events can be predicted by the structured variable . Results on two event prediction tasks : script event prediction and story ending prediction , show that our approach can outperform state - of - the - art baseline methods .", "forward": true, "src_ids": "2022.findings-acl.206_989"}
+{"input": "bert is used for OtherScientificTerm| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "bert", "output": "event connections", "neg_sample": ["bert is used for OtherScientificTerm", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.206", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process . Hence , in the test process , the connection relationship for unseen events can be predicted by the structured variable . Results on two event prediction tasks : script event prediction and story ending prediction , show that our approach can outperform state - of - the - art baseline methods .", "forward": true, "src_ids": "2022.findings-acl.206_990"}
+{"input": "common ground is used for OtherScientificTerm| context: in conversational qa , models have to leverage information in previous turns to answer upcoming questions . current approaches , such as question rewriting , struggle to extract relevant information as the conversation unwinds .", "entity": "common ground", "output": "conversational information", "neg_sample": ["common ground is used for OtherScientificTerm", "in conversational qa , models have to leverage information in previous turns to answer upcoming questions .", "current approaches , such as question rewriting , struggle to extract relevant information as the conversation unwinds ."], "relation": "used for", "id": "2022.nlp4convai-1.7", "year": 2022, "rel_sent": "We show that CG offers a more efficient and human - like way to exploit conversational information compared to existing approaches , leading to improvements on Open Domain Conversational QA .", "forward": true, "src_ids": "2022.nlp4convai-1.7_991"}
+{"input": "conversational information is done by using OtherScientificTerm| context: in conversational qa , models have to leverage information in previous turns to answer upcoming questions . current approaches , such as question rewriting , struggle to extract relevant information as the conversation unwinds .", "entity": "conversational information", "output": "common ground", "neg_sample": ["conversational information is done by using OtherScientificTerm", "in conversational qa , models have to leverage information in previous turns to answer upcoming questions .", "current approaches , such as question rewriting , struggle to extract relevant information as the conversation unwinds ."], "relation": "used for", "id": "2022.nlp4convai-1.7", "year": 2022, "rel_sent": "We show that CG offers a more efficient and human - like way to exploit conversational information compared to existing approaches , leading to improvements on Open Domain Conversational QA .", "forward": false, "src_ids": "2022.nlp4convai-1.7_992"}
+{"input": "higher - level tasks is done by using Method| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "higher - level tasks", "output": "syntactic - level techniques", "neg_sample": ["higher - level tasks is done by using Method", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Finally , we discuss that the results most heavily depend on the task , language pair ( e.g. , syntactic - level techniques mostly benefit higher - level tasks and morphologically richer languages ) , and model type ( e.g. , token - level augmentation provides significant improvements for BPE , while character - level ones give generally higher scores for char and mBERT based models ) .", "forward": false, "src_ids": "2022.cl-1.2_993"}
+{"input": "syntactic - level techniques is used for Generic| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "syntactic - level techniques", "output": "higher - level tasks", "neg_sample": ["syntactic - level techniques is used for Generic", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Finally , we discuss that the results most heavily depend on the task , language pair ( e.g. , syntactic - level techniques mostly benefit higher - level tasks and morphologically richer languages ) , and model type ( e.g. , token - level augmentation provides significant improvements for BPE , while character - level ones give generally higher scores for char and mBERT based models ) .", "forward": true, "src_ids": "2022.cl-1.2_994"}
+{"input": "bpe is done by using Method| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "bpe", "output": "token - level augmentation", "neg_sample": ["bpe is done by using Method", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Finally , we discuss that the results most heavily depend on the task , language pair ( e.g. , syntactic - level techniques mostly benefit higher - level tasks and morphologically richer languages ) , and model type ( e.g. , token - level augmentation provides significant improvements for BPE , while character - level ones give generally higher scores for char and mBERT based models ) .", "forward": false, "src_ids": "2022.cl-1.2_995"}
+{"input": "token - level augmentation is used for Method| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "token - level augmentation", "output": "bpe", "neg_sample": ["token - level augmentation is used for Method", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Finally , we discuss that the results most heavily depend on the task , language pair ( e.g. , syntactic - level techniques mostly benefit higher - level tasks and morphologically richer languages ) , and model type ( e.g. , token - level augmentation provides significant improvements for BPE , while character - level ones give generally higher scores for char and mBERT based models ) .", "forward": true, "src_ids": "2022.cl-1.2_996"}
+{"input": "mbert based models is done by using Method| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "mbert based models", "output": "character - level ones", "neg_sample": ["mbert based models is done by using Method", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Finally , we discuss that the results most heavily depend on the task , language pair ( e.g. , syntactic - level techniques mostly benefit higher - level tasks and morphologically richer languages ) , and model type ( e.g. , token - level augmentation provides significant improvements for BPE , while character - level ones give generally higher scores for char and mBERT based models ) .", "forward": false, "src_ids": "2022.cl-1.2_997"}
+{"input": "character - level ones is used for Method| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "character - level ones", "output": "mbert based models", "neg_sample": ["character - level ones is used for Method", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Finally , we discuss that the results most heavily depend on the task , language pair ( e.g. , syntactic - level techniques mostly benefit higher - level tasks and morphologically richer languages ) , and model type ( e.g. , token - level augmentation provides significant improvements for BPE , while character - level ones give generally higher scores for char and mBERT based models ) .", "forward": true, "src_ids": "2022.cl-1.2_998"}
+{"input": "augmentation is used for Task| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios . one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data . although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks .", "entity": "augmentation", "output": "dependency parsing", "neg_sample": ["augmentation is used for Task", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "one methodology to counterattack this problem is text augmentation , that is , generating new synthetic training data points from existing data .", "although nlp has recently witnessed several new textual augmentation techniques , the field still lacks a systematic performance analysis on a diverse set of languages and sequence tagging tasks ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Augmentation most significantly improves dependency parsing , followed by part - of - speech tagging and semantic role labeling .", "forward": true, "src_ids": "2022.cl-1.2_999"}
+{"input": "dependency parsing is done by using OtherScientificTerm| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "entity": "dependency parsing", "output": "augmentation", "neg_sample": ["dependency parsing is done by using OtherScientificTerm", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Augmentation most significantly improves dependency parsing , followed by part - of - speech tagging and semantic role labeling .", "forward": false, "src_ids": "2022.cl-1.2_1000"}
+{"input": "semantic role labeling is done by using OtherScientificTerm| context: data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones . despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios .", "entity": "semantic role labeling", "output": "augmentation", "neg_sample": ["semantic role labeling is done by using OtherScientificTerm", "data - hungry deep neural networks have established themselves as the de facto standard for many nlp tasks , including the traditional sequence tagging ones .", "despite their state - of - the - art performance on high - resource languages , they still fall behind their statistical counterparts in low - resource scenarios ."], "relation": "used for", "id": "2022.cl-1.2", "year": 2022, "rel_sent": "Augmentation most significantly improves dependency parsing , followed by part - of - speech tagging and semantic role labeling .", "forward": false, "src_ids": "2022.cl-1.2_1001"}
+{"input": "overfitting is done by using OtherScientificTerm| context: fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning . to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed . adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g. , dedicated language and/or task adapters ) . sparse fine - tuning is expressive , as it controls the behavior of all model components .", "entity": "overfitting", "output": "sparsity", "neg_sample": ["overfitting is done by using OtherScientificTerm", "fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning .", "to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed .", "adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g.", ", dedicated language and/or task adapters ) .", "sparse fine - tuning is expressive , as it controls the behavior of all model components ."], "relation": "used for", "id": "2022.acl-long.125", "year": 2022, "rel_sent": "Based on an in - depth analysis , we additionally find that sparsity is crucial to prevent both 1 ) interference between the fine - tunings to be composed and 2 ) overfitting .", "forward": false, "src_ids": "2022.acl-long.125_1002"}
+{"input": "fine - tunings is done by using OtherScientificTerm| context: fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning . to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed . adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g. , dedicated language and/or task adapters ) . sparse fine - tuning is expressive , as it controls the behavior of all model components .", "entity": "fine - tunings", "output": "sparsity", "neg_sample": ["fine - tunings is done by using OtherScientificTerm", "fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning .", "to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed .", "adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g.", ", dedicated language and/or task adapters ) .", "sparse fine - tuning is expressive , as it controls the behavior of all model components ."], "relation": "used for", "id": "2022.acl-long.125", "year": 2022, "rel_sent": "Based on an in - depth analysis , we additionally find that sparsity is crucial to prevent both 1 ) interference between the fine - tunings to be composed and 2 ) overfitting .", "forward": false, "src_ids": "2022.acl-long.125_1003"}
+{"input": "sparsity is used for Method| context: fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning . to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed . adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g. , dedicated language and/or task adapters ) . sparse fine - tuning is expressive , as it controls the behavior of all model components .", "entity": "sparsity", "output": "fine - tunings", "neg_sample": ["sparsity is used for Method", "fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning .", "to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed .", "adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g.", ", dedicated language and/or task adapters ) .", "sparse fine - tuning is expressive , as it controls the behavior of all model components ."], "relation": "used for", "id": "2022.acl-long.125", "year": 2022, "rel_sent": "Based on an in - depth analysis , we additionally find that sparsity is crucial to prevent both 1 ) interference between the fine - tunings to be composed and 2 ) overfitting .", "forward": true, "src_ids": "2022.acl-long.125_1004"}
+{"input": "sparsity is used for OtherScientificTerm| context: fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning . to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed . adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g. , dedicated language and/or task adapters ) . sparse fine - tuning is expressive , as it controls the behavior of all model components .", "entity": "sparsity", "output": "overfitting", "neg_sample": ["sparsity is used for OtherScientificTerm", "fine - tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning .", "to increase its efficiency and prevent catastrophic forgetting and interference , techniques like adapters and sparse fine - tuning have been developed .", "adapters are modular , as they can be combined to adapt a model towards different facets of knowledge ( e.g.", ", dedicated language and/or task adapters ) .", "sparse fine - tuning is expressive , as it controls the behavior of all model components ."], "relation": "used for", "id": "2022.acl-long.125", "year": 2022, "rel_sent": "Based on an in - depth analysis , we additionally find that sparsity is crucial to prevent both 1 ) interference between the fine - tunings to be composed and 2 ) overfitting .", "forward": true, "src_ids": "2022.acl-long.125_1005"}
+{"input": "retrieval approaches is done by using Method| context: statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question . while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets .", "entity": "retrieval approaches", "output": "belgian statutory article retrieval dataset", "neg_sample": ["retrieval approaches is done by using Method", "statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question .", "while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets ."], "relation": "used for", "id": "2022.acl-long.468", "year": 2022, "rel_sent": "Using BSARD , we benchmark several state - of - the - art retrieval approaches , including lexical and dense architectures , both in zero - shot and supervised setups .", "forward": false, "src_ids": "2022.acl-long.468_1006"}
+{"input": "legal information retrieval is done by using Method| context: statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question . while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets .", "entity": "legal information retrieval", "output": "belgian statutory article retrieval dataset", "neg_sample": ["legal information retrieval is done by using Method", "statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question .", "while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets ."], "relation": "used for", "id": "2022.acl-long.468", "year": 2022, "rel_sent": "By the specificity of the domain and addressed task , BSARD presents a unique challenge problem for future research on legal information retrieval .", "forward": false, "src_ids": "2022.acl-long.468_1007"}
+{"input": "belgian statutory article retrieval dataset is used for Method| context: statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question . while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets .", "entity": "belgian statutory article retrieval dataset", "output": "retrieval approaches", "neg_sample": ["belgian statutory article retrieval dataset is used for Method", "statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question .", "while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets ."], "relation": "used for", "id": "2022.acl-long.468", "year": 2022, "rel_sent": "Using BSARD , we benchmark several state - of - the - art retrieval approaches , including lexical and dense architectures , both in zero - shot and supervised setups .", "forward": true, "src_ids": "2022.acl-long.468_1008"}
+{"input": "belgian statutory article retrieval dataset is used for Task| context: statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question . while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets .", "entity": "belgian statutory article retrieval dataset", "output": "legal information retrieval", "neg_sample": ["belgian statutory article retrieval dataset is used for Task", "statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question .", "while recent advances in natural language processing have sparked considerable interest in many legal tasks , statutory article retrieval remains primarily untouched due to the scarcity of large - scale and high - quality annotated datasets ."], "relation": "used for", "id": "2022.acl-long.468", "year": 2022, "rel_sent": "By the specificity of the domain and addressed task , BSARD presents a unique challenge problem for future research on legal information retrieval .", "forward": true, "src_ids": "2022.acl-long.468_1009"}
+{"input": "event - centric reasoning scenarios is done by using Method| context: generating new events given context with correlated ones plays a crucial role in many event - centric reasoning tasks . existing works either limit their scope to specific scenarios or overlook event - level correlations .", "entity": "event - centric reasoning scenarios", "output": "correlation - aware context - to - event transformer", "neg_sample": ["event - centric reasoning scenarios is done by using Method", "generating new events given context with correlated ones plays a crucial role in many event - centric reasoning tasks .", "existing works either limit their scope to specific scenarios or overlook event - level correlations ."], "relation": "used for", "id": "2022.acl-long.183", "year": 2022, "rel_sent": "The proposed ClarET is applicable to a wide range of event - centric reasoning scenarios , considering its versatility of ( i ) event - correlation types ( e.g. , causal , temporal , contrast ) , ( ii ) application formulations ( i.e. , generation and classification ) , and ( iii ) reasoning types ( e.g. , abductive , counterfactual and ending reasoning ) .", "forward": false, "src_ids": "2022.acl-long.183_1010"}
+{"input": "correlation - aware context - to - event transformer is used for Task| context: generating new events given context with correlated ones plays a crucial role in many event - centric reasoning tasks . existing works either limit their scope to specific scenarios or overlook event - level correlations .", "entity": "correlation - aware context - to - event transformer", "output": "event - centric reasoning scenarios", "neg_sample": ["correlation - aware context - to - event transformer is used for Task", "generating new events given context with correlated ones plays a crucial role in many event - centric reasoning tasks .", "existing works either limit their scope to specific scenarios or overlook event - level correlations ."], "relation": "used for", "id": "2022.acl-long.183", "year": 2022, "rel_sent": "The proposed ClarET is applicable to a wide range of event - centric reasoning scenarios , considering its versatility of ( i ) event - correlation types ( e.g. , causal , temporal , contrast ) , ( ii ) application formulations ( i.e. , generation and classification ) , and ( iii ) reasoning types ( e.g. , abductive , counterfactual and ending reasoning ) .", "forward": true, "src_ids": "2022.acl-long.183_1011"}
+{"input": "natural language processing is done by using OtherScientificTerm| context: recent years have witnessed the emergence of a variety of post - hoc interpretations that aim to uncover how natural language processing ( nlp ) models make predictions . despite the surge of new interpretation methods , it remains an open problem how to define and quantitatively measure the faithfulness of interpretations , i.e. , to what extent interpretations reflect the reasoning process by a model .", "entity": "natural language processing", "output": "model interpretations", "neg_sample": ["natural language processing is done by using OtherScientificTerm", "recent years have witnessed the emergence of a variety of post - hoc interpretations that aim to uncover how natural language processing ( nlp ) models make predictions .", "despite the surge of new interpretation methods , it remains an open problem how to define and quantitatively measure the faithfulness of interpretations , i.e.", ", to what extent interpretations reflect the reasoning process by a model ."], "relation": "used for", "id": "2022.acl-long.188", "year": 2022, "rel_sent": "On the Sensitivity and Stability of Model Interpretations in NLP.", "forward": false, "src_ids": "2022.acl-long.188_1012"}
+{"input": "model interpretations is used for Task| context: despite the surge of new interpretation methods , it remains an open problem how to define and quantitatively measure the faithfulness of interpretations , i.e. , to what extent interpretations reflect the reasoning process by a model .", "entity": "model interpretations", "output": "natural language processing", "neg_sample": ["model interpretations is used for Task", "despite the surge of new interpretation methods , it remains an open problem how to define and quantitatively measure the faithfulness of interpretations , i.e.", ", to what extent interpretations reflect the reasoning process by a model ."], "relation": "used for", "id": "2022.acl-long.188", "year": 2022, "rel_sent": "On the Sensitivity and Stability of Model Interpretations in NLP.", "forward": true, "src_ids": "2022.acl-long.188_1013"}
+{"input": "dependency parsing is done by using Method| context: recent years have witnessed the emergence of a variety of post - hoc interpretations that aim to uncover how natural language processing ( nlp ) models make predictions .", "entity": "dependency parsing", "output": "interpretation methods", "neg_sample": ["dependency parsing is done by using Method", "recent years have witnessed the emergence of a variety of post - hoc interpretations that aim to uncover how natural language processing ( nlp ) models make predictions ."], "relation": "used for", "id": "2022.acl-long.188", "year": 2022, "rel_sent": "Besides text classification , we also apply interpretation methods and metrics to dependency parsing .", "forward": false, "src_ids": "2022.acl-long.188_1014"}
+{"input": "interpretation methods is used for Task| context: recent years have witnessed the emergence of a variety of post - hoc interpretations that aim to uncover how natural language processing ( nlp ) models make predictions . despite the surge of new interpretation methods , it remains an open problem how to define and quantitatively measure the faithfulness of interpretations , i.e. , to what extent interpretations reflect the reasoning process by a model .", "entity": "interpretation methods", "output": "dependency parsing", "neg_sample": ["interpretation methods is used for Task", "recent years have witnessed the emergence of a variety of post - hoc interpretations that aim to uncover how natural language processing ( nlp ) models make predictions .", "despite the surge of new interpretation methods , it remains an open problem how to define and quantitatively measure the faithfulness of interpretations , i.e.", ", to what extent interpretations reflect the reasoning process by a model ."], "relation": "used for", "id": "2022.acl-long.188", "year": 2022, "rel_sent": "Besides text classification , we also apply interpretation methods and metrics to dependency parsing .", "forward": true, "src_ids": "2022.acl-long.188_1015"}
+{"input": "automatic sentiment prediction is done by using OtherScientificTerm| context: deep neural networks ( dnn ) models have achieved acceptable performance in sentiment prediction of written text . however , the output of these machine learning ( ml ) models can not be natively interpreted .", "entity": "automatic sentiment prediction", "output": "emotion carriers", "neg_sample": ["automatic sentiment prediction is done by using OtherScientificTerm", "deep neural networks ( dnn ) models have achieved acceptable performance in sentiment prediction of written text .", "however , the output of these machine learning ( ml ) models can not be natively interpreted ."], "relation": "used for", "id": "2022.wassa-1.6", "year": 2022, "rel_sent": "Can Emotion Carriers Explain Automatic Sentiment Prediction ? A Study on Personal Narratives.", "forward": false, "src_ids": "2022.wassa-1.6_1016"}
+{"input": "emotion carriers is used for Task| context: deep neural networks ( dnn ) models have achieved acceptable performance in sentiment prediction of written text . however , the output of these machine learning ( ml ) models can not be natively interpreted .", "entity": "emotion carriers", "output": "automatic sentiment prediction", "neg_sample": ["emotion carriers is used for Task", "deep neural networks ( dnn ) models have achieved acceptable performance in sentiment prediction of written text .", "however , the output of these machine learning ( ml ) models can not be natively interpreted ."], "relation": "used for", "id": "2022.wassa-1.6", "year": 2022, "rel_sent": "Can Emotion Carriers Explain Automatic Sentiment Prediction ? A Study on Personal Narratives.", "forward": true, "src_ids": "2022.wassa-1.6_1017"}
+{"input": "linguistic and visuo - linguistic representations is used for Task| context: noun - noun compounds ( nncs ) occur frequently in the english language . determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks . however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks .", "entity": "linguistic and visuo - linguistic representations", "output": "nnc interpretation", "neg_sample": ["linguistic and visuo - linguistic representations is used for Task", "noun - noun compounds ( nncs ) occur frequently in the english language .", "determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks .", "however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks ."], "relation": "used for", "id": "2022.cmcl-1.3", "year": 2022, "rel_sent": "Our work is a novel comparison of linguistic and visuo - linguistic representations for the task of NNC interpretation .", "forward": true, "src_ids": "2022.cmcl-1.3_1018"}
+{"input": "nnc interpretation is done by using Method| context: noun - noun compounds ( nncs ) occur frequently in the english language . accurate nnc interpretation , i.e. determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks . until now , computational nnc interpretation has been limited to approaches involving linguistic representations only . however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks .", "entity": "nnc interpretation", "output": "linguistic and visuo - linguistic representations", "neg_sample": ["nnc interpretation is done by using Method", "noun - noun compounds ( nncs ) occur frequently in the english language .", "accurate nnc interpretation , i.e.", "determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks .", "until now , computational nnc interpretation has been limited to approaches involving linguistic representations only .", "however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks ."], "relation": "used for", "id": "2022.cmcl-1.3", "year": 2022, "rel_sent": "Our work is a novel comparison of linguistic and visuo - linguistic representations for the task of NNC interpretation .", "forward": false, "src_ids": "2022.cmcl-1.3_1019"}
+{"input": "nnc interpretation systems is done by using OtherScientificTerm| context: noun - noun compounds ( nncs ) occur frequently in the english language . accurate nnc interpretation , i.e. determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks . until now , computational nnc interpretation has been limited to approaches involving linguistic representations only . however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks .", "entity": "nnc interpretation systems", "output": "visual information", "neg_sample": ["nnc interpretation systems is done by using OtherScientificTerm", "noun - noun compounds ( nncs ) occur frequently in the english language .", "accurate nnc interpretation , i.e.", "determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks .", "until now , computational nnc interpretation has been limited to approaches involving linguistic representations only .", "however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks ."], "relation": "used for", "id": "2022.cmcl-1.3", "year": 2022, "rel_sent": "We combine distributional word vectors with image vectors to investigate how visual information can help improve NNC interpretation systems .", "forward": false, "src_ids": "2022.cmcl-1.3_1020"}
+{"input": "visual information is used for Method| context: noun - noun compounds ( nncs ) occur frequently in the english language . accurate nnc interpretation , i.e. determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks . until now , computational nnc interpretation has been limited to approaches involving linguistic representations only . however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks .", "entity": "visual information", "output": "nnc interpretation systems", "neg_sample": ["visual information is used for Method", "noun - noun compounds ( nncs ) occur frequently in the english language .", "accurate nnc interpretation , i.e.", "determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks .", "until now , computational nnc interpretation has been limited to approaches involving linguistic representations only .", "however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks ."], "relation": "used for", "id": "2022.cmcl-1.3", "year": 2022, "rel_sent": "We combine distributional word vectors with image vectors to investigate how visual information can help improve NNC interpretation systems .", "forward": true, "src_ids": "2022.cmcl-1.3_1021"}
+{"input": "classification is done by using OtherScientificTerm| context: noun - noun compounds ( nncs ) occur frequently in the english language . accurate nnc interpretation , i.e. determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks . until now , computational nnc interpretation has been limited to approaches involving linguistic representations only . however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks .", "entity": "classification", "output": "visual vectors", "neg_sample": ["classification is done by using OtherScientificTerm", "noun - noun compounds ( nncs ) occur frequently in the english language .", "accurate nnc interpretation , i.e.", "determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks .", "until now , computational nnc interpretation has been limited to approaches involving linguistic representations only .", "however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks ."], "relation": "used for", "id": "2022.cmcl-1.3", "year": 2022, "rel_sent": "We find that adding visual vectors increases classification performance on our dataset in many cases .", "forward": false, "src_ids": "2022.cmcl-1.3_1022"}
+{"input": "visual vectors is used for Task| context: noun - noun compounds ( nncs ) occur frequently in the english language . accurate nnc interpretation , i.e. determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks . until now , computational nnc interpretation has been limited to approaches involving linguistic representations only . however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks .", "entity": "visual vectors", "output": "classification", "neg_sample": ["visual vectors is used for Task", "noun - noun compounds ( nncs ) occur frequently in the english language .", "accurate nnc interpretation , i.e.", "determining the implicit relationship between the constituents of a nnc , is crucial for the advancement of many natural language processing tasks .", "until now , computational nnc interpretation has been limited to approaches involving linguistic representations only .", "however , much research suggests that grounding linguistic representations in vision or other modalities can increase performance on this and other tasks ."], "relation": "used for", "id": "2022.cmcl-1.3", "year": 2022, "rel_sent": "We find that adding visual vectors increases classification performance on our dataset in many cases .", "forward": true, "src_ids": "2022.cmcl-1.3_1023"}
+{"input": "lexical and syntactical conditions is used for Task| context: however , existing works only highlight a special condition under two indispensable aspects of cpg ( i.e. , lexically and syntactically cpg ) individually , lacking a unified circumstance to explore and analyze their effectiveness .", "entity": "lexical and syntactical conditions", "output": "paraphrase generation", "neg_sample": ["lexical and syntactical conditions is used for Task", "however , existing works only highlight a special condition under two indispensable aspects of cpg ( i.e.", ", lexically and syntactically cpg ) individually , lacking a unified circumstance to explore and analyze their effectiveness ."], "relation": "used for", "id": "2022.findings-acl.318", "year": 2022, "rel_sent": "In addition , the combination of lexical and syntactical conditions shows the significant controllable ability of paraphrase generation , and these empirical results could provide novel insight to user - oriented paraphrasing .", "forward": true, "src_ids": "2022.findings-acl.318_1024"}
+{"input": "paraphrase generation is done by using OtherScientificTerm| context: controllable paraphrase generation ( cpg ) incorporates various external conditions to obtain desirable paraphrases . however , existing works only highlight a special condition under two indispensable aspects of cpg ( i.e. , lexically and syntactically cpg ) individually , lacking a unified circumstance to explore and analyze their effectiveness .", "entity": "paraphrase generation", "output": "lexical and syntactical conditions", "neg_sample": ["paraphrase generation is done by using OtherScientificTerm", "controllable paraphrase generation ( cpg ) incorporates various external conditions to obtain desirable paraphrases .", "however , existing works only highlight a special condition under two indispensable aspects of cpg ( i.e.", ", lexically and syntactically cpg ) individually , lacking a unified circumstance to explore and analyze their effectiveness ."], "relation": "used for", "id": "2022.findings-acl.318", "year": 2022, "rel_sent": "In addition , the combination of lexical and syntactical conditions shows the significant controllable ability of paraphrase generation , and these empirical results could provide novel insight to user - oriented paraphrasing .", "forward": false, "src_ids": "2022.findings-acl.318_1025"}
+{"input": "chinese spelling correction is done by using Method| context: chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts . csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings . many recent works use bert - based language models to directly correct each character of the input sentence . however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters .", "entity": "chinese spelling correction", "output": "detector - corrector multi - task framework", "neg_sample": ["chinese spelling correction is done by using Method", "chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts .", "csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings .", "many recent works use bert - based language models to directly correct each character of the input sentence .", "however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters ."], "relation": "used for", "id": "2022.findings-acl.98", "year": 2022, "rel_sent": "MDCSpell : A Multi - task Detector - Corrector Framework for Chinese Spelling Correction.", "forward": false, "src_ids": "2022.findings-acl.98_1026"}
+{"input": "detector - corrector multi - task framework is used for Task| context: many recent works use bert - based language models to directly correct each character of the input sentence . however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters .", "entity": "detector - corrector multi - task framework", "output": "chinese spelling correction", "neg_sample": ["detector - corrector multi - task framework is used for Task", "many recent works use bert - based language models to directly correct each character of the input sentence .", "however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters ."], "relation": "used for", "id": "2022.findings-acl.98", "year": 2022, "rel_sent": "MDCSpell : A Multi - task Detector - Corrector Framework for Chinese Spelling Correction.", "forward": true, "src_ids": "2022.findings-acl.98_1027"}
+{"input": "correction is done by using Method| context: chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts . csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings . many recent works use bert - based language models to directly correct each character of the input sentence . however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters .", "entity": "correction", "output": "error detector", "neg_sample": ["correction is done by using Method", "chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts .", "csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings .", "many recent works use bert - based language models to directly correct each character of the input sentence .", "however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters ."], "relation": "used for", "id": "2022.findings-acl.98", "year": 2022, "rel_sent": "Some other works propose to use an error detector to guide the correction by masking the detected errors .", "forward": false, "src_ids": "2022.findings-acl.98_1028"}
+{"input": "error detector is used for Task| context: csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings . many recent works use bert - based language models to directly correct each character of the input sentence . however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters .", "entity": "error detector", "output": "correction", "neg_sample": ["error detector is used for Task", "csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings .", "many recent works use bert - based language models to directly correct each character of the input sentence .", "however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters ."], "relation": "used for", "id": "2022.findings-acl.98", "year": 2022, "rel_sent": "Some other works propose to use an error detector to guide the correction by masking the detected errors .", "forward": true, "src_ids": "2022.findings-acl.98_1029"}
+{"input": "multi - hop questions is done by using Method| context: multi - hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage . current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers . however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning .", "entity": "multi - hop questions", "output": "cqg", "neg_sample": ["multi - hop questions is done by using Method", "multi - hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage .", "current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers .", "however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning ."], "relation": "used for", "id": "2022.acl-long.475", "year": 2022, "rel_sent": "CQG employs a simple method to generate the multi - hop questions that contain key entities in multi - hop reasoning chains , which ensure the complexity and quality of the questions .", "forward": false, "src_ids": "2022.acl-long.475_1030"}
+{"input": "multi - hop question generation is done by using Method| context: multi - hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage . current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers . however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning .", "entity": "multi - hop question generation", "output": "controlled generation framework", "neg_sample": ["multi - hop question generation is done by using Method", "multi - hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage .", "current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers .", "however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning ."], "relation": "used for", "id": "2022.acl-long.475", "year": 2022, "rel_sent": "CQG : A Simple and Effective Controlled Generation Framework for Multi - hop Question Generation.", "forward": false, "src_ids": "2022.acl-long.475_1031"}
+{"input": "controlled generation framework is used for Task| context: current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers . however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning .", "entity": "controlled generation framework", "output": "multi - hop question generation", "neg_sample": ["controlled generation framework is used for Task", "current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers .", "however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning ."], "relation": "used for", "id": "2022.acl-long.475", "year": 2022, "rel_sent": "CQG : A Simple and Effective Controlled Generation Framework for Multi - hop Question Generation.", "forward": true, "src_ids": "2022.acl-long.475_1032"}
+{"input": "cqg is used for OtherScientificTerm| context: multi - hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage . current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers . however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning .", "entity": "cqg", "output": "multi - hop questions", "neg_sample": ["cqg is used for OtherScientificTerm", "multi - hop question generation focuses on generating complex questions that require reasoning over multiple pieces of information of the input passage .", "current models with state - of - the - art performance have been able to generate the correct questions corresponding to the answers .", "however , most models can not ensure the complexity of generated questions , so they may generate shallow questions that can be answered without multi - hop reasoning ."], "relation": "used for", "id": "2022.acl-long.475", "year": 2022, "rel_sent": "CQG employs a simple method to generate the multi - hop questions that contain key entities in multi - hop reasoning chains , which ensure the complexity and quality of the questions .", "forward": true, "src_ids": "2022.acl-long.475_1033"}
+{"input": "synthetic training data is used for Task| context: generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions . but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents .", "entity": "synthetic training data", "output": "query - focused summarization", "neg_sample": ["synthetic training data is used for Task", "generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions .", "but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents ."], "relation": "used for", "id": "2022.acl-long.449", "year": 2022, "rel_sent": "Our focus in evaluation is how well existing techniques can generalize to these domains without seeing in - domain training data , so we turn to techniques to construct synthetic training data that have been used in query - focused summarization work .", "forward": true, "src_ids": "2022.acl-long.449_1034"}
+{"input": "generic summary is done by using Generic| context: generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions . but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents .", "entity": "generic summary", "output": "generic summarization system", "neg_sample": ["generic summary is done by using Generic", "generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions .", "but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents ."], "relation": "used for", "id": "2022.acl-long.449", "year": 2022, "rel_sent": "A system producing a single generic summary can not concisely satisfy both aspects .", "forward": false, "src_ids": "2022.acl-long.449_1035"}
+{"input": "generic summarization system is used for OtherScientificTerm| context: generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions . but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents .", "entity": "generic summarization system", "output": "generic summary", "neg_sample": ["generic summarization system is used for OtherScientificTerm", "generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions .", "but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents ."], "relation": "used for", "id": "2022.acl-long.449", "year": 2022, "rel_sent": "A system producing a single generic summary can not concisely satisfy both aspects .", "forward": true, "src_ids": "2022.acl-long.449_1036"}
+{"input": "query - focused summarization is done by using Material| context: generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions . but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents .", "entity": "query - focused summarization", "output": "synthetic training data", "neg_sample": ["query - focused summarization is done by using Material", "generic summaries try to cover an entire document and query - based summaries try to answer document - specific questions .", "but real users ' needs often fall in between these extremes and correspond to aspects , high - level topics discussed among similar types of documents ."], "relation": "used for", "id": "2022.acl-long.449", "year": 2022, "rel_sent": "Our focus in evaluation is how well existing techniques can generalize to these domains without seeing in - domain training data , so we turn to techniques to construct synthetic training data that have been used in query - focused summarization work .", "forward": false, "src_ids": "2022.acl-long.449_1037"}
+{"input": "cross - culture emotion analysis and recognition is done by using Method| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "cross - culture emotion analysis and recognition", "output": "m^3ed", "neg_sample": ["cross - culture emotion analysis and recognition is done by using Method", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "To the best of our knowledge , M^3ED is the first multimodal emotional dialogue dataset in Chinese . It is valuable for cross - culture emotion analysis and recognition .", "forward": false, "src_ids": "2022.acl-long.391_1038"}
+{"input": "m^3ed is used for Task| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "m^3ed", "output": "cross - culture emotion analysis and recognition", "neg_sample": ["m^3ed is used for Task", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "To the best of our knowledge , M^3ED is the first multimodal emotional dialogue dataset in Chinese . It is valuable for cross - culture emotion analysis and recognition .", "forward": true, "src_ids": "2022.acl-long.391_1039"}
+{"input": "dialogue context is done by using Method| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "dialogue context", "output": "multimodal dialogue - aware interaction framework", "neg_sample": ["dialogue context is done by using Method", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": false, "src_ids": "2022.acl-long.391_1040"}
+{"input": "emotion recognition is done by using OtherScientificTerm| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "emotion recognition", "output": "dialogue context", "neg_sample": ["emotion recognition is done by using OtherScientificTerm", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": false, "src_ids": "2022.acl-long.391_1041"}
+{"input": "mdi is used for OtherScientificTerm| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "mdi", "output": "dialogue context", "neg_sample": ["mdi is used for OtherScientificTerm", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": true, "src_ids": "2022.acl-long.391_1042"}
+{"input": "multimodal dialogue - aware interaction framework is used for OtherScientificTerm| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "multimodal dialogue - aware interaction framework", "output": "dialogue context", "neg_sample": ["multimodal dialogue - aware interaction framework is used for OtherScientificTerm", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": true, "src_ids": "2022.acl-long.391_1043"}
+{"input": "dialogue context is used for Task| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "dialogue context", "output": "emotion recognition", "neg_sample": ["dialogue context is used for Task", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.391", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": true, "src_ids": "2022.acl-long.391_1044"}
+{"input": "expressive structures is used for OtherScientificTerm| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "expressive structures", "output": "nested entities", "neg_sample": ["expressive structures is used for OtherScientificTerm", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "In this work , we resort to more expressive structures , lexicalized constituency trees in which constituents are annotated by headwords , to model nested entities .", "forward": true, "src_ids": "2022.acl-long.428_1045"}
+{"input": "lexicalized constituency trees is used for OtherScientificTerm| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "lexicalized constituency trees", "output": "nested entities", "neg_sample": ["lexicalized constituency trees is used for OtherScientificTerm", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "In this work , we resort to more expressive structures , lexicalized constituency trees in which constituents are annotated by headwords , to model nested entities .", "forward": true, "src_ids": "2022.acl-long.428_1046"}
+{"input": "eisner - satta algorithm is used for Task| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "eisner - satta algorithm", "output": "partial marginalization", "neg_sample": ["eisner - satta algorithm is used for Task", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently . In addition , we propose to use ( 1 ) a two - stage strategy ( 2 ) a head regularization loss and ( 3 ) a head - aware labeling loss in order to enhance the performance .", "forward": true, "src_ids": "2022.acl-long.428_1047"}
+{"input": "nested entities is done by using OtherScientificTerm| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "nested entities", "output": "expressive structures", "neg_sample": ["nested entities is done by using OtherScientificTerm", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "In this work , we resort to more expressive structures , lexicalized constituency trees in which constituents are annotated by headwords , to model nested entities .", "forward": false, "src_ids": "2022.acl-long.428_1048"}
+{"input": "partial marginalization is done by using Method| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "partial marginalization", "output": "eisner - satta algorithm", "neg_sample": ["partial marginalization is done by using Method", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently . In addition , we propose to use ( 1 ) a two - stage strategy ( 2 ) a head regularization loss and ( 3 ) a head - aware labeling loss in order to enhance the performance .", "forward": false, "src_ids": "2022.acl-long.428_1049"}
+{"input": "inference is done by using Method| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "inference", "output": "eisner - satta algorithm", "neg_sample": ["inference is done by using Method", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently . In addition , we propose to use ( 1 ) a two - stage strategy ( 2 ) a head regularization loss and ( 3 ) a head - aware labeling loss in order to enhance the performance .", "forward": false, "src_ids": "2022.acl-long.428_1050"}
+{"input": "eisner - satta algorithm is used for Task| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "eisner - satta algorithm", "output": "inference", "neg_sample": ["eisner - satta algorithm is used for Task", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.428", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently . In addition , we propose to use ( 1 ) a two - stage strategy ( 2 ) a head regularization loss and ( 3 ) a head - aware labeling loss in order to enhance the performance .", "forward": true, "src_ids": "2022.acl-long.428_1051"}
+{"input": "natural language processing is done by using Method| context: knowledge in natural language processing ( nlp ) has been a rising trend especially after the advent of large scale pre - trained models . nlp models with attention to knowledge can i ) access unlimited amount of external information ; ii ) delegate the task of storing knowledge from its parameter space to knowledge sources ; iii ) obtain up - to - date information ; iv ) make prediction results more explainable via selected knowledge .", "entity": "natural language processing", "output": "knowledge - augmented methods", "neg_sample": ["natural language processing is done by using Method", "knowledge in natural language processing ( nlp ) has been a rising trend especially after the advent of large scale pre - trained models .", "nlp models with attention to knowledge can i ) access unlimited amount of external information ; ii ) delegate the task of storing knowledge from its parameter space to knowledge sources ; iii ) obtain up - to - date information ; iv ) make prediction results more explainable via selected knowledge ."], "relation": "used for", "id": "2022.acl-tutorials.3", "year": 2022, "rel_sent": "Knowledge - Augmented Methods for Natural Language Processing.", "forward": false, "src_ids": "2022.acl-tutorials.3_1052"}
+{"input": "knowledge - augmented methods is used for Task| context: nlp models with attention to knowledge can i ) access unlimited amount of external information ; ii ) delegate the task of storing knowledge from its parameter space to knowledge sources ; iii ) obtain up - to - date information ; iv ) make prediction results more explainable via selected knowledge .", "entity": "knowledge - augmented methods", "output": "natural language processing", "neg_sample": ["knowledge - augmented methods is used for Task", "nlp models with attention to knowledge can i ) access unlimited amount of external information ; ii ) delegate the task of storing knowledge from its parameter space to knowledge sources ; iii ) obtain up - to - date information ; iv ) make prediction results more explainable via selected knowledge ."], "relation": "used for", "id": "2022.acl-tutorials.3", "year": 2022, "rel_sent": "Knowledge - Augmented Methods for Natural Language Processing.", "forward": true, "src_ids": "2022.acl-tutorials.3_1053"}
+{"input": "multi - document summarization of medical literature is done by using Method| context: although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization . many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital . others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background . despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization .", "entity": "multi - document summarization of medical literature", "output": "discriminative marginalized probabilistic neural method", "neg_sample": ["multi - document summarization of medical literature is done by using Method", "although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization .", "many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital .", "others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background .", "despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization ."], "relation": "used for", "id": "2022.acl-long.15", "year": 2022, "rel_sent": "Discriminative Marginalized Probabilistic Neural Method for Multi - Document Summarization of Medical Literature.", "forward": false, "src_ids": "2022.acl-long.15_1054"}
+{"input": "discriminative marginalized probabilistic neural method is used for Task| context: although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization . many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital . others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background . despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization .", "entity": "discriminative marginalized probabilistic neural method", "output": "multi - document summarization of medical literature", "neg_sample": ["discriminative marginalized probabilistic neural method is used for Task", "although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization .", "many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital .", "others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background .", "despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization ."], "relation": "used for", "id": "2022.acl-long.15", "year": 2022, "rel_sent": "Discriminative Marginalized Probabilistic Neural Method for Multi - Document Summarization of Medical Literature.", "forward": true, "src_ids": "2022.acl-long.15_1055"}
+{"input": "multi - document summary is done by using Method| context: although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization . many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital . others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background . despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization .", "entity": "multi - document summary", "output": "discriminative marginalized probabilistic method ( damen )", "neg_sample": ["multi - document summary is done by using Method", "although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization .", "many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital .", "others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background .", "despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization ."], "relation": "used for", "id": "2022.acl-long.15", "year": 2022, "rel_sent": "For this reason , we propose a novel discriminative marginalized probabilistic method ( DAMEN ) trained to discriminate critical information from a cluster of topic - related medical documents and generate a multi - document summary via token probability marginalization .", "forward": false, "src_ids": "2022.acl-long.15_1056"}
+{"input": "discriminative marginalized probabilistic method ( damen ) is used for OtherScientificTerm| context: although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization . many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital . others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background . despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization .", "entity": "discriminative marginalized probabilistic method ( damen )", "output": "multi - document summary", "neg_sample": ["discriminative marginalized probabilistic method ( damen ) is used for OtherScientificTerm", "although current state - of - the - art transformer - based solutions succeeded in a wide range for single - document nlp tasks , they still struggle to address multi - input tasks such as multi - document summarization .", "many solutions truncate the inputs , thus ignoring potential summary - relevant contents , which is unacceptable in the medical domain where each information can be vital .", "others leverage linear model approximations to apply multi - input concatenation , worsening the results because all information is considered , even if it is conflicting or noisy with respect to a shared background .", "despite the importance and social impact of medicine , there are no ad - hoc solutions for multi - document summarization ."], "relation": "used for", "id": "2022.acl-long.15", "year": 2022, "rel_sent": "For this reason , we propose a novel discriminative marginalized probabilistic method ( DAMEN ) trained to discriminate critical information from a cluster of topic - related medical documents and generate a multi - document summary via token probability marginalization .", "forward": true, "src_ids": "2022.acl-long.15_1057"}
+{"input": "social media framework is used for Task| context: easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular . however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs . detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society .", "entity": "social media framework", "output": "fake news spread", "neg_sample": ["social media framework is used for Task", "easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular .", "however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs .", "detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society ."], "relation": "used for", "id": "2022.acl-long.97", "year": 2022, "rel_sent": "Our experiments over two challenging fake news detection tasks show that using inference operators leads to a better understanding of the social media framework enabling fake news spread , resulting in improved performance .", "forward": true, "src_ids": "2022.acl-long.97_1058"}
+{"input": "graph edges is done by using Method| context: easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular . however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs . detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society .", "entity": "graph edges", "output": "inference operators", "neg_sample": ["graph edges is done by using Method", "easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular .", "however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs .", "detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society ."], "relation": "used for", "id": "2022.acl-long.97", "year": 2022, "rel_sent": "After embedding this information , we formulate inference operators which augment the graph edges by revealing unobserved interactions between its elements , such as similarity between documents ' contents and users ' engagement patterns .", "forward": false, "src_ids": "2022.acl-long.97_1059"}
+{"input": "social media framework is done by using Method| context: easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular . however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs . detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society .", "entity": "social media framework", "output": "inference operators", "neg_sample": ["social media framework is done by using Method", "easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular .", "however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs .", "detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society ."], "relation": "used for", "id": "2022.acl-long.97", "year": 2022, "rel_sent": "Our experiments over two challenging fake news detection tasks show that using inference operators leads to a better understanding of the social media framework enabling fake news spread , resulting in improved performance .", "forward": false, "src_ids": "2022.acl-long.97_1060"}
+{"input": "inference operators is used for OtherScientificTerm| context: easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular . however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs . detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society .", "entity": "inference operators", "output": "graph edges", "neg_sample": ["inference operators is used for OtherScientificTerm", "easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular .", "however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs .", "detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society ."], "relation": "used for", "id": "2022.acl-long.97", "year": 2022, "rel_sent": "After embedding this information , we formulate inference operators which augment the graph edges by revealing unobserved interactions between its elements , such as similarity between documents ' contents and users ' engagement patterns .", "forward": true, "src_ids": "2022.acl-long.97_1061"}
+{"input": "fake news spread is done by using Method| context: easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular . however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs . detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society .", "entity": "fake news spread", "output": "social media framework", "neg_sample": ["fake news spread is done by using Method", "easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular .", "however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs .", "detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society ."], "relation": "used for", "id": "2022.acl-long.97", "year": 2022, "rel_sent": "Our experiments over two challenging fake news detection tasks show that using inference operators leads to a better understanding of the social media framework enabling fake news spread , resulting in improved performance .", "forward": false, "src_ids": "2022.acl-long.97_1062"}
+{"input": "inference operators is used for Method| context: easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular . however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs . detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society .", "entity": "inference operators", "output": "social media framework", "neg_sample": ["inference operators is used for Method", "easy access , variety of content , and fast widespread interactions are some of the reasons making social media increasingly popular .", "however , this rise has also enabled the propagation of fake news , text published by news sources with an intent to spread misinformation and sway beliefs .", "detecting it is an important and challenging problem to prevent large scale misinformation and maintain a healthy society ."], "relation": "used for", "id": "2022.acl-long.97", "year": 2022, "rel_sent": "Our experiments over two challenging fake news detection tasks show that using inference operators leads to a better understanding of the social media framework enabling fake news spread , resulting in improved performance .", "forward": true, "src_ids": "2022.acl-long.97_1063"}
+{"input": "generation models is done by using OtherScientificTerm| context: pre - trained language models ( e.g. bart ) have shown impressive results when fine - tuned on large summarization datasets . however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations .", "entity": "generation models", "output": "training dynamics", "neg_sample": ["generation models is done by using OtherScientificTerm", "pre - trained language models ( e.g.", "bart ) have shown impressive results when fine - tuned on large summarization datasets .", "however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations ."], "relation": "used for", "id": "2022.findings-acl.163", "year": 2022, "rel_sent": "In this work , we analyze the training dynamics for generation models , focusing on summarization .", "forward": false, "src_ids": "2022.findings-acl.163_1064"}
+{"input": "modifying training is done by using Generic| context: pre - trained language models ( e.g. bart ) have shown impressive results when fine - tuned on large summarization datasets . however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations .", "entity": "modifying training", "output": "complementary approaches", "neg_sample": ["modifying training is done by using Generic", "pre - trained language models ( e.g.", "bart ) have shown impressive results when fine - tuned on large summarization datasets .", "however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations ."], "relation": "used for", "id": "2022.findings-acl.163", "year": 2022, "rel_sent": "Based on these observations , we explore complementary approaches for modifying training : first , disregarding high - loss tokens that are challenging to learn and second , disregarding low - loss tokens that are learnt very quickly in the latter stages of the training process .", "forward": false, "src_ids": "2022.findings-acl.163_1065"}
+{"input": "complementary approaches is used for Task| context: pre - trained language models ( e.g. bart ) have shown impressive results when fine - tuned on large summarization datasets . however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations .", "entity": "complementary approaches", "output": "modifying training", "neg_sample": ["complementary approaches is used for Task", "pre - trained language models ( e.g.", "bart ) have shown impressive results when fine - tuned on large summarization datasets .", "however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations ."], "relation": "used for", "id": "2022.findings-acl.163", "year": 2022, "rel_sent": "Based on these observations , we explore complementary approaches for modifying training : first , disregarding high - loss tokens that are challenging to learn and second , disregarding low - loss tokens that are learnt very quickly in the latter stages of the training process .", "forward": true, "src_ids": "2022.findings-acl.163_1066"}
+{"input": "training dynamics is used for Method| context: pre - trained language models ( e.g. bart ) have shown impressive results when fine - tuned on large summarization datasets . however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations .", "entity": "training dynamics", "output": "generation models", "neg_sample": ["training dynamics is used for Method", "pre - trained language models ( e.g.", "bart ) have shown impressive results when fine - tuned on large summarization datasets .", "however , little is understood about this fine - tuning process , including what knowledge is retained from pre - training time or how content selection and generation strategies are learnt across iterations ."], "relation": "used for", "id": "2022.findings-acl.163", "year": 2022, "rel_sent": "In this work , we analyze the training dynamics for generation models , focusing on summarization .", "forward": true, "src_ids": "2022.findings-acl.163_1067"}
+{"input": "generic and robust framework is used for Task| context: existing solutions , however , either ignore external unstructured data completely or devise dataset - specific solutions .", "entity": "generic and robust framework", "output": "enrichment of tabular datasets", "neg_sample": ["generic and robust framework is used for Task", "existing solutions , however , either ignore external unstructured data completely or devise dataset - specific solutions ."], "relation": "used for", "id": "2022.acl-long.111", "year": 2022, "rel_sent": "In this study we proposed Few - Shot Transformer based Enrichment ( FeSTE ) , a generic and robust framework for the enrichment of tabular datasets using unstructured data .", "forward": true, "src_ids": "2022.acl-long.111_1068"}
+{"input": "enrichment of tabular datasets is done by using Method| context: the enrichment of tabular datasets using external sources has gained significant attention in recent years . existing solutions , however , either ignore external unstructured data completely or devise dataset - specific solutions .", "entity": "enrichment of tabular datasets", "output": "generic and robust framework", "neg_sample": ["enrichment of tabular datasets is done by using Method", "the enrichment of tabular datasets using external sources has gained significant attention in recent years .", "existing solutions , however , either ignore external unstructured data completely or devise dataset - specific solutions ."], "relation": "used for", "id": "2022.acl-long.111", "year": 2022, "rel_sent": "In this study we proposed Few - Shot Transformer based Enrichment ( FeSTE ) , a generic and robust framework for the enrichment of tabular datasets using unstructured data .", "forward": false, "src_ids": "2022.acl-long.111_1069"}
+{"input": "inclusive non - expert practice is done by using Method| context: transcribing speech for primarily oral , local languages is often a joint effort involving speakers and outsiders . it is commonly motivated by externally - defined scientific goals , alongside local motivations such as language acquisition and access to heritage materials .", "entity": "inclusive non - expert practice", "output": "situated systems design", "neg_sample": ["inclusive non - expert practice is done by using Method", "transcribing speech for primarily oral , local languages is often a joint effort involving speakers and outsiders .", "it is commonly motivated by externally - defined scientific goals , alongside local motivations such as language acquisition and access to heritage materials ."], "relation": "used for", "id": "2022.computel-1.11", "year": 2022, "rel_sent": "We show that situated systems design for inclusive non - expert practice is a promising new direction for working with speakers of local languages .", "forward": false, "src_ids": "2022.computel-1.11_1070"}
+{"input": "situated systems design is used for Task| context: transcribing speech for primarily oral , local languages is often a joint effort involving speakers and outsiders . it is commonly motivated by externally - defined scientific goals , alongside local motivations such as language acquisition and access to heritage materials .", "entity": "situated systems design", "output": "inclusive non - expert practice", "neg_sample": ["situated systems design is used for Task", "transcribing speech for primarily oral , local languages is often a joint effort involving speakers and outsiders .", "it is commonly motivated by externally - defined scientific goals , alongside local motivations such as language acquisition and access to heritage materials ."], "relation": "used for", "id": "2022.computel-1.11", "year": 2022, "rel_sent": "We show that situated systems design for inclusive non - expert practice is a promising new direction for working with speakers of local languages .", "forward": true, "src_ids": "2022.computel-1.11_1071"}
+{"input": "machine translation is done by using Task| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "machine translation", "output": "morphological segmentation", "neg_sample": ["machine translation is done by using Task", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "BPE vs. Morphological Segmentation : A Case Study on Machine Translation of Four Polysynthetic Languages.", "forward": false, "src_ids": "2022.findings-acl.78_1072"}
+{"input": "morphological segmentation is used for Task| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "morphological segmentation", "output": "machine translation", "neg_sample": ["morphological segmentation is used for Task", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "BPE vs. Morphological Segmentation : A Case Study on Machine Translation of Four Polysynthetic Languages.", "forward": true, "src_ids": "2022.findings-acl.78_1073"}
+{"input": "polysynthetic languages is done by using Method| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "polysynthetic languages", "output": "supervised and unsupervised morphological segmentation methods", "neg_sample": ["polysynthetic languages is done by using Method", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "We investigate a wide variety of supervised and unsupervised morphological segmentation methods for four polysynthetic languages : Nahuatl , Raramuri , Shipibo - Konibo , and Wixarika .", "forward": false, "src_ids": "2022.findings-acl.78_1074"}
+{"input": "morphological segmentation datasets is used for Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "morphological segmentation datasets", "output": "raramuri", "neg_sample": ["morphological segmentation datasets is used for Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri - Spanish .", "forward": true, "src_ids": "2022.findings-acl.78_1075"}
+{"input": "machine translation ( mt ) is done by using Method| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "machine translation ( mt )", "output": "morphologically inspired segmentation methods", "neg_sample": ["machine translation ( mt ) is done by using Method", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Then , we compare the morphologically inspired segmentation methods against Byte - Pair Encodings ( BPEs ) as inputs for machine translation ( MT ) when translating to and from Spanish .", "forward": false, "src_ids": "2022.findings-acl.78_1076"}
+{"input": "byte - pair encodings ( bpes ) is used for Task| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "byte - pair encodings ( bpes )", "output": "machine translation ( mt )", "neg_sample": ["byte - pair encodings ( bpes ) is used for Task", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Then , we compare the morphologically inspired segmentation methods against Byte - Pair Encodings ( BPEs ) as inputs for machine translation ( MT ) when translating to and from Spanish .", "forward": true, "src_ids": "2022.findings-acl.78_1077"}
+{"input": "morphologically inspired segmentation methods is used for Task| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "morphologically inspired segmentation methods", "output": "machine translation ( mt )", "neg_sample": ["morphologically inspired segmentation methods is used for Task", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Then , we compare the morphologically inspired segmentation methods against Byte - Pair Encodings ( BPEs ) as inputs for machine translation ( MT ) when translating to and from Spanish .", "forward": true, "src_ids": "2022.findings-acl.78_1078"}
+{"input": "raramuri is done by using Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "raramuri", "output": "morphological segmentation datasets", "neg_sample": ["raramuri is done by using Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri - Spanish .", "forward": false, "src_ids": "2022.findings-acl.78_1079"}
+{"input": "shipibo - konibo is done by using Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "shipibo - konibo", "output": "morphological segmentation datasets", "neg_sample": ["shipibo - konibo is done by using Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri - Spanish .", "forward": false, "src_ids": "2022.findings-acl.78_1080"}
+{"input": "raramuri - spanish is done by using Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "raramuri - spanish", "output": "parallel corpus", "neg_sample": ["raramuri - spanish is done by using Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri - Spanish .", "forward": false, "src_ids": "2022.findings-acl.78_1081"}
+{"input": "parallel corpus is used for Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "parallel corpus", "output": "raramuri - spanish", "neg_sample": ["parallel corpus is used for Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.78", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri - Spanish .", "forward": true, "src_ids": "2022.findings-acl.78_1082"}
+{"input": "dialogue is done by using Task| context: in communication , a human would recognize the emotion of an interlocutor and respond with an appropriate emotion , such as empathy and comfort .", "entity": "dialogue", "output": "emotion recognition", "neg_sample": ["dialogue is done by using Task", "in communication , a human would recognize the emotion of an interlocutor and respond with an appropriate emotion , such as empathy and comfort ."], "relation": "used for", "id": "2022.acl-srw.3", "year": 2022, "rel_sent": "We hope that the constructed corpus will facilitate the study on emotion recognition in a dialogue and emotion - aware dialogue response generation .", "forward": false, "src_ids": "2022.acl-srw.3_1083"}
+{"input": "emotion recognition is used for Material| context: in communication , a human would recognize the emotion of an interlocutor and respond with an appropriate emotion , such as empathy and comfort .", "entity": "emotion recognition", "output": "dialogue", "neg_sample": ["emotion recognition is used for Material", "in communication , a human would recognize the emotion of an interlocutor and respond with an appropriate emotion , such as empathy and comfort ."], "relation": "used for", "id": "2022.acl-srw.3", "year": 2022, "rel_sent": "We hope that the constructed corpus will facilitate the study on emotion recognition in a dialogue and emotion - aware dialogue response generation .", "forward": true, "src_ids": "2022.acl-srw.3_1084"}
+{"input": "humane and diversity - aware language technology is done by using Material| context: informal social interaction is the primordial home of human language . through the efforts of a worldwide language documentation movement , such corpora are increasingly becoming available .", "entity": "humane and diversity - aware language technology", "output": "conversational corpora", "neg_sample": ["humane and diversity - aware language technology is done by using Material", "informal social interaction is the primordial home of human language .", "through the efforts of a worldwide language documentation movement , such corpora are increasingly becoming available ."], "relation": "used for", "id": "2022.acl-long.385", "year": 2022, "rel_sent": "From text to talk : Harnessing conversational corpora for humane and diversity - aware language technology.", "forward": false, "src_ids": "2022.acl-long.385_1085"}
+{"input": "conversational corpora is used for Task| context: informal social interaction is the primordial home of human language . linguistically diverse conversational corpora are an important and largely untapped resource for computational linguistics and language technology . through the efforts of a worldwide language documentation movement , such corpora are increasingly becoming available .", "entity": "conversational corpora", "output": "humane and diversity - aware language technology", "neg_sample": ["conversational corpora is used for Task", "informal social interaction is the primordial home of human language .", "linguistically diverse conversational corpora are an important and largely untapped resource for computational linguistics and language technology .", "through the efforts of a worldwide language documentation movement , such corpora are increasingly becoming available ."], "relation": "used for", "id": "2022.acl-long.385", "year": 2022, "rel_sent": "From text to talk : Harnessing conversational corpora for humane and diversity - aware language technology.", "forward": true, "src_ids": "2022.acl-long.385_1086"}
+{"input": "detecting signs of depression is done by using Method| context: depression is one of the most common mentalissues faced by people . detecting signs ofdepression early on can help in the treatmentand prevention of extreme outcomes like suicide . since the advent of the internet , peoplehave felt more comfortable discussing topicslike depression online due to the anonymityit provides .", "entity": "detecting signs of depression", "output": "transfer learning", "neg_sample": ["detecting signs of depression is done by using Method", "depression is one of the most common mentalissues faced by people .", "detecting signs ofdepression early on can help in the treatmentand prevention of extreme outcomes like suicide .", "since the advent of the internet , peoplehave felt more comfortable discussing topicslike depression online due to the anonymityit provides ."], "relation": "used for", "id": "2022.ltedi-1.50", "year": 2022, "rel_sent": "SSN@LT - EDI - ACL2022 : Transfer Learning using BERT for Detecting Signs of Depression from Social Media Texts.", "forward": false, "src_ids": "2022.ltedi-1.50_1087"}
+{"input": "transfer learning is used for Task| context: depression is one of the most common mentalissues faced by people .", "entity": "transfer learning", "output": "detecting signs of depression", "neg_sample": ["transfer learning is used for Task", "depression is one of the most common mentalissues faced by people ."], "relation": "used for", "id": "2022.ltedi-1.50", "year": 2022, "rel_sent": "SSN@LT - EDI - ACL2022 : Transfer Learning using BERT for Detecting Signs of Depression from Social Media Texts.", "forward": true, "src_ids": "2022.ltedi-1.50_1088"}
+{"input": "bert is used for Task| context: depression is one of the most common mentalissues faced by people .", "entity": "bert", "output": "detecting signs of depression", "neg_sample": ["bert is used for Task", "depression is one of the most common mentalissues faced by people ."], "relation": "used for", "id": "2022.ltedi-1.50", "year": 2022, "rel_sent": "SSN@LT - EDI - ACL2022 : Transfer Learning using BERT for Detecting Signs of Depression from Social Media Texts.", "forward": true, "src_ids": "2022.ltedi-1.50_1089"}
+{"input": "trustworthy tabular reasoning is done by using Task| context: when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably . however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether .", "entity": "trustworthy tabular reasoning", "output": "evidence extraction", "neg_sample": ["trustworthy tabular reasoning is done by using Task", "when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably .", "however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether ."], "relation": "used for", "id": "2022.acl-long.231", "year": 2022, "rel_sent": "Right for the Right Reason : Evidence Extraction for Trustworthy Tabular Reasoning.", "forward": false, "src_ids": "2022.acl-long.231_1090"}
+{"input": "evidence extraction is used for Task| context: when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably . however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether .", "entity": "evidence extraction", "output": "trustworthy tabular reasoning", "neg_sample": ["evidence extraction is used for Task", "when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably .", "however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether ."], "relation": "used for", "id": "2022.acl-long.231", "year": 2022, "rel_sent": "Right for the Right Reason : Evidence Extraction for Trustworthy Tabular Reasoning.", "forward": true, "src_ids": "2022.acl-long.231_1091"}
+{"input": "infotabs is done by using Method| context: when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably . however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether .", "entity": "infotabs", "output": "unsupervised and supervised evidence extraction strategies", "neg_sample": ["infotabs is done by using Method", "when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably .", "however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether ."], "relation": "used for", "id": "2022.acl-long.231", "year": 2022, "rel_sent": "First , we crowdsource evidence row labels and develop several unsupervised and supervised evidence extraction strategies for InfoTabS , a tabular NLI benchmark .", "forward": false, "src_ids": "2022.acl-long.231_1092"}
+{"input": "unsupervised and supervised evidence extraction strategies is used for Material| context: when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably . however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether .", "entity": "unsupervised and supervised evidence extraction strategies", "output": "infotabs", "neg_sample": ["unsupervised and supervised evidence extraction strategies is used for Material", "when pre - trained contextualized embedding - based models developed for unstructured data are adapted for structured tabular data , they perform admirably .", "however , recent probing studies show that these models use spurious correlations , and often predict inference labels by focusing on false evidence or ignoring it altogether ."], "relation": "used for", "id": "2022.acl-long.231", "year": 2022, "rel_sent": "First , we crowdsource evidence row labels and develop several unsupervised and supervised evidence extraction strategies for InfoTabS , a tabular NLI benchmark .", "forward": true, "src_ids": "2022.acl-long.231_1093"}
+{"input": "multilingual models is done by using Method| context: massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning .", "entity": "multilingual models", "output": "multi task learning", "neg_sample": ["multilingual models is done by using Method", "massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning ."], "relation": "used for", "id": "2022.acl-long.374", "year": 2022, "rel_sent": "Multi Task Learning For Zero Shot Performance Prediction of Multilingual Models.", "forward": false, "src_ids": "2022.acl-long.374_1094"}
+{"input": "multi task learning is used for Method| context: massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning .", "entity": "multi task learning", "output": "multilingual models", "neg_sample": ["multi task learning is used for Method", "massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning ."], "relation": "used for", "id": "2022.acl-long.374", "year": 2022, "rel_sent": "Multi Task Learning For Zero Shot Performance Prediction of Multilingual Models.", "forward": true, "src_ids": "2022.acl-long.374_1095"}
+{"input": "predictors is done by using Method| context: massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning .", "entity": "predictors", "output": "predictive models", "neg_sample": ["predictors is done by using Method", "massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning ."], "relation": "used for", "id": "2022.acl-long.374", "year": 2022, "rel_sent": "We jointly train predictive models for different tasks which helps us build more accurate predictors for tasks where we have test data in very few languages to measure the actual performance of the model .", "forward": false, "src_ids": "2022.acl-long.374_1096"}
+{"input": "predictive models is used for OtherScientificTerm| context: massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning .", "entity": "predictive models", "output": "predictors", "neg_sample": ["predictive models is used for OtherScientificTerm", "massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning ."], "relation": "used for", "id": "2022.acl-long.374", "year": 2022, "rel_sent": "We jointly train predictive models for different tasks which helps us build more accurate predictors for tasks where we have test data in very few languages to measure the actual performance of the model .", "forward": true, "src_ids": "2022.acl-long.374_1097"}
+{"input": "zero - shot is done by using OtherScientificTerm| context: massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning .", "entity": "zero - shot", "output": "features", "neg_sample": ["zero - shot is done by using OtherScientificTerm", "massively multilingual transformer based language models have been observed to be surprisingly effective on zero - shot transfer across languages , though the performance varies from language to language depending on the pivot language(s ) used for fine - tuning ."], "relation": "used for", "id": "2022.acl-long.374", "year": 2022, "rel_sent": "Our approach also lends us the ability to perform a much more robust feature selection , and identify a common set of features that influence zero - shot performance across a variety of tasks .", "forward": false, "src_ids": "2022.acl-long.374_1098"}
+{"input": "under - represented languages is done by using Method| context: the performance of multilingual pretrained models is highly dependent on the availability of monolingual or parallel text present in a target language . thus , the majority of the world 's languages can not benefit from recent progress in nlp as they have no or limited textual data .", "entity": "under - represented languages", "output": "nlp technology", "neg_sample": ["under - represented languages is done by using Method", "the performance of multilingual pretrained models is highly dependent on the availability of monolingual or parallel text present in a target language .", "thus , the majority of the world 's languages can not benefit from recent progress in nlp as they have no or limited textual data ."], "relation": "used for", "id": "2022.acl-long.61", "year": 2022, "rel_sent": "To expand possibilities of using NLP technology in these under - represented languages , we systematically study strategies that relax the reliance on conventional language resources through the use of bilingual lexicons , an alternative resource with much better language coverage .", "forward": false, "src_ids": "2022.acl-long.61_1099"}
+{"input": "nlp technology is used for Material| context: the performance of multilingual pretrained models is highly dependent on the availability of monolingual or parallel text present in a target language . thus , the majority of the world 's languages can not benefit from recent progress in nlp as they have no or limited textual data .", "entity": "nlp technology", "output": "under - represented languages", "neg_sample": ["nlp technology is used for Material", "the performance of multilingual pretrained models is highly dependent on the availability of monolingual or parallel text present in a target language .", "thus , the majority of the world 's languages can not benefit from recent progress in nlp as they have no or limited textual data ."], "relation": "used for", "id": "2022.acl-long.61", "year": 2022, "rel_sent": "To expand possibilities of using NLP technology in these under - represented languages , we systematically study strategies that relax the reliance on conventional language resources through the use of bilingual lexicons , an alternative resource with much better language coverage .", "forward": true, "src_ids": "2022.acl-long.61_1100"}
+{"input": "contextual embedding models is done by using Material| context: lexical substitution is the task of generating meaningful substitutes for a word in a given textual context . contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence .", "entity": "contextual embedding models", "output": "lexical resources", "neg_sample": ["contextual embedding models is done by using Material", "lexical substitution is the task of generating meaningful substitutes for a word in a given textual context .", "contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence ."], "relation": "used for", "id": "2022.acl-long.87", "year": 2022, "rel_sent": "LexSubCon : Integrating Knowledge from Lexical Resources into Contextual Embeddings for Lexical Substitution.", "forward": false, "src_ids": "2022.acl-long.87_1101"}
+{"input": "lexical substitution is done by using Method| context: lexical substitution is the task of generating meaningful substitutes for a word in a given textual context . contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence .", "entity": "lexical substitution", "output": "contextual embedding models", "neg_sample": ["lexical substitution is done by using Method", "lexical substitution is the task of generating meaningful substitutes for a word in a given textual context .", "contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence ."], "relation": "used for", "id": "2022.acl-long.87", "year": 2022, "rel_sent": "LexSubCon : Integrating Knowledge from Lexical Resources into Contextual Embeddings for Lexical Substitution.", "forward": false, "src_ids": "2022.acl-long.87_1102"}
+{"input": "lexical resources is used for Method| context: lexical substitution is the task of generating meaningful substitutes for a word in a given textual context . contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence .", "entity": "lexical resources", "output": "contextual embedding models", "neg_sample": ["lexical resources is used for Method", "lexical substitution is the task of generating meaningful substitutes for a word in a given textual context .", "contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence ."], "relation": "used for", "id": "2022.acl-long.87", "year": 2022, "rel_sent": "LexSubCon : Integrating Knowledge from Lexical Resources into Contextual Embeddings for Lexical Substitution.", "forward": true, "src_ids": "2022.acl-long.87_1103"}
+{"input": "coinco benchmark datasets is used for Task| context: lexical substitution is the task of generating meaningful substitutes for a word in a given textual context .", "entity": "coinco benchmark datasets", "output": "lexical substitution tasks", "neg_sample": ["coinco benchmark datasets is used for Task", "lexical substitution is the task of generating meaningful substitutes for a word in a given textual context ."], "relation": "used for", "id": "2022.acl-long.87", "year": 2022, "rel_sent": "Our experiments show that LexSubCon outperforms previous state - of - the - art methods by at least 2 % over all the official lexical substitution metrics on LS07 and CoInCo benchmark datasets that are widely used for lexical substitution tasks .", "forward": true, "src_ids": "2022.acl-long.87_1104"}
+{"input": "lexical substitution tasks is done by using Material| context: lexical substitution is the task of generating meaningful substitutes for a word in a given textual context . contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence .", "entity": "lexical substitution tasks", "output": "coinco benchmark datasets", "neg_sample": ["lexical substitution tasks is done by using Material", "lexical substitution is the task of generating meaningful substitutes for a word in a given textual context .", "contextual word embedding models have achieved state - of - the - art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence ."], "relation": "used for", "id": "2022.acl-long.87", "year": 2022, "rel_sent": "Our experiments show that LexSubCon outperforms previous state - of - the - art methods by at least 2 % over all the official lexical substitution metrics on LS07 and CoInCo benchmark datasets that are widely used for lexical substitution tasks .", "forward": false, "src_ids": "2022.acl-long.87_1105"}
+{"input": "fact - checking is done by using OtherScientificTerm| context: automatic fake news detection models are ostensibly based on logic , where the truth of a claim made in a headline can be determined by supporting or refuting evidence found in a resulting web query . these models are believed to be reasoning in some way ; however , it has been shown that these same results , or better , can be achieved without considering the claim at all - only the evidence . this implies that other signals are contained within the examined evidence , and could be based on manipulable factors such as emotion , sentiment , or part - of - speech ( pos ) frequencies , which are vulnerable to adversarial inputs .", "entity": "fact - checking", "output": "manipulable features", "neg_sample": ["fact - checking is done by using OtherScientificTerm", "automatic fake news detection models are ostensibly based on logic , where the truth of a claim made in a headline can be determined by supporting or refuting evidence found in a resulting web query .", "these models are believed to be reasoning in some way ; however , it has been shown that these same results , or better , can be achieved without considering the claim at all - only the evidence .", "this implies that other signals are contained within the examined evidence , and could be based on manipulable factors such as emotion , sentiment , or part - of - speech ( pos ) frequencies , which are vulnerable to adversarial inputs ."], "relation": "used for", "id": "2022.fever-1.4", "year": 2022, "rel_sent": "We provide quantifiable results that prove our hypothesis that manipulable features are being used for fact - checking .", "forward": false, "src_ids": "2022.fever-1.4_1106"}
+{"input": "manipulable features is used for Task| context: automatic fake news detection models are ostensibly based on logic , where the truth of a claim made in a headline can be determined by supporting or refuting evidence found in a resulting web query . these models are believed to be reasoning in some way ; however , it has been shown that these same results , or better , can be achieved without considering the claim at all - only the evidence . this implies that other signals are contained within the examined evidence , and could be based on manipulable factors such as emotion , sentiment , or part - of - speech ( pos ) frequencies , which are vulnerable to adversarial inputs .", "entity": "manipulable features", "output": "fact - checking", "neg_sample": ["manipulable features is used for Task", "automatic fake news detection models are ostensibly based on logic , where the truth of a claim made in a headline can be determined by supporting or refuting evidence found in a resulting web query .", "these models are believed to be reasoning in some way ; however , it has been shown that these same results , or better , can be achieved without considering the claim at all - only the evidence .", "this implies that other signals are contained within the examined evidence , and could be based on manipulable factors such as emotion , sentiment , or part - of - speech ( pos ) frequencies , which are vulnerable to adversarial inputs ."], "relation": "used for", "id": "2022.fever-1.4", "year": 2022, "rel_sent": "We provide quantifiable results that prove our hypothesis that manipulable features are being used for fact - checking .", "forward": true, "src_ids": "2022.fever-1.4_1107"}
+{"input": "time features is done by using Method| context: temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information . in this context , tensor decomposition has successfully modeled interactions between entities and relations .", "entity": "time features", "output": "cycle - aware time - encoding scheme", "neg_sample": ["time features is done by using Method", "temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information .", "in this context , tensor decomposition has successfully modeled interactions between entities and relations ."], "relation": "used for", "id": "2022.repl4nlp-1.12", "year": 2022, "rel_sent": "Noting several limitations in current approaches to represent time , we propose a cycle - aware time - encoding scheme for time features , which is model - agnostic and offers a more generalized representation of time .", "forward": false, "src_ids": "2022.repl4nlp-1.12_1108"}
+{"input": "cycle - aware time - encoding scheme is used for OtherScientificTerm| context: temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information . in this context , tensor decomposition has successfully modeled interactions between entities and relations .", "entity": "cycle - aware time - encoding scheme", "output": "time features", "neg_sample": ["cycle - aware time - encoding scheme is used for OtherScientificTerm", "temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information .", "in this context , tensor decomposition has successfully modeled interactions between entities and relations ."], "relation": "used for", "id": "2022.repl4nlp-1.12", "year": 2022, "rel_sent": "Noting several limitations in current approaches to represent time , we propose a cycle - aware time - encoding scheme for time features , which is model - agnostic and offers a more generalized representation of time .", "forward": true, "src_ids": "2022.repl4nlp-1.12_1109"}
+{"input": "time - sensitive data processing is done by using Method| context: temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information . in this context , tensor decomposition has successfully modeled interactions between entities and relations .", "entity": "time - sensitive data processing", "output": "unified temporal knowledge graph embedding framework", "neg_sample": ["time - sensitive data processing is done by using Method", "temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information .", "in this context , tensor decomposition has successfully modeled interactions between entities and relations ."], "relation": "used for", "id": "2022.repl4nlp-1.12", "year": 2022, "rel_sent": "We implement our methods in a unified temporal knowledge graph embedding framework , focusing on time - sensitive data processing .", "forward": false, "src_ids": "2022.repl4nlp-1.12_1110"}
+{"input": "unified temporal knowledge graph embedding framework is used for Task| context: temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information . in this context , tensor decomposition has successfully modeled interactions between entities and relations .", "entity": "unified temporal knowledge graph embedding framework", "output": "time - sensitive data processing", "neg_sample": ["unified temporal knowledge graph embedding framework is used for Task", "temporal knowledge graph completion ( tkgc ) has become a popular approach for reasoning over the event and temporal knowledge graphs , targeting the completion of knowledge with accurate but missing information .", "in this context , tensor decomposition has successfully modeled interactions between entities and relations ."], "relation": "used for", "id": "2022.repl4nlp-1.12", "year": 2022, "rel_sent": "We implement our methods in a unified temporal knowledge graph embedding framework , focusing on time - sensitive data processing .", "forward": true, "src_ids": "2022.repl4nlp-1.12_1111"}
+{"input": "spoilers is done by using Method| context: clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary .", "entity": "spoilers", "output": "question answering model", "neg_sample": ["spoilers is done by using Method", "clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary ."], "relation": "used for", "id": "2022.acl-long.484", "year": 2022, "rel_sent": "A large - scale evaluation and error analysis on a new corpus of 5,000 manually spoiled clickbait posts - the Webis Clickbait Spoiling Corpus 2022 - shows that our spoiler type classifier achieves an accuracy of 80 % , while the question answering model DeBERTa - large outperforms all others in generating spoilers for both types .", "forward": false, "src_ids": "2022.acl-long.484_1112"}
+{"input": "question answering model is used for OtherScientificTerm| context: clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary .", "entity": "question answering model", "output": "spoilers", "neg_sample": ["question answering model is used for OtherScientificTerm", "clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary ."], "relation": "used for", "id": "2022.acl-long.484", "year": 2022, "rel_sent": "A large - scale evaluation and error analysis on a new corpus of 5,000 manually spoiled clickbait posts - the Webis Clickbait Spoiling Corpus 2022 - shows that our spoiler type classifier achieves an accuracy of 80 % , while the question answering model DeBERTa - large outperforms all others in generating spoilers for both types .", "forward": true, "src_ids": "2022.acl-long.484_1113"}
+{"input": "insiders is done by using Method| context: social media is a breeding ground for threat narratives and related conspiracy theories . in these , an outside group threatens the integrity of an inside group , leading to the emergence of sharply defined group identities : insiders - agents with whom the authors identify and outsiders - agents who threaten the insiders . inferring the members of these groups constitutes a challenging new nlp task : ( i ) information is distributed over many poorly - constructed posts ; ( ii ) threats and threat agents are highly contextual , with the same post potentially having multiple agents assigned to membership in either group ; ( iii ) an agent 's identity is often implicit and transitive ; and ( iv ) phrases used to imply outsider status often do not follow common negative sentiment patterns .", "entity": "insiders", "output": "np2io", "neg_sample": ["insiders is done by using Method", "social media is a breeding ground for threat narratives and related conspiracy theories .", "in these , an outside group threatens the integrity of an inside group , leading to the emergence of sharply defined group identities : insiders - agents with whom the authors identify and outsiders - agents who threaten the insiders .", "inferring the members of these groups constitutes a challenging new nlp task : ( i ) information is distributed over many poorly - constructed posts ; ( ii ) threats and threat agents are highly contextual , with the same post potentially having multiple agents assigned to membership in either group ; ( iii ) an agent 's identity is often implicit and transitive ; and ( iv ) phrases used to imply outsider status often do not follow common negative sentiment patterns ."], "relation": "used for", "id": "2022.acl-long.341", "year": 2022, "rel_sent": "NP2IO leverages pretrained language modeling to classify Insiders and Outsiders .", "forward": false, "src_ids": "2022.acl-long.341_1114"}
+{"input": "pretrained language modeling is used for OtherScientificTerm| context: social media is a breeding ground for threat narratives and related conspiracy theories . inferring the members of these groups constitutes a challenging new nlp task : ( i ) information is distributed over many poorly - constructed posts ; ( ii ) threats and threat agents are highly contextual , with the same post potentially having multiple agents assigned to membership in either group ; ( iii ) an agent 's identity is often implicit and transitive ; and ( iv ) phrases used to imply outsider status often do not follow common negative sentiment patterns .", "entity": "pretrained language modeling", "output": "insiders", "neg_sample": ["pretrained language modeling is used for OtherScientificTerm", "social media is a breeding ground for threat narratives and related conspiracy theories .", "inferring the members of these groups constitutes a challenging new nlp task : ( i ) information is distributed over many poorly - constructed posts ; ( ii ) threats and threat agents are highly contextual , with the same post potentially having multiple agents assigned to membership in either group ; ( iii ) an agent 's identity is often implicit and transitive ; and ( iv ) phrases used to imply outsider status often do not follow common negative sentiment patterns ."], "relation": "used for", "id": "2022.acl-long.341", "year": 2022, "rel_sent": "NP2IO leverages pretrained language modeling to classify Insiders and Outsiders .", "forward": true, "src_ids": "2022.acl-long.341_1115"}
+{"input": "np2io is used for OtherScientificTerm| context: social media is a breeding ground for threat narratives and related conspiracy theories . inferring the members of these groups constitutes a challenging new nlp task : ( i ) information is distributed over many poorly - constructed posts ; ( ii ) threats and threat agents are highly contextual , with the same post potentially having multiple agents assigned to membership in either group ; ( iii ) an agent 's identity is often implicit and transitive ; and ( iv ) phrases used to imply outsider status often do not follow common negative sentiment patterns .", "entity": "np2io", "output": "insiders", "neg_sample": ["np2io is used for OtherScientificTerm", "social media is a breeding ground for threat narratives and related conspiracy theories .", "inferring the members of these groups constitutes a challenging new nlp task : ( i ) information is distributed over many poorly - constructed posts ; ( ii ) threats and threat agents are highly contextual , with the same post potentially having multiple agents assigned to membership in either group ; ( iii ) an agent 's identity is often implicit and transitive ; and ( iv ) phrases used to imply outsider status often do not follow common negative sentiment patterns ."], "relation": "used for", "id": "2022.acl-long.341", "year": 2022, "rel_sent": "NP2IO leverages pretrained language modeling to classify Insiders and Outsiders .", "forward": true, "src_ids": "2022.acl-long.341_1116"}
+{"input": "sentence - level functional roles is done by using OtherScientificTerm| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "sentence - level functional roles", "output": "ontology", "neg_sample": ["sentence - level functional roles is done by using OtherScientificTerm", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "We develop an ontology of six sentence - level functional roles for long - form answers , and annotate 3.9k sentences in 640 answer paragraphs .", "forward": false, "src_ids": "2022.acl-long.249_1117"}
+{"input": "long - form answers is done by using OtherScientificTerm| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "long - form answers", "output": "ontology", "neg_sample": ["long - form answers is done by using OtherScientificTerm", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "We develop an ontology of six sentence - level functional roles for long - form answers , and annotate 3.9k sentences in 640 answer paragraphs .", "forward": false, "src_ids": "2022.acl-long.249_1118"}
+{"input": "ontology is used for OtherScientificTerm| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "ontology", "output": "sentence - level functional roles", "neg_sample": ["ontology is used for OtherScientificTerm", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "We develop an ontology of six sentence - level functional roles for long - form answers , and annotate 3.9k sentences in 640 answer paragraphs .", "forward": true, "src_ids": "2022.acl-long.249_1119"}
+{"input": "classifier is done by using Material| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "classifier", "output": "annotated data", "neg_sample": ["classifier is done by using Material", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "Our annotated data enables training a strong classifier that can be used for automatic analysis .", "forward": false, "src_ids": "2022.acl-long.249_1120"}
+{"input": "automatic analysis is done by using Method| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "automatic analysis", "output": "classifier", "neg_sample": ["automatic analysis is done by using Method", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "Our annotated data enables training a strong classifier that can be used for automatic analysis .", "forward": false, "src_ids": "2022.acl-long.249_1121"}
+{"input": "annotated data is used for Method| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "annotated data", "output": "classifier", "neg_sample": ["annotated data is used for Method", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "Our annotated data enables training a strong classifier that can be used for automatic analysis .", "forward": true, "src_ids": "2022.acl-long.249_1122"}
+{"input": "classifier is used for Task| context: long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions .", "entity": "classifier", "output": "automatic analysis", "neg_sample": ["classifier is used for Task", "long - form answers , consisting of multiple sentences , can provide nuanced and comprehensive answers to a broader set of questions ."], "relation": "used for", "id": "2022.acl-long.249", "year": 2022, "rel_sent": "Our annotated data enables training a strong classifier that can be used for automatic analysis .", "forward": true, "src_ids": "2022.acl-long.249_1123"}
+{"input": "scalable dictionary - based method is used for Task| context: this paper presents a new method for automatic detection of gendered terms in large - scale language datasets . currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender . however , manual compilation of lists with lexical gender can lead to static information if lists are not periodically updated and often involve value judgements by individual annotators and researchers .", "entity": "scalable dictionary - based method", "output": "dynamic , up - to - date analysis", "neg_sample": ["scalable dictionary - based method is used for Task", "this paper presents a new method for automatic detection of gendered terms in large - scale language datasets .", "currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender .", "however , manual compilation of lists with lexical gender can lead to static information if lists are not periodically updated and often involve value judgements by individual annotators and researchers ."], "relation": "used for", "id": "2022.ltedi-1.7", "year": 2022, "rel_sent": "Moreover , terms not included in the lexicons fall out of the range of analysis . To address these issues , we devised a scalable dictionary - based method to automatically detect lexical gender that can provide a dynamic , up - to - date analysis with high coverage .", "forward": true, "src_ids": "2022.ltedi-1.7_1124"}
+{"input": "scalable dictionary - based method is used for OtherScientificTerm| context: this paper presents a new method for automatic detection of gendered terms in large - scale language datasets . currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender .", "entity": "scalable dictionary - based method", "output": "lexical gender", "neg_sample": ["scalable dictionary - based method is used for OtherScientificTerm", "this paper presents a new method for automatic detection of gendered terms in large - scale language datasets .", "currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender ."], "relation": "used for", "id": "2022.ltedi-1.7", "year": 2022, "rel_sent": "Moreover , terms not included in the lexicons fall out of the range of analysis . To address these issues , we devised a scalable dictionary - based method to automatically detect lexical gender that can provide a dynamic , up - to - date analysis with high coverage .", "forward": true, "src_ids": "2022.ltedi-1.7_1125"}
+{"input": "lexical gender is done by using Method| context: this paper presents a new method for automatic detection of gendered terms in large - scale language datasets . currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender . however , manual compilation of lists with lexical gender can lead to static information if lists are not periodically updated and often involve value judgements by individual annotators and researchers .", "entity": "lexical gender", "output": "scalable dictionary - based method", "neg_sample": ["lexical gender is done by using Method", "this paper presents a new method for automatic detection of gendered terms in large - scale language datasets .", "currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender .", "however , manual compilation of lists with lexical gender can lead to static information if lists are not periodically updated and often involve value judgements by individual annotators and researchers ."], "relation": "used for", "id": "2022.ltedi-1.7", "year": 2022, "rel_sent": "Moreover , terms not included in the lexicons fall out of the range of analysis . To address these issues , we devised a scalable dictionary - based method to automatically detect lexical gender that can provide a dynamic , up - to - date analysis with high coverage .", "forward": false, "src_ids": "2022.ltedi-1.7_1126"}
+{"input": "dynamic , up - to - date analysis is done by using Method| context: this paper presents a new method for automatic detection of gendered terms in large - scale language datasets . currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender . however , manual compilation of lists with lexical gender can lead to static information if lists are not periodically updated and often involve value judgements by individual annotators and researchers .", "entity": "dynamic , up - to - date analysis", "output": "scalable dictionary - based method", "neg_sample": ["dynamic , up - to - date analysis is done by using Method", "this paper presents a new method for automatic detection of gendered terms in large - scale language datasets .", "currently , the evaluation of gender bias in natural language processing relies on the use of manually compiled lexicons of gendered expressions , such as pronouns and words that imply gender .", "however , manual compilation of lists with lexical gender can lead to static information if lists are not periodically updated and often involve value judgements by individual annotators and researchers ."], "relation": "used for", "id": "2022.ltedi-1.7", "year": 2022, "rel_sent": "Moreover , terms not included in the lexicons fall out of the range of analysis . To address these issues , we devised a scalable dictionary - based method to automatically detect lexical gender that can provide a dynamic , up - to - date analysis with high coverage .", "forward": false, "src_ids": "2022.ltedi-1.7_1127"}
+{"input": "polysemy of korean adverbial postpositions is done by using Method| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "polysemy of korean adverbial postpositions", "output": "transformer - architecture models", "neg_sample": ["polysemy of korean adverbial postpositions is done by using Method", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "How Do Transformer - Architecture Models Address Polysemy of Korean Adverbial Postpositions ?.", "forward": false, "src_ids": "2022.deelio-1.2_1128"}
+{"input": "human interpretation of word - level polysemy is done by using Method| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "human interpretation of word - level polysemy", "output": "transformer - architecture models", "neg_sample": ["human interpretation of word - level polysemy is done by using Method", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "To add more interpretation , for this reason , we devised a classification model by employing two transformer - architecture models - BERT and GPT-2 - and introduces a computational simulation that interactively demonstrates how these transformer - architecture models simulate human interpretation of word - level polysemy involving Korean adverbial postpositions -ey , -eyse , and -(u)lo .", "forward": false, "src_ids": "2022.deelio-1.2_1129"}
+{"input": "transformer - architecture models is used for OtherScientificTerm| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "transformer - architecture models", "output": "polysemy of korean adverbial postpositions", "neg_sample": ["transformer - architecture models is used for OtherScientificTerm", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "How Do Transformer - Architecture Models Address Polysemy of Korean Adverbial Postpositions ?.", "forward": true, "src_ids": "2022.deelio-1.2_1130"}
+{"input": "transformer - architecture models is used for Task| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "transformer - architecture models", "output": "human interpretation of word - level polysemy", "neg_sample": ["transformer - architecture models is used for Task", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "To add more interpretation , for this reason , we devised a classification model by employing two transformer - architecture models - BERT and GPT-2 - and introduces a computational simulation that interactively demonstrates how these transformer - architecture models simulate human interpretation of word - level polysemy involving Korean adverbial postpositions -ey , -eyse , and -(u)lo .", "forward": true, "src_ids": "2022.deelio-1.2_1131"}
+{"input": "function of postpositions is done by using Method| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "function of postpositions", "output": "bert model", "neg_sample": ["function of postpositions is done by using Method", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "Results reveal that ( i ) the BERT model performs better than the GPT-2 model to classify the intended function of postpositions , ( ii ) there is an inverse relationship between the classification accuracy and the number of functions that each postposition manifests , ( iii ) model performance is affected by the corpus size of each function , ( iv ) the models ' performance gradually improves as the epoch proceeds , and ( vi ) the models are affected by the scarcity of input and/or semantic closeness between the items .", "forward": false, "src_ids": "2022.deelio-1.2_1132"}
+{"input": "gpt-2 model is used for OtherScientificTerm| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "gpt-2 model", "output": "function of postpositions", "neg_sample": ["gpt-2 model is used for OtherScientificTerm", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "Results reveal that ( i ) the BERT model performs better than the GPT-2 model to classify the intended function of postpositions , ( ii ) there is an inverse relationship between the classification accuracy and the number of functions that each postposition manifests , ( iii ) model performance is affected by the corpus size of each function , ( iv ) the models ' performance gradually improves as the epoch proceeds , and ( vi ) the models are affected by the scarcity of input and/or semantic closeness between the items .", "forward": true, "src_ids": "2022.deelio-1.2_1133"}
+{"input": "bert model is used for OtherScientificTerm| context: postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage . several studies have used contextualized word - embedding models to reveal the functions of korean postpositions . despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear .", "entity": "bert model", "output": "function of postpositions", "neg_sample": ["bert model is used for OtherScientificTerm", "postpositions , which are characterized as multiple form - function associations and thus polysemous , pose a challenge to automatic identification of their usage .", "several studies have used contextualized word - embedding models to reveal the functions of korean postpositions .", "despite the superior classification performance of previous studies , the particular reason how these models resolve the polysemy of korean postpositions is not enough clear ."], "relation": "used for", "id": "2022.deelio-1.2", "year": 2022, "rel_sent": "Results reveal that ( i ) the BERT model performs better than the GPT-2 model to classify the intended function of postpositions , ( ii ) there is an inverse relationship between the classification accuracy and the number of functions that each postposition manifests , ( iii ) model performance is affected by the corpus size of each function , ( iv ) the models ' performance gradually improves as the epoch proceeds , and ( vi ) the models are affected by the scarcity of input and/or semantic closeness between the items .", "forward": true, "src_ids": "2022.deelio-1.2_1134"}
+{"input": "language models is used for Task| context: we introduce a method for improving the structural understanding abilities of language models .", "entity": "language models", "output": "structure prediction", "neg_sample": ["language models is used for Task", "we introduce a method for improving the structural understanding abilities of language models ."], "relation": "used for", "id": "2022.findings-acl.67", "year": 2022, "rel_sent": "DeepStruct : Pretraining of Language Models for Structure Prediction.", "forward": true, "src_ids": "2022.findings-acl.67_1135"}
+{"input": "long context neural machine translation is done by using Method| context: after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation . a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers . however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length .", "entity": "long context neural machine translation", "output": "locality - sensitive hashing approach", "neg_sample": ["long context neural machine translation is done by using Method", "after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation .", "a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers .", "however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length ."], "relation": "used for", "id": "2022.iwslt-1.4", "year": 2022, "rel_sent": "Locality - Sensitive Hashing for Long Context Neural Machine Translation.", "forward": false, "src_ids": "2022.iwslt-1.4_1136"}
+{"input": "locality - sensitive hashing approach is used for Task| context: after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation . a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers . however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length .", "entity": "locality - sensitive hashing approach", "output": "long context neural machine translation", "neg_sample": ["locality - sensitive hashing approach is used for Task", "after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation .", "a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers .", "however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length ."], "relation": "used for", "id": "2022.iwslt-1.4", "year": 2022, "rel_sent": "Locality - Sensitive Hashing for Long Context Neural Machine Translation.", "forward": true, "src_ids": "2022.iwslt-1.4_1137"}
+{"input": "sentence- and document - level machine translation is done by using Method| context: after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation . a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers . however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length .", "entity": "sentence- and document - level machine translation", "output": "memory efficient framework", "neg_sample": ["sentence- and document - level machine translation is done by using Method", "after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation .", "a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers .", "however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length ."], "relation": "used for", "id": "2022.iwslt-1.4", "year": 2022, "rel_sent": "( 2020 ) to self - attention in the Transformer , we extended it to cross - attention and apply this memory efficient framework to sentence- and document - level machine translation .", "forward": false, "src_ids": "2022.iwslt-1.4_1138"}
+{"input": "memory efficient framework is used for Task| context: after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation . a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers . however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length .", "entity": "memory efficient framework", "output": "sentence- and document - level machine translation", "neg_sample": ["memory efficient framework is used for Task", "after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation .", "a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers .", "however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length ."], "relation": "used for", "id": "2022.iwslt-1.4", "year": 2022, "rel_sent": "( 2020 ) to self - attention in the Transformer , we extended it to cross - attention and apply this memory efficient framework to sentence- and document - level machine translation .", "forward": true, "src_ids": "2022.iwslt-1.4_1139"}
+{"input": "sentence - level is done by using Method| context: after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation . a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers . however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length .", "entity": "sentence - level", "output": "lsh attention scheme", "neg_sample": ["sentence - level is done by using Method", "after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation .", "a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers .", "however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length ."], "relation": "used for", "id": "2022.iwslt-1.4", "year": 2022, "rel_sent": "Our experiments show that the LSH attention scheme for sentence - level comes at the cost of slightly reduced translation quality .", "forward": false, "src_ids": "2022.iwslt-1.4_1140"}
+{"input": "lsh attention scheme is used for OtherScientificTerm| context: after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation . a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers . however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length .", "entity": "lsh attention scheme", "output": "sentence - level", "neg_sample": ["lsh attention scheme is used for OtherScientificTerm", "after its introduction the transformer architecture quickly became the gold standard for the task of neural machine translation .", "a major advantage of the transformer compared to previous architectures is the faster training speed achieved by complete parallelization across timesteps due to the use of attention over recurrent layers .", "however , this also leads to one of the biggest problems of the transformer , namely the quadratic time and memory complexity with respect to the input length ."], "relation": "used for", "id": "2022.iwslt-1.4", "year": 2022, "rel_sent": "Our experiments show that the LSH attention scheme for sentence - level comes at the cost of slightly reduced translation quality .", "forward": true, "src_ids": "2022.iwslt-1.4_1141"}
+{"input": "multimodality is done by using Method| context: indirect speech such as sarcasm achieves a constellation of discourse goals in human communication . while the indirectness of figurative language warrants speakers to achieve certain pragmatic goals , it is challenging for ai agents to comprehend such idiosyncrasies of human communication . though sarcasm identification has been a well - explored topic in dialogue analysis , for conversational systems to truly grasp a conversation 's innate meaning and generate appropriate responses , simply detecting sarcasm is not enough ; it is vital to explain its underlying sarcastic connotation to capture its true essence .", "entity": "multimodality", "output": "multimodal context - aware attention", "neg_sample": ["multimodality is done by using Method", "indirect speech such as sarcasm achieves a constellation of discourse goals in human communication .", "while the indirectness of figurative language warrants speakers to achieve certain pragmatic goals , it is challenging for ai agents to comprehend such idiosyncrasies of human communication .", "though sarcasm identification has been a well - explored topic in dialogue analysis , for conversational systems to truly grasp a conversation 's innate meaning and generate appropriate responses , simply detecting sarcasm is not enough ; it is vital to explain its underlying sarcastic connotation to capture its true essence ."], "relation": "used for", "id": "2022.acl-long.411", "year": 2022, "rel_sent": "We propose MAF ( Modality Aware Fusion ) , a multimodal context - aware attention and global information fusion module to capture multimodality and use it to benchmark WITS .", "forward": false, "src_ids": "2022.acl-long.411_1142"}
+{"input": "multimodal context - aware attention is used for OtherScientificTerm| context: indirect speech such as sarcasm achieves a constellation of discourse goals in human communication . while the indirectness of figurative language warrants speakers to achieve certain pragmatic goals , it is challenging for ai agents to comprehend such idiosyncrasies of human communication . though sarcasm identification has been a well - explored topic in dialogue analysis , for conversational systems to truly grasp a conversation 's innate meaning and generate appropriate responses , simply detecting sarcasm is not enough ; it is vital to explain its underlying sarcastic connotation to capture its true essence .", "entity": "multimodal context - aware attention", "output": "multimodality", "neg_sample": ["multimodal context - aware attention is used for OtherScientificTerm", "indirect speech such as sarcasm achieves a constellation of discourse goals in human communication .", "while the indirectness of figurative language warrants speakers to achieve certain pragmatic goals , it is challenging for ai agents to comprehend such idiosyncrasies of human communication .", "though sarcasm identification has been a well - explored topic in dialogue analysis , for conversational systems to truly grasp a conversation 's innate meaning and generate appropriate responses , simply detecting sarcasm is not enough ; it is vital to explain its underlying sarcastic connotation to capture its true essence ."], "relation": "used for", "id": "2022.acl-long.411", "year": 2022, "rel_sent": "We propose MAF ( Modality Aware Fusion ) , a multimodal context - aware attention and global information fusion module to capture multimodality and use it to benchmark WITS .", "forward": true, "src_ids": "2022.acl-long.411_1143"}
+{"input": "textual outputs is done by using Method| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "textual outputs", "output": "tailor", "neg_sample": ["textual outputs is done by using Method", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "Tailor builds on a pretrained seq2seq model and produces textual outputs conditioned on control codes derived from semantic representations .", "forward": false, "src_ids": "2022.acl-long.228_1144"}
+{"input": "contrast sets is done by using Method| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "contrast sets", "output": "tailor", "neg_sample": ["contrast sets is done by using Method", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "First , we use Tailor to automatically create high - quality contrast sets for four distinct natural language processing ( NLP ) tasks .", "forward": false, "src_ids": "2022.acl-long.228_1145"}
+{"input": "natural language processing ( nlp ) tasks is done by using Method| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "natural language processing ( nlp ) tasks", "output": "tailor", "neg_sample": ["natural language processing ( nlp ) tasks is done by using Method", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "First , we use Tailor to automatically create high - quality contrast sets for four distinct natural language processing ( NLP ) tasks .", "forward": false, "src_ids": "2022.acl-long.228_1146"}
+{"input": "tailor is used for OtherScientificTerm| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "tailor", "output": "textual outputs", "neg_sample": ["tailor is used for OtherScientificTerm", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "Tailor builds on a pretrained seq2seq model and produces textual outputs conditioned on control codes derived from semantic representations .", "forward": true, "src_ids": "2022.acl-long.228_1147"}
+{"input": "natural language processing ( nlp ) tasks is done by using Material| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "natural language processing ( nlp ) tasks", "output": "contrast sets", "neg_sample": ["natural language processing ( nlp ) tasks is done by using Material", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "First , we use Tailor to automatically create high - quality contrast sets for four distinct natural language processing ( NLP ) tasks .", "forward": false, "src_ids": "2022.acl-long.228_1148"}
+{"input": "tailor is used for Material| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "tailor", "output": "contrast sets", "neg_sample": ["tailor is used for Material", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "First , we use Tailor to automatically create high - quality contrast sets for four distinct natural language processing ( NLP ) tasks .", "forward": true, "src_ids": "2022.acl-long.228_1149"}
+{"input": "contrast sets is used for Task| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "contrast sets", "output": "natural language processing ( nlp ) tasks", "neg_sample": ["contrast sets is used for Task", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "First , we use Tailor to automatically create high - quality contrast sets for four distinct natural language processing ( NLP ) tasks .", "forward": true, "src_ids": "2022.acl-long.228_1150"}
+{"input": "tailor is used for Task| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "tailor", "output": "natural language processing ( nlp ) tasks", "neg_sample": ["tailor is used for Task", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "First , we use Tailor to automatically create high - quality contrast sets for four distinct natural language processing ( NLP ) tasks .", "forward": true, "src_ids": "2022.acl-long.228_1151"}
+{"input": "model generalization is done by using Method| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "model generalization", "output": "tailor perturbations", "neg_sample": ["model generalization is done by using Method", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "Second , we show that Tailor perturbations can improve model generalization through data augmentation .", "forward": false, "src_ids": "2022.acl-long.228_1152"}
+{"input": "tailor perturbations is used for Task| context: controlled text perturbation is useful for evaluating and improving model generalizability . however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize .", "entity": "tailor perturbations", "output": "model generalization", "neg_sample": ["tailor perturbations is used for Task", "controlled text perturbation is useful for evaluating and improving model generalizability .", "however , current techniques rely on training a model for every target perturbation , which is expensive and hard to generalize ."], "relation": "used for", "id": "2022.acl-long.228", "year": 2022, "rel_sent": "Second , we show that Tailor perturbations can improve model generalization through data augmentation .", "forward": true, "src_ids": "2022.acl-long.228_1153"}
+{"input": "automatic speech recognition is done by using Method| context: the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family .", "entity": "automatic speech recognition", "output": "fine - tuning pre - trained models", "neg_sample": ["automatic speech recognition is done by using Method", "the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family ."], "relation": "used for", "id": "2022.computel-1.21", "year": 2022, "rel_sent": "Fine - tuning pre - trained models for Automatic Speech Recognition , experiments on a fieldwork corpus of Japhug ( Trans - Himalayan family ).", "forward": false, "src_ids": "2022.computel-1.21_1154"}
+{"input": "fine - tuning pre - trained models is used for Task| context: the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family .", "entity": "fine - tuning pre - trained models", "output": "automatic speech recognition", "neg_sample": ["fine - tuning pre - trained models is used for Task", "the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family ."], "relation": "used for", "id": "2022.computel-1.21", "year": 2022, "rel_sent": "Fine - tuning pre - trained models for Automatic Speech Recognition , experiments on a fieldwork corpus of Japhug ( Trans - Himalayan family ).", "forward": true, "src_ids": "2022.computel-1.21_1155"}
+{"input": "pre - trained representation model is done by using Method| context: the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family .", "entity": "pre - trained representation model", "output": "language - specific tuning", "neg_sample": ["pre - trained representation model is done by using Method", "the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family ."], "relation": "used for", "id": "2022.computel-1.21", "year": 2022, "rel_sent": "The method used is a deep learning approach based on the language - specific tuning of a generic pre - trained representation model , XLS - R , using a Transformer architecture .", "forward": false, "src_ids": "2022.computel-1.21_1156"}
+{"input": "language - specific tuning is used for Method| context: the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family .", "entity": "language - specific tuning", "output": "pre - trained representation model", "neg_sample": ["language - specific tuning is used for Method", "the test case is an extensive fieldwork corpus of japhug , an endangered language of the trans - himalayan ( sino - tibetan ) family ."], "relation": "used for", "id": "2022.computel-1.21", "year": 2022, "rel_sent": "The method used is a deep learning approach based on the language - specific tuning of a generic pre - trained representation model , XLS - R , using a Transformer architecture .", "forward": true, "src_ids": "2022.computel-1.21_1157"}
+{"input": "evaluation metrics is used for OtherScientificTerm| context: model - based , reference - free evaluation metricshave been proposed as a fast and cost - effectiveapproach to evaluate natural language generation(nlg ) systems . despite promising recentresults , we find evidence that reference - freeevaluation metrics of summarization and dialoggeneration may be relying on spuriouscorrelations with measures such as word overlap , perplexity , and length .", "entity": "evaluation metrics", "output": "spurious features", "neg_sample": ["evaluation metrics is used for OtherScientificTerm", "model - based , reference - free evaluation metricshave been proposed as a fast and cost - effectiveapproach to evaluate natural language generation(nlg ) systems .", "despite promising recentresults , we find evidence that reference - freeevaluation metrics of summarization and dialoggeneration may be relying on spuriouscorrelations with measures such as word overlap , perplexity , and length ."], "relation": "used for", "id": "2022.acl-long.102", "year": 2022, "rel_sent": "Wedemonstrate that these errors can be mitigatedby explicitly designing evaluation metrics toavoid spurious features in reference - free evaluation .", "forward": true, "src_ids": "2022.acl-long.102_1158"}
+{"input": "evaluation metrics is used for Task| context: despite promising recentresults , we find evidence that reference - freeevaluation metrics of summarization and dialoggeneration may be relying on spuriouscorrelations with measures such as word overlap , perplexity , and length .", "entity": "evaluation metrics", "output": "reference - free evaluation", "neg_sample": ["evaluation metrics is used for Task", "despite promising recentresults , we find evidence that reference - freeevaluation metrics of summarization and dialoggeneration may be relying on spuriouscorrelations with measures such as word overlap , perplexity , and length ."], "relation": "used for", "id": "2022.acl-long.102", "year": 2022, "rel_sent": "Wedemonstrate that these errors can be mitigatedby explicitly designing evaluation metrics toavoid spurious features in reference - free evaluation .", "forward": true, "src_ids": "2022.acl-long.102_1159"}
+{"input": "pretrained language models is used for Method| context: thanks to the effectiveness and wide availability of modern pretrained language models ( plms ) , recently proposed approaches have achieved remarkable results in dependency- and span - based , multilingual and cross - lingual semantic role labeling ( srl ) .", "entity": "pretrained language models", "output": "contextualized representation of a predicate", "neg_sample": ["pretrained language models is used for Method", "thanks to the effectiveness and wide availability of modern pretrained language models ( plms ) , recently proposed approaches have achieved remarkable results in dependency- and span - based , multilingual and cross - lingual semantic role labeling ( srl ) ."], "relation": "used for", "id": "2022.acl-long.316", "year": 2022, "rel_sent": "Our study shows that PLMs do encode semantic structures directly into the contextualized representation of a predicate , and also provides insights into the correlation between predicate senses and their structures , the degree of transferability between nominal and verbal structures , and how such structures are encoded across languages .", "forward": true, "src_ids": "2022.acl-long.316_1160"}
+{"input": "few - shot named entity recognition is done by using Method| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "few - shot named entity recognition", "output": "decomposed meta - learning approach", "neg_sample": ["few - shot named entity recognition is done by using Method", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.124", "year": 2022, "rel_sent": "Decomposed Meta - Learning for Few - Shot Named Entity Recognition.", "forward": false, "src_ids": "2022.findings-acl.124_1161"}
+{"input": "few - shot ner is done by using Method| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "few - shot ner", "output": "decomposed meta - learning approach", "neg_sample": ["few - shot ner is done by using Method", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.124", "year": 2022, "rel_sent": "In this paper , we present a decomposed meta - learning approach which addresses the problem of few - shot NER by sequentially tackling few - shot span detection and few - shot entity typing using meta - learning .", "forward": false, "src_ids": "2022.findings-acl.124_1162"}
+{"input": "decomposed meta - learning approach is used for Task| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "decomposed meta - learning approach", "output": "few - shot ner", "neg_sample": ["decomposed meta - learning approach is used for Task", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.124", "year": 2022, "rel_sent": "In this paper , we present a decomposed meta - learning approach which addresses the problem of few - shot NER by sequentially tackling few - shot span detection and few - shot entity typing using meta - learning .", "forward": true, "src_ids": "2022.findings-acl.124_1163"}
+{"input": "few - shot learning is done by using Method| context: under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much .", "entity": "few - shot learning", "output": "data augmentation", "neg_sample": ["few - shot learning is done by using Method", "under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much ."], "relation": "used for", "id": "2022.acl-long.592", "year": 2022, "rel_sent": "FlipDA : Effective and Robust Data Augmentation for Few - Shot Learning.", "forward": false, "src_ids": "2022.acl-long.592_1164"}
+{"input": "data augmentation is used for Method| context: most previous methods for text data augmentation are limited to simple tasks and weak baselines . we explore data augmentation on hard tasks ( i.e. , few - shot natural language understanding ) and strong baselines ( i.e. , pretrained models with over one billion parameters ) . under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much .", "entity": "data augmentation", "output": "few - shot learning", "neg_sample": ["data augmentation is used for Method", "most previous methods for text data augmentation are limited to simple tasks and weak baselines .", "we explore data augmentation on hard tasks ( i.e.", ", few - shot natural language understanding ) and strong baselines ( i.e.", ", pretrained models with over one billion parameters ) .", "under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much ."], "relation": "used for", "id": "2022.acl-long.592", "year": 2022, "rel_sent": "FlipDA : Effective and Robust Data Augmentation for Few - Shot Learning.", "forward": true, "src_ids": "2022.acl-long.592_1165"}
+{"input": "label - flipped data is done by using Method| context: most previous methods for text data augmentation are limited to simple tasks and weak baselines . we explore data augmentation on hard tasks ( i.e. , few - shot natural language understanding ) and strong baselines ( i.e. , pretrained models with over one billion parameters ) . under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much .", "entity": "label - flipped data", "output": "generative model", "neg_sample": ["label - flipped data is done by using Method", "most previous methods for text data augmentation are limited to simple tasks and weak baselines .", "we explore data augmentation on hard tasks ( i.e.", ", few - shot natural language understanding ) and strong baselines ( i.e.", ", pretrained models with over one billion parameters ) .", "under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much ."], "relation": "used for", "id": "2022.acl-long.592", "year": 2022, "rel_sent": "To address this challenge , we propose a novel data augmentation method FlipDA that jointly uses a generative model and a classifier to generate label - flipped data .", "forward": false, "src_ids": "2022.acl-long.592_1166"}
+{"input": "classifier is used for Material| context: most previous methods for text data augmentation are limited to simple tasks and weak baselines . we explore data augmentation on hard tasks ( i.e. , few - shot natural language understanding ) and strong baselines ( i.e. , pretrained models with over one billion parameters ) . under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much .", "entity": "classifier", "output": "label - flipped data", "neg_sample": ["classifier is used for Material", "most previous methods for text data augmentation are limited to simple tasks and weak baselines .", "we explore data augmentation on hard tasks ( i.e.", ", few - shot natural language understanding ) and strong baselines ( i.e.", ", pretrained models with over one billion parameters ) .", "under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much ."], "relation": "used for", "id": "2022.acl-long.592", "year": 2022, "rel_sent": "To address this challenge , we propose a novel data augmentation method FlipDA that jointly uses a generative model and a classifier to generate label - flipped data .", "forward": true, "src_ids": "2022.acl-long.592_1167"}
+{"input": "generative model is used for Material| context: most previous methods for text data augmentation are limited to simple tasks and weak baselines . we explore data augmentation on hard tasks ( i.e. , few - shot natural language understanding ) and strong baselines ( i.e. , pretrained models with over one billion parameters ) . under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much .", "entity": "generative model", "output": "label - flipped data", "neg_sample": ["generative model is used for Material", "most previous methods for text data augmentation are limited to simple tasks and weak baselines .", "we explore data augmentation on hard tasks ( i.e.", ", few - shot natural language understanding ) and strong baselines ( i.e.", ", pretrained models with over one billion parameters ) .", "under this setting , we reproduced a large number of previous augmentation methods and found that these methods bring marginal gains at best and sometimes degrade the performance much ."], "relation": "used for", "id": "2022.acl-long.592", "year": 2022, "rel_sent": "To address this challenge , we propose a novel data augmentation method FlipDA that jointly uses a generative model and a classifier to generate label - flipped data .", "forward": true, "src_ids": "2022.acl-long.592_1168"}
+{"input": "encoders is done by using Method| context: these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "encoders", "output": "multilingual training", "neg_sample": ["encoders is done by using Method", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "We show that multilingual training is beneficial to encoders in general , while it only benefits decoders for low - resource languages ( LRLs ) .", "forward": false, "src_ids": "2022.findings-acl.218_1169"}
+{"input": "multilingual training is used for Method| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "multilingual training", "output": "encoders", "neg_sample": ["multilingual training is used for Method", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "We show that multilingual training is beneficial to encoders in general , while it only benefits decoders for low - resource languages ( LRLs ) .", "forward": true, "src_ids": "2022.findings-acl.218_1170"}
+{"input": "low - resource languages ( lrls ) is done by using Method| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "low - resource languages ( lrls )", "output": "decoders", "neg_sample": ["low - resource languages ( lrls ) is done by using Method", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "We show that multilingual training is beneficial to encoders in general , while it only benefits decoders for low - resource languages ( LRLs ) .", "forward": false, "src_ids": "2022.findings-acl.218_1171"}
+{"input": "decoders is used for Material| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "decoders", "output": "low - resource languages ( lrls )", "neg_sample": ["decoders is used for Material", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "We show that multilingual training is beneficial to encoders in general , while it only benefits decoders for low - resource languages ( LRLs ) .", "forward": true, "src_ids": "2022.findings-acl.218_1172"}
+{"input": "high - resource languages is done by using Method| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "high - resource languages", "output": "many - to - one models", "neg_sample": ["high - resource languages is done by using Method", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "Our many - to - one models for high - resource languages and one - to - many models for LRL outperform the best results reported by Aharoni et al .", "forward": false, "src_ids": "2022.findings-acl.218_1173"}
+{"input": "lrl is done by using Method| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "lrl", "output": "many - to - one models", "neg_sample": ["lrl is done by using Method", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "Our many - to - one models for high - resource languages and one - to - many models for LRL outperform the best results reported by Aharoni et al .", "forward": false, "src_ids": "2022.findings-acl.218_1174"}
+{"input": "many - to - one models is used for Material| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "many - to - one models", "output": "high - resource languages", "neg_sample": ["many - to - one models is used for Material", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "Our many - to - one models for high - resource languages and one - to - many models for LRL outperform the best results reported by Aharoni et al .", "forward": true, "src_ids": "2022.findings-acl.218_1175"}
+{"input": "lrl is done by using Generic| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "lrl", "output": "one - to - many models", "neg_sample": ["lrl is done by using Generic", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "Our many - to - one models for high - resource languages and one - to - many models for LRL outperform the best results reported by Aharoni et al .", "forward": false, "src_ids": "2022.findings-acl.218_1176"}
+{"input": "one - to - many models is used for OtherScientificTerm| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "one - to - many models", "output": "lrl", "neg_sample": ["one - to - many models is used for OtherScientificTerm", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "Our many - to - one models for high - resource languages and one - to - many models for LRL outperform the best results reported by Aharoni et al .", "forward": true, "src_ids": "2022.findings-acl.218_1177"}
+{"input": "many - to - one models is used for OtherScientificTerm| context: while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning . these training settings expose the encoder and the decoder in a machine translation model with different data distributions .", "entity": "many - to - one models", "output": "lrl", "neg_sample": ["many - to - one models is used for OtherScientificTerm", "while multilingual training is now an essential ingredient in machine translation ( mt ) systems , recent work has demonstrated that it has different effects in different multilingual settings , such as many - to - one , one - to - many , and many - to - many learning .", "these training settings expose the encoder and the decoder in a machine translation model with different data distributions ."], "relation": "used for", "id": "2022.findings-acl.218", "year": 2022, "rel_sent": "Our many - to - one models for high - resource languages and one - to - many models for LRL outperform the best results reported by Aharoni et al .", "forward": true, "src_ids": "2022.findings-acl.218_1178"}
+{"input": "length - aware framework is used for OtherScientificTerm| context: simultaneous machine translation ( simt ) starts translating while receiving the streaming source inputs , and hence the source sentence is always incomplete during translating . different from the full - sentence mt using the conventional seq - to - seq architecture , simt often applies prefix - to - prefix architecture , which forces each target word to only align with a partial source prefix to adapt to the incomplete source in streaming inputs .", "entity": "length - aware framework", "output": "position bias", "neg_sample": ["length - aware framework is used for OtherScientificTerm", "simultaneous machine translation ( simt ) starts translating while receiving the streaming source inputs , and hence the source sentence is always incomplete during translating .", "different from the full - sentence mt using the conventional seq - to - seq architecture , simt often applies prefix - to - prefix architecture , which forces each target word to only align with a partial source prefix to adapt to the incomplete source in streaming inputs ."], "relation": "used for", "id": "2022.acl-long.467", "year": 2022, "rel_sent": "In this paper , we first analyze the phenomenon of position bias in SiMT , and develop a Length - Aware Framework to reduce the position bias by bridging the structural gap between SiMT and full - sentence MT .", "forward": true, "src_ids": "2022.acl-long.467_1179"}
+{"input": "simultaneous machine translation is used for OtherScientificTerm| context: simultaneous machine translation ( simt ) starts translating while receiving the streaming source inputs , and hence the source sentence is always incomplete during translating . different from the full - sentence mt using the conventional seq - to - seq architecture , simt often applies prefix - to - prefix architecture , which forces each target word to only align with a partial source prefix to adapt to the incomplete source in streaming inputs .", "entity": "simultaneous machine translation", "output": "position bias", "neg_sample": ["simultaneous machine translation is used for OtherScientificTerm", "simultaneous machine translation ( simt ) starts translating while receiving the streaming source inputs , and hence the source sentence is always incomplete during translating .", "different from the full - sentence mt using the conventional seq - to - seq architecture , simt often applies prefix - to - prefix architecture , which forces each target word to only align with a partial source prefix to adapt to the incomplete source in streaming inputs ."], "relation": "used for", "id": "2022.acl-long.467", "year": 2022, "rel_sent": "Experiments on two representative SiMT methods , including the state - of - the - art adaptive policy , show that our method successfully reduces the position bias and thereby achieves better SiMT performance .", "forward": true, "src_ids": "2022.acl-long.467_1180"}
+{"input": "position bias is done by using Task| context: however , the source words in the front positions are always illusoryly considered more important since they appear in more prefixes , resulting in position bias , which makes the model pay more attention on the front source positions in testing .", "entity": "position bias", "output": "simultaneous machine translation", "neg_sample": ["position bias is done by using Task", "however , the source words in the front positions are always illusoryly considered more important since they appear in more prefixes , resulting in position bias , which makes the model pay more attention on the front source positions in testing ."], "relation": "used for", "id": "2022.acl-long.467", "year": 2022, "rel_sent": "Experiments on two representative SiMT methods , including the state - of - the - art adaptive policy , show that our method successfully reduces the position bias and thereby achieves better SiMT performance .", "forward": false, "src_ids": "2022.acl-long.467_1181"}
+{"input": "position bias is done by using Method| context: simultaneous machine translation ( simt ) starts translating while receiving the streaming source inputs , and hence the source sentence is always incomplete during translating . different from the full - sentence mt using the conventional seq - to - seq architecture , simt often applies prefix - to - prefix architecture , which forces each target word to only align with a partial source prefix to adapt to the incomplete source in streaming inputs . however , the source words in the front positions are always illusoryly considered more important since they appear in more prefixes , resulting in position bias , which makes the model pay more attention on the front source positions in testing .", "entity": "position bias", "output": "length - aware framework", "neg_sample": ["position bias is done by using Method", "simultaneous machine translation ( simt ) starts translating while receiving the streaming source inputs , and hence the source sentence is always incomplete during translating .", "different from the full - sentence mt using the conventional seq - to - seq architecture , simt often applies prefix - to - prefix architecture , which forces each target word to only align with a partial source prefix to adapt to the incomplete source in streaming inputs .", "however , the source words in the front positions are always illusoryly considered more important since they appear in more prefixes , resulting in position bias , which makes the model pay more attention on the front source positions in testing ."], "relation": "used for", "id": "2022.acl-long.467", "year": 2022, "rel_sent": "In this paper , we first analyze the phenomenon of position bias in SiMT , and develop a Length - Aware Framework to reduce the position bias by bridging the structural gap between SiMT and full - sentence MT .", "forward": false, "src_ids": "2022.acl-long.467_1182"}
+{"input": "baseline classification system is done by using Method| context: increased use of online social media sites has given rise to tremendous amounts of user generated data . social media sites have become a platform where users express and voice their opinions in a real - time environment . social media sites such as twitter limit the number of characters used to express a thought in a tweet , leading to increased use of creative , humorous and confusing language in order to convey the message . due to this , automatic humor detection has become a difficult task , especially for low - resource languages such as the dravidian languages . humor detection has been a well studied area for resource rich languages due to the availability of rich and accurate data .", "entity": "baseline classification system", "output": "transformer models", "neg_sample": ["baseline classification system is done by using Method", "increased use of online social media sites has given rise to tremendous amounts of user generated data .", "social media sites have become a platform where users express and voice their opinions in a real - time environment .", "social media sites such as twitter limit the number of characters used to express a thought in a tweet , leading to increased use of creative , humorous and confusing language in order to convey the message .", "due to this , automatic humor detection has become a difficult task , especially for low - resource languages such as the dravidian languages .", "humor detection has been a well studied area for resource rich languages due to the availability of rich and accurate data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.2", "year": 2022, "rel_sent": "We experimented on the corpus using various transformer models such as Multilingual BERT , Multilingual DistillBERT and XLM - RoBERTa to establish a baseline classification system .", "forward": false, "src_ids": "2022.dravidianlangtech-1.2_1183"}
+{"input": "xlm - roberta is used for Method| context: increased use of online social media sites has given rise to tremendous amounts of user generated data . social media sites have become a platform where users express and voice their opinions in a real - time environment . social media sites such as twitter limit the number of characters used to express a thought in a tweet , leading to increased use of creative , humorous and confusing language in order to convey the message . due to this , automatic humor detection has become a difficult task , especially for low - resource languages such as the dravidian languages . humor detection has been a well studied area for resource rich languages due to the availability of rich and accurate data .", "entity": "xlm - roberta", "output": "baseline classification system", "neg_sample": ["xlm - roberta is used for Method", "increased use of online social media sites has given rise to tremendous amounts of user generated data .", "social media sites have become a platform where users express and voice their opinions in a real - time environment .", "social media sites such as twitter limit the number of characters used to express a thought in a tweet , leading to increased use of creative , humorous and confusing language in order to convey the message .", "due to this , automatic humor detection has become a difficult task , especially for low - resource languages such as the dravidian languages .", "humor detection has been a well studied area for resource rich languages due to the availability of rich and accurate data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.2", "year": 2022, "rel_sent": "We experimented on the corpus using various transformer models such as Multilingual BERT , Multilingual DistillBERT and XLM - RoBERTa to establish a baseline classification system .", "forward": true, "src_ids": "2022.dravidianlangtech-1.2_1184"}
+{"input": "transformer models is used for Method| context: increased use of online social media sites has given rise to tremendous amounts of user generated data . social media sites have become a platform where users express and voice their opinions in a real - time environment . social media sites such as twitter limit the number of characters used to express a thought in a tweet , leading to increased use of creative , humorous and confusing language in order to convey the message . due to this , automatic humor detection has become a difficult task , especially for low - resource languages such as the dravidian languages . humor detection has been a well studied area for resource rich languages due to the availability of rich and accurate data .", "entity": "transformer models", "output": "baseline classification system", "neg_sample": ["transformer models is used for Method", "increased use of online social media sites has given rise to tremendous amounts of user generated data .", "social media sites have become a platform where users express and voice their opinions in a real - time environment .", "social media sites such as twitter limit the number of characters used to express a thought in a tweet , leading to increased use of creative , humorous and confusing language in order to convey the message .", "due to this , automatic humor detection has become a difficult task , especially for low - resource languages such as the dravidian languages .", "humor detection has been a well studied area for resource rich languages due to the availability of rich and accurate data ."], "relation": "used for", "id": "2022.dravidianlangtech-1.2", "year": 2022, "rel_sent": "We experimented on the corpus using various transformer models such as Multilingual BERT , Multilingual DistillBERT and XLM - RoBERTa to establish a baseline classification system .", "forward": true, "src_ids": "2022.dravidianlangtech-1.2_1185"}
+{"input": "generalizable coherence modeling is done by using OtherScientificTerm| context: given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated . prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task .", "entity": "generalizable coherence modeling", "output": "self - supervision objectives", "neg_sample": ["generalizable coherence modeling is done by using OtherScientificTerm", "given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated .", "prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task ."], "relation": "used for", "id": "2022.acl-long.418", "year": 2022, "rel_sent": "Rethinking Self - Supervision Objectives for Generalizable Coherence Modeling.", "forward": false, "src_ids": "2022.acl-long.418_1186"}
+{"input": "self - supervision objectives is used for Task| context: given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated . prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task .", "entity": "self - supervision objectives", "output": "generalizable coherence modeling", "neg_sample": ["self - supervision objectives is used for Task", "given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated .", "prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task ."], "relation": "used for", "id": "2022.acl-long.418", "year": 2022, "rel_sent": "Rethinking Self - Supervision Objectives for Generalizable Coherence Modeling.", "forward": true, "src_ids": "2022.acl-long.418_1187"}
+{"input": "basic model is done by using OtherScientificTerm| context: given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated . prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task .", "entity": "basic model", "output": "density of negative samples", "neg_sample": ["basic model is done by using OtherScientificTerm", "given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated .", "prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task ."], "relation": "used for", "id": "2022.acl-long.418", "year": 2022, "rel_sent": "We show empirically that increasing the density of negative samples improves the basic model , and using a global negative queue further improves and stabilizes the model while training with hard negative samples .", "forward": false, "src_ids": "2022.acl-long.418_1188"}
+{"input": "density of negative samples is used for Generic| context: given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated . prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task .", "entity": "density of negative samples", "output": "basic model", "neg_sample": ["density of negative samples is used for Generic", "given the claims of improved text generation quality across various pre - trained neural models , we consider the coherence evaluation of machine generated text to be one of the principal applications of coherence models that needs to be investigated .", "prior work in neural coherence modeling has primarily focused on devising new architectures for solving the permuted document task ."], "relation": "used for", "id": "2022.acl-long.418", "year": 2022, "rel_sent": "We show empirically that increasing the density of negative samples improves the basic model , and using a global negative queue further improves and stabilizes the model while training with hard negative samples .", "forward": true, "src_ids": "2022.acl-long.418_1189"}
+{"input": "question answering is done by using Method| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "question answering", "output": "enhanced language representation", "neg_sample": ["question answering is done by using Method", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "ANNA ' : ' Enhanced Language Representation for Question Answering.", "forward": false, "src_ids": "2022.repl4nlp-1.13_1190"}
+{"input": "enhanced language representation is used for Task| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "enhanced language representation", "output": "question answering", "neg_sample": ["enhanced language representation is used for Task", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "ANNA ' : ' Enhanced Language Representation for Question Answering.", "forward": true, "src_ids": "2022.repl4nlp-1.13_1191"}
+{"input": "question answering task is done by using Method| context: most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "question answering task", "output": "language model", "neg_sample": ["question answering task is done by using Method", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "In this paper , we demonstrate how the approaches affect performance individually , and that the language model performs the best results on a specific question answering task when those approaches are jointly considered in pre - training models .", "forward": false, "src_ids": "2022.repl4nlp-1.13_1192"}
+{"input": "language model is used for Task| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "language model", "output": "question answering task", "neg_sample": ["language model is used for Task", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "In this paper , we demonstrate how the approaches affect performance individually , and that the language model performs the best results on a specific question answering task when those approaches are jointly considered in pre - training models .", "forward": true, "src_ids": "2022.repl4nlp-1.13_1193"}
+{"input": "pre - training language modeling is done by using Method| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "pre - training language modeling", "output": "neighbor - aware mechanism", "neg_sample": ["pre - training language modeling is done by using Method", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "In particular , we propose an extended pre - training task , and a new neighbor - aware mechanism that attends neighboring tokens more to capture the richness of context for pre - training language modeling .", "forward": false, "src_ids": "2022.repl4nlp-1.13_1194"}
+{"input": "richness of context is done by using Method| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "richness of context", "output": "neighbor - aware mechanism", "neg_sample": ["richness of context is done by using Method", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "In particular , we propose an extended pre - training task , and a new neighbor - aware mechanism that attends neighboring tokens more to capture the richness of context for pre - training language modeling .", "forward": false, "src_ids": "2022.repl4nlp-1.13_1195"}
+{"input": "neighbor - aware mechanism is used for OtherScientificTerm| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "neighbor - aware mechanism", "output": "richness of context", "neg_sample": ["neighbor - aware mechanism is used for OtherScientificTerm", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "In particular , we propose an extended pre - training task , and a new neighbor - aware mechanism that attends neighboring tokens more to capture the richness of context for pre - training language modeling .", "forward": true, "src_ids": "2022.repl4nlp-1.13_1196"}
+{"input": "neighbor - aware mechanism is used for Task| context: pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks . most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning .", "entity": "neighbor - aware mechanism", "output": "pre - training language modeling", "neg_sample": ["neighbor - aware mechanism is used for Task", "pre - trained language models have brought significant improvements in performance in a variety of natural language processing tasks .", "most existing models performing state - of - the - art results have shown their approaches in the separate perspectives of data processing , pre - training tasks , neural network modeling , or fine - tuning ."], "relation": "used for", "id": "2022.repl4nlp-1.13", "year": 2022, "rel_sent": "In particular , we propose an extended pre - training task , and a new neighbor - aware mechanism that attends neighboring tokens more to capture the richness of context for pre - training language modeling .", "forward": true, "src_ids": "2022.repl4nlp-1.13_1197"}
+{"input": "relation extraction is done by using Task| context: this paper presents a new task of predicting the coverage of a text document for relation extraction ( re ): does the document contain many relational tuples for a given entity ? coverage predictions are useful in selecting the best documents for knowledge base construction with large input corpora .", "entity": "relation extraction", "output": "predicting document coverage", "neg_sample": ["relation extraction is done by using Task", "this paper presents a new task of predicting the coverage of a text document for relation extraction ( re ): does the document contain many relational tuples for a given entity ?", "coverage predictions are useful in selecting the best documents for knowledge base construction with large input corpora ."], "relation": "used for", "id": "2022.tacl-1.12", "year": 2022, "rel_sent": "Predicting Document Coverage for Relation Extraction.", "forward": false, "src_ids": "2022.tacl-1.12_1198"}
+{"input": "predicting document coverage is used for Task| context: coverage predictions are useful in selecting the best documents for knowledge base construction with large input corpora .", "entity": "predicting document coverage", "output": "relation extraction", "neg_sample": ["predicting document coverage is used for Task", "coverage predictions are useful in selecting the best documents for knowledge base construction with large input corpora ."], "relation": "used for", "id": "2022.tacl-1.12", "year": 2022, "rel_sent": "Predicting Document Coverage for Relation Extraction.", "forward": true, "src_ids": "2022.tacl-1.12_1199"}
+{"input": "commonsense - based dialogue reasoning is done by using Material| context: this paper addresses the problem of dialogue reasoning with contextualized commonsense inference .", "entity": "commonsense - based dialogue reasoning", "output": "cicero", "neg_sample": ["commonsense - based dialogue reasoning is done by using Material", "this paper addresses the problem of dialogue reasoning with contextualized commonsense inference ."], "relation": "used for", "id": "2022.acl-long.344", "year": 2022, "rel_sent": "It is our hope that CICERO will open new research avenues into commonsense - based dialogue reasoning .", "forward": false, "src_ids": "2022.acl-long.344_1200"}
+{"input": "utterance - level reasoning - based inferences is done by using Material| context: this paper addresses the problem of dialogue reasoning with contextualized commonsense inference .", "entity": "utterance - level reasoning - based inferences", "output": "dyadic conversations", "neg_sample": ["utterance - level reasoning - based inferences is done by using Material", "this paper addresses the problem of dialogue reasoning with contextualized commonsense inference ."], "relation": "used for", "id": "2022.acl-long.344", "year": 2022, "rel_sent": "We curate CICERO , a dataset of dyadic conversations with five types of utterance - level reasoning - based inferences : cause , subsequent event , prerequisite , motivation , and emotional reaction .", "forward": false, "src_ids": "2022.acl-long.344_1201"}
+{"input": "dyadic conversations is used for OtherScientificTerm| context: this paper addresses the problem of dialogue reasoning with contextualized commonsense inference .", "entity": "dyadic conversations", "output": "utterance - level reasoning - based inferences", "neg_sample": ["dyadic conversations is used for OtherScientificTerm", "this paper addresses the problem of dialogue reasoning with contextualized commonsense inference ."], "relation": "used for", "id": "2022.acl-long.344", "year": 2022, "rel_sent": "We curate CICERO , a dataset of dyadic conversations with five types of utterance - level reasoning - based inferences : cause , subsequent event , prerequisite , motivation , and emotional reaction .", "forward": true, "src_ids": "2022.acl-long.344_1202"}
+{"input": "cicero is used for Task| context: this paper addresses the problem of dialogue reasoning with contextualized commonsense inference .", "entity": "cicero", "output": "commonsense - based dialogue reasoning", "neg_sample": ["cicero is used for Task", "this paper addresses the problem of dialogue reasoning with contextualized commonsense inference ."], "relation": "used for", "id": "2022.acl-long.344", "year": 2022, "rel_sent": "It is our hope that CICERO will open new research avenues into commonsense - based dialogue reasoning .", "forward": true, "src_ids": "2022.acl-long.344_1203"}
+{"input": "hopeful comments is done by using Method| context: hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings . non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating .", "entity": "hopeful comments", "output": "deep learning based models", "neg_sample": ["hopeful comments is done by using Method", "hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings .", "non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating ."], "relation": "used for", "id": "2022.ltedi-1.25", "year": 2022, "rel_sent": "We employed several deep learning based models such as DNN ( dense or fully connected neural network ) , CNN ( Convolutional Neural Network ) , Bi - LSTM ( Bidirectional Long Short Term Memory Network ) , and GRU(Gated Recurrent Unit ) to identify the hopeful comments .", "forward": false, "src_ids": "2022.ltedi-1.25_1204"}
+{"input": "stacked lstm - lstm network is used for Method| context: hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings . non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating .", "entity": "stacked lstm - lstm network", "output": "deep learning based models", "neg_sample": ["stacked lstm - lstm network is used for Method", "hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings .", "non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating ."], "relation": "used for", "id": "2022.ltedi-1.25", "year": 2022, "rel_sent": "We also used Stacked LSTM - CNN and Stacked LSTM - LSTM network to train the model .", "forward": true, "src_ids": "2022.ltedi-1.25_1205"}
+{"input": "stacked lstm - cnn is used for Method| context: hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings . non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating .", "entity": "stacked lstm - cnn", "output": "deep learning based models", "neg_sample": ["stacked lstm - cnn is used for Method", "hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings .", "non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating ."], "relation": "used for", "id": "2022.ltedi-1.25", "year": 2022, "rel_sent": "We also used Stacked LSTM - CNN and Stacked LSTM - LSTM network to train the model .", "forward": true, "src_ids": "2022.ltedi-1.25_1206"}
+{"input": "deep learning based models is used for Material| context: hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings . non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating .", "entity": "deep learning based models", "output": "hopeful comments", "neg_sample": ["deep learning based models is used for Material", "hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings .", "non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating ."], "relation": "used for", "id": "2022.ltedi-1.25", "year": 2022, "rel_sent": "We employed several deep learning based models such as DNN ( dense or fully connected neural network ) , CNN ( Convolutional Neural Network ) , Bi - LSTM ( Bidirectional Long Short Term Memory Network ) , and GRU(Gated Recurrent Unit ) to identify the hopeful comments .", "forward": true, "src_ids": "2022.ltedi-1.25_1207"}
+{"input": "gru(gated recurrent unit ) is used for Material| context: hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings . non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating .", "entity": "gru(gated recurrent unit )", "output": "hopeful comments", "neg_sample": ["gru(gated recurrent unit ) is used for Material", "hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings .", "non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating ."], "relation": "used for", "id": "2022.ltedi-1.25", "year": 2022, "rel_sent": "We employed several deep learning based models such as DNN ( dense or fully connected neural network ) , CNN ( Convolutional Neural Network ) , Bi - LSTM ( Bidirectional Long Short Term Memory Network ) , and GRU(Gated Recurrent Unit ) to identify the hopeful comments .", "forward": true, "src_ids": "2022.ltedi-1.25_1208"}
+{"input": "deep learning based models is done by using Method| context: hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings . non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating .", "entity": "deep learning based models", "output": "stacked lstm - cnn", "neg_sample": ["deep learning based models is done by using Method", "hope speech are positive terms that help to promote or criticise a point of view without hurting the user 's or community 's feelings .", "non - hope speech , on the other side , includes expressions that are harsh , ridiculing , or demotivating ."], "relation": "used for", "id": "2022.ltedi-1.25", "year": 2022, "rel_sent": "We also used Stacked LSTM - CNN and Stacked LSTM - LSTM network to train the model .", "forward": false, "src_ids": "2022.ltedi-1.25_1209"}
+{"input": "dynamic schema graph is done by using Method| context: in dst , modelling the relations among domains and slots is still an under - studied problem . existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains .", "entity": "dynamic schema graph", "output": "dynamic schema graph fusion network ( dsgfnet )", "neg_sample": ["dynamic schema graph is done by using Method", "in dst , modelling the relations among domains and slots is still an under - studied problem .", "existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains ."], "relation": "used for", "id": "2022.acl-long.10", "year": 2022, "rel_sent": "To address these issues , we propose a novel Dynamic Schema Graph Fusion Network ( DSGFNet ) , which generates a dynamic schema graph to explicitly fuse the prior slot - domain membership relations and dialogue - aware dynamic slot relations .", "forward": false, "src_ids": "2022.acl-long.10_1210"}
+{"input": "knowledge transfer is done by using OtherScientificTerm| context: in dst , modelling the relations among domains and slots is still an under - studied problem . existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains .", "entity": "knowledge transfer", "output": "schemata", "neg_sample": ["knowledge transfer is done by using OtherScientificTerm", "in dst , modelling the relations among domains and slots is still an under - studied problem .", "existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains ."], "relation": "used for", "id": "2022.acl-long.10", "year": 2022, "rel_sent": "It also uses the schemata tofacilitate knowledge transfer to new domains .", "forward": false, "src_ids": "2022.acl-long.10_1211"}
+{"input": "schemata is used for Task| context: in dst , modelling the relations among domains and slots is still an under - studied problem . existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains .", "entity": "schemata", "output": "knowledge transfer", "neg_sample": ["schemata is used for Task", "in dst , modelling the relations among domains and slots is still an under - studied problem .", "existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains ."], "relation": "used for", "id": "2022.acl-long.10", "year": 2022, "rel_sent": "It also uses the schemata tofacilitate knowledge transfer to new domains .", "forward": true, "src_ids": "2022.acl-long.10_1212"}
+{"input": "dynamic schema graph fusion network ( dsgfnet ) is used for Method| context: in dst , modelling the relations among domains and slots is still an under - studied problem . existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains .", "entity": "dynamic schema graph fusion network ( dsgfnet )", "output": "dynamic schema graph", "neg_sample": ["dynamic schema graph fusion network ( dsgfnet ) is used for Method", "in dst , modelling the relations among domains and slots is still an under - studied problem .", "existing approaches that have considered such relations generally fall short in : ( 1 ) fusing prior slot - domain membership relations and dialogue - aware dynamic slot relations explicitly , and ( 2 ) generalizing to unseen domains ."], "relation": "used for", "id": "2022.acl-long.10", "year": 2022, "rel_sent": "To address these issues , we propose a novel Dynamic Schema Graph Fusion Network ( DSGFNet ) , which generates a dynamic schema graph to explicitly fuse the prior slot - domain membership relations and dialogue - aware dynamic slot relations .", "forward": true, "src_ids": "2022.acl-long.10_1213"}
+{"input": "data - to - text generation is done by using Task| context: data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) . such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs .", "entity": "data - to - text generation", "output": "compositional generalization", "neg_sample": ["data - to - text generation is done by using Task", "data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) .", "such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs ."], "relation": "used for", "id": "2022.acl-long.289", "year": 2022, "rel_sent": "Improving Compositional Generalization with Self - Training for Data - to - Text Generation.", "forward": false, "src_ids": "2022.acl-long.289_1214"}
+{"input": "data - to - text generation is done by using Method| context: data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) . such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs .", "entity": "data - to - text generation", "output": "self - training approach", "neg_sample": ["data - to - text generation is done by using Method", "data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) .", "such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs ."], "relation": "used for", "id": "2022.acl-long.289", "year": 2022, "rel_sent": "Improving Compositional Generalization with Self - Training for Data - to - Text Generation.", "forward": false, "src_ids": "2022.acl-long.289_1215"}
+{"input": "self - training approach is used for Task| context: such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs .", "entity": "self - training approach", "output": "data - to - text generation", "neg_sample": ["self - training approach is used for Task", "such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs ."], "relation": "used for", "id": "2022.acl-long.289", "year": 2022, "rel_sent": "Improving Compositional Generalization with Self - Training for Data - to - Text Generation.", "forward": true, "src_ids": "2022.acl-long.289_1216"}
+{"input": "compositional generalization is used for Task| context: such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs .", "entity": "compositional generalization", "output": "data - to - text generation", "neg_sample": ["compositional generalization is used for Task", "such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs ."], "relation": "used for", "id": "2022.acl-long.289", "year": 2022, "rel_sent": "Improving Compositional Generalization with Self - Training for Data - to - Text Generation.", "forward": true, "src_ids": "2022.acl-long.289_1217"}
+{"input": "promotional tone detection is done by using OtherScientificTerm| context: detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric .", "entity": "promotional tone detection", "output": "wikipedia article evolution", "neg_sample": ["promotional tone detection is done by using OtherScientificTerm", "detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric ."], "relation": "used for", "id": "2022.acl-long.384", "year": 2022, "rel_sent": "Leveraging Wikipedia article evolution for promotional tone detection.", "forward": false, "src_ids": "2022.acl-long.384_1218"}
+{"input": "wikipedia article evolution is used for Task| context: detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric .", "entity": "wikipedia article evolution", "output": "promotional tone detection", "neg_sample": ["wikipedia article evolution is used for Task", "detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric ."], "relation": "used for", "id": "2022.acl-long.384", "year": 2022, "rel_sent": "Leveraging Wikipedia article evolution for promotional tone detection.", "forward": true, "src_ids": "2022.acl-long.384_1219"}
+{"input": "wikievolve is used for Task| context: detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric .", "entity": "wikievolve", "output": "document - level promotional tone detection", "neg_sample": ["wikievolve is used for Task", "detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric ."], "relation": "used for", "id": "2022.acl-long.384", "year": 2022, "rel_sent": "In this work we introduce WikiEvolve , a dataset for document - level promotional tone detection .", "forward": true, "src_ids": "2022.acl-long.384_1220"}
+{"input": "document - level promotional tone detection is done by using Material| context: detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric .", "entity": "document - level promotional tone detection", "output": "wikievolve", "neg_sample": ["document - level promotional tone detection is done by using Material", "detecting biased language is useful for a variety of applications , such as identifying hyperpartisan news sources or flagging one - sided rhetoric ."], "relation": "used for", "id": "2022.acl-long.384", "year": 2022, "rel_sent": "In this work we introduce WikiEvolve , a dataset for document - level promotional tone detection .", "forward": false, "src_ids": "2022.acl-long.384_1221"}
+{"input": "flow - adapter architecture is used for Task| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "flow - adapter architecture", "output": "unsupervised machine translation", "neg_sample": ["flow - adapter architecture is used for Task", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "Flow - Adapter Architecture for Unsupervised Machine Translation.", "forward": true, "src_ids": "2022.acl-long.89_1222"}
+{"input": "sentence - level latent representations is done by using OtherScientificTerm| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "sentence - level latent representations", "output": "normalizing flows", "neg_sample": ["sentence - level latent representations is done by using OtherScientificTerm", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "It leverages normalizing flows to explicitly model the distributions of sentence - level latent representations , which are subsequently used in conjunction with the attention mechanism for the translation task .", "forward": false, "src_ids": "2022.acl-long.89_1223"}
+{"input": "unsupervised mt is done by using OtherScientificTerm| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "unsupervised mt", "output": "normalizing flows", "neg_sample": ["unsupervised mt is done by using OtherScientificTerm", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "While there is prior work on latent variables for supervised MT , to the best of our knowledge , this is the first work that uses latent variables and normalizing flows for unsupervised MT .", "forward": false, "src_ids": "2022.acl-long.89_1224"}
+{"input": "normalizing flows is used for OtherScientificTerm| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "normalizing flows", "output": "sentence - level latent representations", "neg_sample": ["normalizing flows is used for OtherScientificTerm", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "It leverages normalizing flows to explicitly model the distributions of sentence - level latent representations , which are subsequently used in conjunction with the attention mechanism for the translation task .", "forward": true, "src_ids": "2022.acl-long.89_1225"}
+{"input": "translation task is done by using Method| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "translation task", "output": "attention mechanism", "neg_sample": ["translation task is done by using Method", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "It leverages normalizing flows to explicitly model the distributions of sentence - level latent representations , which are subsequently used in conjunction with the attention mechanism for the translation task .", "forward": false, "src_ids": "2022.acl-long.89_1226"}
+{"input": "attention mechanism is used for Task| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "attention mechanism", "output": "translation task", "neg_sample": ["attention mechanism is used for Task", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "It leverages normalizing flows to explicitly model the distributions of sentence - level latent representations , which are subsequently used in conjunction with the attention mechanism for the translation task .", "forward": true, "src_ids": "2022.acl-long.89_1227"}
+{"input": "translating is done by using Method| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "translating", "output": "latent representations", "neg_sample": ["translating is done by using Method", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "The primary novelties of our model are : ( a ) capturing language - specific sentence representations separately for each language using normalizing flows and ( b ) using a simple transformation of these latent representations for translating from one language to another .", "forward": false, "src_ids": "2022.acl-long.89_1228"}
+{"input": "latent representations is used for Task| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "latent representations", "output": "translating", "neg_sample": ["latent representations is used for Task", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "The primary novelties of our model are : ( a ) capturing language - specific sentence representations separately for each language using normalizing flows and ( b ) using a simple transformation of these latent representations for translating from one language to another .", "forward": true, "src_ids": "2022.acl-long.89_1229"}
+{"input": "supervised mt is done by using OtherScientificTerm| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "supervised mt", "output": "latent variables", "neg_sample": ["supervised mt is done by using OtherScientificTerm", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "While there is prior work on latent variables for supervised MT , to the best of our knowledge , this is the first work that uses latent variables and normalizing flows for unsupervised MT .", "forward": false, "src_ids": "2022.acl-long.89_1230"}
+{"input": "latent variables is used for Method| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "latent variables", "output": "supervised mt", "neg_sample": ["latent variables is used for Method", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "While there is prior work on latent variables for supervised MT , to the best of our knowledge , this is the first work that uses latent variables and normalizing flows for unsupervised MT .", "forward": true, "src_ids": "2022.acl-long.89_1231"}
+{"input": "normalizing flows is used for Task| context: in this work , we propose a flow - adapter architecture for unsupervised nmt .", "entity": "normalizing flows", "output": "unsupervised mt", "neg_sample": ["normalizing flows is used for Task", "in this work , we propose a flow - adapter architecture for unsupervised nmt ."], "relation": "used for", "id": "2022.acl-long.89", "year": 2022, "rel_sent": "While there is prior work on latent variables for supervised MT , to the best of our knowledge , this is the first work that uses latent variables and normalizing flows for unsupervised MT .", "forward": true, "src_ids": "2022.acl-long.89_1232"}
+{"input": "syntactic task is done by using Task| context: although transformer - based neural language models demonstrate impressive performance on a variety of tasks , their generalization abilities are not well understood . they have been shown to perform strongly on subject - verb number agreement in a wide array of settings , suggesting that they learned to track syntactic dependencies during their training even without explicit supervision .", "entity": "syntactic task", "output": "fine - grained analysis of lexical dependence", "neg_sample": ["syntactic task is done by using Task", "although transformer - based neural language models demonstrate impressive performance on a variety of tasks , their generalization abilities are not well understood .", "they have been shown to perform strongly on subject - verb number agreement in a wide array of settings , suggesting that they learned to track syntactic dependencies during their training even without explicit supervision ."], "relation": "used for", "id": "2022.findings-acl.181", "year": 2022, "rel_sent": "Does BERT really agree ? Fine - grained Analysis of Lexical Dependence on a Syntactic Task.", "forward": false, "src_ids": "2022.findings-acl.181_1233"}
+{"input": "fine - grained analysis of lexical dependence is used for Task| context: although transformer - based neural language models demonstrate impressive performance on a variety of tasks , their generalization abilities are not well understood . they have been shown to perform strongly on subject - verb number agreement in a wide array of settings , suggesting that they learned to track syntactic dependencies during their training even without explicit supervision .", "entity": "fine - grained analysis of lexical dependence", "output": "syntactic task", "neg_sample": ["fine - grained analysis of lexical dependence is used for Task", "although transformer - based neural language models demonstrate impressive performance on a variety of tasks , their generalization abilities are not well understood .", "they have been shown to perform strongly on subject - verb number agreement in a wide array of settings , suggesting that they learned to track syntactic dependencies during their training even without explicit supervision ."], "relation": "used for", "id": "2022.findings-acl.181", "year": 2022, "rel_sent": "Does BERT really agree ? Fine - grained Analysis of Lexical Dependence on a Syntactic Task.", "forward": true, "src_ids": "2022.findings-acl.181_1234"}
+{"input": "coherence is done by using OtherScientificTerm| context: we present a memory - augmented approach to condition an autoregressive language model on a knowledge graph .", "entity": "coherence", "output": "relational memory", "neg_sample": ["coherence is done by using OtherScientificTerm", "we present a memory - augmented approach to condition an autoregressive language model on a knowledge graph ."], "relation": "used for", "id": "2022.tacl-1.32", "year": 2022, "rel_sent": "We also show that relational memory improves coherence , is complementary to token - based memory , and enables causal interventions .", "forward": false, "src_ids": "2022.tacl-1.32_1235"}
+{"input": "relational memory is used for OtherScientificTerm| context: we present a memory - augmented approach to condition an autoregressive language model on a knowledge graph .", "entity": "relational memory", "output": "coherence", "neg_sample": ["relational memory is used for OtherScientificTerm", "we present a memory - augmented approach to condition an autoregressive language model on a knowledge graph ."], "relation": "used for", "id": "2022.tacl-1.32", "year": 2022, "rel_sent": "We also show that relational memory improves coherence , is complementary to token - based memory , and enables causal interventions .", "forward": true, "src_ids": "2022.tacl-1.32_1236"}
+{"input": "knowledge graph is used for Task| context: we present a memory - augmented approach to condition an autoregressive language model on a knowledge graph .", "entity": "knowledge graph", "output": "coherent and logical generation", "neg_sample": ["knowledge graph is used for Task", "we present a memory - augmented approach to condition an autoregressive language model on a knowledge graph ."], "relation": "used for", "id": "2022.tacl-1.32", "year": 2022, "rel_sent": "Our model provides a simple yet effective way to combine an autoregressive language model and a knowledge graph for more coherent and logical generation .", "forward": true, "src_ids": "2022.tacl-1.32_1237"}
+{"input": "non - literal translation is done by using Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "non - literal translation", "output": "transformer", "neg_sample": ["non - literal translation is done by using Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "In this work , we investigate whether the non - compositionality of idioms is reflected in the mechanics of the dominant NMT model , Transformer , by analysing the hidden states and attention patterns for models with English as source language and one of seven European languages as target language . When Transformer emits a non - literal translation - i.e.", "forward": false, "src_ids": "2022.acl-long.252_1238"}
+{"input": "literal translations of idioms is done by using Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "literal translations of idioms", "output": "transformer", "neg_sample": ["literal translations of idioms is done by using Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "These results suggest that Transformer 's tendency to process idioms as compositional expressions contributes to literal translations of idioms .", "forward": false, "src_ids": "2022.acl-long.252_1239"}
+{"input": "idioms is done by using Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "idioms", "output": "transformer", "neg_sample": ["idioms is done by using Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "These results suggest that Transformer 's tendency to process idioms as compositional expressions contributes to literal translations of idioms .", "forward": false, "src_ids": "2022.acl-long.252_1240"}
+{"input": "neural machine translation is done by using Task| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "neural machine translation", "output": "analysing idiom processing", "neg_sample": ["neural machine translation is done by using Task", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "Can Transformer be Too Compositional ? Analysing Idiom Processing in Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.252_1241"}
+{"input": "analysing idiom processing is used for Task| context: nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "analysing idiom processing", "output": "neural machine translation", "neg_sample": ["analysing idiom processing is used for Task", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "Can Transformer be Too Compositional ? Analysing Idiom Processing in Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.252_1242"}
+{"input": "transformer is used for Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "transformer", "output": "non - literal translation", "neg_sample": ["transformer is used for Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "In this work , we investigate whether the non - compositionality of idioms is reflected in the mechanics of the dominant NMT model , Transformer , by analysing the hidden states and attention patterns for models with English as source language and one of seven European languages as target language . When Transformer emits a non - literal translation - i.e.", "forward": true, "src_ids": "2022.acl-long.252_1243"}
+{"input": "idioms is done by using Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "idioms", "output": "encoder", "neg_sample": ["idioms is done by using Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "identifies the expression as idiomatic - the encoder processes idioms more strongly as single lexical units compared to literal expressions .", "forward": false, "src_ids": "2022.acl-long.252_1244"}
+{"input": "transformer is used for Task| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "transformer", "output": "literal translations of idioms", "neg_sample": ["transformer is used for Task", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.252", "year": 2022, "rel_sent": "These results suggest that Transformer 's tendency to process idioms as compositional expressions contributes to literal translations of idioms .", "forward": true, "src_ids": "2022.acl-long.252_1245"}
+{"input": "summarization is done by using Method| context: the availability of large - scale datasets has driven the development of neural models that create generic summaries for single or multiple documents . for query - focused summarization ( qfs ) , labeled training data in the form of queries , documents , and summaries is not readily available .", "entity": "summarization", "output": "unified modeling framework", "neg_sample": ["summarization is done by using Method", "the availability of large - scale datasets has driven the development of neural models that create generic summaries for single or multiple documents .", "for query - focused summarization ( qfs ) , labeled training data in the form of queries , documents , and summaries is not readily available ."], "relation": "used for", "id": "2022.tacl-1.36", "year": 2022, "rel_sent": "We provide a unified modeling framework for any kind of summarization , under the assumption that all summaries are a response to a query , which is observed in the case of QFS and latent in the case of generic summarization .", "forward": false, "src_ids": "2022.tacl-1.36_1246"}
+{"input": "unified modeling framework is used for Task| context: the availability of large - scale datasets has driven the development of neural models that create generic summaries for single or multiple documents .", "entity": "unified modeling framework", "output": "summarization", "neg_sample": ["unified modeling framework is used for Task", "the availability of large - scale datasets has driven the development of neural models that create generic summaries for single or multiple documents ."], "relation": "used for", "id": "2022.tacl-1.36", "year": 2022, "rel_sent": "We provide a unified modeling framework for any kind of summarization , under the assumption that all summaries are a response to a query , which is observed in the case of QFS and latent in the case of generic summarization .", "forward": true, "src_ids": "2022.tacl-1.36_1247"}
+{"input": "temporal knowledge graphs is done by using Method| context: temporal knowledge graphs store the dynamics of entities and relations during a time period . however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios . hence , modeling temporal knowledge graphs to complete the missing facts is important .", "entity": "temporal knowledge graphs", "output": "tempcaps", "neg_sample": ["temporal knowledge graphs is done by using Method", "temporal knowledge graphs store the dynamics of entities and relations during a time period .", "however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios .", "hence , modeling temporal knowledge graphs to complete the missing facts is important ."], "relation": "used for", "id": "2022.spnlp-1.3", "year": 2022, "rel_sent": "TempCaps models temporal knowledge graphs by introducing a novel dynamic routing aggregator inspired by Capsule Networks .", "forward": false, "src_ids": "2022.spnlp-1.3_1248"}
+{"input": "temporal knowledge graph completion task is done by using Method| context: temporal knowledge graphs store the dynamics of entities and relations during a time period . however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios . hence , modeling temporal knowledge graphs to complete the missing facts is important .", "entity": "temporal knowledge graph completion task", "output": "tempcaps", "neg_sample": ["temporal knowledge graph completion task is done by using Method", "temporal knowledge graphs store the dynamics of entities and relations during a time period .", "however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios .", "hence , modeling temporal knowledge graphs to complete the missing facts is important ."], "relation": "used for", "id": "2022.spnlp-1.3", "year": 2022, "rel_sent": "Experimental results demonstrate that TempCaps reaches state - of - the - art performance for temporal knowledge graph completion .", "forward": false, "src_ids": "2022.spnlp-1.3_1249"}
+{"input": "temporal knowledge graph completion task is done by using Method| context: temporal knowledge graphs store the dynamics of entities and relations during a time period . however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios . hence , modeling temporal knowledge graphs to complete the missing facts is important .", "entity": "temporal knowledge graph completion task", "output": "capsule network - based embedding model", "neg_sample": ["temporal knowledge graph completion task is done by using Method", "temporal knowledge graphs store the dynamics of entities and relations during a time period .", "however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios .", "hence , modeling temporal knowledge graphs to complete the missing facts is important ."], "relation": "used for", "id": "2022.spnlp-1.3", "year": 2022, "rel_sent": "In this paper , we tackle the temporal knowledge graph completion task by proposing TempCaps , which is a Capsule network - based embedding model for Temporal knowledge graph completion .", "forward": false, "src_ids": "2022.spnlp-1.3_1250"}
+{"input": "capsule network - based embedding model is used for Task| context: temporal knowledge graphs store the dynamics of entities and relations during a time period . however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios . hence , modeling temporal knowledge graphs to complete the missing facts is important .", "entity": "capsule network - based embedding model", "output": "temporal knowledge graph completion task", "neg_sample": ["capsule network - based embedding model is used for Task", "temporal knowledge graphs store the dynamics of entities and relations during a time period .", "however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios .", "hence , modeling temporal knowledge graphs to complete the missing facts is important ."], "relation": "used for", "id": "2022.spnlp-1.3", "year": 2022, "rel_sent": "In this paper , we tackle the temporal knowledge graph completion task by proposing TempCaps , which is a Capsule network - based embedding model for Temporal knowledge graph completion .", "forward": true, "src_ids": "2022.spnlp-1.3_1251"}
+{"input": "tempcaps is used for Task| context: temporal knowledge graphs store the dynamics of entities and relations during a time period . however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios . hence , modeling temporal knowledge graphs to complete the missing facts is important .", "entity": "tempcaps", "output": "temporal knowledge graph completion task", "neg_sample": ["tempcaps is used for Task", "temporal knowledge graphs store the dynamics of entities and relations during a time period .", "however , typical temporal knowledge graphs often suffer from incomplete dynamics with missing facts in real - world scenarios .", "hence , modeling temporal knowledge graphs to complete the missing facts is important ."], "relation": "used for", "id": "2022.spnlp-1.3", "year": 2022, "rel_sent": "Experimental results demonstrate that TempCaps reaches state - of - the - art performance for temporal knowledge graph completion .", "forward": true, "src_ids": "2022.spnlp-1.3_1252"}
+{"input": "complex word identification is done by using Task| context: complex word identification ( cwi ) is a cornerstone process towards proper text simplification . cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages . as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples .", "entity": "complex word identification", "output": "domain adaptation", "neg_sample": ["complex word identification is done by using Task", "complex word identification ( cwi ) is a cornerstone process towards proper text simplification .", "cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages .", "as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples ."], "relation": "used for", "id": "2022.acl-long.6", "year": 2022, "rel_sent": "Domain Adaptation in Multilingual and Multi - Domain Monolingual Settings for Complex Word Identification.", "forward": false, "src_ids": "2022.acl-long.6_1253"}
+{"input": "multilingual and multi - domain monolingual settings is done by using Task| context: complex word identification ( cwi ) is a cornerstone process towards proper text simplification . cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages . as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples .", "entity": "multilingual and multi - domain monolingual settings", "output": "domain adaptation", "neg_sample": ["multilingual and multi - domain monolingual settings is done by using Task", "complex word identification ( cwi ) is a cornerstone process towards proper text simplification .", "cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages .", "as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples ."], "relation": "used for", "id": "2022.acl-long.6", "year": 2022, "rel_sent": "Domain Adaptation in Multilingual and Multi - Domain Monolingual Settings for Complex Word Identification.", "forward": false, "src_ids": "2022.acl-long.6_1254"}
+{"input": "domain adaptation is used for Material| context: complex word identification ( cwi ) is a cornerstone process towards proper text simplification . cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages . as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples .", "entity": "domain adaptation", "output": "multilingual and multi - domain monolingual settings", "neg_sample": ["domain adaptation is used for Material", "complex word identification ( cwi ) is a cornerstone process towards proper text simplification .", "cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages .", "as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples ."], "relation": "used for", "id": "2022.acl-long.6", "year": 2022, "rel_sent": "Domain Adaptation in Multilingual and Multi - Domain Monolingual Settings for Complex Word Identification.", "forward": true, "src_ids": "2022.acl-long.6_1255"}
+{"input": "domain adaptation is used for Task| context: as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples .", "entity": "domain adaptation", "output": "complex word identification", "neg_sample": ["domain adaptation is used for Task", "as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples ."], "relation": "used for", "id": "2022.acl-long.6", "year": 2022, "rel_sent": "Domain Adaptation in Multilingual and Multi - Domain Monolingual Settings for Complex Word Identification.", "forward": true, "src_ids": "2022.acl-long.6_1256"}
+{"input": "cwi task is done by using Method| context: complex word identification ( cwi ) is a cornerstone process towards proper text simplification . cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages . as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples .", "entity": "cwi task", "output": "training technique", "neg_sample": ["cwi task is done by using Method", "complex word identification ( cwi ) is a cornerstone process towards proper text simplification .", "cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages .", "as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples ."], "relation": "used for", "id": "2022.acl-long.6", "year": 2022, "rel_sent": "In this paper , we propose a novel training technique for the CWI task based on domain adaptation to improve the target character and context representations .", "forward": false, "src_ids": "2022.acl-long.6_1257"}
+{"input": "training technique is used for Task| context: complex word identification ( cwi ) is a cornerstone process towards proper text simplification . cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages . as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples .", "entity": "training technique", "output": "cwi task", "neg_sample": ["training technique is used for Task", "complex word identification ( cwi ) is a cornerstone process towards proper text simplification .", "cwi is highly dependent on context , whereas its difficulty is augmented by the scarcity of available datasets which vary greatly in terms of domains and languages .", "as such , it becomes increasingly more difficult to develop a robust model that generalizes across a wide array of input examples ."], "relation": "used for", "id": "2022.acl-long.6", "year": 2022, "rel_sent": "In this paper , we propose a novel training technique for the CWI task based on domain adaptation to improve the target character and context representations .", "forward": true, "src_ids": "2022.acl-long.6_1258"}
+{"input": "pretrained language models is done by using Method| context: the popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down - stream tasks , which have a higher potential for societal impact . the evaluation of such systems usually focuses on accuracy measures . our findings in this paper call for attention to be paid tofairness measures as well .", "entity": "pretrained language models", "output": "post - processing methods", "neg_sample": ["pretrained language models is done by using Method", "the popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down - stream tasks , which have a higher potential for societal impact .", "the evaluation of such systems usually focuses on accuracy measures .", "our findings in this paper call for attention to be paid tofairness measures as well ."], "relation": "used for", "id": "2022.findings-acl.176", "year": 2022, "rel_sent": "To improve model fairness without retraining , we show that two post - processing methods developed for structured , tabular data can be successfully applied to a range of pretrained language models .", "forward": false, "src_ids": "2022.findings-acl.176_1259"}
+{"input": "structured , tabular data is done by using Method| context: the popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down - stream tasks , which have a higher potential for societal impact . the evaluation of such systems usually focuses on accuracy measures . our findings in this paper call for attention to be paid tofairness measures as well .", "entity": "structured , tabular data", "output": "post - processing methods", "neg_sample": ["structured , tabular data is done by using Method", "the popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down - stream tasks , which have a higher potential for societal impact .", "the evaluation of such systems usually focuses on accuracy measures .", "our findings in this paper call for attention to be paid tofairness measures as well ."], "relation": "used for", "id": "2022.findings-acl.176", "year": 2022, "rel_sent": "To improve model fairness without retraining , we show that two post - processing methods developed for structured , tabular data can be successfully applied to a range of pretrained language models .", "forward": false, "src_ids": "2022.findings-acl.176_1260"}
+{"input": "post - processing methods is used for Method| context: the evaluation of such systems usually focuses on accuracy measures . our findings in this paper call for attention to be paid tofairness measures as well .", "entity": "post - processing methods", "output": "pretrained language models", "neg_sample": ["post - processing methods is used for Method", "the evaluation of such systems usually focuses on accuracy measures .", "our findings in this paper call for attention to be paid tofairness measures as well ."], "relation": "used for", "id": "2022.findings-acl.176", "year": 2022, "rel_sent": "To improve model fairness without retraining , we show that two post - processing methods developed for structured , tabular data can be successfully applied to a range of pretrained language models .", "forward": true, "src_ids": "2022.findings-acl.176_1261"}
+{"input": "post - processing methods is used for Material| context: the popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down - stream tasks , which have a higher potential for societal impact . the evaluation of such systems usually focuses on accuracy measures . our findings in this paper call for attention to be paid tofairness measures as well .", "entity": "post - processing methods", "output": "structured , tabular data", "neg_sample": ["post - processing methods is used for Material", "the popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down - stream tasks , which have a higher potential for societal impact .", "the evaluation of such systems usually focuses on accuracy measures .", "our findings in this paper call for attention to be paid tofairness measures as well ."], "relation": "used for", "id": "2022.findings-acl.176", "year": 2022, "rel_sent": "To improve model fairness without retraining , we show that two post - processing methods developed for structured , tabular data can be successfully applied to a range of pretrained language models .", "forward": true, "src_ids": "2022.findings-acl.176_1262"}
+{"input": "zero - shot dialogue comprehension is done by using Method| context: comprehending a dialogue requires a model to capture diverse kinds of key information in the utterances , which are either scattered around or implicitly implied in different turns of conversations . therefore , dialogue comprehension requires diverse capabilities such as paraphrasing , summarizing , and commonsense reasoning . however , the dialogue - narrative parallel corpus for such a pre - training strategy is currently unavailable .", "entity": "zero - shot dialogue comprehension", "output": "narrative pre - training", "neg_sample": ["zero - shot dialogue comprehension is done by using Method", "comprehending a dialogue requires a model to capture diverse kinds of key information in the utterances , which are either scattered around or implicitly implied in different turns of conversations .", "therefore , dialogue comprehension requires diverse capabilities such as paraphrasing , summarizing , and commonsense reasoning .", "however , the dialogue - narrative parallel corpus for such a pre - training strategy is currently unavailable ."], "relation": "used for", "id": "2022.acl-short.23", "year": 2022, "rel_sent": "Learning - by - Narrating : Narrative Pre - Training for Zero - Shot Dialogue Comprehension.", "forward": false, "src_ids": "2022.acl-short.23_1263"}
+{"input": "narrative pre - training is used for Task| context: comprehending a dialogue requires a model to capture diverse kinds of key information in the utterances , which are either scattered around or implicitly implied in different turns of conversations . therefore , dialogue comprehension requires diverse capabilities such as paraphrasing , summarizing , and commonsense reasoning . however , the dialogue - narrative parallel corpus for such a pre - training strategy is currently unavailable .", "entity": "narrative pre - training", "output": "zero - shot dialogue comprehension", "neg_sample": ["narrative pre - training is used for Task", "comprehending a dialogue requires a model to capture diverse kinds of key information in the utterances , which are either scattered around or implicitly implied in different turns of conversations .", "therefore , dialogue comprehension requires diverse capabilities such as paraphrasing , summarizing , and commonsense reasoning .", "however , the dialogue - narrative parallel corpus for such a pre - training strategy is currently unavailable ."], "relation": "used for", "id": "2022.acl-short.23", "year": 2022, "rel_sent": "Learning - by - Narrating : Narrative Pre - Training for Zero - Shot Dialogue Comprehension.", "forward": true, "src_ids": "2022.acl-short.23_1264"}
+{"input": "language model generalization is done by using Method| context: the allure of superhuman - level capabilities has led to considerable interest in language models like gpt-3 and t5 , wherein the research has , by and large , revolved around new model architectures , training tasks , and loss objectives , along with substantial engineering efforts to scale up model capacity and dataset size . comparatively little work has been done to improve the generalization of these models through better optimization .", "entity": "language model generalization", "output": "sharpness - aware minimization ( sam )", "neg_sample": ["language model generalization is done by using Method", "the allure of superhuman - level capabilities has led to considerable interest in language models like gpt-3 and t5 , wherein the research has , by and large , revolved around new model architectures , training tasks , and loss objectives , along with substantial engineering efforts to scale up model capacity and dataset size .", "comparatively little work has been done to improve the generalization of these models through better optimization ."], "relation": "used for", "id": "2022.acl-long.508", "year": 2022, "rel_sent": "Sharpness - Aware Minimization Improves Language Model Generalization.", "forward": false, "src_ids": "2022.acl-long.508_1265"}
+{"input": "sharpness - aware minimization ( sam ) is used for Task| context: the allure of superhuman - level capabilities has led to considerable interest in language models like gpt-3 and t5 , wherein the research has , by and large , revolved around new model architectures , training tasks , and loss objectives , along with substantial engineering efforts to scale up model capacity and dataset size . comparatively little work has been done to improve the generalization of these models through better optimization .", "entity": "sharpness - aware minimization ( sam )", "output": "language model generalization", "neg_sample": ["sharpness - aware minimization ( sam ) is used for Task", "the allure of superhuman - level capabilities has led to considerable interest in language models like gpt-3 and t5 , wherein the research has , by and large , revolved around new model architectures , training tasks , and loss objectives , along with substantial engineering efforts to scale up model capacity and dataset size .", "comparatively little work has been done to improve the generalization of these models through better optimization ."], "relation": "used for", "id": "2022.acl-long.508", "year": 2022, "rel_sent": "Sharpness - Aware Minimization Improves Language Model Generalization.", "forward": true, "src_ids": "2022.acl-long.508_1266"}
+{"input": "plains cree is done by using Task| context: the composition of richly - inflected words in morphologically complex languages can be a challenge for language learners developing literacy . accordingly , lane and bird ( 2020 ) proposed a finite state approach which maps prefixes in a language to a set of possible completions up to the next morpheme boundary , for the incremental building of complex words .", "entity": "plains cree", "output": "interactive word completion", "neg_sample": ["plains cree is done by using Task", "the composition of richly - inflected words in morphologically complex languages can be a challenge for language learners developing literacy .", "accordingly , lane and bird ( 2020 ) proposed a finite state approach which maps prefixes in a language to a set of possible completions up to the next morpheme boundary , for the incremental building of complex words ."], "relation": "used for", "id": "2022.acl-long.232", "year": 2022, "rel_sent": "Interactive Word Completion for Plains Cree.", "forward": false, "src_ids": "2022.acl-long.232_1267"}
+{"input": "interactive word completion is used for Material| context: the composition of richly - inflected words in morphologically complex languages can be a challenge for language learners developing literacy . accordingly , lane and bird ( 2020 ) proposed a finite state approach which maps prefixes in a language to a set of possible completions up to the next morpheme boundary , for the incremental building of complex words .", "entity": "interactive word completion", "output": "plains cree", "neg_sample": ["interactive word completion is used for Material", "the composition of richly - inflected words in morphologically complex languages can be a challenge for language learners developing literacy .", "accordingly , lane and bird ( 2020 ) proposed a finite state approach which maps prefixes in a language to a set of possible completions up to the next morpheme boundary , for the incremental building of complex words ."], "relation": "used for", "id": "2022.acl-long.232", "year": 2022, "rel_sent": "Interactive Word Completion for Plains Cree.", "forward": true, "src_ids": "2022.acl-long.232_1268"}
+{"input": "mt - ready glossaries is done by using Method| context: commercial machine translation ( mt ) providers offer functionalities that allow users to leverage bilingual glossaries . this poses the question of how to turn glossaries that were intended to be used by a human translator into mt - ready ones , removing entries that could harm the mt output .", "entity": "mt - ready glossaries", "output": "automatic and manual filtering methods", "neg_sample": ["mt - ready glossaries is done by using Method", "commercial machine translation ( mt ) providers offer functionalities that allow users to leverage bilingual glossaries .", "this poses the question of how to turn glossaries that were intended to be used by a human translator into mt - ready ones , removing entries that could harm the mt output ."], "relation": "used for", "id": "2022.eamt-1.25", "year": 2022, "rel_sent": "Investigating automatic and manual filtering methods to produce MT - ready glossaries from existing ones.", "forward": false, "src_ids": "2022.eamt-1.25_1269"}
+{"input": "automatic and manual filtering methods is used for Material| context: commercial machine translation ( mt ) providers offer functionalities that allow users to leverage bilingual glossaries . this poses the question of how to turn glossaries that were intended to be used by a human translator into mt - ready ones , removing entries that could harm the mt output .", "entity": "automatic and manual filtering methods", "output": "mt - ready glossaries", "neg_sample": ["automatic and manual filtering methods is used for Material", "commercial machine translation ( mt ) providers offer functionalities that allow users to leverage bilingual glossaries .", "this poses the question of how to turn glossaries that were intended to be used by a human translator into mt - ready ones , removing entries that could harm the mt output ."], "relation": "used for", "id": "2022.eamt-1.25", "year": 2022, "rel_sent": "Investigating automatic and manual filtering methods to produce MT - ready glossaries from existing ones.", "forward": true, "src_ids": "2022.eamt-1.25_1270"}
+{"input": "pre - trained clm is done by using Method| context: the introduction of immensely large causal language models ( clms ) has rejuvenated the interest in open - ended text generation . however , controlling the generative process for these transformer - based models is at large an unsolved problem . earlier work has explored either plug - and - play decoding strategies , or more powerful but blunt approaches such as prompting . there hence currently exists a trade - off between fine - grained control , and the capability for more expressive high - level instructions .", "entity": "pre - trained clm", "output": "resource - efficient method", "neg_sample": ["pre - trained clm is done by using Method", "the introduction of immensely large causal language models ( clms ) has rejuvenated the interest in open - ended text generation .", "however , controlling the generative process for these transformer - based models is at large an unsolved problem .", "earlier work has explored either plug - and - play decoding strategies , or more powerful but blunt approaches such as prompting .", "there hence currently exists a trade - off between fine - grained control , and the capability for more expressive high - level instructions ."], "relation": "used for", "id": "2022.acl-long.471", "year": 2022, "rel_sent": "We propose a resource - efficient method for converting a pre - trained CLM into this architecture , and demonstrate its potential on various experiments , including the novel task of contextualized word inclusion .", "forward": false, "src_ids": "2022.acl-long.471_1271"}
+{"input": "resource - efficient method is used for Method| context: the introduction of immensely large causal language models ( clms ) has rejuvenated the interest in open - ended text generation . however , controlling the generative process for these transformer - based models is at large an unsolved problem . earlier work has explored either plug - and - play decoding strategies , or more powerful but blunt approaches such as prompting . there hence currently exists a trade - off between fine - grained control , and the capability for more expressive high - level instructions .", "entity": "resource - efficient method", "output": "pre - trained clm", "neg_sample": ["resource - efficient method is used for Method", "the introduction of immensely large causal language models ( clms ) has rejuvenated the interest in open - ended text generation .", "however , controlling the generative process for these transformer - based models is at large an unsolved problem .", "earlier work has explored either plug - and - play decoding strategies , or more powerful but blunt approaches such as prompting .", "there hence currently exists a trade - off between fine - grained control , and the capability for more expressive high - level instructions ."], "relation": "used for", "id": "2022.acl-long.471", "year": 2022, "rel_sent": "We propose a resource - efficient method for converting a pre - trained CLM into this architecture , and demonstrate its potential on various experiments , including the novel task of contextualized word inclusion .", "forward": true, "src_ids": "2022.acl-long.471_1272"}
+{"input": "petrogold annotation is done by using Method| context: the corpus follows the dependencies approach of the universal dependencies multilingual project .", "entity": "petrogold annotation", "output": "semantic - discursively oriented approach", "neg_sample": ["petrogold annotation is done by using Method", "the corpus follows the dependencies approach of the universal dependencies multilingual project ."], "relation": "used for", "id": "2022.udfestbr-1.5", "year": 2022, "rel_sent": "After consulting Portuguese grammars to assist in the annotation of the adverbial adjunct and indirect object relations , we propose a semantic - discursively oriented approach , which was used in the PetroGold annotation and affected 14.8 % of the sentences in the treebank .", "forward": false, "src_ids": "2022.udfestbr-1.5_1273"}
+{"input": "semantic - discursively oriented approach is used for Task| context: the corpus follows the dependencies approach of the universal dependencies multilingual project .", "entity": "semantic - discursively oriented approach", "output": "petrogold annotation", "neg_sample": ["semantic - discursively oriented approach is used for Task", "the corpus follows the dependencies approach of the universal dependencies multilingual project ."], "relation": "used for", "id": "2022.udfestbr-1.5", "year": 2022, "rel_sent": "After consulting Portuguese grammars to assist in the annotation of the adverbial adjunct and indirect object relations , we propose a semantic - discursively oriented approach , which was used in the PetroGold annotation and affected 14.8 % of the sentences in the treebank .", "forward": true, "src_ids": "2022.udfestbr-1.5_1274"}
+{"input": "data augmentation is done by using OtherScientificTerm| context: end - to - end speech translation relies on data that pair source - language speech inputs with corresponding translations into a target language . such data are notoriously scarce , making synthetic data augmentation by back - translation or knowledge distillation a necessary ingredient of end - to - end training .", "entity": "data augmentation", "output": "audio alignments", "neg_sample": ["data augmentation is done by using OtherScientificTerm", "end - to - end speech translation relies on data that pair source - language speech inputs with corresponding translations into a target language .", "such data are notoriously scarce , making synthetic data augmentation by back - translation or knowledge distillation a necessary ingredient of end - to - end training ."], "relation": "used for", "id": "2022.acl-short.27", "year": 2022, "rel_sent": "Sample , Translate , Recombine : Leveraging Audio Alignments for Data Augmentation in End - to - end Speech Translation.", "forward": false, "src_ids": "2022.acl-short.27_1275"}
+{"input": "end - to - end speech translation is done by using Method| context: end - to - end speech translation relies on data that pair source - language speech inputs with corresponding translations into a target language .", "entity": "end - to - end speech translation", "output": "data augmentation", "neg_sample": ["end - to - end speech translation is done by using Method", "end - to - end speech translation relies on data that pair source - language speech inputs with corresponding translations into a target language ."], "relation": "used for", "id": "2022.acl-short.27", "year": 2022, "rel_sent": "Sample , Translate , Recombine : Leveraging Audio Alignments for Data Augmentation in End - to - end Speech Translation.", "forward": false, "src_ids": "2022.acl-short.27_1276"}
+{"input": "audio alignments is used for Method| context: end - to - end speech translation relies on data that pair source - language speech inputs with corresponding translations into a target language .", "entity": "audio alignments", "output": "data augmentation", "neg_sample": ["audio alignments is used for Method", "end - to - end speech translation relies on data that pair source - language speech inputs with corresponding translations into a target language ."], "relation": "used for", "id": "2022.acl-short.27", "year": 2022, "rel_sent": "Sample , Translate , Recombine : Leveraging Audio Alignments for Data Augmentation in End - to - end Speech Translation.", "forward": true, "src_ids": "2022.acl-short.27_1277"}
+{"input": "data augmentation is used for Task| context: such data are notoriously scarce , making synthetic data augmentation by back - translation or knowledge distillation a necessary ingredient of end - to - end training .", "entity": "data augmentation", "output": "end - to - end speech translation", "neg_sample": ["data augmentation is used for Task", "such data are notoriously scarce , making synthetic data augmentation by back - translation or knowledge distillation a necessary ingredient of end - to - end training ."], "relation": "used for", "id": "2022.acl-short.27", "year": 2022, "rel_sent": "Sample , Translate , Recombine : Leveraging Audio Alignments for Data Augmentation in End - to - end Speech Translation.", "forward": true, "src_ids": "2022.acl-short.27_1278"}
+{"input": "contextualization of tabular data is done by using Method| context: the robustness of text - to - sql parsers against adversarial perturbations plays a crucial role in delivering highly reliable applications . previous studies along this line primarily focused on perturbations in the natural language question side , neglecting the variability of tables .", "entity": "contextualization of tabular data", "output": "systematic adversarial training example generation framework", "neg_sample": ["contextualization of tabular data is done by using Method", "the robustness of text - to - sql parsers against adversarial perturbations plays a crucial role in delivering highly reliable applications .", "previous studies along this line primarily focused on perturbations in the natural language question side , neglecting the variability of tables ."], "relation": "used for", "id": "2022.acl-long.142", "year": 2022, "rel_sent": "To defense against ATP , we build a systematic adversarial training example generation framework tailored for better contextualization of tabular data .", "forward": false, "src_ids": "2022.acl-long.142_1279"}
+{"input": "systematic adversarial training example generation framework is used for Material| context: the robustness of text - to - sql parsers against adversarial perturbations plays a crucial role in delivering highly reliable applications . previous studies along this line primarily focused on perturbations in the natural language question side , neglecting the variability of tables .", "entity": "systematic adversarial training example generation framework", "output": "contextualization of tabular data", "neg_sample": ["systematic adversarial training example generation framework is used for Material", "the robustness of text - to - sql parsers against adversarial perturbations plays a crucial role in delivering highly reliable applications .", "previous studies along this line primarily focused on perturbations in the natural language question side , neglecting the variability of tables ."], "relation": "used for", "id": "2022.acl-long.142", "year": 2022, "rel_sent": "To defense against ATP , we build a systematic adversarial training example generation framework tailored for better contextualization of tabular data .", "forward": true, "src_ids": "2022.acl-long.142_1280"}
+{"input": "signs of depression is done by using Method| context: depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act . detecting depression is essential as it must be treated early to avoid painful consequences . nowadays , people are broadcasting how they feel via posts and comments .", "entity": "signs of depression", "output": "random undersampling", "neg_sample": ["signs of depression is done by using Method", "depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act .", "detecting depression is essential as it must be treated early to avoid painful consequences .", "nowadays , people are broadcasting how they feel via posts and comments ."], "relation": "used for", "id": "2022.ltedi-1.45", "year": 2022, "rel_sent": "DepressionOne@LT - EDI - ACL2022 : Using Machine Learning with SMOTE and Random UnderSampling to Detect Signs of Depression on Social Media Text ..", "forward": false, "src_ids": "2022.ltedi-1.45_1281"}
+{"input": "random undersampling is used for OtherScientificTerm| context: depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act . detecting depression is essential as it must be treated early to avoid painful consequences . nowadays , people are broadcasting how they feel via posts and comments .", "entity": "random undersampling", "output": "signs of depression", "neg_sample": ["random undersampling is used for OtherScientificTerm", "depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act .", "detecting depression is essential as it must be treated early to avoid painful consequences .", "nowadays , people are broadcasting how they feel via posts and comments ."], "relation": "used for", "id": "2022.ltedi-1.45", "year": 2022, "rel_sent": "DepressionOne@LT - EDI - ACL2022 : Using Machine Learning with SMOTE and Random UnderSampling to Detect Signs of Depression on Social Media Text ..", "forward": true, "src_ids": "2022.ltedi-1.45_1282"}
+{"input": "machine learning methods is used for OtherScientificTerm| context: depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act . detecting depression is essential as it must be treated early to avoid painful consequences . nowadays , people are broadcasting how they feel via posts and comments .", "entity": "machine learning methods", "output": "signs of depression", "neg_sample": ["machine learning methods is used for OtherScientificTerm", "depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act .", "detecting depression is essential as it must be treated early to avoid painful consequences .", "nowadays , people are broadcasting how they feel via posts and comments ."], "relation": "used for", "id": "2022.ltedi-1.45", "year": 2022, "rel_sent": "Later , we used machine learning methods to train and detect the signs of depression .", "forward": true, "src_ids": "2022.ltedi-1.45_1283"}
+{"input": "nlp techniques is used for OtherScientificTerm| context: nowadays , people are broadcasting how they feel via posts and comments .", "entity": "nlp techniques", "output": "depression", "neg_sample": ["nlp techniques is used for OtherScientificTerm", "nowadays , people are broadcasting how they feel via posts and comments ."], "relation": "used for", "id": "2022.ltedi-1.45", "year": 2022, "rel_sent": "Using social media , we can extract many comments related to depression and use NLP techniques to train and detect depression .", "forward": true, "src_ids": "2022.ltedi-1.45_1284"}
+{"input": "depression is done by using Method| context: depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act . detecting depression is essential as it must be treated early to avoid painful consequences . nowadays , people are broadcasting how they feel via posts and comments .", "entity": "depression", "output": "nlp techniques", "neg_sample": ["depression is done by using Method", "depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act .", "detecting depression is essential as it must be treated early to avoid painful consequences .", "nowadays , people are broadcasting how they feel via posts and comments ."], "relation": "used for", "id": "2022.ltedi-1.45", "year": 2022, "rel_sent": "Using social media , we can extract many comments related to depression and use NLP techniques to train and detect depression .", "forward": false, "src_ids": "2022.ltedi-1.45_1285"}
+{"input": "signs of depression is done by using Method| context: depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act . detecting depression is essential as it must be treated early to avoid painful consequences . nowadays , people are broadcasting how they feel via posts and comments .", "entity": "signs of depression", "output": "machine learning methods", "neg_sample": ["signs of depression is done by using Method", "depression is a common and serious medical illness that negatively affects how you feel , the way you think , and how you act .", "detecting depression is essential as it must be treated early to avoid painful consequences .", "nowadays , people are broadcasting how they feel via posts and comments ."], "relation": "used for", "id": "2022.ltedi-1.45", "year": 2022, "rel_sent": "Later , we used machine learning methods to train and detect the signs of depression .", "forward": false, "src_ids": "2022.ltedi-1.45_1286"}
+{"input": "dv - ngrams - cosine is done by using Method| context: the current state - of - the - art test accuracy ( 97.42 % ) on the imdb movie reviews dataset was reported by thongtan and phienthrakul ( 2019 ) and achieved by the logistic regression classifier trained on the document vectors using cosine similarity ( dv - ngrams - cosine ) proposed in their paper and the bag - of - n - grams ( bon ) vectors scaled by naive bayesian weights . while large pre - trained transformer - based models have shown sota results across many datasets and tasks , the aforementioned model has not been surpassed by them , despite being much simpler and pre - trained on the imdb dataset only .", "entity": "dv - ngrams - cosine", "output": "sub - sampling scheme", "neg_sample": ["dv - ngrams - cosine is done by using Method", "the current state - of - the - art test accuracy ( 97.42 % ) on the imdb movie reviews dataset was reported by thongtan and phienthrakul ( 2019 ) and achieved by the logistic regression classifier trained on the document vectors using cosine similarity ( dv - ngrams - cosine ) proposed in their paper and the bag - of - n - grams ( bon ) vectors scaled by naive bayesian weights .", "while large pre - trained transformer - based models have shown sota results across many datasets and tasks , the aforementioned model has not been surpassed by them , despite being much simpler and pre - trained on the imdb dataset only ."], "relation": "used for", "id": "2022.insights-1.17", "year": 2022, "rel_sent": "Finally , we introduce a sub - sampling scheme based on Naive Bayesian weights for the training process of the DV - ngrams - cosine , which leads tofaster training and better quality .", "forward": false, "src_ids": "2022.insights-1.17_1287"}
+{"input": "sub - sampling scheme is used for Method| context: while large pre - trained transformer - based models have shown sota results across many datasets and tasks , the aforementioned model has not been surpassed by them , despite being much simpler and pre - trained on the imdb dataset only .", "entity": "sub - sampling scheme", "output": "dv - ngrams - cosine", "neg_sample": ["sub - sampling scheme is used for Method", "while large pre - trained transformer - based models have shown sota results across many datasets and tasks , the aforementioned model has not been surpassed by them , despite being much simpler and pre - trained on the imdb dataset only ."], "relation": "used for", "id": "2022.insights-1.17", "year": 2022, "rel_sent": "Finally , we introduce a sub - sampling scheme based on Naive Bayesian weights for the training process of the DV - ngrams - cosine , which leads tofaster training and better quality .", "forward": true, "src_ids": "2022.insights-1.17_1288"}
+{"input": "spatiotemporal quantity extraction is done by using Method| context: news events are often associated with quantities ( e.g. , the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events .", "entity": "spatiotemporal quantity extraction", "output": "meta - framework", "neg_sample": ["spatiotemporal quantity extraction is done by using Method", "news events are often associated with quantities ( e.g.", ", the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events ."], "relation": "used for", "id": "2022.acl-long.195", "year": 2022, "rel_sent": "A Meta - framework for Spatiotemporal Quantity Extraction from Text.", "forward": false, "src_ids": "2022.acl-long.195_1289"}
+{"input": "meta - framework is done by using Method| context: news events are often associated with quantities ( e.g. , the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events .", "entity": "meta - framework", "output": "meta - framework", "neg_sample": ["meta - framework is done by using Method", "news events are often associated with quantities ( e.g.", ", the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events ."], "relation": "used for", "id": "2022.acl-long.195", "year": 2022, "rel_sent": "We demonstrate the meta - framework in three domains - the COVID-19 pandemic , Black Lives Matter protests , and 2020 California wildfires - to show that the formalism is general and extensible , the crowdsourcing pipeline facilitates fast and high - quality data annotation , and the baseline system can handle spatiotemporal quantity extraction well enough to be practically useful .", "forward": false, "src_ids": "2022.acl-long.195_1290"}
+{"input": "meta - framework is used for Method| context: news events are often associated with quantities ( e.g. , the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events .", "entity": "meta - framework", "output": "meta - framework", "neg_sample": ["meta - framework is used for Method", "news events are often associated with quantities ( e.g.", ", the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events ."], "relation": "used for", "id": "2022.acl-long.195", "year": 2022, "rel_sent": "We demonstrate the meta - framework in three domains - the COVID-19 pandemic , Black Lives Matter protests , and 2020 California wildfires - to show that the formalism is general and extensible , the crowdsourcing pipeline facilitates fast and high - quality data annotation , and the baseline system can handle spatiotemporal quantity extraction well enough to be practically useful .", "forward": true, "src_ids": "2022.acl-long.195_1291"}
+{"input": "meta - framework is used for Task| context: news events are often associated with quantities ( e.g. , the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events .", "entity": "meta - framework", "output": "spatiotemporal quantity extraction", "neg_sample": ["meta - framework is used for Task", "news events are often associated with quantities ( e.g.", ", the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events ."], "relation": "used for", "id": "2022.acl-long.195", "year": 2022, "rel_sent": "A Meta - framework for Spatiotemporal Quantity Extraction from Text.", "forward": true, "src_ids": "2022.acl-long.195_1292"}
+{"input": "noise - based negatives is used for OtherScientificTerm| context: recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations .", "entity": "noise - based negatives", "output": "uniformity of the whole representation space", "neg_sample": ["noise - based negatives is used for OtherScientificTerm", "recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations ."], "relation": "used for", "id": "2022.acl-long.423", "year": 2022, "rel_sent": "Such a way may cause the sampling bias that improper negatives ( false negatives and anisotropy representations ) are used to learn sentence representations , which will hurt the uniformity of the representation space . To address it , we present a new framework DCLR ( Debiased Contrastive Learning of unsupervised sentence Representations ) to alleviate the influence of these improper negatives . In DCLR , we design an instance weighting method to punish false negatives and generate noise - based negatives to guarantee the uniformity of the representation space . Experiments on seven semantic textual similarity tasks show that our approach is more effective than competitive baselines .", "forward": true, "src_ids": "2022.acl-long.423_1293"}
+{"input": "instance weighting method is used for OtherScientificTerm| context: recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations . it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space . however , previous works mostly adopt in - batch negatives or sample from training data at random .", "entity": "instance weighting method", "output": "false negatives", "neg_sample": ["instance weighting method is used for OtherScientificTerm", "recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations .", "it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space .", "however , previous works mostly adopt in - batch negatives or sample from training data at random ."], "relation": "used for", "id": "2022.acl-long.423", "year": 2022, "rel_sent": "Such a way may cause the sampling bias that improper negatives ( false negatives and anisotropy representations ) are used to learn sentence representations , which will hurt the uniformity of the representation space . To address it , we present a new framework DCLR ( Debiased Contrastive Learning of unsupervised sentence Representations ) to alleviate the influence of these improper negatives . In DCLR , we design an instance weighting method to punish false negatives and generate noise - based negatives to guarantee the uniformity of the representation space . Experiments on seven semantic textual similarity tasks show that our approach is more effective than competitive baselines .", "forward": true, "src_ids": "2022.acl-long.423_1294"}
+{"input": "false negatives is done by using Method| context: recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations . it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space . however , previous works mostly adopt in - batch negatives or sample from training data at random .", "entity": "false negatives", "output": "instance weighting method", "neg_sample": ["false negatives is done by using Method", "recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations .", "it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space .", "however , previous works mostly adopt in - batch negatives or sample from training data at random ."], "relation": "used for", "id": "2022.acl-long.423", "year": 2022, "rel_sent": "Such a way may cause the sampling bias that improper negatives ( false negatives and anisotropy representations ) are used to learn sentence representations , which will hurt the uniformity of the representation space . To address it , we present a new framework DCLR ( Debiased Contrastive Learning of unsupervised sentence Representations ) to alleviate the influence of these improper negatives . In DCLR , we design an instance weighting method to punish false negatives and generate noise - based negatives to guarantee the uniformity of the representation space . Experiments on seven semantic textual similarity tasks show that our approach is more effective than competitive baselines .", "forward": false, "src_ids": "2022.acl-long.423_1295"}
+{"input": "noise - based negatives is done by using Method| context: recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations . it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space . however , previous works mostly adopt in - batch negatives or sample from training data at random .", "entity": "noise - based negatives", "output": "instance weighting method", "neg_sample": ["noise - based negatives is done by using Method", "recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations .", "it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space .", "however , previous works mostly adopt in - batch negatives or sample from training data at random ."], "relation": "used for", "id": "2022.acl-long.423", "year": 2022, "rel_sent": "Such a way may cause the sampling bias that improper negatives ( false negatives and anisotropy representations ) are used to learn sentence representations , which will hurt the uniformity of the representation space . To address it , we present a new framework DCLR ( Debiased Contrastive Learning of unsupervised sentence Representations ) to alleviate the influence of these improper negatives . In DCLR , we design an instance weighting method to punish false negatives and generate noise - based negatives to guarantee the uniformity of the representation space . Experiments on seven semantic textual similarity tasks show that our approach is more effective than competitive baselines .", "forward": false, "src_ids": "2022.acl-long.423_1296"}
+{"input": "uniformity of the whole representation space is done by using OtherScientificTerm| context: recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations . it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space . however , previous works mostly adopt in - batch negatives or sample from training data at random .", "entity": "uniformity of the whole representation space", "output": "noise - based negatives", "neg_sample": ["uniformity of the whole representation space is done by using OtherScientificTerm", "recently , contrastive learning has been shown to be effective in improving pre - trained language models ( plm ) to derive high - quality sentence representations .", "it aims to pull close positive examples to enhance the alignment while push apart irrelevant negatives for the uniformity of the whole representation space .", "however , previous works mostly adopt in - batch negatives or sample from training data at random ."], "relation": "used for", "id": "2022.acl-long.423", "year": 2022, "rel_sent": "Such a way may cause the sampling bias that improper negatives ( false negatives and anisotropy representations ) are used to learn sentence representations , which will hurt the uniformity of the representation space . To address it , we present a new framework DCLR ( Debiased Contrastive Learning of unsupervised sentence Representations ) to alleviate the influence of these improper negatives . In DCLR , we design an instance weighting method to punish false negatives and generate noise - based negatives to guarantee the uniformity of the representation space . Experiments on seven semantic textual similarity tasks show that our approach is more effective than competitive baselines .", "forward": false, "src_ids": "2022.acl-long.423_1297"}
+{"input": "expression of emotions is done by using Material| context: while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "expression of emotions", "output": "images", "neg_sample": ["expression of emotions is done by using Material", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "On the Complementarity of Images and Text for the Expression of Emotions in Social Media.", "forward": false, "src_ids": "2022.wassa-1.1_1298"}
+{"input": "images is used for OtherScientificTerm| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "images", "output": "expression of emotions", "neg_sample": ["images is used for OtherScientificTerm", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "On the Complementarity of Images and Text for the Expression of Emotions in Social Media.", "forward": true, "src_ids": "2022.wassa-1.1_1299"}
+{"input": "disgust is done by using Material| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "disgust", "output": "text alone", "neg_sample": ["disgust is done by using Material", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "The emotions of anger and sadness are best predicted with a multimodal model , while text alone is sufficient for disgust , joy , and surprise .", "forward": false, "src_ids": "2022.wassa-1.1_1300"}
+{"input": "joy is done by using Material| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "joy", "output": "text alone", "neg_sample": ["joy is done by using Material", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "The emotions of anger and sadness are best predicted with a multimodal model , while text alone is sufficient for disgust , joy , and surprise .", "forward": false, "src_ids": "2022.wassa-1.1_1301"}
+{"input": "surprise is done by using Material| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "surprise", "output": "text alone", "neg_sample": ["surprise is done by using Material", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "The emotions of anger and sadness are best predicted with a multimodal model , while text alone is sufficient for disgust , joy , and surprise .", "forward": false, "src_ids": "2022.wassa-1.1_1302"}
+{"input": "text alone is used for OtherScientificTerm| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "text alone", "output": "disgust", "neg_sample": ["text alone is used for OtherScientificTerm", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "The emotions of anger and sadness are best predicted with a multimodal model , while text alone is sufficient for disgust , joy , and surprise .", "forward": true, "src_ids": "2022.wassa-1.1_1303"}
+{"input": "events is done by using Method| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "events", "output": "multimodal mod- els", "neg_sample": ["events is done by using Method", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "Stimuli depicted by objects , animals , food , or a person are best predicted by image - only models , while multimodal mod- els are most effective on art , events , memes , places , or screenshots .", "forward": false, "src_ids": "2022.wassa-1.1_1304"}
+{"input": "memes is done by using Method| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "memes", "output": "multimodal mod- els", "neg_sample": ["memes is done by using Method", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "Stimuli depicted by objects , animals , food , or a person are best predicted by image - only models , while multimodal mod- els are most effective on art , events , memes , places , or screenshots .", "forward": false, "src_ids": "2022.wassa-1.1_1305"}
+{"input": "screenshots is done by using Method| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "screenshots", "output": "multimodal mod- els", "neg_sample": ["screenshots is done by using Method", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "Stimuli depicted by objects , animals , food , or a person are best predicted by image - only models , while multimodal mod- els are most effective on art , events , memes , places , or screenshots .", "forward": false, "src_ids": "2022.wassa-1.1_1306"}
+{"input": "multimodal mod- els is used for OtherScientificTerm| context: authors of posts in social media communicate their emotions and what causes them with text and images . while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media .", "entity": "multimodal mod- els", "output": "events", "neg_sample": ["multimodal mod- els is used for OtherScientificTerm", "authors of posts in social media communicate their emotions and what causes them with text and images .", "while there is work on emotion and stimulus detection for each modality separately , it is yet unknown if the modalities contain complementary emotion information in social media ."], "relation": "used for", "id": "2022.wassa-1.1", "year": 2022, "rel_sent": "Stimuli depicted by objects , animals , food , or a person are best predicted by image - only models , while multimodal mod- els are most effective on art , events , memes , places , or screenshots .", "forward": true, "src_ids": "2022.wassa-1.1_1307"}
+{"input": "iterative text revisions is done by using Method| context: writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process . a crucial part of writing is editing and revising the text . previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles .", "entity": "iterative text revisions", "output": "iterater", "neg_sample": ["iterative text revisions is done by using Method", "writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process .", "a crucial part of writing is editing and revising the text .", "previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles ."], "relation": "used for", "id": "2022.acl-long.250", "year": 2022, "rel_sent": "In particular , IteraTeR is collected based on a new framework to comprehensively model the iterative text revisions that generalizes to a variety of domains , edit intentions , revision depths , and granularities .", "forward": false, "src_ids": "2022.acl-long.250_1308"}
+{"input": "iterater is used for Task| context: writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process . a crucial part of writing is editing and revising the text . previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles .", "entity": "iterater", "output": "iterative text revisions", "neg_sample": ["iterater is used for Task", "writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process .", "a crucial part of writing is editing and revising the text .", "previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles ."], "relation": "used for", "id": "2022.acl-long.250", "year": 2022, "rel_sent": "In particular , IteraTeR is collected based on a new framework to comprehensively model the iterative text revisions that generalizes to a variety of domains , edit intentions , revision depths , and granularities .", "forward": true, "src_ids": "2022.acl-long.250_1309"}
+{"input": "generative and action - based text revision models is done by using OtherScientificTerm| context: writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process . a crucial part of writing is editing and revising the text . previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles .", "entity": "generative and action - based text revision models", "output": "annotated edit intentions", "neg_sample": ["generative and action - based text revision models is done by using OtherScientificTerm", "writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process .", "a crucial part of writing is editing and revising the text .", "previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles ."], "relation": "used for", "id": "2022.acl-long.250", "year": 2022, "rel_sent": "When we incorporate our annotated edit intentions , both generative and action - based text revision models significantly improve automatic evaluations .", "forward": false, "src_ids": "2022.acl-long.250_1310"}
+{"input": "annotated edit intentions is used for Method| context: writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process . a crucial part of writing is editing and revising the text . previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles .", "entity": "annotated edit intentions", "output": "generative and action - based text revision models", "neg_sample": ["annotated edit intentions is used for Method", "writing is , by nature , a strategic , adaptive , and , more importantly , an iterative process .", "a crucial part of writing is editing and revising the text .", "previous works on text revision have focused on defining edit intention taxonomies within a single domain or developing computational models with a single level of edit granularity , such as sentence - level edits , which differ from human 's revision cycles ."], "relation": "used for", "id": "2022.acl-long.250", "year": 2022, "rel_sent": "When we incorporate our annotated edit intentions , both generative and action - based text revision models significantly improve automatic evaluations .", "forward": true, "src_ids": "2022.acl-long.250_1311"}
+{"input": "unintended bias mitigation is done by using Method| context: natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability . e.g. , neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance . most mitigation techniques use lists of identity terms or samples from the target domain during training .", "entity": "unintended bias mitigation", "output": "entropy - based attention regularization", "neg_sample": ["unintended bias mitigation is done by using Method", "natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability .", "e.g.", ", neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance .", "most mitigation techniques use lists of identity terms or samples from the target domain during training ."], "relation": "used for", "id": "2022.findings-acl.88", "year": 2022, "rel_sent": "Entropy - based Attention Regularization Frees Unintended Bias Mitigation from Lists.", "forward": false, "src_ids": "2022.findings-acl.88_1312"}
+{"input": "overfitting terms is done by using Method| context: natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability . e.g. , neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance . most mitigation techniques use lists of identity terms or samples from the target domain during training .", "entity": "overfitting terms", "output": "entropy - based attention regularization", "neg_sample": ["overfitting terms is done by using Method", "natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability .", "e.g.", ", neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance .", "most mitigation techniques use lists of identity terms or samples from the target domain during training ."], "relation": "used for", "id": "2022.findings-acl.88", "year": 2022, "rel_sent": "An additional objective function penalizes tokens with low self - attention entropy . We fine - tune BERT via EAR : the resulting model matches or exceeds state - of - the - art performance for hate speech classification and bias metrics on three benchmark corpora in English and Italian . EAR also reveals overfitting terms , i.e. , terms most likely to induce bias , to help identify their effect on the model , task , and predictions .", "forward": false, "src_ids": "2022.findings-acl.88_1313"}
+{"input": "entropy - based attention regularization is used for Task| context: natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability . e.g. , neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance . most mitigation techniques use lists of identity terms or samples from the target domain during training .", "entity": "entropy - based attention regularization", "output": "unintended bias mitigation", "neg_sample": ["entropy - based attention regularization is used for Task", "natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability .", "e.g.", ", neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance .", "most mitigation techniques use lists of identity terms or samples from the target domain during training ."], "relation": "used for", "id": "2022.findings-acl.88", "year": 2022, "rel_sent": "Entropy - based Attention Regularization Frees Unintended Bias Mitigation from Lists.", "forward": true, "src_ids": "2022.findings-acl.88_1314"}
+{"input": "entropy - based attention regularization is used for OtherScientificTerm| context: natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability . e.g. , neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance . most mitigation techniques use lists of identity terms or samples from the target domain during training .", "entity": "entropy - based attention regularization", "output": "overfitting terms", "neg_sample": ["entropy - based attention regularization is used for OtherScientificTerm", "natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability .", "e.g.", ", neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance .", "most mitigation techniques use lists of identity terms or samples from the target domain during training ."], "relation": "used for", "id": "2022.findings-acl.88", "year": 2022, "rel_sent": "An additional objective function penalizes tokens with low self - attention entropy . We fine - tune BERT via EAR : the resulting model matches or exceeds state - of - the - art performance for hate speech classification and bias metrics on three benchmark corpora in English and Italian . EAR also reveals overfitting terms , i.e. , terms most likely to induce bias , to help identify their effect on the model , task , and predictions .", "forward": true, "src_ids": "2022.findings-acl.88_1315"}
+{"input": "neural language models is used for Method| context: conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system . moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously .", "entity": "neural language models", "output": "rots", "neg_sample": ["neural language models is used for Method", "conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system .", "moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously ."], "relation": "used for", "id": "2022.acl-long.261", "year": 2022, "rel_sent": "Most importantly , we show that current neural language models can automatically generate new RoTs that reasonably describe previously unseen interactions , but they still struggle with certain scenarios .", "forward": true, "src_ids": "2022.acl-long.261_1316"}
+{"input": "integrity of conversational agents is done by using Method| context: conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system . moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously .", "entity": "integrity of conversational agents", "output": "mic", "neg_sample": ["integrity of conversational agents is done by using Method", "conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system .", "moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously ."], "relation": "used for", "id": "2022.acl-long.261", "year": 2022, "rel_sent": "Our findings suggest that MIC will be a useful resource for understanding and language models ' implicit moral assumptions and flexibly benchmarking the integrity of conversational agents .", "forward": false, "src_ids": "2022.acl-long.261_1317"}
+{"input": "rots is done by using Method| context: conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system . moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously .", "entity": "rots", "output": "neural language models", "neg_sample": ["rots is done by using Method", "conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system .", "moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously ."], "relation": "used for", "id": "2022.acl-long.261", "year": 2022, "rel_sent": "Most importantly , we show that current neural language models can automatically generate new RoTs that reasonably describe previously unseen interactions , but they still struggle with certain scenarios .", "forward": false, "src_ids": "2022.acl-long.261_1318"}
+{"input": "mic is used for Task| context: conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system . moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously .", "entity": "mic", "output": "integrity of conversational agents", "neg_sample": ["mic is used for Task", "conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system .", "moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously ."], "relation": "used for", "id": "2022.acl-long.261", "year": 2022, "rel_sent": "Our findings suggest that MIC will be a useful resource for understanding and language models ' implicit moral assumptions and flexibly benchmarking the integrity of conversational agents .", "forward": true, "src_ids": "2022.acl-long.261_1319"}
+{"input": "temporal knowledge graph reasoning is done by using Method| context: a temporal knowledge graph ( tkg ) is a sequence of kgs corresponding to different timestamps . the evolutional patterns are complex in two aspects , length - diversity and time - variability . existing models for tkg reasoning focus on modeling fact sequences of a fixed length , which can not discover complex evolutional patterns that vary in length .", "entity": "temporal knowledge graph reasoning", "output": "complex evolutional pattern learning", "neg_sample": ["temporal knowledge graph reasoning is done by using Method", "a temporal knowledge graph ( tkg ) is a sequence of kgs corresponding to different timestamps .", "the evolutional patterns are complex in two aspects , length - diversity and time - variability .", "existing models for tkg reasoning focus on modeling fact sequences of a fixed length , which can not discover complex evolutional patterns that vary in length ."], "relation": "used for", "id": "2022.acl-short.32", "year": 2022, "rel_sent": "Complex Evolutional Pattern Learning for Temporal Knowledge Graph Reasoning.", "forward": false, "src_ids": "2022.acl-short.32_1320"}
+{"input": "complex evolutional pattern learning is used for Task| context: a temporal knowledge graph ( tkg ) is a sequence of kgs corresponding to different timestamps . the evolutional patterns are complex in two aspects , length - diversity and time - variability . existing models for tkg reasoning focus on modeling fact sequences of a fixed length , which can not discover complex evolutional patterns that vary in length .", "entity": "complex evolutional pattern learning", "output": "temporal knowledge graph reasoning", "neg_sample": ["complex evolutional pattern learning is used for Task", "a temporal knowledge graph ( tkg ) is a sequence of kgs corresponding to different timestamps .", "the evolutional patterns are complex in two aspects , length - diversity and time - variability .", "existing models for tkg reasoning focus on modeling fact sequences of a fixed length , which can not discover complex evolutional patterns that vary in length ."], "relation": "used for", "id": "2022.acl-short.32", "year": 2022, "rel_sent": "Complex Evolutional Pattern Learning for Temporal Knowledge Graph Reasoning.", "forward": true, "src_ids": "2022.acl-short.32_1321"}
+{"input": "linguistic uncertainty is done by using Method| context: recent advances in human language processing research have suggested that the predictive power of large language models ( llms ) can serve as cognitive models of human language processing . evidence for this comes from llms ' close fit to human psychophysical data , such as reaction times or brain responses in language comprehension experiments . those adopting llm architectures as models of human language processing frame the problem of language comprehension as prediction of the next linguistic event ( goodkind and bicknell , 2018 ; eisape et al . , 2020 ) , in particular focusing on lexical or syntactic surprisal . however , this approach fails to consider that comprehenders make predictions using some representation of the content of an utterance . that is , in contrast to surprisal , readers make use of a mental model that creates an evolving understanding of who is doing what to whom and how . in contrast to comprehenders , surprisal measures do not make predictions about the content , as surprisal simply measures the conditional probability of some linguistic event given the surrounding context . many convergent cues in the upstream context , such as the frequencies of words in a sentence sofar , will affect hidden state representations of models , which may then influence the predictability of upcoming words .", "entity": "linguistic uncertainty", "output": "masked language models", "neg_sample": ["linguistic uncertainty is done by using Method", "recent advances in human language processing research have suggested that the predictive power of large language models ( llms ) can serve as cognitive models of human language processing .", "evidence for this comes from llms ' close fit to human psychophysical data , such as reaction times or brain responses in language comprehension experiments .", "those adopting llm architectures as models of human language processing frame the problem of language comprehension as prediction of the next linguistic event ( goodkind and bicknell , 2018 ; eisape et al .", ", 2020 ) , in particular focusing on lexical or syntactic surprisal .", "however , this approach fails to consider that comprehenders make predictions using some representation of the content of an utterance .", "that is , in contrast to surprisal , readers make use of a mental model that creates an evolving understanding of who is doing what to whom and how .", "in contrast to comprehenders , surprisal measures do not make predictions about the content , as surprisal simply measures the conditional probability of some linguistic event given the surrounding context .", "many convergent cues in the upstream context , such as the frequencies of words in a sentence sofar , will affect hidden state representations of models , which may then influence the predictability of upcoming words ."], "relation": "used for", "id": "2022.scil-1.22", "year": 2022, "rel_sent": "Masked language models directly encode linguistic uncertainty.", "forward": false, "src_ids": "2022.scil-1.22_1322"}
+{"input": "masked language models is used for OtherScientificTerm| context: recent advances in human language processing research have suggested that the predictive power of large language models ( llms ) can serve as cognitive models of human language processing . evidence for this comes from llms ' close fit to human psychophysical data , such as reaction times or brain responses in language comprehension experiments . those adopting llm architectures as models of human language processing frame the problem of language comprehension as prediction of the next linguistic event ( goodkind and bicknell , 2018 ; eisape et al . , 2020 ) , in particular focusing on lexical or syntactic surprisal . however , this approach fails to consider that comprehenders make predictions using some representation of the content of an utterance . that is , in contrast to surprisal , readers make use of a mental model that creates an evolving understanding of who is doing what to whom and how . in contrast to comprehenders , surprisal measures do not make predictions about the content , as surprisal simply measures the conditional probability of some linguistic event given the surrounding context . many convergent cues in the upstream context , such as the frequencies of words in a sentence sofar , will affect hidden state representations of models , which may then influence the predictability of upcoming words .", "entity": "masked language models", "output": "linguistic uncertainty", "neg_sample": ["masked language models is used for OtherScientificTerm", "recent advances in human language processing research have suggested that the predictive power of large language models ( llms ) can serve as cognitive models of human language processing .", "evidence for this comes from llms ' close fit to human psychophysical data , such as reaction times or brain responses in language comprehension experiments .", "those adopting llm architectures as models of human language processing frame the problem of language comprehension as prediction of the next linguistic event ( goodkind and bicknell , 2018 ; eisape et al .", ", 2020 ) , in particular focusing on lexical or syntactic surprisal .", "however , this approach fails to consider that comprehenders make predictions using some representation of the content of an utterance .", "that is , in contrast to surprisal , readers make use of a mental model that creates an evolving understanding of who is doing what to whom and how .", "in contrast to comprehenders , surprisal measures do not make predictions about the content , as surprisal simply measures the conditional probability of some linguistic event given the surrounding context .", "many convergent cues in the upstream context , such as the frequencies of words in a sentence sofar , will affect hidden state representations of models , which may then influence the predictability of upcoming words ."], "relation": "used for", "id": "2022.scil-1.22", "year": 2022, "rel_sent": "Masked language models directly encode linguistic uncertainty.", "forward": true, "src_ids": "2022.scil-1.22_1323"}
+{"input": "etransla- tion services is done by using Material| context: the work in progress on the cef action curlicat is presented .", "entity": "etransla- tion services", "output": "curated datasets", "neg_sample": ["etransla- tion services is done by using Material", "the work in progress on the cef action curlicat is presented ."], "relation": "used for", "id": "2022.eamt-1.60", "year": 2022, "rel_sent": "The general aim of the Action is to compile curated datasets in seven languages of the con- sortium in domains of relevance to Euro- pean Digital Service Infrastructures ( DSIs ) in order to enhance the eTransla- tion services .", "forward": false, "src_ids": "2022.eamt-1.60_1324"}
+{"input": "curated datasets is used for Task| context: the work in progress on the cef action curlicat is presented .", "entity": "curated datasets", "output": "etransla- tion services", "neg_sample": ["curated datasets is used for Task", "the work in progress on the cef action curlicat is presented ."], "relation": "used for", "id": "2022.eamt-1.60", "year": 2022, "rel_sent": "The general aim of the Action is to compile curated datasets in seven languages of the con- sortium in domains of relevance to Euro- pean Digital Service Infrastructures ( DSIs ) in order to enhance the eTransla- tion services .", "forward": true, "src_ids": "2022.eamt-1.60_1325"}
+{"input": "continuous - output neural machine translation is done by using Task| context: continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation . however , continuous models for text generation have received limited attention from the community .", "entity": "continuous - output neural machine translation", "output": "target representation", "neg_sample": ["continuous - output neural machine translation is done by using Task", "continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation .", "however , continuous models for text generation have received limited attention from the community ."], "relation": "used for", "id": "2022.repl4nlp-1.24", "year": 2022, "rel_sent": "On Target Representation in Continuous - output Neural Machine Translation.", "forward": false, "src_ids": "2022.repl4nlp-1.24_1326"}
+{"input": "target representation is used for Task| context: continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation . however , continuous models for text generation have received limited attention from the community .", "entity": "target representation", "output": "continuous - output neural machine translation", "neg_sample": ["target representation is used for Task", "continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation .", "however , continuous models for text generation have received limited attention from the community ."], "relation": "used for", "id": "2022.repl4nlp-1.24", "year": 2022, "rel_sent": "On Target Representation in Continuous - output Neural Machine Translation.", "forward": true, "src_ids": "2022.repl4nlp-1.24_1327"}
+{"input": "transfer is used for Method| context: continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation .", "entity": "transfer", "output": "continuous models", "neg_sample": ["transfer is used for Method", "continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation ."], "relation": "used for", "id": "2022.repl4nlp-1.24", "year": 2022, "rel_sent": "Our results on the WMT Romanian - English and English - Turkish benchmarks show such transfer leads to the best - performing continuous model .", "forward": true, "src_ids": "2022.repl4nlp-1.24_1328"}
+{"input": "neural machine translation ( nmt ) is done by using Method| context: continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation . however , continuous models for text generation have received limited attention from the community .", "entity": "neural machine translation ( nmt )", "output": "transformers", "neg_sample": ["neural machine translation ( nmt ) is done by using Method", "continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation .", "however , continuous models for text generation have received limited attention from the community ."], "relation": "used for", "id": "2022.repl4nlp-1.24", "year": 2022, "rel_sent": "In this work , we study continuous text generation using Transformers for neural machine translation ( NMT ) .", "forward": false, "src_ids": "2022.repl4nlp-1.24_1329"}
+{"input": "transformers is used for Task| context: continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation . however , continuous models for text generation have received limited attention from the community .", "entity": "transformers", "output": "neural machine translation ( nmt )", "neg_sample": ["transformers is used for Task", "continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation .", "however , continuous models for text generation have received limited attention from the community ."], "relation": "used for", "id": "2022.repl4nlp-1.24", "year": 2022, "rel_sent": "In this work , we study continuous text generation using Transformers for neural machine translation ( NMT ) .", "forward": true, "src_ids": "2022.repl4nlp-1.24_1330"}
+{"input": "continuous models is done by using Task| context: continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation . however , continuous models for text generation have received limited attention from the community .", "entity": "continuous models", "output": "transfer", "neg_sample": ["continuous models is done by using Task", "continuous generative models proved their usefulness in high - dimensional data , such as image and audio generation .", "however , continuous models for text generation have received limited attention from the community ."], "relation": "used for", "id": "2022.repl4nlp-1.24", "year": 2022, "rel_sent": "Our results on the WMT Romanian - English and English - Turkish benchmarks show such transfer leads to the best - performing continuous model .", "forward": false, "src_ids": "2022.repl4nlp-1.24_1331"}
+{"input": "causal facts is done by using Task| context: understanding causality has vital importance for various natural language processing ( nlp ) applications . beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process . however , such explanation information still remains absent in existing causal reasoning resources .", "entity": "causal facts", "output": "generating valid explanations", "neg_sample": ["causal facts is done by using Task", "understanding causality has vital importance for various natural language processing ( nlp ) applications .", "beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process .", "however , such explanation information still remains absent in existing causal reasoning resources ."], "relation": "used for", "id": "2022.acl-long.33", "year": 2022, "rel_sent": "Experimental results show that generating valid explanations for causal facts still remains especially challenging for the state - of - the - art models , and the explanation information can be helpful for promoting the accuracy and stability of causal reasoning models .", "forward": false, "src_ids": "2022.acl-long.33_1332"}
+{"input": "generating valid explanations is used for OtherScientificTerm| context: understanding causality has vital importance for various natural language processing ( nlp ) applications . beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process . however , such explanation information still remains absent in existing causal reasoning resources .", "entity": "generating valid explanations", "output": "causal facts", "neg_sample": ["generating valid explanations is used for OtherScientificTerm", "understanding causality has vital importance for various natural language processing ( nlp ) applications .", "beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process .", "however , such explanation information still remains absent in existing causal reasoning resources ."], "relation": "used for", "id": "2022.acl-long.33", "year": 2022, "rel_sent": "Experimental results show that generating valid explanations for causal facts still remains especially challenging for the state - of - the - art models , and the explanation information can be helpful for promoting the accuracy and stability of causal reasoning models .", "forward": true, "src_ids": "2022.acl-long.33_1333"}
+{"input": "explanation information is used for Method| context: understanding causality has vital importance for various natural language processing ( nlp ) applications . beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process . however , such explanation information still remains absent in existing causal reasoning resources .", "entity": "explanation information", "output": "causal reasoning models", "neg_sample": ["explanation information is used for Method", "understanding causality has vital importance for various natural language processing ( nlp ) applications .", "beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process .", "however , such explanation information still remains absent in existing causal reasoning resources ."], "relation": "used for", "id": "2022.acl-long.33", "year": 2022, "rel_sent": "Experimental results show that generating valid explanations for causal facts still remains especially challenging for the state - of - the - art models , and the explanation information can be helpful for promoting the accuracy and stability of causal reasoning models .", "forward": true, "src_ids": "2022.acl-long.33_1334"}
+{"input": "causal reasoning models is done by using OtherScientificTerm| context: understanding causality has vital importance for various natural language processing ( nlp ) applications . beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process .", "entity": "causal reasoning models", "output": "explanation information", "neg_sample": ["causal reasoning models is done by using OtherScientificTerm", "understanding causality has vital importance for various natural language processing ( nlp ) applications .", "beyond the labeled instances , conceptual explanations of the causality can provide deep understanding of the causal fact tofacilitate the causal reasoning process ."], "relation": "used for", "id": "2022.acl-long.33", "year": 2022, "rel_sent": "Experimental results show that generating valid explanations for causal facts still remains especially challenging for the state - of - the - art models , and the explanation information can be helpful for promoting the accuracy and stability of causal reasoning models .", "forward": false, "src_ids": "2022.acl-long.33_1335"}
+{"input": "vision - and - language inference is done by using Method| context: analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms . while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored . in this paper , we present sdro , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference . experiments on benchmark datasets with images ( nlvr^2 ) and video ( violin ) demonstrate performance improvements as well as robustness to adversarial attacks . experiments on binary vqa explore the generalizability of this method to other v&l tasks .", "entity": "vision - and - language inference", "output": "semantically distributed robust optimization", "neg_sample": ["vision - and - language inference is done by using Method", "analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms .", "while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored .", "in this paper , we present sdro , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference .", "experiments on benchmark datasets with images ( nlvr^2 ) and video ( violin ) demonstrate performance improvements as well as robustness to adversarial attacks .", "experiments on binary vqa explore the generalizability of this method to other v&l tasks ."], "relation": "used for", "id": "2022.findings-acl.118", "year": 2022, "rel_sent": "Semantically Distributed Robust Optimization for Vision - and - Language Inference.", "forward": false, "src_ids": "2022.findings-acl.118_1336"}
+{"input": "semantically distributed robust optimization is used for Task| context: analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms . while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored . in this paper , we present sdro , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference . experiments on benchmark datasets with images ( nlvr^2 ) and video ( violin ) demonstrate performance improvements as well as robustness to adversarial attacks . experiments on binary vqa explore the generalizability of this method to other v&l tasks .", "entity": "semantically distributed robust optimization", "output": "vision - and - language inference", "neg_sample": ["semantically distributed robust optimization is used for Task", "analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms .", "while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored .", "in this paper , we present sdro , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference .", "experiments on benchmark datasets with images ( nlvr^2 ) and video ( violin ) demonstrate performance improvements as well as robustness to adversarial attacks .", "experiments on binary vqa explore the generalizability of this method to other v&l tasks ."], "relation": "used for", "id": "2022.findings-acl.118", "year": 2022, "rel_sent": "Semantically Distributed Robust Optimization for Vision - and - Language Inference.", "forward": true, "src_ids": "2022.findings-acl.118_1337"}
+{"input": "multidoc2dial is done by using Task| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "multidoc2dial", "output": "refined retriever - reader pipeline", "neg_sample": ["multidoc2dial is done by using Task", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "R3 : Refined Retriever - Reader pipeline for Multidoc2dial.", "forward": false, "src_ids": "2022.dialdoc-1.17_1338"}
+{"input": "refined retriever - reader pipeline is used for Task| context: the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "refined retriever - reader pipeline", "output": "multidoc2dial", "neg_sample": ["refined retriever - reader pipeline is used for Task", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "R3 : Refined Retriever - Reader pipeline for Multidoc2dial.", "forward": true, "src_ids": "2022.dialdoc-1.17_1339"}
+{"input": "goal - oriented dialogues is done by using Method| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "goal - oriented dialogues", "output": "baseline 's retriever - reader architecture", "neg_sample": ["goal - oriented dialogues is done by using Method", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "We propose several improvements over the baseline 's retriever - reader architecture to aid in modeling goal - oriented dialogues grounded in multiple documents .", "forward": false, "src_ids": "2022.dialdoc-1.17_1340"}
+{"input": "baseline 's retriever - reader architecture is used for Task| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "baseline 's retriever - reader architecture", "output": "goal - oriented dialogues", "neg_sample": ["baseline 's retriever - reader architecture is used for Task", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "We propose several improvements over the baseline 's retriever - reader architecture to aid in modeling goal - oriented dialogues grounded in multiple documents .", "forward": true, "src_ids": "2022.dialdoc-1.17_1341"}
+{"input": "passage retrieval is done by using Method| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "passage retrieval", "output": "sparse representations", "neg_sample": ["passage retrieval is done by using Method", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "Our proposed approach employs sparse representations for passage retrieval , a passage re - ranker , the fusion - in - decoder architecture for generation , and a curriculum learning training paradigm .", "forward": false, "src_ids": "2022.dialdoc-1.17_1342"}
+{"input": "sparse representations is used for Task| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "sparse representations", "output": "passage retrieval", "neg_sample": ["sparse representations is used for Task", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "Our proposed approach employs sparse representations for passage retrieval , a passage re - ranker , the fusion - in - decoder architecture for generation , and a curriculum learning training paradigm .", "forward": true, "src_ids": "2022.dialdoc-1.17_1343"}
+{"input": "generation is done by using Method| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "generation", "output": "fusion - in - decoder architecture", "neg_sample": ["generation is done by using Method", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "Our proposed approach employs sparse representations for passage retrieval , a passage re - ranker , the fusion - in - decoder architecture for generation , and a curriculum learning training paradigm .", "forward": false, "src_ids": "2022.dialdoc-1.17_1344"}
+{"input": "fusion - in - decoder architecture is used for Task| context: multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents . the task involves grounding a user 's query in a document followed by generating an appropriate response .", "entity": "fusion - in - decoder architecture", "output": "generation", "neg_sample": ["fusion - in - decoder architecture is used for Task", "multidoc2dial is a conversational question answering dataset that grounds dialogues in multiple documents .", "the task involves grounding a user 's query in a document followed by generating an appropriate response ."], "relation": "used for", "id": "2022.dialdoc-1.17", "year": 2022, "rel_sent": "Our proposed approach employs sparse representations for passage retrieval , a passage re - ranker , the fusion - in - decoder architecture for generation , and a curriculum learning training paradigm .", "forward": true, "src_ids": "2022.dialdoc-1.17_1345"}
+{"input": "siamese networks is done by using Method| context: we study the problem of building text classifiers with little or no training data , commonly known as zero and few - shot text classification . in recent years , an approach based on neural textual entailment models has been found to give strong results on a diverse range of tasks .", "entity": "siamese networks", "output": "pre - training", "neg_sample": ["siamese networks is done by using Method", "we study the problem of building text classifiers with little or no training data , commonly known as zero and few - shot text classification .", "in recent years , an approach based on neural textual entailment models has been found to give strong results on a diverse range of tasks ."], "relation": "used for", "id": "2022.acl-long.584", "year": 2022, "rel_sent": "In this work , we show that with proper pre - training , Siamese Networks that embed texts and labels offer a competitive alternative .", "forward": false, "src_ids": "2022.acl-long.584_1346"}
+{"input": "pre - training is used for Method| context: we study the problem of building text classifiers with little or no training data , commonly known as zero and few - shot text classification . in recent years , an approach based on neural textual entailment models has been found to give strong results on a diverse range of tasks .", "entity": "pre - training", "output": "siamese networks", "neg_sample": ["pre - training is used for Method", "we study the problem of building text classifiers with little or no training data , commonly known as zero and few - shot text classification .", "in recent years , an approach based on neural textual entailment models has been found to give strong results on a diverse range of tasks ."], "relation": "used for", "id": "2022.acl-long.584", "year": 2022, "rel_sent": "In this work , we show that with proper pre - training , Siamese Networks that embed texts and labels offer a competitive alternative .", "forward": true, "src_ids": "2022.acl-long.584_1347"}
+{"input": "word importance is done by using OtherScientificTerm| context: deaf and hard of hearing individuals regularly rely on captioning while watching live tv . live tv captioning is evaluated by regulatory agencies using various caption evaluation metrics . however , caption evaluation metrics are often not informed by preferences of dhh users or how meaningful the captions are . there is a need to construct caption evaluation metrics that take the relative importance of words in transcript into account .", "entity": "word importance", "output": "bert embeddings", "neg_sample": ["word importance is done by using OtherScientificTerm", "deaf and hard of hearing individuals regularly rely on captioning while watching live tv .", "live tv captioning is evaluated by regulatory agencies using various caption evaluation metrics .", "however , caption evaluation metrics are often not informed by preferences of dhh users or how meaningful the captions are .", "there is a need to construct caption evaluation metrics that take the relative importance of words in transcript into account ."], "relation": "used for", "id": "2022.ltedi-1.5", "year": 2022, "rel_sent": "Using BERT Embeddings to Model Word Importance in Conversational Transcripts for Deaf and Hard of Hearing Users.", "forward": false, "src_ids": "2022.ltedi-1.5_1348"}
+{"input": "bert embeddings is used for OtherScientificTerm| context: deaf and hard of hearing individuals regularly rely on captioning while watching live tv . live tv captioning is evaluated by regulatory agencies using various caption evaluation metrics . however , caption evaluation metrics are often not informed by preferences of dhh users or how meaningful the captions are . there is a need to construct caption evaluation metrics that take the relative importance of words in transcript into account .", "entity": "bert embeddings", "output": "word importance", "neg_sample": ["bert embeddings is used for OtherScientificTerm", "deaf and hard of hearing individuals regularly rely on captioning while watching live tv .", "live tv captioning is evaluated by regulatory agencies using various caption evaluation metrics .", "however , caption evaluation metrics are often not informed by preferences of dhh users or how meaningful the captions are .", "there is a need to construct caption evaluation metrics that take the relative importance of words in transcript into account ."], "relation": "used for", "id": "2022.ltedi-1.5", "year": 2022, "rel_sent": "Using BERT Embeddings to Model Word Importance in Conversational Transcripts for Deaf and Hard of Hearing Users.", "forward": true, "src_ids": "2022.ltedi-1.5_1349"}
+{"input": "low - resource neural machine translation is done by using OtherScientificTerm| context: we explore the roles and interactions of the hyper - parameters governing regularization , and propose a range of values applicable to low - resource neural machine translation . we demonstrate that default or recommended values for high - resource settings are not optimal for low - resource ones , and that more aggressive regularization is needed when resources are scarce , in proportion to their scarcity .", "entity": "low - resource neural machine translation", "output": "regularization factors", "neg_sample": ["low - resource neural machine translation is done by using OtherScientificTerm", "we explore the roles and interactions of the hyper - parameters governing regularization , and propose a range of values applicable to low - resource neural machine translation .", "we demonstrate that default or recommended values for high - resource settings are not optimal for low - resource ones , and that more aggressive regularization is needed when resources are scarce , in proportion to their scarcity ."], "relation": "used for", "id": "2022.eamt-1.14", "year": 2022, "rel_sent": "On the Interaction of Regularization Factors in Low - resource Neural Machine Translation.", "forward": false, "src_ids": "2022.eamt-1.14_1350"}
+{"input": "regularization factors is used for Task| context: we demonstrate that default or recommended values for high - resource settings are not optimal for low - resource ones , and that more aggressive regularization is needed when resources are scarce , in proportion to their scarcity .", "entity": "regularization factors", "output": "low - resource neural machine translation", "neg_sample": ["regularization factors is used for Task", "we demonstrate that default or recommended values for high - resource settings are not optimal for low - resource ones , and that more aggressive regularization is needed when resources are scarce , in proportion to their scarcity ."], "relation": "used for", "id": "2022.eamt-1.14", "year": 2022, "rel_sent": "On the Interaction of Regularization Factors in Low - resource Neural Machine Translation.", "forward": true, "src_ids": "2022.eamt-1.14_1351"}
+{"input": "fine - tuning is done by using Method| context: we present a comprehensive study of sparse attention patterns in transformer models .", "entity": "fine - tuning", "output": "adaptive axis attention method", "neg_sample": ["fine - tuning is done by using Method", "we present a comprehensive study of sparse attention patterns in transformer models ."], "relation": "used for", "id": "2022.findings-acl.74", "year": 2022, "rel_sent": "Learning Adaptive Axis Attentions in Fine - tuning : Beyond Fixed Sparse Attention Patterns.", "forward": false, "src_ids": "2022.findings-acl.74_1352"}
+{"input": "adaptive axis attention method is used for Method| context: we present a comprehensive study of sparse attention patterns in transformer models .", "entity": "adaptive axis attention method", "output": "fine - tuning", "neg_sample": ["adaptive axis attention method is used for Method", "we present a comprehensive study of sparse attention patterns in transformer models ."], "relation": "used for", "id": "2022.findings-acl.74", "year": 2022, "rel_sent": "Learning Adaptive Axis Attentions in Fine - tuning : Beyond Fixed Sparse Attention Patterns.", "forward": true, "src_ids": "2022.findings-acl.74_1353"}
+{"input": "attention patterns is used for OtherScientificTerm| context: we present a comprehensive study of sparse attention patterns in transformer models .", "entity": "attention patterns", "output": "transformer layer", "neg_sample": ["attention patterns is used for OtherScientificTerm", "we present a comprehensive study of sparse attention patterns in transformer models ."], "relation": "used for", "id": "2022.findings-acl.74", "year": 2022, "rel_sent": "Drawing on this insight , we propose a novel Adaptive Axis Attention method , which learns - during fine - tuning - different attention patterns for each Transformer layer depending on the downstream task .", "forward": true, "src_ids": "2022.findings-acl.74_1354"}
+{"input": "sparse patterns is done by using Method| context: we present a comprehensive study of sparse attention patterns in transformer models .", "entity": "sparse patterns", "output": "pre - training", "neg_sample": ["sparse patterns is done by using Method", "we present a comprehensive study of sparse attention patterns in transformer models ."], "relation": "used for", "id": "2022.findings-acl.74", "year": 2022, "rel_sent": "It does not require pre - training to accommodate the sparse patterns and demonstrates competitive and sometimes better performance against fixed sparse attention patterns that require resource - intensive pre - training .", "forward": false, "src_ids": "2022.findings-acl.74_1355"}
+{"input": "pre - training is used for OtherScientificTerm| context: we present a comprehensive study of sparse attention patterns in transformer models .", "entity": "pre - training", "output": "sparse patterns", "neg_sample": ["pre - training is used for OtherScientificTerm", "we present a comprehensive study of sparse attention patterns in transformer models ."], "relation": "used for", "id": "2022.findings-acl.74", "year": 2022, "rel_sent": "It does not require pre - training to accommodate the sparse patterns and demonstrates competitive and sometimes better performance against fixed sparse attention patterns that require resource - intensive pre - training .", "forward": true, "src_ids": "2022.findings-acl.74_1356"}
+{"input": "commonsense reasoning is done by using Method| context: it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models .", "entity": "commonsense reasoning", "output": "generated knowledge prompting", "neg_sample": ["commonsense reasoning is done by using Method", "it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models ."], "relation": "used for", "id": "2022.acl-long.225", "year": 2022, "rel_sent": "Generated Knowledge Prompting for Commonsense Reasoning.", "forward": false, "src_ids": "2022.acl-long.225_1357"}
+{"input": "knowledge integration is done by using OtherScientificTerm| context: it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models .", "entity": "knowledge integration", "output": "task - specific supervision", "neg_sample": ["knowledge integration is done by using OtherScientificTerm", "it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models ."], "relation": "used for", "id": "2022.acl-long.225", "year": 2022, "rel_sent": "Our method does not require task - specific supervision for knowledge integration , or access to a structured knowledge base , yet it improves performance of large - scale , state - of - the - art models on four commonsense reasoning tasks , achieving state - of - the - art results on numerical commonsense ( NumerSense ) , general commonsense ( CommonsenseQA 2.0 ) , and scientific commonsense ( QASC ) benchmarks .", "forward": false, "src_ids": "2022.acl-long.225_1358"}
+{"input": "task - specific supervision is used for Task| context: it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models .", "entity": "task - specific supervision", "output": "knowledge integration", "neg_sample": ["task - specific supervision is used for Task", "it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models ."], "relation": "used for", "id": "2022.acl-long.225", "year": 2022, "rel_sent": "Our method does not require task - specific supervision for knowledge integration , or access to a structured knowledge base , yet it improves performance of large - scale , state - of - the - art models on four commonsense reasoning tasks , achieving state - of - the - art results on numerical commonsense ( NumerSense ) , general commonsense ( CommonsenseQA 2.0 ) , and scientific commonsense ( QASC ) benchmarks .", "forward": true, "src_ids": "2022.acl-long.225_1359"}
+{"input": "commonsense reasoning is done by using Method| context: it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models .", "entity": "commonsense reasoning", "output": "large - scale language models", "neg_sample": ["commonsense reasoning is done by using Method", "it remains an open question whether incorporating external knowledge benefits commonsense reasoning while maintaining the flexibility of pretrained sequence models ."], "relation": "used for", "id": "2022.acl-long.225", "year": 2022, "rel_sent": "Generated knowledge prompting highlights large - scale language models as flexible sources of external knowledge for improving commonsense reasoning .", "forward": false, "src_ids": "2022.acl-long.225_1360"}
+{"input": "language models is done by using OtherScientificTerm| context: language models ( lms ) have shown great potential as implicit knowledge bases ( kbs ) . and for their practical use , knowledge in lms need to be updated periodically .", "entity": "language models", "output": "multiple large - scale updates", "neg_sample": ["language models is done by using OtherScientificTerm", "language models ( lms ) have shown great potential as implicit knowledge bases ( kbs ) .", "and for their practical use , knowledge in lms need to be updated periodically ."], "relation": "used for", "id": "2022.findings-acl.37", "year": 2022, "rel_sent": "To this end , we first propose a novel task - Continuously - updated QA ( CuQA)-in which multiple large - scale updates are made to LMs , and the performance is measured with respect to the success in adding and updating knowledge while retaining existing knowledge .", "forward": false, "src_ids": "2022.findings-acl.37_1361"}
+{"input": "embedding composition is used for OtherScientificTerm| context: to alleviate runtime complexity of such inference , previous work has adopted a late interaction architecture with pre - computed contextual token representations at the cost of a large online storage .", "entity": "embedding composition", "output": "search relevance", "neg_sample": ["embedding composition is used for OtherScientificTerm", "to alleviate runtime complexity of such inference , previous work has adopted a late interaction architecture with pre - computed contextual token representations at the cost of a large online storage ."], "relation": "used for", "id": "2022.acl-long.51", "year": 2022, "rel_sent": "This allows effective online decompression and embedding composition for better search relevance .", "forward": true, "src_ids": "2022.acl-long.51_1362"}
+{"input": "online decompression is used for OtherScientificTerm| context: to alleviate runtime complexity of such inference , previous work has adopted a late interaction architecture with pre - computed contextual token representations at the cost of a large online storage .", "entity": "online decompression", "output": "search relevance", "neg_sample": ["online decompression is used for OtherScientificTerm", "to alleviate runtime complexity of such inference , previous work has adopted a late interaction architecture with pre - computed contextual token representations at the cost of a large online storage ."], "relation": "used for", "id": "2022.acl-long.51", "year": 2022, "rel_sent": "This allows effective online decompression and embedding composition for better search relevance .", "forward": true, "src_ids": "2022.acl-long.51_1363"}
+{"input": "search relevance is done by using Task| context: transformer based re - ranking models can achieve high search relevance through context- aware soft matching of query tokens with document tokens . to alleviate runtime complexity of such inference , previous work has adopted a late interaction architecture with pre - computed contextual token representations at the cost of a large online storage .", "entity": "search relevance", "output": "online decompression", "neg_sample": ["search relevance is done by using Task", "transformer based re - ranking models can achieve high search relevance through context- aware soft matching of query tokens with document tokens .", "to alleviate runtime complexity of such inference , previous work has adopted a late interaction architecture with pre - computed contextual token representations at the cost of a large online storage ."], "relation": "used for", "id": "2022.acl-long.51", "year": 2022, "rel_sent": "This allows effective online decompression and embedding composition for better search relevance .", "forward": false, "src_ids": "2022.acl-long.51_1364"}
+{"input": "pretrained chinese bert is used for OtherScientificTerm| context: large - scale pretrained language models have achieved sota results on nlp tasks .", "entity": "pretrained chinese bert", "output": "adversarial attacks", "neg_sample": ["pretrained chinese bert is used for OtherScientificTerm", "large - scale pretrained language models have achieved sota results on nlp tasks ."], "relation": "used for", "id": "2022.acl-long.65", "year": 2022, "rel_sent": "In this work , we propose RoCBert : a pretrained Chinese Bert that is robust to various forms of adversarial attacks like word perturbation , synonyms , typos , etc .", "forward": true, "src_ids": "2022.acl-long.65_1365"}
+{"input": "adversarial attacks is done by using Method| context: large - scale pretrained language models have achieved sota results on nlp tasks . however , they have been shown vulnerable to adversarial attacks especially for logographic languages like chinese .", "entity": "adversarial attacks", "output": "pretrained chinese bert", "neg_sample": ["adversarial attacks is done by using Method", "large - scale pretrained language models have achieved sota results on nlp tasks .", "however , they have been shown vulnerable to adversarial attacks especially for logographic languages like chinese ."], "relation": "used for", "id": "2022.acl-long.65", "year": 2022, "rel_sent": "In this work , we propose RoCBert : a pretrained Chinese Bert that is robust to various forms of adversarial attacks like word perturbation , synonyms , typos , etc .", "forward": false, "src_ids": "2022.acl-long.65_1366"}
+{"input": "clinical section classification is done by using Metric| context: identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks . most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance . however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity .", "entity": "clinical section classification", "output": "task transferability", "neg_sample": ["clinical section classification is done by using Metric", "identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks .", "most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance .", "however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity ."], "relation": "used for", "id": "2022.acl-long.461", "year": 2022, "rel_sent": "Leveraging Task Transferability to Meta - learning for Clinical Section Classification with Limited Data.", "forward": false, "src_ids": "2022.acl-long.461_1367"}
+{"input": "meta - learning algorithms is done by using Metric| context: identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks . most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance . however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity .", "entity": "meta - learning algorithms", "output": "task transferability", "neg_sample": ["meta - learning algorithms is done by using Metric", "identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks .", "most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance .", "however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity ."], "relation": "used for", "id": "2022.acl-long.461", "year": 2022, "rel_sent": "Leveraging Task Transferability to Meta - learning for Clinical Section Classification with Limited Data.", "forward": false, "src_ids": "2022.acl-long.461_1368"}
+{"input": "clinical section classification is done by using Method| context: identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks . most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance . however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity .", "entity": "clinical section classification", "output": "meta - learning algorithms", "neg_sample": ["clinical section classification is done by using Method", "identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks .", "most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance .", "however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity ."], "relation": "used for", "id": "2022.acl-long.461", "year": 2022, "rel_sent": "Leveraging Task Transferability to Meta - learning for Clinical Section Classification with Limited Data.", "forward": false, "src_ids": "2022.acl-long.461_1369"}
+{"input": "task transferability is used for Method| context: identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks . most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance . however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity .", "entity": "task transferability", "output": "meta - learning algorithms", "neg_sample": ["task transferability is used for Method", "identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks .", "most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance .", "however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity ."], "relation": "used for", "id": "2022.acl-long.461", "year": 2022, "rel_sent": "Leveraging Task Transferability to Meta - learning for Clinical Section Classification with Limited Data.", "forward": true, "src_ids": "2022.acl-long.461_1370"}
+{"input": "meta - learning algorithms is used for Task| context: identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks . most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance . however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity .", "entity": "meta - learning algorithms", "output": "clinical section classification", "neg_sample": ["meta - learning algorithms is used for Task", "identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks .", "most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance .", "however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity ."], "relation": "used for", "id": "2022.acl-long.461", "year": 2022, "rel_sent": "Leveraging Task Transferability to Meta - learning for Clinical Section Classification with Limited Data.", "forward": true, "src_ids": "2022.acl-long.461_1371"}
+{"input": "task transferability is used for Task| context: identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks . most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance . however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity .", "entity": "task transferability", "output": "clinical section classification", "neg_sample": ["task transferability is used for Task", "identifying sections is one of the critical components of understanding medical information from unstructured clinical notes and developing assistive technologies for clinical note - writing tasks .", "most state - of - the - art text classification systems require thousands of in - domain text data to achieve high performance .", "however , collecting in - domain and recent clinical note data with section labels is challenging given the high level of privacy and sensitivity ."], "relation": "used for", "id": "2022.acl-long.461", "year": 2022, "rel_sent": "Leveraging Task Transferability to Meta - learning for Clinical Section Classification with Limited Data.", "forward": true, "src_ids": "2022.acl-long.461_1372"}
+{"input": "nlp technology is used for Material| context: nlp research is impeded by a lack of resources and awareness of the challenges presented by underrepresented languages and dialects .", "entity": "nlp technology", "output": "languages of indonesia", "neg_sample": ["nlp technology is used for Material", "nlp research is impeded by a lack of resources and awareness of the challenges presented by underrepresented languages and dialects ."], "relation": "used for", "id": "2022.acl-long.500", "year": 2022, "rel_sent": "Finally , we provide general recommendations to help develop NLP technology not only for languages of Indonesia but also other underrepresented languages .", "forward": true, "src_ids": "2022.acl-long.500_1373"}
+{"input": "languages of indonesia is done by using Method| context: nlp research is impeded by a lack of resources and awareness of the challenges presented by underrepresented languages and dialects .", "entity": "languages of indonesia", "output": "nlp technology", "neg_sample": ["languages of indonesia is done by using Method", "nlp research is impeded by a lack of resources and awareness of the challenges presented by underrepresented languages and dialects ."], "relation": "used for", "id": "2022.acl-long.500", "year": 2022, "rel_sent": "Finally , we provide general recommendations to help develop NLP technology not only for languages of Indonesia but also other underrepresented languages .", "forward": false, "src_ids": "2022.acl-long.500_1374"}
+{"input": "synthetic translations is used for OtherScientificTerm| context: synthetic translations have been used for a wide range of nlp tasks primarily as a means of data augmentation .", "entity": "synthetic translations", "output": "imperfect reference translations", "neg_sample": ["synthetic translations is used for OtherScientificTerm", "synthetic translations have been used for a wide range of nlp tasks primarily as a means of data augmentation ."], "relation": "used for", "id": "2022.acl-long.326", "year": 2022, "rel_sent": "This work explores , instead , how synthetic translations can be used to revise potentially imperfect reference translations in mined bitext .", "forward": true, "src_ids": "2022.acl-long.326_1375"}
+{"input": "nmt noise is done by using Method| context: synthetic translations have been used for a wide range of nlp tasks primarily as a means of data augmentation .", "entity": "nmt noise", "output": "semantic equivalence classifier", "neg_sample": ["nmt noise is done by using Method", "synthetic translations have been used for a wide range of nlp tasks primarily as a means of data augmentation ."], "relation": "used for", "id": "2022.acl-long.326", "year": 2022, "rel_sent": "We find that synthetic samples can improve bitext quality without any additional bilingual supervision when they replace the originals based on a semantic equivalence classifier that helps mitigate NMT noise .", "forward": false, "src_ids": "2022.acl-long.326_1376"}
+{"input": "semantic equivalence classifier is used for OtherScientificTerm| context: synthetic translations have been used for a wide range of nlp tasks primarily as a means of data augmentation .", "entity": "semantic equivalence classifier", "output": "nmt noise", "neg_sample": ["semantic equivalence classifier is used for OtherScientificTerm", "synthetic translations have been used for a wide range of nlp tasks primarily as a means of data augmentation ."], "relation": "used for", "id": "2022.acl-long.326", "year": 2022, "rel_sent": "We find that synthetic samples can improve bitext quality without any additional bilingual supervision when they replace the originals based on a semantic equivalence classifier that helps mitigate NMT noise .", "forward": true, "src_ids": "2022.acl-long.326_1377"}
+{"input": "reading comprehension assessment is done by using Task| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "entity": "reading comprehension assessment", "output": "question generation", "neg_sample": ["reading comprehension assessment is done by using Task", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "Question Generation for Reading Comprehension Assessment by Modeling How and What to Ask.", "forward": false, "src_ids": "2022.findings-acl.168_1378"}
+{"input": "reading comprehension is done by using Task| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "entity": "reading comprehension", "output": "question generation", "neg_sample": ["reading comprehension is done by using Task", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "In this paper , we study QG for reading comprehension where inferential questions are critical and extractive techniques can not be used .", "forward": false, "src_ids": "2022.findings-acl.168_1379"}
+{"input": "question generation is used for Task| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions . however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question .", "entity": "question generation", "output": "reading comprehension assessment", "neg_sample": ["question generation is used for Task", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "Question Generation for Reading Comprehension Assessment by Modeling How and What to Ask.", "forward": true, "src_ids": "2022.findings-acl.168_1380"}
+{"input": "question generation is used for Task| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions . however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question .", "entity": "question generation", "output": "reading comprehension", "neg_sample": ["question generation is used for Task", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "In this paper , we study QG for reading comprehension where inferential questions are critical and extractive techniques can not be used .", "forward": true, "src_ids": "2022.findings-acl.168_1381"}
+{"input": "reader assessment is done by using Material| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions . however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question .", "entity": "reader assessment", "output": "reading comprehension dataset", "neg_sample": ["reader assessment is done by using Material", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "We propose a new reading comprehension dataset that contains questions annotated with story - based reading comprehension skills ( SBRCS ) , allowing for a more complete reader assessment .", "forward": false, "src_ids": "2022.findings-acl.168_1382"}
+{"input": "reading comprehension dataset is used for Task| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions . however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question .", "entity": "reading comprehension dataset", "output": "reader assessment", "neg_sample": ["reading comprehension dataset is used for Task", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "We propose a new reading comprehension dataset that contains questions annotated with story - based reading comprehension skills ( SBRCS ) , allowing for a more complete reader assessment .", "forward": true, "src_ids": "2022.findings-acl.168_1383"}
+{"input": "scrs is done by using Method| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions . however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question .", "entity": "scrs", "output": "hta - wta model", "neg_sample": ["scrs is done by using Method", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "We show that the HTA - WTA model tests for strong SCRS by asking deep inferential questions .", "forward": false, "src_ids": "2022.findings-acl.168_1384"}
+{"input": "hta - wta model is used for OtherScientificTerm| context: reading is integral to everyday life , and yet learning to read is a struggle for many young learners . during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention . historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions . however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question .", "entity": "hta - wta model", "output": "scrs", "neg_sample": ["hta - wta model is used for OtherScientificTerm", "reading is integral to everyday life , and yet learning to read is a struggle for many young learners .", "during lessons , teachers can use comprehension questions to increase engagement , test reading skills , and improve retention .", "historically such questions were written by skilled teachers , but recently language models have been used to generate comprehension questions .", "however , many existing question generation ( qg ) systems focus on generating extractive questions from the text , and have no way to control the type of the generated question ."], "relation": "used for", "id": "2022.findings-acl.168", "year": 2022, "rel_sent": "We show that the HTA - WTA model tests for strong SCRS by asking deep inferential questions .", "forward": true, "src_ids": "2022.findings-acl.168_1385"}
+{"input": "recognizing offensive spans is done by using Method| context: toxic span detection is the task of recognizing offensive spans in a text snippet . although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet .", "entity": "recognizing offensive spans", "output": "transfer learning", "neg_sample": ["recognizing offensive spans is done by using Method", "toxic span detection is the task of recognizing offensive spans in a text snippet .", "although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet ."], "relation": "used for", "id": "2022.findings-acl.128", "year": 2022, "rel_sent": "Transfer Learning and Prediction Consistency for Detecting Offensive Spans of Text.", "forward": false, "src_ids": "2022.findings-acl.128_1386"}
+{"input": "transfer learning is used for Task| context: although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet .", "entity": "transfer learning", "output": "recognizing offensive spans", "neg_sample": ["transfer learning is used for Task", "although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet ."], "relation": "used for", "id": "2022.findings-acl.128", "year": 2022, "rel_sent": "Transfer Learning and Prediction Consistency for Detecting Offensive Spans of Text.", "forward": true, "src_ids": "2022.findings-acl.128_1387"}
+{"input": "toxic span detection is done by using Method| context: toxic span detection is the task of recognizing offensive spans in a text snippet . although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet .", "entity": "toxic span detection", "output": "multi - task framework", "neg_sample": ["toxic span detection is done by using Method", "toxic span detection is the task of recognizing offensive spans in a text snippet .", "although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet ."], "relation": "used for", "id": "2022.findings-acl.128", "year": 2022, "rel_sent": "In this work , we introduce a novel multi - task framework for toxic span detection in which the model seeks to simultaneously predict offensive words and opinion phrases to leverage their inter - dependencies and improve the performance .", "forward": false, "src_ids": "2022.findings-acl.128_1388"}
+{"input": "multi - task framework is used for Task| context: although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet .", "entity": "multi - task framework", "output": "toxic span detection", "neg_sample": ["multi - task framework is used for Task", "although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet ."], "relation": "used for", "id": "2022.findings-acl.128", "year": 2022, "rel_sent": "In this work , we introduce a novel multi - task framework for toxic span detection in which the model seeks to simultaneously predict offensive words and opinion phrases to leverage their inter - dependencies and improve the performance .", "forward": true, "src_ids": "2022.findings-acl.128_1389"}
+{"input": "regularization mechanism is used for Task| context: although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet .", "entity": "regularization mechanism", "output": "toxic span detection", "neg_sample": ["regularization mechanism is used for Task", "although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet ."], "relation": "used for", "id": "2022.findings-acl.128", "year": 2022, "rel_sent": "Moreover , we introduce a novel regularization mechanism to encourage the consistency of the model predictions across similar inputs for toxic span detection .", "forward": true, "src_ids": "2022.findings-acl.128_1390"}
+{"input": "toxic span detection is done by using Method| context: toxic span detection is the task of recognizing offensive spans in a text snippet . although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet .", "entity": "toxic span detection", "output": "regularization mechanism", "neg_sample": ["toxic span detection is done by using Method", "toxic span detection is the task of recognizing offensive spans in a text snippet .", "although there has been prior work on classifying text snippets as offensive or not , the task of recognizing spans responsible for the toxicity of a text is not explored yet ."], "relation": "used for", "id": "2022.findings-acl.128", "year": 2022, "rel_sent": "Moreover , we introduce a novel regularization mechanism to encourage the consistency of the model predictions across similar inputs for toxic span detection .", "forward": false, "src_ids": "2022.findings-acl.128_1391"}
+{"input": "reweighting mechanism is used for Generic| context: most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples . however , the augmented adversarial examples may not be natural , which might distort the training distribution , resulting in inferior performance both in clean accuracy and adversarial robustness .", "entity": "reweighting mechanism", "output": "robust models", "neg_sample": ["reweighting mechanism is used for Generic", "most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples .", "however , the augmented adversarial examples may not be natural , which might distort the training distribution , resulting in inferior performance both in clean accuracy and adversarial robustness ."], "relation": "used for", "id": "2022.findings-acl.134", "year": 2022, "rel_sent": "In this study , we explore the feasibility of introducing a reweighting mechanism to calibrate the training distribution to obtain robust models .", "forward": true, "src_ids": "2022.findings-acl.134_1392"}
+{"input": "reweighting mechanism is used for OtherScientificTerm| context: most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples .", "entity": "reweighting mechanism", "output": "training distribution", "neg_sample": ["reweighting mechanism is used for OtherScientificTerm", "most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples ."], "relation": "used for", "id": "2022.findings-acl.134", "year": 2022, "rel_sent": "In this study , we explore the feasibility of introducing a reweighting mechanism to calibrate the training distribution to obtain robust models .", "forward": true, "src_ids": "2022.findings-acl.134_1393"}
+{"input": "robust models is done by using Method| context: most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples . however , the augmented adversarial examples may not be natural , which might distort the training distribution , resulting in inferior performance both in clean accuracy and adversarial robustness .", "entity": "robust models", "output": "reweighting mechanism", "neg_sample": ["robust models is done by using Method", "most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples .", "however , the augmented adversarial examples may not be natural , which might distort the training distribution , resulting in inferior performance both in clean accuracy and adversarial robustness ."], "relation": "used for", "id": "2022.findings-acl.134", "year": 2022, "rel_sent": "In this study , we explore the feasibility of introducing a reweighting mechanism to calibrate the training distribution to obtain robust models .", "forward": false, "src_ids": "2022.findings-acl.134_1394"}
+{"input": "training distribution is done by using Method| context: most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples . however , the augmented adversarial examples may not be natural , which might distort the training distribution , resulting in inferior performance both in clean accuracy and adversarial robustness .", "entity": "training distribution", "output": "reweighting mechanism", "neg_sample": ["training distribution is done by using Method", "most of the existing defense methods improve the adversarial robustness by making the models adapt to the training set augmented with some adversarial examples .", "however , the augmented adversarial examples may not be natural , which might distort the training distribution , resulting in inferior performance both in clean accuracy and adversarial robustness ."], "relation": "used for", "id": "2022.findings-acl.134", "year": 2022, "rel_sent": "In this study , we explore the feasibility of introducing a reweighting mechanism to calibrate the training distribution to obtain robust models .", "forward": false, "src_ids": "2022.findings-acl.134_1395"}
+{"input": "pre - trained language models is done by using Method| context: recent work has shown pre - trained language models capture social biases from the large amounts of text they are trained on . this has attracted attention to developing techniques that mitigate such biases .", "entity": "pre - trained language models", "output": "debiasing techniques", "neg_sample": ["pre - trained language models is done by using Method", "recent work has shown pre - trained language models capture social biases from the large amounts of text they are trained on .", "this has attracted attention to developing techniques that mitigate such biases ."], "relation": "used for", "id": "2022.acl-long.132", "year": 2022, "rel_sent": "An Empirical Survey of the Effectiveness of Debiasing Techniques for Pre - trained Language Models.", "forward": false, "src_ids": "2022.acl-long.132_1396"}
+{"input": "non - gender biases is done by using Method| context: recent work has shown pre - trained language models capture social biases from the large amounts of text they are trained on . this has attracted attention to developing techniques that mitigate such biases .", "entity": "non - gender biases", "output": "debiasing techniques", "neg_sample": ["non - gender biases is done by using Method", "recent work has shown pre - trained language models capture social biases from the large amounts of text they are trained on .", "this has attracted attention to developing techniques that mitigate such biases ."], "relation": "used for", "id": "2022.acl-long.132", "year": 2022, "rel_sent": "We experimentally find that : ( 1 ) Self - Debias is the strongest debiasing technique , obtaining improved scores on all bias benchmarks ; ( 2 ) Current debiasing techniques perform less consistently when mitigating non - gender biases ; And ( 3 ) improvements on bias benchmarks such as StereoSet and CrowS - Pairs by using debiasing strategies are often accompanied by a decrease in language modeling ability , making it difficult to determine whether the bias mitigation was effective .", "forward": false, "src_ids": "2022.acl-long.132_1397"}
+{"input": "debiasing techniques is used for Method| context: this has attracted attention to developing techniques that mitigate such biases .", "entity": "debiasing techniques", "output": "pre - trained language models", "neg_sample": ["debiasing techniques is used for Method", "this has attracted attention to developing techniques that mitigate such biases ."], "relation": "used for", "id": "2022.acl-long.132", "year": 2022, "rel_sent": "An Empirical Survey of the Effectiveness of Debiasing Techniques for Pre - trained Language Models.", "forward": true, "src_ids": "2022.acl-long.132_1398"}
+{"input": "debiasing techniques is used for OtherScientificTerm| context: recent work has shown pre - trained language models capture social biases from the large amounts of text they are trained on . this has attracted attention to developing techniques that mitigate such biases .", "entity": "debiasing techniques", "output": "non - gender biases", "neg_sample": ["debiasing techniques is used for OtherScientificTerm", "recent work has shown pre - trained language models capture social biases from the large amounts of text they are trained on .", "this has attracted attention to developing techniques that mitigate such biases ."], "relation": "used for", "id": "2022.acl-long.132", "year": 2022, "rel_sent": "We experimentally find that : ( 1 ) Self - Debias is the strongest debiasing technique , obtaining improved scores on all bias benchmarks ; ( 2 ) Current debiasing techniques perform less consistently when mitigating non - gender biases ; And ( 3 ) improvements on bias benchmarks such as StereoSet and CrowS - Pairs by using debiasing strategies are often accompanied by a decrease in language modeling ability , making it difficult to determine whether the bias mitigation was effective .", "forward": true, "src_ids": "2022.acl-long.132_1399"}
+{"input": "neural machine translation is done by using Method| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "neural machine translation", "output": "ciphertext based data augmentation", "neg_sample": ["neural machine translation is done by using Method", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "CipherDAug : Ciphertext based Data Augmentation for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.17_1400"}
+{"input": "ciphertext based data augmentation is used for Task| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "ciphertext based data augmentation", "output": "neural machine translation", "neg_sample": ["ciphertext based data augmentation is used for Task", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "CipherDAug : Ciphertext based Data Augmentation for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.17_1401"}
+{"input": "data - augmentation technique is used for Task| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "data - augmentation technique", "output": "neural machine translation", "neg_sample": ["data - augmentation technique is used for Task", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "We propose a novel data - augmentation technique for neural machine translation based on ROT - k ciphertexts .", "forward": true, "src_ids": "2022.acl-long.17_1402"}
+{"input": "multi - source training is used for Task| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "multi - source training", "output": "neural machine translation", "neg_sample": ["multi - source training is used for Task", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "We then leverage this enciphered training data along with the original parallel data via multi - source training to improve neural machine translation .", "forward": true, "src_ids": "2022.acl-long.17_1403"}
+{"input": "neural machine translation is done by using Method| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "neural machine translation", "output": "data - augmentation technique", "neg_sample": ["neural machine translation is done by using Method", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "We propose a novel data - augmentation technique for neural machine translation based on ROT - k ciphertexts .", "forward": false, "src_ids": "2022.acl-long.17_1404"}
+{"input": "data augmentation is done by using Method| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "data augmentation", "output": "data - augmentation technique", "neg_sample": ["data augmentation is done by using Method", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "This technique combines easily with existing approaches to data augmentation , and yields particularly strong results in low - resource settings .", "forward": false, "src_ids": "2022.acl-long.17_1405"}
+{"input": "neural machine translation is done by using Method| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "neural machine translation", "output": "multi - source training", "neg_sample": ["neural machine translation is done by using Method", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "We then leverage this enciphered training data along with the original parallel data via multi - source training to improve neural machine translation .", "forward": false, "src_ids": "2022.acl-long.17_1406"}
+{"input": "data - augmentation technique is used for Method| context: rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet .", "entity": "data - augmentation technique", "output": "data augmentation", "neg_sample": ["data - augmentation technique is used for Method", "rot - k is a simple letter substitution cipher that replaces a letter in the plaintext with the kth letter after it in the alphabet ."], "relation": "used for", "id": "2022.acl-long.17", "year": 2022, "rel_sent": "This technique combines easily with existing approaches to data augmentation , and yields particularly strong results in low - resource settings .", "forward": true, "src_ids": "2022.acl-long.17_1407"}
+{"input": "continual relation extraction is done by using Method| context: some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting . however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets .", "entity": "continual relation extraction", "output": "consistent representation learning method", "neg_sample": ["continual relation extraction is done by using Method", "some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting .", "however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets ."], "relation": "used for", "id": "2022.findings-acl.268", "year": 2022, "rel_sent": "Consistent Representation Learning for Continual Relation Extraction.", "forward": false, "src_ids": "2022.findings-acl.268_1408"}
+{"input": "consistent representation learning method is used for Task| context: some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting . however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets .", "entity": "consistent representation learning method", "output": "continual relation extraction", "neg_sample": ["consistent representation learning method is used for Task", "some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting .", "however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets ."], "relation": "used for", "id": "2022.findings-acl.268", "year": 2022, "rel_sent": "Consistent Representation Learning for Continual Relation Extraction.", "forward": true, "src_ids": "2022.findings-acl.268_1409"}
+{"input": "consistent representations is used for OtherScientificTerm| context: however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets .", "entity": "consistent representations", "output": "forgetting", "neg_sample": ["consistent representations is used for OtherScientificTerm", "however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets ."], "relation": "used for", "id": "2022.findings-acl.268", "year": 2022, "rel_sent": "The proposed method can better learn consistent representations to alleviate forgetting effectively .", "forward": true, "src_ids": "2022.findings-acl.268_1410"}
+{"input": "forgetting is done by using Method| context: some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting . however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets .", "entity": "forgetting", "output": "consistent representations", "neg_sample": ["forgetting is done by using Method", "some previous work has proved that storing a few typical samples of old relations and replaying them when learning new relations can effectively avoid forgetting .", "however , these memory - based methods tend to overfit the memory samples and perform poorly on imbalanced datasets ."], "relation": "used for", "id": "2022.findings-acl.268", "year": 2022, "rel_sent": "The proposed method can better learn consistent representations to alleviate forgetting effectively .", "forward": false, "src_ids": "2022.findings-acl.268_1411"}
+{"input": "indic natural language generation is done by using Method| context: in this paper , we study pre - trained sequence - to - sequence models for a group of related languages , with a focus on indic languages .", "entity": "indic natural language generation", "output": "pre - trained model", "neg_sample": ["indic natural language generation is done by using Method", "in this paper , we study pre - trained sequence - to - sequence models for a group of related languages , with a focus on indic languages ."], "relation": "used for", "id": "2022.findings-acl.145", "year": 2022, "rel_sent": "IndicBART : A Pre - trained Model for Indic Natural Language Generation.", "forward": false, "src_ids": "2022.findings-acl.145_1412"}
+{"input": "pre - trained model is used for Task| context: in this paper , we study pre - trained sequence - to - sequence models for a group of related languages , with a focus on indic languages .", "entity": "pre - trained model", "output": "indic natural language generation", "neg_sample": ["pre - trained model is used for Task", "in this paper , we study pre - trained sequence - to - sequence models for a group of related languages , with a focus on indic languages ."], "relation": "used for", "id": "2022.findings-acl.145", "year": 2022, "rel_sent": "IndicBART : A Pre - trained Model for Indic Natural Language Generation.", "forward": true, "src_ids": "2022.findings-acl.145_1413"}
+{"input": "targeted structures is done by using Method| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "targeted structures", "output": "unified text - to - structure generation framework", "neg_sample": ["targeted structures is done by using Method", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "In this paper , we propose a unified text - to - structure generation framework , namely UIE , which can universally model different IE tasks , adaptively generate targeted structures , and collaboratively learn general IE abilities from different knowledge sources .", "forward": false, "src_ids": "2022.acl-long.395_1414"}
+{"input": "ie tasks is done by using Method| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "ie tasks", "output": "uie", "neg_sample": ["ie tasks is done by using Method", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "In this paper , we propose a unified text - to - structure generation framework , namely UIE , which can universally model different IE tasks , adaptively generate targeted structures , and collaboratively learn general IE abilities from different knowledge sources .", "forward": false, "src_ids": "2022.acl-long.395_1415"}
+{"input": "extraction structures is done by using Method| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "extraction structures", "output": "uie", "neg_sample": ["extraction structures is done by using Method", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "Specifically , UIE uniformly encodes different extraction structures via a structured extraction language , adaptively generates target extractions via a schema - based prompt mechanism - structural schema instructor , and captures the common IE abilities via a large - scale pretrained text - to - structure model .", "forward": false, "src_ids": "2022.acl-long.395_1416"}
+{"input": "target extractions is done by using Method| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "target extractions", "output": "uie", "neg_sample": ["target extractions is done by using Method", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "Specifically , UIE uniformly encodes different extraction structures via a structured extraction language , adaptively generates target extractions via a schema - based prompt mechanism - structural schema instructor , and captures the common IE abilities via a large - scale pretrained text - to - structure model .", "forward": false, "src_ids": "2022.acl-long.395_1417"}
+{"input": "general ie abilities is done by using Method| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "general ie abilities", "output": "uie", "neg_sample": ["general ie abilities is done by using Method", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "Specifically , UIE uniformly encodes different extraction structures via a structured extraction language , adaptively generates target extractions via a schema - based prompt mechanism - structural schema instructor , and captures the common IE abilities via a large - scale pretrained text - to - structure model .", "forward": false, "src_ids": "2022.acl-long.395_1418"}
+{"input": "uie is used for Task| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "uie", "output": "ie tasks", "neg_sample": ["uie is used for Task", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "In this paper , we propose a unified text - to - structure generation framework , namely UIE , which can universally model different IE tasks , adaptively generate targeted structures , and collaboratively learn general IE abilities from different knowledge sources .", "forward": true, "src_ids": "2022.acl-long.395_1419"}
+{"input": "unified text - to - structure generation framework is used for OtherScientificTerm| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "unified text - to - structure generation framework", "output": "targeted structures", "neg_sample": ["unified text - to - structure generation framework is used for OtherScientificTerm", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "In this paper , we propose a unified text - to - structure generation framework , namely UIE , which can universally model different IE tasks , adaptively generate targeted structures , and collaboratively learn general IE abilities from different knowledge sources .", "forward": true, "src_ids": "2022.acl-long.395_1420"}
+{"input": "uie is used for OtherScientificTerm| context: information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas .", "entity": "uie", "output": "general ie abilities", "neg_sample": ["uie is used for OtherScientificTerm", "information extraction suffers from its varying targets , heterogeneous structures , and demand - specific schemas ."], "relation": "used for", "id": "2022.acl-long.395", "year": 2022, "rel_sent": "Specifically , UIE uniformly encodes different extraction structures via a structured extraction language , adaptively generates target extractions via a schema - based prompt mechanism - structural schema instructor , and captures the common IE abilities via a large - scale pretrained text - to - structure model .", "forward": true, "src_ids": "2022.acl-long.395_1421"}
+{"input": "partial observability is done by using Method| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "partial observability", "output": "holm", "neg_sample": ["partial observability is done by using Method", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "In this paper , we introduce HOLM , Hallucinating Objects with Language Models , to address the challenge of partial observability .", "forward": false, "src_ids": "2022.acl-long.373_1422"}
+{"input": "hallucinating objects is done by using Method| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "hallucinating objects", "output": "holm", "neg_sample": ["hallucinating objects is done by using Method", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "In this paper , we introduce HOLM , Hallucinating Objects with Language Models , to address the challenge of partial observability .", "forward": false, "src_ids": "2022.acl-long.373_1423"}
+{"input": "object hallucinations is done by using Method| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "object hallucinations", "output": "holm", "neg_sample": ["object hallucinations is done by using Method", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM uses large pre - trained language models ( LMs ) to infer object hallucinations for the unobserved part of the environment .", "forward": false, "src_ids": "2022.acl-long.373_1424"}
+{"input": "generalization is done by using Method| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "generalization", "output": "holm", "neg_sample": ["generalization is done by using Method", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "Our experiments show that HOLM performs better than the state - of - the - art approaches on two datasets for dRER ; allowing to study generalization for both indoor and outdoor settings .", "forward": false, "src_ids": "2022.acl-long.373_1425"}
+{"input": "referring expression recognition is done by using Task| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "referring expression recognition", "output": "hallucinating objects", "neg_sample": ["referring expression recognition is done by using Task", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM : Hallucinating Objects with Language Models for Referring Expression Recognition in Partially - Observed Scenes.", "forward": false, "src_ids": "2022.acl-long.373_1426"}
+{"input": "holm is used for Task| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "holm", "output": "hallucinating objects", "neg_sample": ["holm is used for Task", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "In this paper , we introduce HOLM , Hallucinating Objects with Language Models , to address the challenge of partial observability .", "forward": true, "src_ids": "2022.acl-long.373_1427"}
+{"input": "referring expression recognition is done by using Method| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "referring expression recognition", "output": "language models", "neg_sample": ["referring expression recognition is done by using Method", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM : Hallucinating Objects with Language Models for Referring Expression Recognition in Partially - Observed Scenes.", "forward": false, "src_ids": "2022.acl-long.373_1428"}
+{"input": "object affinities is done by using Method| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "object affinities", "output": "language models", "neg_sample": ["object affinities is done by using Method", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "Based on this intuition , we prompt language models to extract knowledge about object affinities which gives us a proxy for spatial relationships of objects .", "forward": false, "src_ids": "2022.acl-long.373_1429"}
+{"input": "language models is used for Task| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view .", "entity": "language models", "output": "referring expression recognition", "neg_sample": ["language models is used for Task", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM : Hallucinating Objects with Language Models for Referring Expression Recognition in Partially - Observed Scenes.", "forward": true, "src_ids": "2022.acl-long.373_1430"}
+{"input": "hallucinating objects is used for Task| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view .", "entity": "hallucinating objects", "output": "referring expression recognition", "neg_sample": ["hallucinating objects is used for Task", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM : Hallucinating Objects with Language Models for Referring Expression Recognition in Partially - Observed Scenes.", "forward": true, "src_ids": "2022.acl-long.373_1431"}
+{"input": "holm is used for Task| context: this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "holm", "output": "partial observability", "neg_sample": ["holm is used for Task", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "In this paper , we introduce HOLM , Hallucinating Objects with Language Models , to address the challenge of partial observability .", "forward": true, "src_ids": "2022.acl-long.373_1432"}
+{"input": "holm is used for OtherScientificTerm| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "holm", "output": "object hallucinations", "neg_sample": ["holm is used for OtherScientificTerm", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM uses large pre - trained language models ( LMs ) to infer object hallucinations for the unobserved part of the environment .", "forward": true, "src_ids": "2022.acl-long.373_1433"}
+{"input": "pre - trained language models ( lms ) is used for OtherScientificTerm| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "pre - trained language models ( lms )", "output": "object hallucinations", "neg_sample": ["pre - trained language models ( lms ) is used for OtherScientificTerm", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "HOLM uses large pre - trained language models ( LMs ) to infer object hallucinations for the unobserved part of the environment .", "forward": true, "src_ids": "2022.acl-long.373_1434"}
+{"input": "language models is used for OtherScientificTerm| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "language models", "output": "object affinities", "neg_sample": ["language models is used for OtherScientificTerm", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "Based on this intuition , we prompt language models to extract knowledge about object affinities which gives us a proxy for spatial relationships of objects .", "forward": true, "src_ids": "2022.acl-long.373_1435"}
+{"input": "indoor and outdoor settings is done by using Task| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "indoor and outdoor settings", "output": "generalization", "neg_sample": ["indoor and outdoor settings is done by using Task", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "Our experiments show that HOLM performs better than the state - of - the - art approaches on two datasets for dRER ; allowing to study generalization for both indoor and outdoor settings .", "forward": false, "src_ids": "2022.acl-long.373_1436"}
+{"input": "holm is used for Task| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "holm", "output": "generalization", "neg_sample": ["holm is used for Task", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "Our experiments show that HOLM performs better than the state - of - the - art approaches on two datasets for dRER ; allowing to study generalization for both indoor and outdoor settings .", "forward": true, "src_ids": "2022.acl-long.373_1437"}
+{"input": "generalization is used for OtherScientificTerm| context: ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment . this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g. giving many instructions ) are not immediately visible . actions by the ai system may be required to bring these objects in view . a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes .", "entity": "generalization", "output": "indoor and outdoor settings", "neg_sample": ["generalization is used for OtherScientificTerm", "ai systems embodied in the physical world face a fundamental challenge of partial observability ; operating with only a limited view and knowledge of the environment .", "this creates challenges when ai systems try to reason about language and its relationship with the environment : objects referred to through language ( e.g.", "giving many instructions ) are not immediately visible .", "actions by the ai system may be required to bring these objects in view .", "a good benchmark to study this challenge is dynamic referring expression recognition ( drer ) task , where the goal is tofind a target location by dynamically adjusting the field of view ( fov ) in a partially observed 360 scenes ."], "relation": "used for", "id": "2022.acl-long.373", "year": 2022, "rel_sent": "Our experiments show that HOLM performs better than the state - of - the - art approaches on two datasets for dRER ; allowing to study generalization for both indoor and outdoor settings .", "forward": true, "src_ids": "2022.acl-long.373_1438"}
+{"input": "extractive question answering is done by using Method| context: we study learning from user feedback for extractive question answering by simulating feedback using supervised data .", "entity": "extractive question answering", "output": "simulating bandit learning", "neg_sample": ["extractive question answering is done by using Method", "we study learning from user feedback for extractive question answering by simulating feedback using supervised data ."], "relation": "used for", "id": "2022.acl-long.355", "year": 2022, "rel_sent": "Simulating Bandit Learning from User Feedback for Extractive Question Answering.", "forward": false, "src_ids": "2022.acl-long.355_1439"}
+{"input": "comment is used for Method| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "comment", "output": "unixcoder", "neg_sample": ["comment is used for Method", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "Results show that our model achieves state - of - the - art performance on most tasks and analysis reveals that comment and AST can both enhance UniXcoder .", "forward": true, "src_ids": "2022.acl-long.499_1440"}
+{"input": "ast is used for Method| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "ast", "output": "unixcoder", "neg_sample": ["ast is used for Method", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "Results show that our model achieves state - of - the - art performance on most tasks and analysis reveals that comment and AST can both enhance UniXcoder .", "forward": true, "src_ids": "2022.acl-long.499_1441"}
+{"input": "programming languages is done by using Method| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "programming languages", "output": "unified cross - modal pre - trained model", "neg_sample": ["programming languages is done by using Method", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "In this paper , we present UniXcoder , a unified cross - modal pre - trained model for programming language .", "forward": false, "src_ids": "2022.acl-long.499_1442"}
+{"input": "unified cross - modal pre - trained model is used for OtherScientificTerm| context: to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "unified cross - modal pre - trained model", "output": "programming languages", "neg_sample": ["unified cross - modal pre - trained model is used for OtherScientificTerm", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "In this paper , we present UniXcoder , a unified cross - modal pre - trained model for programming language .", "forward": true, "src_ids": "2022.acl-long.499_1443"}
+{"input": "unixcoder is done by using OtherScientificTerm| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "unixcoder", "output": "comment", "neg_sample": ["unixcoder is done by using OtherScientificTerm", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "Results show that our model achieves state - of - the - art performance on most tasks and analysis reveals that comment and AST can both enhance UniXcoder .", "forward": false, "src_ids": "2022.acl-long.499_1444"}
+{"input": "code representation is done by using OtherScientificTerm| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "code representation", "output": "cross - modal contents", "neg_sample": ["code representation is done by using OtherScientificTerm", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "The model utilizes mask attention matrices with prefix adapters to control the behavior of the model and leverages cross - modal contents like AST and code comment to enhance code representation .", "forward": false, "src_ids": "2022.acl-long.499_1445"}
+{"input": "one - to - one mapping method is used for OtherScientificTerm| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "one - to - one mapping method", "output": "ast", "neg_sample": ["one - to - one mapping method is used for OtherScientificTerm", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "To encode AST that is represented as a tree in parallel , we propose a one - to - one mapping method to transform AST in a sequence structure that retains all structural information from the tree .", "forward": true, "src_ids": "2022.acl-long.499_1446"}
+{"input": "cross - modal contents is used for Method| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "cross - modal contents", "output": "code representation", "neg_sample": ["cross - modal contents is used for Method", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "The model utilizes mask attention matrices with prefix adapters to control the behavior of the model and leverages cross - modal contents like AST and code comment to enhance code representation .", "forward": true, "src_ids": "2022.acl-long.499_1447"}
+{"input": "ast is done by using Method| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "ast", "output": "one - to - one mapping method", "neg_sample": ["ast is done by using Method", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "To encode AST that is represented as a tree in parallel , we propose a one - to - one mapping method to transform AST in a sequence structure that retains all structural information from the tree .", "forward": false, "src_ids": "2022.acl-long.499_1448"}
+{"input": "sequence structure is done by using Method| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "sequence structure", "output": "one - to - one mapping method", "neg_sample": ["sequence structure is done by using Method", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "To encode AST that is represented as a tree in parallel , we propose a one - to - one mapping method to transform AST in a sequence structure that retains all structural information from the tree .", "forward": false, "src_ids": "2022.acl-long.499_1449"}
+{"input": "representation of code fragment is done by using OtherScientificTerm| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "representation of code fragment", "output": "multi - modal contents", "neg_sample": ["representation of code fragment is done by using OtherScientificTerm", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "Furthermore , we propose to utilize multi - modal contents to learn representation of code fragment with contrastive learning , and then align representations among programming languages using a cross - modal generation task .", "forward": false, "src_ids": "2022.acl-long.499_1450"}
+{"input": "multi - modal contents is used for OtherScientificTerm| context: pre - trained models for programming languages have recently demonstrated great success on code intelligence . to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models . however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference .", "entity": "multi - modal contents", "output": "representation of code fragment", "neg_sample": ["multi - modal contents is used for OtherScientificTerm", "pre - trained models for programming languages have recently demonstrated great success on code intelligence .", "to support both code - related understanding and generation tasks , recent works attempt to pre - train unified encoder - decoder models .", "however , such encoder - decoder framework is sub - optimal for auto - regressive tasks , especially code completion that requires a decoder - only manner for efficient inference ."], "relation": "used for", "id": "2022.acl-long.499", "year": 2022, "rel_sent": "Furthermore , we propose to utilize multi - modal contents to learn representation of code fragment with contrastive learning , and then align representations among programming languages using a cross - modal generation task .", "forward": true, "src_ids": "2022.acl-long.499_1451"}
+{"input": "copy mechanism is used for OtherScientificTerm| context: with the rapid development of deep learning , seq2seq paradigm has become prevalent for end - to - end data - to - text generation , and the bleu scores have been increasing in recent years . however , it is widely recognized that there is still a gap between the quality of the texts generated by models and the texts written by human .", "entity": "copy mechanism", "output": "omission and inaccuracy extrinsic errors", "neg_sample": ["copy mechanism is used for OtherScientificTerm", "with the rapid development of deep learning , seq2seq paradigm has become prevalent for end - to - end data - to - text generation , and the bleu scores have been increasing in recent years .", "however , it is widely recognized that there is still a gap between the quality of the texts generated by models and the texts written by human ."], "relation": "used for", "id": "2022.acl-long.531", "year": 2022, "rel_sent": "We annotate the outputs of five models on four datasets with eight error types and find that 1 ) copy mechanism is helpful for the improvement in Omission and Inaccuracy Extrinsic errors but it increases other types of errors such as Addition ; 2 ) pre - training techniques are highly effective , and pre - training strategy and model size are very significant ; 3 ) the structure of the dataset also influences the model 's performance greatly ; 4 ) some specific types of errors are generally challenging for seq2seq models .", "forward": true, "src_ids": "2022.acl-long.531_1452"}
+{"input": "omission and inaccuracy extrinsic errors is done by using Method| context: with the rapid development of deep learning , seq2seq paradigm has become prevalent for end - to - end data - to - text generation , and the bleu scores have been increasing in recent years . however , it is widely recognized that there is still a gap between the quality of the texts generated by models and the texts written by human .", "entity": "omission and inaccuracy extrinsic errors", "output": "copy mechanism", "neg_sample": ["omission and inaccuracy extrinsic errors is done by using Method", "with the rapid development of deep learning , seq2seq paradigm has become prevalent for end - to - end data - to - text generation , and the bleu scores have been increasing in recent years .", "however , it is widely recognized that there is still a gap between the quality of the texts generated by models and the texts written by human ."], "relation": "used for", "id": "2022.acl-long.531", "year": 2022, "rel_sent": "We annotate the outputs of five models on four datasets with eight error types and find that 1 ) copy mechanism is helpful for the improvement in Omission and Inaccuracy Extrinsic errors but it increases other types of errors such as Addition ; 2 ) pre - training techniques are highly effective , and pre - training strategy and model size are very significant ; 3 ) the structure of the dataset also influences the model 's performance greatly ; 4 ) some specific types of errors are generally challenging for seq2seq models .", "forward": false, "src_ids": "2022.acl-long.531_1453"}
+{"input": "automatic metrics is used for Task| context: although text style transfer has witnessed rapid development in recent years , there is as yet no established standard for evaluation , which is performed using several automatic metrics , lacking the possibility of always resorting to human judgement .", "entity": "automatic metrics", "output": "formality transfer", "neg_sample": ["automatic metrics is used for Task", "although text style transfer has witnessed rapid development in recent years , there is as yet no established standard for evaluation , which is performed using several automatic metrics , lacking the possibility of always resorting to human judgement ."], "relation": "used for", "id": "2022.humeval-1.9", "year": 2022, "rel_sent": "Human Judgement as a Compass to Navigate Automatic Metrics for Formality Transfer.", "forward": true, "src_ids": "2022.humeval-1.9_1454"}
+{"input": "automatic metrics is used for Task| context: although text style transfer has witnessed rapid development in recent years , there is as yet no established standard for evaluation , which is performed using several automatic metrics , lacking the possibility of always resorting to human judgement .", "entity": "automatic metrics", "output": "formality transfer", "neg_sample": ["automatic metrics is used for Task", "although text style transfer has witnessed rapid development in recent years , there is as yet no established standard for evaluation , which is performed using several automatic metrics , lacking the possibility of always resorting to human judgement ."], "relation": "used for", "id": "2022.humeval-1.9", "year": 2022, "rel_sent": "We are then able to offer some recommendations on the use of such metrics in formality transfer , also with an eye to their generalisability ( or not ) to related tasks .", "forward": true, "src_ids": "2022.humeval-1.9_1455"}
+{"input": "prototype vectors is done by using Method| context: prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning . typically , prompt - based tuning wraps the input text into a cloze question . to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built .", "entity": "prototype vectors", "output": "prototypical verbalizer ( protoverb )", "neg_sample": ["prototype vectors is done by using Method", "prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning .", "typically , prompt - based tuning wraps the input text into a cloze question .", "to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built ."], "relation": "used for", "id": "2022.acl-long.483", "year": 2022, "rel_sent": "Specifically , ProtoVerb learns prototype vectors as verbalizers by contrastive learning .", "forward": false, "src_ids": "2022.acl-long.483_1456"}
+{"input": "prompt - based tuning is done by using Method| context: prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning . typically , prompt - based tuning wraps the input text into a cloze question . to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built .", "entity": "prompt - based tuning", "output": "prototypical verbalizer ( protoverb )", "neg_sample": ["prompt - based tuning is done by using Method", "prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning .", "typically , prompt - based tuning wraps the input text into a cloze question .", "to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built ."], "relation": "used for", "id": "2022.acl-long.483", "year": 2022, "rel_sent": "More surprisingly , ProtoVerb consistently boosts prompt - based tuning even on untuned PLMs , indicating an elegant non - tuning way to utilize PLMs .", "forward": false, "src_ids": "2022.acl-long.483_1457"}
+{"input": "prototypical verbalizer ( protoverb ) is used for Method| context: to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built .", "entity": "prototypical verbalizer ( protoverb )", "output": "prompt - based tuning", "neg_sample": ["prototypical verbalizer ( protoverb ) is used for Method", "to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built ."], "relation": "used for", "id": "2022.acl-long.483", "year": 2022, "rel_sent": "More surprisingly , ProtoVerb consistently boosts prompt - based tuning even on untuned PLMs , indicating an elegant non - tuning way to utilize PLMs .", "forward": true, "src_ids": "2022.acl-long.483_1458"}
+{"input": "prototype vectors is used for Method| context: prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning . typically , prompt - based tuning wraps the input text into a cloze question .", "entity": "prototype vectors", "output": "verbalizers", "neg_sample": ["prototype vectors is used for Method", "prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning .", "typically , prompt - based tuning wraps the input text into a cloze question ."], "relation": "used for", "id": "2022.acl-long.483", "year": 2022, "rel_sent": "Specifically , ProtoVerb learns prototype vectors as verbalizers by contrastive learning .", "forward": true, "src_ids": "2022.acl-long.483_1459"}
+{"input": "verbalizers is done by using OtherScientificTerm| context: prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning . typically , prompt - based tuning wraps the input text into a cloze question . to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built .", "entity": "verbalizers", "output": "prototype vectors", "neg_sample": ["verbalizers is done by using OtherScientificTerm", "prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning .", "typically , prompt - based tuning wraps the input text into a cloze question .", "to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built ."], "relation": "used for", "id": "2022.acl-long.483", "year": 2022, "rel_sent": "Specifically , ProtoVerb learns prototype vectors as verbalizers by contrastive learning .", "forward": false, "src_ids": "2022.acl-long.483_1460"}
+{"input": "prototypical verbalizer ( protoverb ) is used for OtherScientificTerm| context: prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning . typically , prompt - based tuning wraps the input text into a cloze question . to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built .", "entity": "prototypical verbalizer ( protoverb )", "output": "prototype vectors", "neg_sample": ["prototypical verbalizer ( protoverb ) is used for OtherScientificTerm", "prompt - based tuning for pre - trained language models ( plms ) has shown its effectiveness in few - shot learning .", "typically , prompt - based tuning wraps the input text into a cloze question .", "to make predictions , the model maps the output words to labels via a verbalizer , which is either manually designed or automatically built ."], "relation": "used for", "id": "2022.acl-long.483", "year": 2022, "rel_sent": "Specifically , ProtoVerb learns prototype vectors as verbalizers by contrastive learning .", "forward": true, "src_ids": "2022.acl-long.483_1461"}
+{"input": "emotions is done by using Task| context: in the field of sentiment analysis , several studies have highlighted that a single sentence may express multiple , sometimes contrasting , sentiments and emotions , each with its own experiencer , target and/or cause . to this end , over the past few years researchers have started to collect and annotate data manually , in order to investigate the capabilities of automatic systems not only to distinguish between emotions , but also to capture their semantic constituents . however , currently available gold datasets are heterogeneous in size , domain , format , splits , emotion categories and role labels , making comparisons across different works difficult and hampering progress in the area .", "entity": "emotions", "output": "semantic role labeling", "neg_sample": ["emotions is done by using Task", "in the field of sentiment analysis , several studies have highlighted that a single sentence may express multiple , sometimes contrasting , sentiments and emotions , each with its own experiencer , target and/or cause .", "to this end , over the past few years researchers have started to collect and annotate data manually , in order to investigate the capabilities of automatic systems not only to distinguish between emotions , but also to capture their semantic constituents .", "however , currently available gold datasets are heterogeneous in size , domain , format , splits , emotion categories and role labels , making comparisons across different works difficult and hampering progress in the area ."], "relation": "used for", "id": "2022.acl-long.314", "year": 2022, "rel_sent": "SRL4E - Semantic Role Labeling for Emotions : A Unified Evaluation Framework.", "forward": false, "src_ids": "2022.acl-long.314_1462"}
+{"input": "semantic role labeling is used for OtherScientificTerm| context: however , currently available gold datasets are heterogeneous in size , domain , format , splits , emotion categories and role labels , making comparisons across different works difficult and hampering progress in the area .", "entity": "semantic role labeling", "output": "emotions", "neg_sample": ["semantic role labeling is used for OtherScientificTerm", "however , currently available gold datasets are heterogeneous in size , domain , format , splits , emotion categories and role labels , making comparisons across different works difficult and hampering progress in the area ."], "relation": "used for", "id": "2022.acl-long.314", "year": 2022, "rel_sent": "SRL4E - Semantic Role Labeling for Emotions : A Unified Evaluation Framework.", "forward": true, "src_ids": "2022.acl-long.314_1463"}
+{"input": "dialogue state tracking is done by using Method| context: in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models . however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated . apparently , it requires different dialogue history to update different slots in different turns . therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance .", "entity": "dialogue state tracking", "output": "multi - perspective dialogue collaborative selection", "neg_sample": ["dialogue state tracking is done by using Method", "in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models .", "however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated .", "apparently , it requires different dialogue history to update different slots in different turns .", "therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance ."], "relation": "used for", "id": "2022.acl-long.165", "year": 2022, "rel_sent": "Beyond the Granularity : Multi - Perspective Dialogue Collaborative Selection for Dialogue State Tracking.", "forward": false, "src_ids": "2022.acl-long.165_1464"}
+{"input": "multi - perspective dialogue collaborative selection is used for Task| context: however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated . apparently , it requires different dialogue history to update different slots in different turns . therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance .", "entity": "multi - perspective dialogue collaborative selection", "output": "dialogue state tracking", "neg_sample": ["multi - perspective dialogue collaborative selection is used for Task", "however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated .", "apparently , it requires different dialogue history to update different slots in different turns .", "therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance ."], "relation": "used for", "id": "2022.acl-long.165", "year": 2022, "rel_sent": "Beyond the Granularity : Multi - Perspective Dialogue Collaborative Selection for Dialogue State Tracking.", "forward": true, "src_ids": "2022.acl-long.165_1465"}
+{"input": "dialogue contents is done by using Method| context: in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models . however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated . apparently , it requires different dialogue history to update different slots in different turns . therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance .", "entity": "dialogue contents", "output": "dicos - dst", "neg_sample": ["dialogue contents is done by using Method", "in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models .", "however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated .", "apparently , it requires different dialogue history to update different slots in different turns .", "therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance ."], "relation": "used for", "id": "2022.acl-long.165", "year": 2022, "rel_sent": "To address this problem , we devise DiCoS - DST to dynamically select the relevant dialogue contents corresponding to each slot for state updating .", "forward": false, "src_ids": "2022.acl-long.165_1466"}
+{"input": "state updating is done by using OtherScientificTerm| context: in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models . however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated . apparently , it requires different dialogue history to update different slots in different turns .", "entity": "state updating", "output": "dialogue contents", "neg_sample": ["state updating is done by using OtherScientificTerm", "in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models .", "however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated .", "apparently , it requires different dialogue history to update different slots in different turns ."], "relation": "used for", "id": "2022.acl-long.165", "year": 2022, "rel_sent": "To address this problem , we devise DiCoS - DST to dynamically select the relevant dialogue contents corresponding to each slot for state updating .", "forward": false, "src_ids": "2022.acl-long.165_1467"}
+{"input": "dicos - dst is used for OtherScientificTerm| context: in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models . however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated . apparently , it requires different dialogue history to update different slots in different turns .", "entity": "dicos - dst", "output": "dialogue contents", "neg_sample": ["dicos - dst is used for OtherScientificTerm", "in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models .", "however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated .", "apparently , it requires different dialogue history to update different slots in different turns ."], "relation": "used for", "id": "2022.acl-long.165", "year": 2022, "rel_sent": "To address this problem , we devise DiCoS - DST to dynamically select the relevant dialogue contents corresponding to each slot for state updating .", "forward": true, "src_ids": "2022.acl-long.165_1468"}
+{"input": "dialogue contents is used for Task| context: in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models . however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated . apparently , it requires different dialogue history to update different slots in different turns . therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance .", "entity": "dialogue contents", "output": "state updating", "neg_sample": ["dialogue contents is used for Task", "in dialogue state tracking , dialogue history is a crucial material , and its utilization varies between different models .", "however , no matter how the dialogue history is used , each existing model uses its own consistent dialogue history during the entire state tracking process , regardless of which slot is updated .", "apparently , it requires different dialogue history to update different slots in different turns .", "therefore , using consistent dialogue contents may lead to insufficient or redundant information for different slots , which affects the overall performance ."], "relation": "used for", "id": "2022.acl-long.165", "year": 2022, "rel_sent": "To address this problem , we devise DiCoS - DST to dynamically select the relevant dialogue contents corresponding to each slot for state updating .", "forward": true, "src_ids": "2022.acl-long.165_1469"}
+{"input": "discriminative pre - trained language models is done by using Method| context: however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert . it is still unknown whether and how discriminative plms , e.g. , electra , can be effectively prompt - tuned .", "entity": "discriminative pre - trained language models", "output": "prompt tuning", "neg_sample": ["discriminative pre - trained language models is done by using Method", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "it is still unknown whether and how discriminative plms , e.g.", ", electra , can be effectively prompt - tuned ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "Prompt Tuning for Discriminative Pre - trained Language Models.", "forward": false, "src_ids": "2022.findings-acl.273_1470"}
+{"input": "prompt tuning is used for Method| context: recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks . however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert . it is still unknown whether and how discriminative plms , e.g. , electra , can be effectively prompt - tuned .", "entity": "prompt tuning", "output": "discriminative pre - trained language models", "neg_sample": ["prompt tuning is used for Method", "recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks .", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "it is still unknown whether and how discriminative plms , e.g.", ", electra , can be effectively prompt - tuned ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "Prompt Tuning for Discriminative Pre - trained Language Models.", "forward": true, "src_ids": "2022.findings-acl.273_1471"}
+{"input": "prompt tuning framework is used for Method| context: recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks . however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "entity": "prompt tuning framework", "output": "discriminative plms", "neg_sample": ["prompt tuning framework is used for Method", "recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks .", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "In this work , we present DPT , the first prompt tuning framework for discriminative PLMs , which reformulates NLP tasks into a discriminative language modeling problem .", "forward": true, "src_ids": "2022.findings-acl.273_1472"}
+{"input": "discriminative language modeling problem is done by using Method| context: recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks . however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert . it is still unknown whether and how discriminative plms , e.g. , electra , can be effectively prompt - tuned .", "entity": "discriminative language modeling problem", "output": "dpt", "neg_sample": ["discriminative language modeling problem is done by using Method", "recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks .", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "it is still unknown whether and how discriminative plms , e.g.", ", electra , can be effectively prompt - tuned ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "In this work , we present DPT , the first prompt tuning framework for discriminative PLMs , which reformulates NLP tasks into a discriminative language modeling problem .", "forward": false, "src_ids": "2022.findings-acl.273_1473"}
+{"input": "nlp tasks is done by using Method| context: recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks . however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert . it is still unknown whether and how discriminative plms , e.g. , electra , can be effectively prompt - tuned .", "entity": "nlp tasks", "output": "dpt", "neg_sample": ["nlp tasks is done by using Method", "recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks .", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "it is still unknown whether and how discriminative plms , e.g.", ", electra , can be effectively prompt - tuned ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "In this work , we present DPT , the first prompt tuning framework for discriminative PLMs , which reformulates NLP tasks into a discriminative language modeling problem .", "forward": false, "src_ids": "2022.findings-acl.273_1474"}
+{"input": "discriminative plms is done by using Method| context: recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks . however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert . it is still unknown whether and how discriminative plms , e.g. , electra , can be effectively prompt - tuned .", "entity": "discriminative plms", "output": "prompt tuning framework", "neg_sample": ["discriminative plms is done by using Method", "recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks .", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "it is still unknown whether and how discriminative plms , e.g.", ", electra , can be effectively prompt - tuned ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "In this work , we present DPT , the first prompt tuning framework for discriminative PLMs , which reformulates NLP tasks into a discriminative language modeling problem .", "forward": false, "src_ids": "2022.findings-acl.273_1475"}
+{"input": "dpt is used for Task| context: recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks . however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert . it is still unknown whether and how discriminative plms , e.g. , electra , can be effectively prompt - tuned .", "entity": "dpt", "output": "nlp tasks", "neg_sample": ["dpt is used for Task", "recent works have shown promising results of prompt tuning in stimulating pre - trained language models ( plms ) for natural language processing ( nlp ) tasks .", "however , to the best of our knowledge , existing works focus on prompt - tuning generative plms that are pre - trained to generate target tokens , such as bert .", "it is still unknown whether and how discriminative plms , e.g.", ", electra , can be effectively prompt - tuned ."], "relation": "used for", "id": "2022.findings-acl.273", "year": 2022, "rel_sent": "In this work , we present DPT , the first prompt tuning framework for discriminative PLMs , which reformulates NLP tasks into a discriminative language modeling problem .", "forward": true, "src_ids": "2022.findings-acl.273_1476"}
+{"input": "prompt engineering is done by using Method| context: pre - trained language models derive substantial linguistic and factual knowledge from the massive corpora on which they are trained , and prompt engineering seeks to align these models to specific tasks . unfortunately , existing prompt engineering methods require significant amounts of labeled data , access to model parameters , or both .", "entity": "prompt engineering", "output": "information - theoretic approach", "neg_sample": ["prompt engineering is done by using Method", "pre - trained language models derive substantial linguistic and factual knowledge from the massive corpora on which they are trained , and prompt engineering seeks to align these models to specific tasks .", "unfortunately , existing prompt engineering methods require significant amounts of labeled data , access to model parameters , or both ."], "relation": "used for", "id": "2022.acl-long.60", "year": 2022, "rel_sent": "An Information - theoretic Approach to Prompt Engineering Without Ground Truth Labels.", "forward": false, "src_ids": "2022.acl-long.60_1477"}
+{"input": "cross - lingual spoken language understanding is done by using Method| context: due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort . however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "cross - lingual spoken language understanding", "output": "global - local contrastive learning framework", "neg_sample": ["cross - lingual spoken language understanding is done by using Method", "due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort .", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.191", "year": 2022, "rel_sent": "GL - CLeF : A Global - Local Contrastive Learning Framework for Cross - lingual Spoken Language Understanding.", "forward": false, "src_ids": "2022.acl-long.191_1478"}
+{"input": "global - local contrastive learning framework is used for Task| context: however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "global - local contrastive learning framework", "output": "cross - lingual spoken language understanding", "neg_sample": ["global - local contrastive learning framework is used for Task", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.191", "year": 2022, "rel_sent": "GL - CLeF : A Global - Local Contrastive Learning Framework for Cross - lingual Spoken Language Understanding.", "forward": true, "src_ids": "2022.acl-long.191_1479"}
+{"input": "multilingual views is done by using Material| context: due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort . however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "multilingual views", "output": "bilingual dictionaries", "neg_sample": ["multilingual views is done by using Material", "due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort .", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.191", "year": 2022, "rel_sent": "Specifically , we employ contrastive learning , leveraging bilingual dictionaries to construct multilingual views of the same utterance , then encourage their representations to be more similar than negative example pairs , which achieves to explicitly align representations of similar sentences across languages .", "forward": false, "src_ids": "2022.acl-long.191_1480"}
+{"input": "bilingual dictionaries is used for OtherScientificTerm| context: due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort . however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "bilingual dictionaries", "output": "multilingual views", "neg_sample": ["bilingual dictionaries is used for OtherScientificTerm", "due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort .", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.191", "year": 2022, "rel_sent": "Specifically , we employ contrastive learning , leveraging bilingual dictionaries to construct multilingual views of the same utterance , then encourage their representations to be more similar than negative example pairs , which achieves to explicitly align representations of similar sentences across languages .", "forward": true, "src_ids": "2022.acl-long.191_1481"}
+{"input": "fine - grained cross - lingual transfer is done by using Method| context: due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort . however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "fine - grained cross - lingual transfer", "output": "local and global component", "neg_sample": ["fine - grained cross - lingual transfer is done by using Method", "due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort .", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.191", "year": 2022, "rel_sent": "In addition , a key step in GL - CLeF is a proposed Local and Global component , which achieves a fine - grained cross - lingual transfer ( i.e. , sentence - level Local intent transfer , token - level Local slot transfer , and semantic - level Global transfer across intent and slot ) .", "forward": false, "src_ids": "2022.acl-long.191_1482"}
+{"input": "local and global component is used for Task| context: due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort . however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "local and global component", "output": "fine - grained cross - lingual transfer", "neg_sample": ["local and global component is used for Task", "due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort .", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.191", "year": 2022, "rel_sent": "In addition , a key step in GL - CLeF is a proposed Local and Global component , which achieves a fine - grained cross - lingual transfer ( i.e. , sentence - level Local intent transfer , token - level Local slot transfer , and semantic - level Global transfer across intent and slot ) .", "forward": true, "src_ids": "2022.acl-long.191_1483"}
+{"input": "cross - device federated learning is done by using Method| context: most studies in cross - device federated learning focus on small models , due to the server - client communication and on - device computation bottlenecks .", "entity": "cross - device federated learning", "output": "larger language models", "neg_sample": ["cross - device federated learning is done by using Method", "most studies in cross - device federated learning focus on small models , due to the server - client communication and on - device computation bottlenecks ."], "relation": "used for", "id": "2022.fl4nlp-1.2", "year": 2022, "rel_sent": "In this work , we leverage various techniques for mitigating these bottlenecks to train larger language models in cross - device federated learning .", "forward": false, "src_ids": "2022.fl4nlp-1.2_1484"}
+{"input": "vector arithmetic is used for Task| context: prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective . we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model .", "entity": "vector arithmetic", "output": "unsupervised sentiment transfer", "neg_sample": ["vector arithmetic is used for Task", "prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective .", "we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model ."], "relation": "used for", "id": "2022.findings-acl.48", "year": 2022, "rel_sent": "We show that vector arithmetic can be used for unsupervised sentiment transfer on the Yelp sentiment benchmark , with performance comparable to models tailored to this task .", "forward": true, "src_ids": "2022.findings-acl.48_1485"}
+{"input": "sentence similarity is done by using OtherScientificTerm| context: prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective . we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model .", "entity": "sentence similarity", "output": "steering vectors", "neg_sample": ["sentence similarity is done by using OtherScientificTerm", "prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective .", "we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model ."], "relation": "used for", "id": "2022.findings-acl.48", "year": 2022, "rel_sent": "We find that distances between steering vectors reflect sentence similarity when evaluated on a textual similarity benchmark ( STS - B ) , outperforming pooled hidden states of models .", "forward": false, "src_ids": "2022.findings-acl.48_1486"}
+{"input": "steering vectors is used for OtherScientificTerm| context: prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective . we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model .", "entity": "steering vectors", "output": "sentence similarity", "neg_sample": ["steering vectors is used for OtherScientificTerm", "prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective .", "we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model ."], "relation": "used for", "id": "2022.findings-acl.48", "year": 2022, "rel_sent": "We find that distances between steering vectors reflect sentence similarity when evaluated on a textual similarity benchmark ( STS - B ) , outperforming pooled hidden states of models .", "forward": true, "src_ids": "2022.findings-acl.48_1487"}
+{"input": "unsupervised sentiment transfer is done by using Method| context: prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective . we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model .", "entity": "unsupervised sentiment transfer", "output": "vector arithmetic", "neg_sample": ["unsupervised sentiment transfer is done by using Method", "prior work on controllable text generation has focused on learning how to control language models through trainable decoding , smart - prompt design , or fine - tuning based on a desired objective .", "we hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model ."], "relation": "used for", "id": "2022.findings-acl.48", "year": 2022, "rel_sent": "We show that vector arithmetic can be used for unsupervised sentiment transfer on the Yelp sentiment benchmark , with performance comparable to models tailored to this task .", "forward": false, "src_ids": "2022.findings-acl.48_1488"}
+{"input": "public administration is done by using Method| context: this paper is about a multilingual chatbot developed for public administration within the cef funded project enrich4all .", "entity": "public administration", "output": "machine translation - powered chatbot", "neg_sample": ["public administration is done by using Method", "this paper is about a multilingual chatbot developed for public administration within the cef funded project enrich4all ."], "relation": "used for", "id": "2022.eamt-1.54", "year": 2022, "rel_sent": "A Machine Translation - Powered Chatbot for Public Administration.", "forward": false, "src_ids": "2022.eamt-1.54_1489"}
+{"input": "chatbot solution is done by using Method| context: this paper is about a multilingual chatbot developed for public administration within the cef funded project enrich4all .", "entity": "chatbot solution", "output": "cef etranslation service", "neg_sample": ["chatbot solution is done by using Method", "this paper is about a multilingual chatbot developed for public administration within the cef funded project enrich4all ."], "relation": "used for", "id": "2022.eamt-1.54", "year": 2022, "rel_sent": "We argue for multi - lingual chatbots empowered through MT and discuss the integration of the CEF eTranslation service in a chatbot solution .", "forward": false, "src_ids": "2022.eamt-1.54_1490"}
+{"input": "cef etranslation service is used for Method| context: this paper is about a multilingual chatbot developed for public administration within the cef funded project enrich4all .", "entity": "cef etranslation service", "output": "chatbot solution", "neg_sample": ["cef etranslation service is used for Method", "this paper is about a multilingual chatbot developed for public administration within the cef funded project enrich4all ."], "relation": "used for", "id": "2022.eamt-1.54", "year": 2022, "rel_sent": "We argue for multi - lingual chatbots empowered through MT and discuss the integration of the CEF eTranslation service in a chatbot solution .", "forward": true, "src_ids": "2022.eamt-1.54_1491"}
+{"input": "automated wolof universal dependency parsing is done by using OtherScientificTerm| context: focus on language - specific properties with insights from formal minimalist syntax can improve universal dependency ( ud ) parsing . such improvements are especially sensitive for low - resource african languages , like wolof , which have fewer ud treebanks in number and amount of annotations , and fewer contributing annotators .", "entity": "automated wolof universal dependency parsing", "output": "syntactic insights", "neg_sample": ["automated wolof universal dependency parsing is done by using OtherScientificTerm", "focus on language - specific properties with insights from formal minimalist syntax can improve universal dependency ( ud ) parsing .", "such improvements are especially sensitive for low - resource african languages , like wolof , which have fewer ud treebanks in number and amount of annotations , and fewer contributing annotators ."], "relation": "used for", "id": "2022.computel-1.2", "year": 2022, "rel_sent": "New syntactic insights for automated Wolof Universal Dependency parsing.", "forward": false, "src_ids": "2022.computel-1.2_1492"}
+{"input": "syntactic insights is used for Task| context: focus on language - specific properties with insights from formal minimalist syntax can improve universal dependency ( ud ) parsing . such improvements are especially sensitive for low - resource african languages , like wolof , which have fewer ud treebanks in number and amount of annotations , and fewer contributing annotators .", "entity": "syntactic insights", "output": "automated wolof universal dependency parsing", "neg_sample": ["syntactic insights is used for Task", "focus on language - specific properties with insights from formal minimalist syntax can improve universal dependency ( ud ) parsing .", "such improvements are especially sensitive for low - resource african languages , like wolof , which have fewer ud treebanks in number and amount of annotations , and fewer contributing annotators ."], "relation": "used for", "id": "2022.computel-1.2", "year": 2022, "rel_sent": "New syntactic insights for automated Wolof Universal Dependency parsing.", "forward": true, "src_ids": "2022.computel-1.2_1493"}
+{"input": "free - form multiple choice question answering is done by using Task| context: pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks . this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem . while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration .", "entity": "free - form multiple choice question answering", "output": "answer - level calibration", "neg_sample": ["free - form multiple choice question answering is done by using Task", "pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks .", "this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem .", "while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration ."], "relation": "used for", "id": "2022.acl-long.49", "year": 2022, "rel_sent": "Answer - level Calibration for Free - form Multiple Choice Question Answering.", "forward": false, "src_ids": "2022.acl-long.49_1494"}
+{"input": "biases is done by using Task| context: pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks . this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem . while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration .", "entity": "biases", "output": "answer - level calibration", "neg_sample": ["biases is done by using Task", "pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks .", "this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem .", "while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration ."], "relation": "used for", "id": "2022.acl-long.49", "year": 2022, "rel_sent": "Our analysis indicates that answer - level calibration is able to remove such biases and leads to a more robust measure of model capability .", "forward": false, "src_ids": "2022.acl-long.49_1495"}
+{"input": "model capability is done by using Task| context: pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks . this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem . while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration .", "entity": "model capability", "output": "answer - level calibration", "neg_sample": ["model capability is done by using Task", "pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks .", "this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem .", "while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration ."], "relation": "used for", "id": "2022.acl-long.49", "year": 2022, "rel_sent": "Our analysis indicates that answer - level calibration is able to remove such biases and leads to a more robust measure of model capability .", "forward": false, "src_ids": "2022.acl-long.49_1496"}
+{"input": "answer - level calibration is used for Task| context: pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks . this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem . while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration .", "entity": "answer - level calibration", "output": "free - form multiple choice question answering", "neg_sample": ["answer - level calibration is used for Task", "pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks .", "this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem .", "while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration ."], "relation": "used for", "id": "2022.acl-long.49", "year": 2022, "rel_sent": "Answer - level Calibration for Free - form Multiple Choice Question Answering.", "forward": true, "src_ids": "2022.acl-long.49_1497"}
+{"input": "answer - level calibration is used for OtherScientificTerm| context: pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks . this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem . while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration .", "entity": "answer - level calibration", "output": "biases", "neg_sample": ["answer - level calibration is used for OtherScientificTerm", "pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks .", "this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem .", "while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration ."], "relation": "used for", "id": "2022.acl-long.49", "year": 2022, "rel_sent": "Our analysis indicates that answer - level calibration is able to remove such biases and leads to a more robust measure of model capability .", "forward": true, "src_ids": "2022.acl-long.49_1498"}
+{"input": "answer - level calibration is used for Metric| context: pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks . this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem . while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration .", "entity": "answer - level calibration", "output": "model capability", "neg_sample": ["answer - level calibration is used for Metric", "pre - trained language models have recently shown that training on large corpora using the language modeling objective enables few - shot and zero - shot capabilities on a variety of nlp tasks , including commonsense reasoning tasks .", "this is achieved using text interactions with the model , usually by posing the task as a natural language text completion problem .", "while using language model probabilities to obtain task specific scores has been generally useful , it often requires task - specific heuristics such as length normalization , or probability calibration ."], "relation": "used for", "id": "2022.acl-long.49", "year": 2022, "rel_sent": "Our analysis indicates that answer - level calibration is able to remove such biases and leads to a more robust measure of model capability .", "forward": true, "src_ids": "2022.acl-long.49_1499"}
+{"input": "regularizers is used for Method| context: dominant approaches to disentangle a sensitive attribute from textual representations rely on learning simultaneously a penalization term that involves either an adversary loss ( e.g. , a discriminator ) or an information measure ( e.g. , mutual information ) . as a matter of fact , the resulting nested optimization loop is both times consuming , adding complexity to the optimization dynamic , and requires a fine hyperparameter selection ( e.g. , learning rates , architecture ) .", "entity": "regularizers", "output": "disentangled representations", "neg_sample": ["regularizers is used for Method", "dominant approaches to disentangle a sensitive attribute from textual representations rely on learning simultaneously a penalization term that involves either an adversary loss ( e.g.", ", a discriminator ) or an information measure ( e.g.", ", mutual information ) .", "as a matter of fact , the resulting nested optimization loop is both times consuming , adding complexity to the optimization dynamic , and requires a fine hyperparameter selection ( e.g.", ", learning rates , architecture ) ."], "relation": "used for", "id": "2022.acl-long.187", "year": 2022, "rel_sent": "In this work , we introduce a family of regularizers for learning disentangled representations that do not require training .", "forward": true, "src_ids": "2022.acl-long.187_1500"}
+{"input": "disentangled representations is done by using Method| context: when working with textual data , a natural application of disentangled representations is the fair classification where the goal is to make predictions without being biased ( or influenced ) by sensible attributes that may be present in the data ( e.g. , age , gender or race ) . dominant approaches to disentangle a sensitive attribute from textual representations rely on learning simultaneously a penalization term that involves either an adversary loss ( e.g. , a discriminator ) or an information measure ( e.g. , mutual information ) . as a matter of fact , the resulting nested optimization loop is both times consuming , adding complexity to the optimization dynamic , and requires a fine hyperparameter selection ( e.g. , learning rates , architecture ) .", "entity": "disentangled representations", "output": "regularizers", "neg_sample": ["disentangled representations is done by using Method", "when working with textual data , a natural application of disentangled representations is the fair classification where the goal is to make predictions without being biased ( or influenced ) by sensible attributes that may be present in the data ( e.g.", ", age , gender or race ) .", "dominant approaches to disentangle a sensitive attribute from textual representations rely on learning simultaneously a penalization term that involves either an adversary loss ( e.g.", ", a discriminator ) or an information measure ( e.g.", ", mutual information ) .", "as a matter of fact , the resulting nested optimization loop is both times consuming , adding complexity to the optimization dynamic , and requires a fine hyperparameter selection ( e.g.", ", learning rates , architecture ) ."], "relation": "used for", "id": "2022.acl-long.187", "year": 2022, "rel_sent": "In this work , we introduce a family of regularizers for learning disentangled representations that do not require training .", "forward": false, "src_ids": "2022.acl-long.187_1501"}
+{"input": "vanilla pseudo - labeling based methods is used for Task| context: recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive .", "entity": "vanilla pseudo - labeling based methods", "output": "sequence - to - sequence model distillation", "neg_sample": ["vanilla pseudo - labeling based methods is used for Task", "recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive ."], "relation": "used for", "id": "2022.acl-long.11", "year": 2022, "rel_sent": "Pseudo - labeling based methods are popular in sequence - to - sequence model distillation .", "forward": true, "src_ids": "2022.acl-long.11_1502"}
+{"input": "student models is done by using OtherScientificTerm| context: recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive .", "entity": "student models", "output": "pseudo labels", "neg_sample": ["student models is done by using OtherScientificTerm", "recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive ."], "relation": "used for", "id": "2022.acl-long.11", "year": 2022, "rel_sent": "In this paper , we find simply manipulating attention temperatures in Transformers can make pseudo labels easier to learn for student models .", "forward": false, "src_ids": "2022.acl-long.11_1503"}
+{"input": "pseudo labels is used for Method| context: recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive .", "entity": "pseudo labels", "output": "student models", "neg_sample": ["pseudo labels is used for Method", "recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive ."], "relation": "used for", "id": "2022.acl-long.11", "year": 2022, "rel_sent": "In this paper , we find simply manipulating attention temperatures in Transformers can make pseudo labels easier to learn for student models .", "forward": true, "src_ids": "2022.acl-long.11_1504"}
+{"input": "sequence - to - sequence model distillation is done by using Method| context: recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive .", "entity": "sequence - to - sequence model distillation", "output": "vanilla pseudo - labeling based methods", "neg_sample": ["sequence - to - sequence model distillation is done by using Method", "recent progress of abstractive text summarization largely relies on large pre - trained sequence - to - sequence transformer models , which are computationally expensive ."], "relation": "used for", "id": "2022.acl-long.11", "year": 2022, "rel_sent": "Pseudo - labeling based methods are popular in sequence - to - sequence model distillation .", "forward": false, "src_ids": "2022.acl-long.11_1505"}
+{"input": "static - dynamic model is used for Task| context: empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation . existing work for empathetic dialogue generation concentrates on the two - party conversation scenario . multi - party dialogues , however , are pervasive in reality . furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings .", "entity": "static - dynamic model", "output": "multi - party empathetic dialogue generation", "neg_sample": ["static - dynamic model is used for Task", "empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation .", "existing work for empathetic dialogue generation concentrates on the two - party conversation scenario .", "multi - party dialogues , however , are pervasive in reality .", "furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings ."], "relation": "used for", "id": "2022.acl-long.24", "year": 2022, "rel_sent": "Additionally , a Static - Dynamic model for Multi - Party Empathetic Dialogue Generation , SDMPED , is introduced as a baseline by exploring the static sensibility and dynamic emotion for the multi - party empathetic dialogue learning , the aspects that help SDMPED achieve the state - of - the - art performance .", "forward": true, "src_ids": "2022.acl-long.24_1506"}
+{"input": "multi - party empathetic dialogue generation is done by using Method| context: empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation . existing work for empathetic dialogue generation concentrates on the two - party conversation scenario . multi - party dialogues , however , are pervasive in reality . furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings .", "entity": "multi - party empathetic dialogue generation", "output": "static - dynamic model", "neg_sample": ["multi - party empathetic dialogue generation is done by using Method", "empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation .", "existing work for empathetic dialogue generation concentrates on the two - party conversation scenario .", "multi - party dialogues , however , are pervasive in reality .", "furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings ."], "relation": "used for", "id": "2022.acl-long.24", "year": 2022, "rel_sent": "Additionally , a Static - Dynamic model for Multi - Party Empathetic Dialogue Generation , SDMPED , is introduced as a baseline by exploring the static sensibility and dynamic emotion for the multi - party empathetic dialogue learning , the aspects that help SDMPED achieve the state - of - the - art performance .", "forward": false, "src_ids": "2022.acl-long.24_1507"}
+{"input": "multi - party empathetic dialogue learning is done by using OtherScientificTerm| context: empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation . existing work for empathetic dialogue generation concentrates on the two - party conversation scenario . multi - party dialogues , however , are pervasive in reality . furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings .", "entity": "multi - party empathetic dialogue learning", "output": "static sensibility", "neg_sample": ["multi - party empathetic dialogue learning is done by using OtherScientificTerm", "empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation .", "existing work for empathetic dialogue generation concentrates on the two - party conversation scenario .", "multi - party dialogues , however , are pervasive in reality .", "furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings ."], "relation": "used for", "id": "2022.acl-long.24", "year": 2022, "rel_sent": "Additionally , a Static - Dynamic model for Multi - Party Empathetic Dialogue Generation , SDMPED , is introduced as a baseline by exploring the static sensibility and dynamic emotion for the multi - party empathetic dialogue learning , the aspects that help SDMPED achieve the state - of - the - art performance .", "forward": false, "src_ids": "2022.acl-long.24_1508"}
+{"input": "dynamic emotion is used for Method| context: empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation . existing work for empathetic dialogue generation concentrates on the two - party conversation scenario . multi - party dialogues , however , are pervasive in reality . furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings .", "entity": "dynamic emotion", "output": "multi - party empathetic dialogue learning", "neg_sample": ["dynamic emotion is used for Method", "empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation .", "existing work for empathetic dialogue generation concentrates on the two - party conversation scenario .", "multi - party dialogues , however , are pervasive in reality .", "furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings ."], "relation": "used for", "id": "2022.acl-long.24", "year": 2022, "rel_sent": "Additionally , a Static - Dynamic model for Multi - Party Empathetic Dialogue Generation , SDMPED , is introduced as a baseline by exploring the static sensibility and dynamic emotion for the multi - party empathetic dialogue learning , the aspects that help SDMPED achieve the state - of - the - art performance .", "forward": true, "src_ids": "2022.acl-long.24_1509"}
+{"input": "static sensibility is used for Method| context: empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation . existing work for empathetic dialogue generation concentrates on the two - party conversation scenario . multi - party dialogues , however , are pervasive in reality . furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings .", "entity": "static sensibility", "output": "multi - party empathetic dialogue learning", "neg_sample": ["static sensibility is used for Method", "empathetic dialogue assembles emotion understanding , feeling projection , and appropriate response generation .", "existing work for empathetic dialogue generation concentrates on the two - party conversation scenario .", "multi - party dialogues , however , are pervasive in reality .", "furthermore , emotion and sensibility are typically confused ; a refined empathy analysis is needed for comprehending fragile and nuanced human feelings ."], "relation": "used for", "id": "2022.acl-long.24", "year": 2022, "rel_sent": "Additionally , a Static - Dynamic model for Multi - Party Empathetic Dialogue Generation , SDMPED , is introduced as a baseline by exploring the static sensibility and dynamic emotion for the multi - party empathetic dialogue learning , the aspects that help SDMPED achieve the state - of - the - art performance .", "forward": true, "src_ids": "2022.acl-long.24_1510"}
+{"input": "human label confidence scores is done by using Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "human label confidence scores", "output": "contrastive meta - learning framework ( cml )", "neg_sample": ["human label confidence scores is done by using Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "CML : A Contrastive Meta Learning Method to Estimate Human Label Confidence Scores and Reduce Data Collection Cost.", "forward": false, "src_ids": "2022.ecnlp-1.5_1511"}
+{"input": "natural language processing is done by using Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "natural language processing", "output": "contrastive meta - learning framework ( cml )", "neg_sample": ["natural language processing is done by using Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "In this paper , we propose a contrastive meta - learning framework ( CML ) to address the challenges introduced by noisy annotated data , specifically in the context of natural language processing .", "forward": false, "src_ids": "2022.ecnlp-1.5_1512"}
+{"input": "text feature representations is done by using Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "text feature representations", "output": "contrastive meta - learning framework ( cml )", "neg_sample": ["text feature representations is done by using Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "CML combines contrastive and meta learning to improve the quality of text feature representations .", "forward": false, "src_ids": "2022.ecnlp-1.5_1513"}
+{"input": "contrastive meta - learning framework ( cml ) is used for Metric| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "contrastive meta - learning framework ( cml )", "output": "human label confidence scores", "neg_sample": ["contrastive meta - learning framework ( cml ) is used for Metric", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "CML : A Contrastive Meta Learning Method to Estimate Human Label Confidence Scores and Reduce Data Collection Cost.", "forward": true, "src_ids": "2022.ecnlp-1.5_1514"}
+{"input": "meta - learning is used for Metric| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "entity": "meta - learning", "output": "label quality", "neg_sample": ["meta - learning is used for Metric", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "Meta - learning is also used to generate confidence scores to assess label quality .", "forward": true, "src_ids": "2022.ecnlp-1.5_1515"}
+{"input": "contrastive meta - learning framework ( cml ) is used for Task| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "contrastive meta - learning framework ( cml )", "output": "natural language processing", "neg_sample": ["contrastive meta - learning framework ( cml ) is used for Task", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "In this paper , we propose a contrastive meta - learning framework ( CML ) to address the challenges introduced by noisy annotated data , specifically in the context of natural language processing .", "forward": true, "src_ids": "2022.ecnlp-1.5_1516"}
+{"input": "contrastive meta - learning framework ( cml ) is used for Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "contrastive meta - learning framework ( cml )", "output": "text feature representations", "neg_sample": ["contrastive meta - learning framework ( cml ) is used for Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "CML combines contrastive and meta learning to improve the quality of text feature representations .", "forward": true, "src_ids": "2022.ecnlp-1.5_1517"}
+{"input": "contrastive and meta learning is used for Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "contrastive and meta learning", "output": "text feature representations", "neg_sample": ["contrastive and meta learning is used for Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "CML combines contrastive and meta learning to improve the quality of text feature representations .", "forward": true, "src_ids": "2022.ecnlp-1.5_1518"}
+{"input": "confidence scores is done by using Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "confidence scores", "output": "meta - learning", "neg_sample": ["confidence scores is done by using Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "Meta - learning is also used to generate confidence scores to assess label quality .", "forward": false, "src_ids": "2022.ecnlp-1.5_1519"}
+{"input": "label quality is done by using Method| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "label quality", "output": "meta - learning", "neg_sample": ["label quality is done by using Method", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "Meta - learning is also used to generate confidence scores to assess label quality .", "forward": false, "src_ids": "2022.ecnlp-1.5_1520"}
+{"input": "meta - learning is used for OtherScientificTerm| context: deep neural network models are especially susceptible to noise in annotated labels . in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias . label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly .", "entity": "meta - learning", "output": "confidence scores", "neg_sample": ["meta - learning is used for OtherScientificTerm", "deep neural network models are especially susceptible to noise in annotated labels .", "in the real world , annotated data typically contains noise caused by a variety of factors such as task difficulty , annotator experience , and annotator bias .", "label quality is critical for label validation tasks ; however , correcting for noise by collecting more data is often costly ."], "relation": "used for", "id": "2022.ecnlp-1.5", "year": 2022, "rel_sent": "Meta - learning is also used to generate confidence scores to assess label quality .", "forward": true, "src_ids": "2022.ecnlp-1.5_1521"}
+{"input": "culture - related research is done by using Method| context: while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language .", "entity": "culture - related research", "output": "encbp", "neg_sample": ["culture - related research is done by using Method", "while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language ."], "relation": "used for", "id": "2022.findings-acl.221", "year": 2022, "rel_sent": "Our findings strongly support the importance of cultural background modeling to a wide variety of NLP tasks and demonstrate the applicability of EnCBP in culture - related research .", "forward": false, "src_ids": "2022.findings-acl.221_1522"}
+{"input": "nlp tasks is done by using Method| context: while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language .", "entity": "nlp tasks", "output": "cultural background modeling", "neg_sample": ["nlp tasks is done by using Method", "while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language ."], "relation": "used for", "id": "2022.findings-acl.221", "year": 2022, "rel_sent": "Our findings strongly support the importance of cultural background modeling to a wide variety of NLP tasks and demonstrate the applicability of EnCBP in culture - related research .", "forward": false, "src_ids": "2022.findings-acl.221_1523"}
+{"input": "cultural background modeling is used for Task| context: while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language .", "entity": "cultural background modeling", "output": "nlp tasks", "neg_sample": ["cultural background modeling is used for Task", "while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language ."], "relation": "used for", "id": "2022.findings-acl.221", "year": 2022, "rel_sent": "Our findings strongly support the importance of cultural background modeling to a wide variety of NLP tasks and demonstrate the applicability of EnCBP in culture - related research .", "forward": true, "src_ids": "2022.findings-acl.221_1524"}
+{"input": "encbp is used for Task| context: while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language .", "entity": "encbp", "output": "culture - related research", "neg_sample": ["encbp is used for Task", "while cultural backgrounds have been shown to affect linguistic expressions , existing natural language processing ( nlp ) research on culture modeling is overly coarse - grained and does not examine cultural differences among speakers of the same language ."], "relation": "used for", "id": "2022.findings-acl.221", "year": 2022, "rel_sent": "Our findings strongly support the importance of cultural background modeling to a wide variety of NLP tasks and demonstrate the applicability of EnCBP in culture - related research .", "forward": true, "src_ids": "2022.findings-acl.221_1525"}
+{"input": "text classifiers is done by using OtherScientificTerm| context: graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art .", "entity": "text classifiers", "output": "synthetic graphs", "neg_sample": ["text classifiers is done by using OtherScientificTerm", "graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art ."], "relation": "used for", "id": "2022.acl-long.279", "year": 2022, "rel_sent": "These results question the importance of synthetic graphs used in modern text classifiers .", "forward": false, "src_ids": "2022.acl-long.279_1526"}
+{"input": "synthetic graphs is used for Method| context: graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art .", "entity": "synthetic graphs", "output": "text classifiers", "neg_sample": ["synthetic graphs is used for Method", "graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art ."], "relation": "used for", "id": "2022.acl-long.279", "year": 2022, "rel_sent": "These results question the importance of synthetic graphs used in modern text classifiers .", "forward": true, "src_ids": "2022.acl-long.279_1527"}
+{"input": "inductive text classification setting is done by using Method| context: graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art .", "entity": "inductive text classification setting", "output": "hetegcn", "neg_sample": ["inductive text classification setting is done by using Method", "graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art ."], "relation": "used for", "id": "2022.acl-long.279", "year": 2022, "rel_sent": "We show that a wide multi - layer perceptron ( MLP ) using a Bag - of - Words ( BoW ) outperforms the recent graph - based models TextGCN and HeteGCN in an inductive text classification setting and is comparable with HyperGAT .", "forward": false, "src_ids": "2022.acl-long.279_1528"}
+{"input": "hetegcn is used for Task| context: graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art .", "entity": "hetegcn", "output": "inductive text classification setting", "neg_sample": ["hetegcn is used for Task", "graph neural networks have triggered a resurgence of graph - based text classification methods , defining today 's state of the art ."], "relation": "used for", "id": "2022.acl-long.279", "year": 2022, "rel_sent": "We show that a wide multi - layer perceptron ( MLP ) using a Bag - of - Words ( BoW ) outperforms the recent graph - based models TextGCN and HeteGCN in an inductive text classification setting and is comparable with HyperGAT .", "forward": true, "src_ids": "2022.acl-long.279_1529"}
+{"input": "language model training is done by using Method| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "language model training", "output": "adaptive differential privacy", "neg_sample": ["language model training is done by using Method", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "Adaptive Differential Privacy for Language Model Training.", "forward": false, "src_ids": "2022.fl4nlp-1.3_1530"}
+{"input": "differentially private language modeling is done by using Method| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "differentially private language modeling", "output": "adaptive differential privacy", "neg_sample": ["differentially private language modeling is done by using Method", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "Experiments demonstrate that our ADP improves differentially private language modeling to achieve good protection from canary attackers .", "forward": false, "src_ids": "2022.fl4nlp-1.3_1531"}
+{"input": "adaptive differential privacy is used for Method| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "adaptive differential privacy", "output": "language model training", "neg_sample": ["adaptive differential privacy is used for Method", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "Adaptive Differential Privacy for Language Model Training.", "forward": true, "src_ids": "2022.fl4nlp-1.3_1532"}
+{"input": "language modeling is done by using Method| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "language modeling", "output": "adaptive differential privacy ( adp ) framework", "neg_sample": ["language modeling is done by using Method", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "In this paper , we propose an Adaptive Differential Privacy ( ADP ) framework for language modeling without resorting to prior privacy information .", "forward": false, "src_ids": "2022.fl4nlp-1.3_1533"}
+{"input": "adaptive differential privacy ( adp ) framework is used for Task| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "adaptive differential privacy ( adp ) framework", "output": "language modeling", "neg_sample": ["adaptive differential privacy ( adp ) framework is used for Task", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "In this paper , we propose an Adaptive Differential Privacy ( ADP ) framework for language modeling without resorting to prior privacy information .", "forward": true, "src_ids": "2022.fl4nlp-1.3_1534"}
+{"input": "differential privacy noise is done by using Method| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "differential privacy noise", "output": "adam algorithm", "neg_sample": ["differential privacy noise is done by using Method", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "We further propose a new Adam algorithm that adjusts the degree of differential privacy noise injected to the language model according to the estimated privacy probabilities .", "forward": false, "src_ids": "2022.fl4nlp-1.3_1535"}
+{"input": "adam algorithm is used for OtherScientificTerm| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "adam algorithm", "output": "differential privacy noise", "neg_sample": ["adam algorithm is used for OtherScientificTerm", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "We further propose a new Adam algorithm that adjusts the degree of differential privacy noise injected to the language model according to the estimated privacy probabilities .", "forward": true, "src_ids": "2022.fl4nlp-1.3_1536"}
+{"input": "adaptive differential privacy is used for Method| context: although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility . previous works improve dp training by discriminating privacy and non - privacy data . but these works rely on datasets with prior privacy information , which is not available in real - world scenarios .", "entity": "adaptive differential privacy", "output": "differentially private language modeling", "neg_sample": ["adaptive differential privacy is used for Method", "although differential privacy ( dp ) can protect language models from leaking privacy , its indiscriminative protection on all data points reduces its practical utility .", "previous works improve dp training by discriminating privacy and non - privacy data .", "but these works rely on datasets with prior privacy information , which is not available in real - world scenarios ."], "relation": "used for", "id": "2022.fl4nlp-1.3", "year": 2022, "rel_sent": "Experiments demonstrate that our ADP improves differentially private language modeling to achieve good protection from canary attackers .", "forward": true, "src_ids": "2022.fl4nlp-1.3_1537"}
+{"input": "question answering is done by using Task| context: relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems . existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "question answering", "output": "implicit relation linking", "neg_sample": ["question answering is done by using Task", "relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems .", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Implicit Relation Linking for Question Answering over Knowledge Graph.", "forward": false, "src_ids": "2022.findings-acl.312_1538"}
+{"input": "implicit relation linking is used for Task| context: existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "implicit relation linking", "output": "question answering", "neg_sample": ["implicit relation linking is used for Task", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Implicit Relation Linking for Question Answering over Knowledge Graph.", "forward": true, "src_ids": "2022.findings-acl.312_1539"}
+{"input": "implicit rl is done by using Method| context: relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems . existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "implicit rl", "output": "imrl", "neg_sample": ["implicit rl is done by using Method", "relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems .", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Our experiments on two benchmark and a newly - created datasets show that ImRL significantly outperforms several state - of - the - art methods , especially for implicit RL .", "forward": false, "src_ids": "2022.findings-acl.312_1540"}
+{"input": "gated mechanism is used for OtherScientificTerm| context: relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems . existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "gated mechanism", "output": "relation phrases", "neg_sample": ["gated mechanism is used for OtherScientificTerm", "relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems .", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Besides , we leverage a gated mechanism with attention to inject prior knowledge from external paraphrase dictionaries to address the relation phrases with vague meaning .", "forward": true, "src_ids": "2022.findings-acl.312_1541"}
+{"input": "prior knowledge is done by using Method| context: relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems . existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "prior knowledge", "output": "gated mechanism", "neg_sample": ["prior knowledge is done by using Method", "relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems .", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Besides , we leverage a gated mechanism with attention to inject prior knowledge from external paraphrase dictionaries to address the relation phrases with vague meaning .", "forward": false, "src_ids": "2022.findings-acl.312_1542"}
+{"input": "relation phrases is done by using Method| context: relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems . existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "relation phrases", "output": "gated mechanism", "neg_sample": ["relation phrases is done by using Method", "relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems .", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Besides , we leverage a gated mechanism with attention to inject prior knowledge from external paraphrase dictionaries to address the relation phrases with vague meaning .", "forward": false, "src_ids": "2022.findings-acl.312_1543"}
+{"input": "imrl is used for Method| context: relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems . existing methods mainly rely on the textual similarities between nl and kg to build relation links . due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods .", "entity": "imrl", "output": "implicit rl", "neg_sample": ["imrl is used for Method", "relation linking ( rl ) is a vital module in knowledge - based question answering ( kbqa ) systems .", "existing methods mainly rely on the textual similarities between nl and kg to build relation links .", "due to the ambiguity of nl and the incompleteness of kg , many relations in nl are implicitly expressed , and may not link to a single relation in kg , which challenges the current methods ."], "relation": "used for", "id": "2022.findings-acl.312", "year": 2022, "rel_sent": "Our experiments on two benchmark and a newly - created datasets show that ImRL significantly outperforms several state - of - the - art methods , especially for implicit RL .", "forward": true, "src_ids": "2022.findings-acl.312_1544"}
+{"input": "pii masking systems is done by using Task| context: many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals . pii masking is commonly used to redact personal information such as names , addresses , and phone numbers from text data . most modern pii masking pipelines involve machine learning algorithms . however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed .", "entity": "pii masking systems", "output": "name detection", "neg_sample": ["pii masking systems is done by using Task", "many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals .", "pii masking is commonly used to redact personal information such as names , addresses , and phone numbers from text data .", "most modern pii masking pipelines involve machine learning algorithms .", "however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed ."], "relation": "used for", "id": "2022.ltedi-1.10", "year": 2022, "rel_sent": "Behind the Mask : Demographic bias in name detection for PII masking.", "forward": false, "src_ids": "2022.ltedi-1.10_1545"}
+{"input": "pii masking systems is used for Task| context: many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals . pii masking is commonly used to redact personal information such as names , addresses , and phone numbers from text data . most modern pii masking pipelines involve machine learning algorithms . however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed .", "entity": "pii masking systems", "output": "name detection", "neg_sample": ["pii masking systems is used for Task", "many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals .", "pii masking is commonly used to redact personal information such as names , addresses , and phone numbers from text data .", "most modern pii masking pipelines involve machine learning algorithms .", "however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed ."], "relation": "used for", "id": "2022.ltedi-1.10", "year": 2022, "rel_sent": "In this paper , we evaluate the performance of three off - the - shelf PII masking systems on name detection and redaction .", "forward": true, "src_ids": "2022.ltedi-1.10_1546"}
+{"input": "name detection is done by using Method| context: many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals . most modern pii masking pipelines involve machine learning algorithms . however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed .", "entity": "name detection", "output": "pii masking systems", "neg_sample": ["name detection is done by using Method", "many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals .", "most modern pii masking pipelines involve machine learning algorithms .", "however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed ."], "relation": "used for", "id": "2022.ltedi-1.10", "year": 2022, "rel_sent": "In this paper , we evaluate the performance of three off - the - shelf PII masking systems on name detection and redaction .", "forward": false, "src_ids": "2022.ltedi-1.10_1547"}
+{"input": "redaction is done by using Method| context: many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals . most modern pii masking pipelines involve machine learning algorithms . however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed .", "entity": "redaction", "output": "pii masking systems", "neg_sample": ["redaction is done by using Method", "many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals .", "most modern pii masking pipelines involve machine learning algorithms .", "however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed ."], "relation": "used for", "id": "2022.ltedi-1.10", "year": 2022, "rel_sent": "In this paper , we evaluate the performance of three off - the - shelf PII masking systems on name detection and redaction .", "forward": false, "src_ids": "2022.ltedi-1.10_1548"}
+{"input": "name detection is used for Method| context: many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals . most modern pii masking pipelines involve machine learning algorithms . however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed .", "entity": "name detection", "output": "pii masking systems", "neg_sample": ["name detection is used for Method", "many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals .", "most modern pii masking pipelines involve machine learning algorithms .", "however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed ."], "relation": "used for", "id": "2022.ltedi-1.10", "year": 2022, "rel_sent": "Behind the Mask : Demographic bias in name detection for PII masking.", "forward": true, "src_ids": "2022.ltedi-1.10_1549"}
+{"input": "pii masking systems is used for OtherScientificTerm| context: many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals . pii masking is commonly used to redact personal information such as names , addresses , and phone numbers from text data . most modern pii masking pipelines involve machine learning algorithms . however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed .", "entity": "pii masking systems", "output": "redaction", "neg_sample": ["pii masking systems is used for OtherScientificTerm", "many datasets contain personally identifiable information , or pii , which poses privacy risks to individuals .", "pii masking is commonly used to redact personal information such as names , addresses , and phone numbers from text data .", "most modern pii masking pipelines involve machine learning algorithms .", "however , these systems may vary in performance , such that individuals from particular demographic groups bear a higher risk for having their personal information exposed ."], "relation": "used for", "id": "2022.ltedi-1.10", "year": 2022, "rel_sent": "In this paper , we evaluate the performance of three off - the - shelf PII masking systems on name detection and redaction .", "forward": true, "src_ids": "2022.ltedi-1.10_1550"}
+{"input": "neural machine translation is done by using Task| context: building effective neural machine translation models often implies accommodating diverse sets of heterogeneous data so as to optimize performance for the domain(s ) of interest .", "entity": "neural machine translation", "output": "multi - domain adaptation", "neg_sample": ["neural machine translation is done by using Task", "building effective neural machine translation models often implies accommodating diverse sets of heterogeneous data so as to optimize performance for the domain(s ) of interest ."], "relation": "used for", "id": "2022.eamt-1.4", "year": 2022, "rel_sent": "Multi - Domain Adaptation in Neural Machine Translation with Dynamic Sampling Strategies.", "forward": false, "src_ids": "2022.eamt-1.4_1551"}
+{"input": "multi - domain adaptation is used for Task| context: such multi - source / multi - domain adaptation problems are typically approached through instance selection or reweighting strategies , based on a static assessment of the relevance of training instances with respect to the task at hand .", "entity": "multi - domain adaptation", "output": "neural machine translation", "neg_sample": ["multi - domain adaptation is used for Task", "such multi - source / multi - domain adaptation problems are typically approached through instance selection or reweighting strategies , based on a static assessment of the relevance of training instances with respect to the task at hand ."], "relation": "used for", "id": "2022.eamt-1.4", "year": 2022, "rel_sent": "Multi - Domain Adaptation in Neural Machine Translation with Dynamic Sampling Strategies.", "forward": true, "src_ids": "2022.eamt-1.4_1552"}
+{"input": "data selection policy is done by using Method| context: building effective neural machine translation models often implies accommodating diverse sets of heterogeneous data so as to optimize performance for the domain(s ) of interest . such multi - source / multi - domain adaptation problems are typically approached through instance selection or reweighting strategies , based on a static assessment of the relevance of training instances with respect to the task at hand .", "entity": "data selection policy", "output": "dynamic data selection strategies", "neg_sample": ["data selection policy is done by using Method", "building effective neural machine translation models often implies accommodating diverse sets of heterogeneous data so as to optimize performance for the domain(s ) of interest .", "such multi - source / multi - domain adaptation problems are typically approached through instance selection or reweighting strategies , based on a static assessment of the relevance of training instances with respect to the task at hand ."], "relation": "used for", "id": "2022.eamt-1.4", "year": 2022, "rel_sent": "In this paper , we study dynamic data selection strategies that are able to automatically re - evaluate the usefulness of data samples and to evolve a data selection policy in the course of training .", "forward": false, "src_ids": "2022.eamt-1.4_1553"}
+{"input": "dynamic data selection strategies is used for Method| context: building effective neural machine translation models often implies accommodating diverse sets of heterogeneous data so as to optimize performance for the domain(s ) of interest . such multi - source / multi - domain adaptation problems are typically approached through instance selection or reweighting strategies , based on a static assessment of the relevance of training instances with respect to the task at hand .", "entity": "dynamic data selection strategies", "output": "data selection policy", "neg_sample": ["dynamic data selection strategies is used for Method", "building effective neural machine translation models often implies accommodating diverse sets of heterogeneous data so as to optimize performance for the domain(s ) of interest .", "such multi - source / multi - domain adaptation problems are typically approached through instance selection or reweighting strategies , based on a static assessment of the relevance of training instances with respect to the task at hand ."], "relation": "used for", "id": "2022.eamt-1.4", "year": 2022, "rel_sent": "In this paper , we study dynamic data selection strategies that are able to automatically re - evaluate the usefulness of data samples and to evolve a data selection policy in the course of training .", "forward": true, "src_ids": "2022.eamt-1.4_1554"}
+{"input": "data curation method is used for Task| context: over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages . however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages .", "entity": "data curation method", "output": "globalwoz", "neg_sample": ["data curation method is used for Task", "over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages .", "however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages ."], "relation": "used for", "id": "2022.acl-long.115", "year": 2022, "rel_sent": "To tackle these limitations , we introduce a novel data curation method that generates GlobalWoZ - a large - scale multilingual ToD dataset globalized from an English ToD dataset for three unexplored use cases of multilingual ToD systems .", "forward": true, "src_ids": "2022.acl-long.115_1555"}
+{"input": "multilingual task - oriented dialogue systems is done by using Task| context: over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages . however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages .", "entity": "multilingual task - oriented dialogue systems", "output": "globalizing multiwoz", "neg_sample": ["multilingual task - oriented dialogue systems is done by using Task", "over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages .", "however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages ."], "relation": "used for", "id": "2022.acl-long.115", "year": 2022, "rel_sent": "GlobalWoZ : Globalizing MultiWoZ to Develop Multilingual Task - Oriented Dialogue Systems.", "forward": false, "src_ids": "2022.acl-long.115_1556"}
+{"input": "globalizing multiwoz is used for Task| context: over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages . however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages .", "entity": "globalizing multiwoz", "output": "multilingual task - oriented dialogue systems", "neg_sample": ["globalizing multiwoz is used for Task", "over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages .", "however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages ."], "relation": "used for", "id": "2022.acl-long.115", "year": 2022, "rel_sent": "GlobalWoZ : Globalizing MultiWoZ to Develop Multilingual Task - Oriented Dialogue Systems.", "forward": true, "src_ids": "2022.acl-long.115_1557"}
+{"input": "large - scale multilingual tod dataset is done by using Method| context: over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages . however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages .", "entity": "large - scale multilingual tod dataset", "output": "data curation method", "neg_sample": ["large - scale multilingual tod dataset is done by using Method", "over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages .", "however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages ."], "relation": "used for", "id": "2022.acl-long.115", "year": 2022, "rel_sent": "To tackle these limitations , we introduce a novel data curation method that generates GlobalWoZ - a large - scale multilingual ToD dataset globalized from an English ToD dataset for three unexplored use cases of multilingual ToD systems .", "forward": false, "src_ids": "2022.acl-long.115_1558"}
+{"input": "globalwoz is done by using Method| context: over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages . however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages .", "entity": "globalwoz", "output": "data curation method", "neg_sample": ["globalwoz is done by using Method", "over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages .", "however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages ."], "relation": "used for", "id": "2022.acl-long.115", "year": 2022, "rel_sent": "To tackle these limitations , we introduce a novel data curation method that generates GlobalWoZ - a large - scale multilingual ToD dataset globalized from an English ToD dataset for three unexplored use cases of multilingual ToD systems .", "forward": false, "src_ids": "2022.acl-long.115_1559"}
+{"input": "data curation method is used for Material| context: over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages . however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages .", "entity": "data curation method", "output": "large - scale multilingual tod dataset", "neg_sample": ["data curation method is used for Material", "over the last few years , there has been a move towards data curation for multilingual task - oriented dialogue ( tod ) systems that can serve people speaking different languages .", "however , existing multilingual tod datasets either have a limited coverage of languages due to the high cost of data curation , or ignore the fact that dialogue entities barely exist in countries speaking these languages ."], "relation": "used for", "id": "2022.acl-long.115", "year": 2022, "rel_sent": "To tackle these limitations , we introduce a novel data curation method that generates GlobalWoZ - a large - scale multilingual ToD dataset globalized from an English ToD dataset for three unexplored use cases of multilingual ToD systems .", "forward": true, "src_ids": "2022.acl-long.115_1560"}
+{"input": "memorization is used for Task| context: a recent study by feldman ( 2020 ) proposed a long - tail theory to explain the memorization behavior of deep learning models .", "entity": "memorization", "output": "nlp", "neg_sample": ["memorization is used for Task", "a recent study by feldman ( 2020 ) proposed a long - tail theory to explain the memorization behavior of deep learning models ."], "relation": "used for", "id": "2022.acl-long.434", "year": 2022, "rel_sent": "An Empirical Study of Memorization in NLP.", "forward": true, "src_ids": "2022.acl-long.434_1561"}
+{"input": "neural machine translation is done by using Method| context: the principal task in supervised neural machine translation ( nmt ) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs , and thus produce a model capable of generalizing to unseen instances . however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training . although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples .", "entity": "neural machine translation", "output": "continuous semantic augmentation ( csanmt )", "neg_sample": ["neural machine translation is done by using Method", "the principal task in supervised neural machine translation ( nmt ) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs , and thus produce a model capable of generalizing to unseen instances .", "however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training .", "although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples ."], "relation": "used for", "id": "2022.acl-long.546", "year": 2022, "rel_sent": "Learning to Generalize to More : Continuous Semantic Augmentation for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.546_1562"}
+{"input": "continuous semantic augmentation ( csanmt ) is used for Task| context: however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training . although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples .", "entity": "continuous semantic augmentation ( csanmt )", "output": "neural machine translation", "neg_sample": ["continuous semantic augmentation ( csanmt ) is used for Task", "however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training .", "although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples ."], "relation": "used for", "id": "2022.acl-long.546", "year": 2022, "rel_sent": "Learning to Generalize to More : Continuous Semantic Augmentation for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.546_1563"}
+{"input": "literal expression is done by using OtherScientificTerm| context: the principal task in supervised neural machine translation ( nmt ) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs , and thus produce a model capable of generalizing to unseen instances . however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training . although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples .", "entity": "literal expression", "output": "adjacency semantic region", "neg_sample": ["literal expression is done by using OtherScientificTerm", "the principal task in supervised neural machine translation ( nmt ) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs , and thus produce a model capable of generalizing to unseen instances .", "however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training .", "although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples ."], "relation": "used for", "id": "2022.acl-long.546", "year": 2022, "rel_sent": "In this paper , we present a novel data augmentation paradigm termed Continuous Semantic Augmentation ( CsaNMT ) , which augments each training instance with an adjacency semantic region that could cover adequate variants of literal expression under the same meaning .", "forward": false, "src_ids": "2022.acl-long.546_1564"}
+{"input": "adjacency semantic region is used for OtherScientificTerm| context: the principal task in supervised neural machine translation ( nmt ) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs , and thus produce a model capable of generalizing to unseen instances . however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training . although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples .", "entity": "adjacency semantic region", "output": "literal expression", "neg_sample": ["adjacency semantic region is used for OtherScientificTerm", "the principal task in supervised neural machine translation ( nmt ) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs , and thus produce a model capable of generalizing to unseen instances .", "however , it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training .", "although data augmentation is widely used to enrich the training data , conventional methods with discrete manipulations fail to generate diverse and faithful training samples ."], "relation": "used for", "id": "2022.acl-long.546", "year": 2022, "rel_sent": "In this paper , we present a novel data augmentation paradigm termed Continuous Semantic Augmentation ( CsaNMT ) , which augments each training instance with an adjacency semantic region that could cover adequate variants of literal expression under the same meaning .", "forward": true, "src_ids": "2022.acl-long.546_1565"}
+{"input": "chemistry is done by using Method| context: foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g. , law , healthcare , education , etc . however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications .", "entity": "chemistry", "output": "foundation models of scientific knowledge", "neg_sample": ["chemistry is done by using Method", "foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g.", ", law , healthcare , education , etc .", "however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications ."], "relation": "used for", "id": "2022.bigscience-1.12", "year": 2022, "rel_sent": "Foundation Models of Scientific Knowledge for Chemistry : Opportunities , Challenges and Lessons Learned.", "forward": false, "src_ids": "2022.bigscience-1.12_1566"}
+{"input": "chemistry is done by using Method| context: foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g. , law , healthcare , education , etc . however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications .", "entity": "chemistry", "output": "foundation models of scientific knowledge", "neg_sample": ["chemistry is done by using Method", "foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g.", ", law , healthcare , education , etc .", "however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications ."], "relation": "used for", "id": "2022.bigscience-1.12", "year": 2022, "rel_sent": "In this work , we develop foundation models of scientific knowledge for chemistry to augment scientists with the advanced ability to perceive and reason at scale previously unimagined .", "forward": false, "src_ids": "2022.bigscience-1.12_1567"}
+{"input": "foundation models of scientific knowledge is used for Material| context: foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g. , law , healthcare , education , etc . however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications .", "entity": "foundation models of scientific knowledge", "output": "chemistry", "neg_sample": ["foundation models of scientific knowledge is used for Material", "foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g.", ", law , healthcare , education , etc .", "however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications ."], "relation": "used for", "id": "2022.bigscience-1.12", "year": 2022, "rel_sent": "Foundation Models of Scientific Knowledge for Chemistry : Opportunities , Challenges and Lessons Learned.", "forward": true, "src_ids": "2022.bigscience-1.12_1568"}
+{"input": "foundation models of scientific knowledge is used for Material| context: foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g. , law , healthcare , education , etc . however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications .", "entity": "foundation models of scientific knowledge", "output": "chemistry", "neg_sample": ["foundation models of scientific knowledge is used for Material", "foundation models pre - trained on large corpora demonstrate significant gains across many natural language processing tasks and domains e.g.", ", law , healthcare , education , etc .", "however , only limited efforts have investigated the opportunities and limitations of applying these powerful models to science and security applications ."], "relation": "used for", "id": "2022.bigscience-1.12", "year": 2022, "rel_sent": "In this work , we develop foundation models of scientific knowledge for chemistry to augment scientists with the advanced ability to perceive and reason at scale previously unimagined .", "forward": true, "src_ids": "2022.bigscience-1.12_1569"}
+{"input": "category - aligned representations is done by using OtherScientificTerm| context: query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query . taking click - through logs as training data in deep learning methods is a common and effective approach for query classification . however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries . the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries .", "entity": "category - aligned representations", "output": "contrastive loss", "neg_sample": ["category - aligned representations is done by using OtherScientificTerm", "query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query .", "taking click - through logs as training data in deep learning methods is a common and effective approach for query classification .", "however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries .", "the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries ."], "relation": "used for", "id": "2022.ecnlp-1.17", "year": 2022, "rel_sent": "The long - tail queries are guided by the contrastive loss to obtain category - aligned representations in the auxiliary module , where the variant frequent queries serve as anchors in the representation space .", "forward": false, "src_ids": "2022.ecnlp-1.17_1570"}
+{"input": "representations of long - tail queries is done by using Method| context: query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query . taking click - through logs as training data in deep learning methods is a common and effective approach for query classification . however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries . the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries .", "entity": "representations of long - tail queries", "output": "auxiliary module", "neg_sample": ["representations of long - tail queries is done by using Method", "query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query .", "taking click - through logs as training data in deep learning methods is a common and effective approach for query classification .", "however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries .", "the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries ."], "relation": "used for", "id": "2022.ecnlp-1.17", "year": 2022, "rel_sent": "To solve the above problem , we propose a novel method that leverages an auxiliary module to enhance the representations of long - tail queries by taking advantage of reliable supervised information of variant frequent queries .", "forward": false, "src_ids": "2022.ecnlp-1.17_1571"}
+{"input": "auxiliary module is used for Method| context: query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query . taking click - through logs as training data in deep learning methods is a common and effective approach for query classification . however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries . the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries .", "entity": "auxiliary module", "output": "representations of long - tail queries", "neg_sample": ["auxiliary module is used for Method", "query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query .", "taking click - through logs as training data in deep learning methods is a common and effective approach for query classification .", "however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries .", "the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries ."], "relation": "used for", "id": "2022.ecnlp-1.17", "year": 2022, "rel_sent": "To solve the above problem , we propose a novel method that leverages an auxiliary module to enhance the representations of long - tail queries by taking advantage of reliable supervised information of variant frequent queries .", "forward": true, "src_ids": "2022.ecnlp-1.17_1572"}
+{"input": "variant frequent queries is done by using OtherScientificTerm| context: query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query . taking click - through logs as training data in deep learning methods is a common and effective approach for query classification . however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries . the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries .", "entity": "variant frequent queries", "output": "supervised information", "neg_sample": ["variant frequent queries is done by using OtherScientificTerm", "query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query .", "taking click - through logs as training data in deep learning methods is a common and effective approach for query classification .", "however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries .", "the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries ."], "relation": "used for", "id": "2022.ecnlp-1.17", "year": 2022, "rel_sent": "To solve the above problem , we propose a novel method that leverages an auxiliary module to enhance the representations of long - tail queries by taking advantage of reliable supervised information of variant frequent queries .", "forward": false, "src_ids": "2022.ecnlp-1.17_1573"}
+{"input": "supervised information is used for OtherScientificTerm| context: query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query . taking click - through logs as training data in deep learning methods is a common and effective approach for query classification . however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries . the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries .", "entity": "supervised information", "output": "variant frequent queries", "neg_sample": ["supervised information is used for OtherScientificTerm", "query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query .", "taking click - through logs as training data in deep learning methods is a common and effective approach for query classification .", "however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries .", "the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries ."], "relation": "used for", "id": "2022.ecnlp-1.17", "year": 2022, "rel_sent": "To solve the above problem , we propose a novel method that leverages an auxiliary module to enhance the representations of long - tail queries by taking advantage of reliable supervised information of variant frequent queries .", "forward": true, "src_ids": "2022.ecnlp-1.17_1574"}
+{"input": "contrastive loss is used for Method| context: query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query . taking click - through logs as training data in deep learning methods is a common and effective approach for query classification . however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries . the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries .", "entity": "contrastive loss", "output": "category - aligned representations", "neg_sample": ["contrastive loss is used for Method", "query classification is a fundamental task in an e - commerce search engine , which assigns one or multiple predefined product categories in response to each search query .", "taking click - through logs as training data in deep learning methods is a common and effective approach for query classification .", "however , the frequency distribution of queries typically has long - tail property , which means that there are few logs for most of the queries .", "the lack of reliable user feedback information results in worse performance of long - tail queries compared with frequent queries ."], "relation": "used for", "id": "2022.ecnlp-1.17", "year": 2022, "rel_sent": "The long - tail queries are guided by the contrastive loss to obtain category - aligned representations in the auxiliary module , where the variant frequent queries serve as anchors in the representation space .", "forward": true, "src_ids": "2022.ecnlp-1.17_1575"}
+{"input": "knowledgeable dialogue generation is done by using Method| context: existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases . these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed .", "entity": "knowledgeable dialogue generation", "output": "multi - stage prompting approach", "neg_sample": ["knowledgeable dialogue generation is done by using Method", "existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases .", "these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed ."], "relation": "used for", "id": "2022.findings-acl.104", "year": 2022, "rel_sent": "Multi - Stage Prompting for Knowledgeable Dialogue Generation.", "forward": false, "src_ids": "2022.findings-acl.104_1576"}
+{"input": "knowledgeable responses is done by using Method| context: existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases . these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed .", "entity": "knowledgeable responses", "output": "multi - stage prompting approach", "neg_sample": ["knowledgeable responses is done by using Method", "existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases .", "these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed ."], "relation": "used for", "id": "2022.findings-acl.104", "year": 2022, "rel_sent": "We propose a multi - stage prompting approach to generate knowledgeable responses from a single pretrained LM .", "forward": false, "src_ids": "2022.findings-acl.104_1577"}
+{"input": "multi - stage prompting approach is used for Task| context: existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases . these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed .", "entity": "multi - stage prompting approach", "output": "knowledgeable dialogue generation", "neg_sample": ["multi - stage prompting approach is used for Task", "existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases .", "these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed ."], "relation": "used for", "id": "2022.findings-acl.104", "year": 2022, "rel_sent": "Multi - Stage Prompting for Knowledgeable Dialogue Generation.", "forward": true, "src_ids": "2022.findings-acl.104_1578"}
+{"input": "multi - stage prompting approach is used for OtherScientificTerm| context: existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases . these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed .", "entity": "multi - stage prompting approach", "output": "knowledgeable responses", "neg_sample": ["multi - stage prompting approach is used for OtherScientificTerm", "existing knowledge - grounded dialogue systems typically use finetuned versions of a pretrained language model ( lm ) and large - scale knowledge bases .", "these models typically fail to generalize on topics outside of the knowledge base , and require maintaining separate potentially large checkpoints each time finetuning is needed ."], "relation": "used for", "id": "2022.findings-acl.104", "year": 2022, "rel_sent": "We propose a multi - stage prompting approach to generate knowledgeable responses from a single pretrained LM .", "forward": true, "src_ids": "2022.findings-acl.104_1579"}
+{"input": "pre - trained language model is done by using Task| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "pre - trained language model", "output": "domain knowledge transferring", "neg_sample": ["pre - trained language model is done by using Task", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "Domain Knowledge Transferring for Pre - trained Language Model via Calibrated Activation Boundary Distillation.", "forward": false, "src_ids": "2022.acl-long.116_1580"}
+{"input": "domain knowledge transferring is used for Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "domain knowledge transferring", "output": "pre - trained language model", "neg_sample": ["domain knowledge transferring is used for Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "Domain Knowledge Transferring for Pre - trained Language Model via Calibrated Activation Boundary Distillation.", "forward": true, "src_ids": "2022.acl-long.116_1581"}
+{"input": "domain knowledge transferring ( doktra ) framework is used for Method| context: however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "domain knowledge transferring ( doktra ) framework", "output": "pretrained language models", "neg_sample": ["domain knowledge transferring ( doktra ) framework is used for Method", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "In this study , we propose a domain knowledge transferring ( DoKTra ) framework for PLMs without additional in - domain pretraining .", "forward": true, "src_ids": "2022.acl-long.116_1582"}
+{"input": "pretrained language models is done by using Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "pretrained language models", "output": "domain knowledge transferring ( doktra ) framework", "neg_sample": ["pretrained language models is done by using Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "In this study , we propose a domain knowledge transferring ( DoKTra ) framework for PLMs without additional in - domain pretraining .", "forward": false, "src_ids": "2022.acl-long.116_1583"}
+{"input": "activation of hidden neurons is done by using Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "activation of hidden neurons", "output": "activation boundary distillation", "neg_sample": ["activation of hidden neurons is done by using Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "In particular , we employ activation boundary distillation , which focuses on the activation of hidden neurons .", "forward": false, "src_ids": "2022.acl-long.116_1584"}
+{"input": "activation boundary distillation is used for OtherScientificTerm| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "activation boundary distillation", "output": "activation of hidden neurons", "neg_sample": ["activation boundary distillation is used for OtherScientificTerm", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "In particular , we employ activation boundary distillation , which focuses on the activation of hidden neurons .", "forward": true, "src_ids": "2022.acl-long.116_1585"}
+{"input": "teacher training is done by using OtherScientificTerm| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "teacher training", "output": "entropy regularization term", "neg_sample": ["teacher training is done by using OtherScientificTerm", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "We also apply an entropy regularization term in both teacher training and distillation to encourage the model to generate reliable output probabilities , and thus aid the distillation .", "forward": false, "src_ids": "2022.acl-long.116_1586"}
+{"input": "distillation is done by using OtherScientificTerm| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "distillation", "output": "entropy regularization term", "neg_sample": ["distillation is done by using OtherScientificTerm", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "We also apply an entropy regularization term in both teacher training and distillation to encourage the model to generate reliable output probabilities , and thus aid the distillation .", "forward": false, "src_ids": "2022.acl-long.116_1587"}
+{"input": "entropy regularization term is used for Task| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "entropy regularization term", "output": "teacher training", "neg_sample": ["entropy regularization term is used for Task", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "We also apply an entropy regularization term in both teacher training and distillation to encourage the model to generate reliable output probabilities , and thus aid the distillation .", "forward": true, "src_ids": "2022.acl-long.116_1588"}
+{"input": "entropy regularization term is used for Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "entropy regularization term", "output": "distillation", "neg_sample": ["entropy regularization term is used for Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "We also apply an entropy regularization term in both teacher training and distillation to encourage the model to generate reliable output probabilities , and thus aid the distillation .", "forward": true, "src_ids": "2022.acl-long.116_1589"}
+{"input": "downstream tasks is done by using Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "downstream tasks", "output": "doktra framework", "neg_sample": ["downstream tasks is done by using Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "By applying the proposed DoKTra framework to downstream tasks in the biomedical , clinical , and financial domains , our student models can retain a high percentage of teacher performance and even outperform the teachers in certain tasks .", "forward": false, "src_ids": "2022.acl-long.116_1590"}
+{"input": "student models is done by using Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "student models", "output": "doktra framework", "neg_sample": ["student models is done by using Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "By applying the proposed DoKTra framework to downstream tasks in the biomedical , clinical , and financial domains , our student models can retain a high percentage of teacher performance and even outperform the teachers in certain tasks .", "forward": false, "src_ids": "2022.acl-long.116_1591"}
+{"input": "doktra framework is used for Task| context: additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "doktra framework", "output": "downstream tasks", "neg_sample": ["doktra framework is used for Task", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "By applying the proposed DoKTra framework to downstream tasks in the biomedical , clinical , and financial domains , our student models can retain a high percentage of teacher performance and even outperform the teachers in certain tasks .", "forward": true, "src_ids": "2022.acl-long.116_1592"}
+{"input": "doktra framework is used for Method| context: since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains . additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms . however , these pre - training methods require considerable in - domain data and training resources and a longer training time . moreover , the training must be re - performed whenever a new plm emerges .", "entity": "doktra framework", "output": "student models", "neg_sample": ["doktra framework is used for Method", "since the development and wide use of pretrained language models ( plms ) , several approaches have been applied to boost their performance on downstream tasks in specific domains , such as biomedical or scientific domains .", "additional pre - training with in - domain texts is the most common approach for providing domain - specific knowledge to plms .", "however , these pre - training methods require considerable in - domain data and training resources and a longer training time .", "moreover , the training must be re - performed whenever a new plm emerges ."], "relation": "used for", "id": "2022.acl-long.116", "year": 2022, "rel_sent": "By applying the proposed DoKTra framework to downstream tasks in the biomedical , clinical , and financial domains , our student models can retain a high percentage of teacher performance and even outperform the teachers in certain tasks .", "forward": true, "src_ids": "2022.acl-long.116_1593"}
+{"input": "natural language understanding is done by using Task| context: natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc . recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters .", "entity": "natural language understanding", "output": "canary extraction", "neg_sample": ["natural language understanding is done by using Task", "natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc .", "recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters ."], "relation": "used for", "id": "2022.acl-short.61", "year": 2022, "rel_sent": "Canary Extraction in Natural Language Understanding Models.", "forward": false, "src_ids": "2022.acl-short.61_1594"}
+{"input": "canary extraction is used for Task| context: recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters .", "entity": "canary extraction", "output": "natural language understanding", "neg_sample": ["canary extraction is used for Task", "recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters ."], "relation": "used for", "id": "2022.acl-short.61", "year": 2022, "rel_sent": "Canary Extraction in Natural Language Understanding Models.", "forward": true, "src_ids": "2022.acl-short.61_1595"}
+{"input": "text completion is used for OtherScientificTerm| context: natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc . recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters .", "entity": "text completion", "output": "canaries", "neg_sample": ["text completion is used for OtherScientificTerm", "natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc .", "recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters ."], "relation": "used for", "id": "2022.acl-short.61", "year": 2022, "rel_sent": "We evaluate our approach by performing text completion on canaries and demonstrate that by using the prefix ( non - sensitive ) tokens of the canary , we can generate the full canary .", "forward": true, "src_ids": "2022.acl-short.61_1596"}
+{"input": "canaries is done by using Task| context: natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc . recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters .", "entity": "canaries", "output": "text completion", "neg_sample": ["canaries is done by using Task", "natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc .", "recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters ."], "relation": "used for", "id": "2022.acl-short.61", "year": 2022, "rel_sent": "We evaluate our approach by performing text completion on canaries and demonstrate that by using the prefix ( non - sensitive ) tokens of the canary , we can generate the full canary .", "forward": false, "src_ids": "2022.acl-short.61_1597"}
+{"input": "canary is done by using OtherScientificTerm| context: natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc . recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters .", "entity": "canary", "output": "prefix ( non - sensitive ) tokens", "neg_sample": ["canary is done by using OtherScientificTerm", "natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc .", "recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters ."], "relation": "used for", "id": "2022.acl-short.61", "year": 2022, "rel_sent": "We evaluate our approach by performing text completion on canaries and demonstrate that by using the prefix ( non - sensitive ) tokens of the canary , we can generate the full canary .", "forward": false, "src_ids": "2022.acl-short.61_1598"}
+{"input": "prefix ( non - sensitive ) tokens is used for Material| context: natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc . recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters .", "entity": "prefix ( non - sensitive ) tokens", "output": "canary", "neg_sample": ["prefix ( non - sensitive ) tokens is used for Material", "natural language understanding ( nlu ) models can be trained on sensitive information such as phone numbers , zip - codes etc .", "recent literature has focused on model inversion attacks ( modiva ) that can extract training data from model parameters ."], "relation": "used for", "id": "2022.acl-short.61", "year": 2022, "rel_sent": "We evaluate our approach by performing text completion on canaries and demonstrate that by using the prefix ( non - sensitive ) tokens of the canary , we can generate the full canary .", "forward": true, "src_ids": "2022.acl-short.61_1599"}
+{"input": "domain adaptation of question answering is done by using Task| context: synthesizing qa pairs with a question generator ( qg ) on the target domain has become a popular approach for domain adaptation of question answering ( qa ) models . since synthetic questions are often noisy in practice , existing work adapts scores from a pretrained qa ( or qg ) model as criteria to select high - quality questions . however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain .", "entity": "domain adaptation of question answering", "output": "synthetic question value estimation", "neg_sample": ["domain adaptation of question answering is done by using Task", "synthesizing qa pairs with a question generator ( qg ) on the target domain has become a popular approach for domain adaptation of question answering ( qa ) models .", "since synthetic questions are often noisy in practice , existing work adapts scores from a pretrained qa ( or qg ) model as criteria to select high - quality questions .", "however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain ."], "relation": "used for", "id": "2022.acl-long.95", "year": 2022, "rel_sent": "Synthetic Question Value Estimation for Domain Adaptation of Question Answering.", "forward": false, "src_ids": "2022.acl-long.95_1600"}
+{"input": "synthetic question value estimation is used for Task| context: since synthetic questions are often noisy in practice , existing work adapts scores from a pretrained qa ( or qg ) model as criteria to select high - quality questions . however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain .", "entity": "synthetic question value estimation", "output": "domain adaptation of question answering", "neg_sample": ["synthetic question value estimation is used for Task", "since synthetic questions are often noisy in practice , existing work adapts scores from a pretrained qa ( or qg ) model as criteria to select high - quality questions .", "however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain ."], "relation": "used for", "id": "2022.acl-long.95", "year": 2022, "rel_sent": "Synthetic Question Value Estimation for Domain Adaptation of Question Answering.", "forward": true, "src_ids": "2022.acl-long.95_1601"}
+{"input": "target - domain qa is done by using OtherScientificTerm| context: synthesizing qa pairs with a question generator ( qg ) on the target domain has become a popular approach for domain adaptation of question answering ( qa ) models . however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain .", "entity": "target - domain qa", "output": "synthetic questions", "neg_sample": ["target - domain qa is done by using OtherScientificTerm", "synthesizing qa pairs with a question generator ( qg ) on the target domain has become a popular approach for domain adaptation of question answering ( qa ) models .", "however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain ."], "relation": "used for", "id": "2022.acl-long.95", "year": 2022, "rel_sent": "By conducting comprehensive experiments , we show that the synthetic questions selected by QVE can help achieve better target - domain QA performance , in comparison with existing techniques .", "forward": false, "src_ids": "2022.acl-long.95_1602"}
+{"input": "synthetic questions is used for Task| context: synthesizing qa pairs with a question generator ( qg ) on the target domain has become a popular approach for domain adaptation of question answering ( qa ) models . since synthetic questions are often noisy in practice , existing work adapts scores from a pretrained qa ( or qg ) model as criteria to select high - quality questions . however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain .", "entity": "synthetic questions", "output": "target - domain qa", "neg_sample": ["synthetic questions is used for Task", "synthesizing qa pairs with a question generator ( qg ) on the target domain has become a popular approach for domain adaptation of question answering ( qa ) models .", "since synthetic questions are often noisy in practice , existing work adapts scores from a pretrained qa ( or qg ) model as criteria to select high - quality questions .", "however , these scores do not directly serve the ultimate goal of improving qa performance on the target domain ."], "relation": "used for", "id": "2022.acl-long.95", "year": 2022, "rel_sent": "By conducting comprehensive experiments , we show that the synthetic questions selected by QVE can help achieve better target - domain QA performance , in comparison with existing techniques .", "forward": true, "src_ids": "2022.acl-long.95_1603"}
+{"input": "fine tuning deep nets is done by using Method| context: the first half of this tutorial will make deep nets more accessible to a broader audience , following ' deep nets for poets ' and ' a gentle introduction tofine - tuning . '", "entity": "fine tuning deep nets", "output": "gft ( general fine tuning )", "neg_sample": ["fine tuning deep nets is done by using Method", "the first half of this tutorial will make deep nets more accessible to a broader audience , following ' deep nets for poets ' and ' a gentle introduction tofine - tuning . '"], "relation": "used for", "id": "2022.acl-tutorials.1", "year": 2022, "rel_sent": "We will also introduce GFT ( general fine tuning ) , a little language for fine tuning deep nets with short ( one line ) programs that are as easy to code as regression in statistics packages such as R using glm ( general linear models ) .", "forward": false, "src_ids": "2022.acl-tutorials.1_1604"}
+{"input": "gft ( general fine tuning ) is used for Task| context: the first half of this tutorial will make deep nets more accessible to a broader audience , following ' deep nets for poets ' and ' a gentle introduction tofine - tuning . '", "entity": "gft ( general fine tuning )", "output": "fine tuning deep nets", "neg_sample": ["gft ( general fine tuning ) is used for Task", "the first half of this tutorial will make deep nets more accessible to a broader audience , following ' deep nets for poets ' and ' a gentle introduction tofine - tuning . '"], "relation": "used for", "id": "2022.acl-tutorials.1", "year": 2022, "rel_sent": "We will also introduce GFT ( general fine tuning ) , a little language for fine tuning deep nets with short ( one line ) programs that are as easy to code as regression in statistics packages such as R using glm ( general linear models ) .", "forward": true, "src_ids": "2022.acl-tutorials.1_1605"}
+{"input": "rare token embeddings is done by using Method| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "rare token embeddings", "output": "adaptive gradient gating(agg )", "neg_sample": ["rare token embeddings is done by using Method", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "Rare Tokens Degenerate All Tokens : Improving Neural Text Generation via Adaptive Gradient Gating for Rare Token Embeddings.", "forward": false, "src_ids": "2022.acl-long.3_1606"}
+{"input": "adaptive gradient gating(agg ) is used for OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "adaptive gradient gating(agg )", "output": "rare token embeddings", "neg_sample": ["adaptive gradient gating(agg ) is used for OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "Rare Tokens Degenerate All Tokens : Improving Neural Text Generation via Adaptive Gradient Gating for Rare Token Embeddings.", "forward": true, "src_ids": "2022.acl-long.3_1607"}
+{"input": "token embeddings is used for OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "token embeddings", "output": "rare token embeddings", "neg_sample": ["token embeddings is used for OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "In this study , we analyze the training dynamics of the token embeddings focusing on rare token embedding .", "forward": true, "src_ids": "2022.acl-long.3_1608"}
+{"input": "gradient is used for OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "gradient", "output": "rare token embeddings", "neg_sample": ["gradient is used for OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "We demonstrate that the specific part of the gradient for rare token embeddings is the key cause of the degeneration problem for all tokens during training stage .", "forward": true, "src_ids": "2022.acl-long.3_1609"}
+{"input": "gradient is used for OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "gradient", "output": "rare token embeddings", "neg_sample": ["gradient is used for OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "AGG addresses the degeneration problem by gating the specific part of the gradient for rare token embeddings .", "forward": true, "src_ids": "2022.acl-long.3_1610"}
+{"input": "gradient is used for Task| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "entity": "gradient", "output": "degeneration problem", "neg_sample": ["gradient is used for Task", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "We demonstrate that the specific part of the gradient for rare token embeddings is the key cause of the degeneration problem for all tokens during training stage .", "forward": true, "src_ids": "2022.acl-long.3_1611"}
+{"input": "agg is used for Task| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "entity": "agg", "output": "degeneration problem", "neg_sample": ["agg is used for Task", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "AGG addresses the degeneration problem by gating the specific part of the gradient for rare token embeddings .", "forward": true, "src_ids": "2022.acl-long.3_1612"}
+{"input": "degeneration problem is done by using OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "degeneration problem", "output": "gradient", "neg_sample": ["degeneration problem is done by using OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "We demonstrate that the specific part of the gradient for rare token embeddings is the key cause of the degeneration problem for all tokens during training stage .", "forward": false, "src_ids": "2022.acl-long.3_1613"}
+{"input": "rare token embeddings is done by using OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "rare token embeddings", "output": "gradient", "neg_sample": ["rare token embeddings is done by using OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "We demonstrate that the specific part of the gradient for rare token embeddings is the key cause of the degeneration problem for all tokens during training stage .", "forward": false, "src_ids": "2022.acl-long.3_1614"}
+{"input": "rare token embeddings is done by using OtherScientificTerm| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "rare token embeddings", "output": "gradient", "neg_sample": ["rare token embeddings is done by using OtherScientificTerm", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "AGG addresses the degeneration problem by gating the specific part of the gradient for rare token embeddings .", "forward": false, "src_ids": "2022.acl-long.3_1615"}
+{"input": "degeneration problem is done by using Method| context: recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape . this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models . although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored .", "entity": "degeneration problem", "output": "agg", "neg_sample": ["degeneration problem is done by using Method", "recent studies have determined that the learned token embeddings of large - scale neural language models are degenerated to be anisotropic with a narrow - cone shape .", "this phenomenon , called the representation degeneration problem , facilitates an increase in the overall similarity between token embeddings that negatively affect the performance of the models .", "although the existing methods that address the degeneration problem based on observations of the phenomenon triggered by the problem improves the performance of the text generation , the training dynamics of token embeddings behind the degeneration problem are still not explored ."], "relation": "used for", "id": "2022.acl-long.3", "year": 2022, "rel_sent": "AGG addresses the degeneration problem by gating the specific part of the gradient for rare token embeddings .", "forward": false, "src_ids": "2022.acl-long.3_1616"}
+{"input": "singing voice beautification is done by using OtherScientificTerm| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "singing voice beautification", "output": "singing voice", "neg_sample": ["singing voice beautification is done by using OtherScientificTerm", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Given the singing voice of an amateur singer , SVB aims to improve the intonation and vocal tone of the voice , while keeping the content and vocal timbre .", "forward": false, "src_ids": "2022.acl-long.549_1617"}
+{"input": "intonation is done by using Method| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "intonation", "output": "singing voice beautification", "neg_sample": ["intonation is done by using Method", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Given the singing voice of an amateur singer , SVB aims to improve the intonation and vocal tone of the voice , while keeping the content and vocal timbre .", "forward": false, "src_ids": "2022.acl-long.549_1618"}
+{"input": "singing voice is used for Method| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "singing voice", "output": "singing voice beautification", "neg_sample": ["singing voice is used for Method", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Given the singing voice of an amateur singer , SVB aims to improve the intonation and vocal tone of the voice , while keeping the content and vocal timbre .", "forward": true, "src_ids": "2022.acl-long.549_1619"}
+{"input": "svb task is done by using Method| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "svb task", "output": "neural singing voice beautifier ( nsvb )", "neg_sample": ["svb task is done by using Method", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Hence , we introduce Neural Singing Voice Beautifier ( NSVB ) , the first generative model to solve the SVB task , which adopts a conditional variational autoencoder as the backbone and learns the latent representations of vocal tone .", "forward": false, "src_ids": "2022.acl-long.549_1620"}
+{"input": "neural singing voice beautifier ( nsvb ) is used for Task| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "neural singing voice beautifier ( nsvb )", "output": "svb task", "neg_sample": ["neural singing voice beautifier ( nsvb ) is used for Task", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Hence , we introduce Neural Singing Voice Beautifier ( NSVB ) , the first generative model to solve the SVB task , which adopts a conditional variational autoencoder as the backbone and learns the latent representations of vocal tone .", "forward": true, "src_ids": "2022.acl-long.549_1621"}
+{"input": "generative model is used for Task| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "generative model", "output": "svb task", "neg_sample": ["generative model is used for Task", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Hence , we introduce Neural Singing Voice Beautifier ( NSVB ) , the first generative model to solve the SVB task , which adopts a conditional variational autoencoder as the backbone and learns the latent representations of vocal tone .", "forward": true, "src_ids": "2022.acl-long.549_1622"}
+{"input": "pitch correction is done by using Method| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "pitch correction", "output": "time - warping approaches", "neg_sample": ["pitch correction is done by using Method", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "In NSVB , we propose a novel time - warping approach for pitch correction : Shape - Aware Dynamic Time Warping ( SADTW ) , which ameliorates the robustness of existing time - warping approaches , to synchronize the amateur recording with the template pitch curve .", "forward": false, "src_ids": "2022.acl-long.549_1623"}
+{"input": "amateur vocal tone is done by using Method| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "amateur vocal tone", "output": "latent - mapping algorithm", "neg_sample": ["amateur vocal tone is done by using Method", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Furthermore , we propose a latent - mapping algorithm in the latent space to convert the amateur vocal tone to the professional one .", "forward": false, "src_ids": "2022.acl-long.549_1624"}
+{"input": "latent - mapping algorithm is used for OtherScientificTerm| context: current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality .", "entity": "latent - mapping algorithm", "output": "amateur vocal tone", "neg_sample": ["latent - mapping algorithm is used for OtherScientificTerm", "current automatic pitch correction techniques are immature , and most of them are restricted to intonation but ignore the overall aesthetic quality ."], "relation": "used for", "id": "2022.acl-long.549", "year": 2022, "rel_sent": "Furthermore , we propose a latent - mapping algorithm in the latent space to convert the amateur vocal tone to the professional one .", "forward": true, "src_ids": "2022.acl-long.549_1625"}
+{"input": "evaluating scientific paper embeddings is done by using Method| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "evaluating scientific paper embeddings", "output": "textomics", "neg_sample": ["evaluating scientific paper embeddings is done by using Method", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "We further illustrate how Textomics can be used to advance other applications , including evaluating scientific paper embeddings and generating masked templates for scientific paper understanding .", "forward": false, "src_ids": "2022.acl-long.335_1626"}
+{"input": "generating masked templates is done by using Method| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "generating masked templates", "output": "textomics", "neg_sample": ["generating masked templates is done by using Method", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "We further illustrate how Textomics can be used to advance other applications , including evaluating scientific paper embeddings and generating masked templates for scientific paper understanding .", "forward": false, "src_ids": "2022.acl-long.335_1627"}
+{"input": "textual summaries is done by using Method| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "textual summaries", "output": "textomics", "neg_sample": ["textual summaries is done by using Method", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "Textomics serves as the first benchmark for generating textual summaries for genomics data and we envision it will be broadly applied to other biomedical and natural language processing applications .", "forward": false, "src_ids": "2022.acl-long.335_1628"}
+{"input": "modeling genomics data is done by using Method| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "modeling genomics data", "output": "k nearest neighbors", "neg_sample": ["modeling genomics data is done by using Method", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "Inspired by the successful applications of k nearest neighbors in modeling genomics data , we propose a kNN - Vec2Text model to address these tasks and observe substantial improvement on our dataset .", "forward": false, "src_ids": "2022.acl-long.335_1629"}
+{"input": "k nearest neighbors is used for Task| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "k nearest neighbors", "output": "modeling genomics data", "neg_sample": ["k nearest neighbors is used for Task", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "Inspired by the successful applications of k nearest neighbors in modeling genomics data , we propose a kNN - Vec2Text model to address these tasks and observe substantial improvement on our dataset .", "forward": true, "src_ids": "2022.acl-long.335_1630"}
+{"input": "textomics is used for Task| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "textomics", "output": "evaluating scientific paper embeddings", "neg_sample": ["textomics is used for Task", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "We further illustrate how Textomics can be used to advance other applications , including evaluating scientific paper embeddings and generating masked templates for scientific paper understanding .", "forward": true, "src_ids": "2022.acl-long.335_1631"}
+{"input": "scientific paper understanding is done by using Task| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "scientific paper understanding", "output": "generating masked templates", "neg_sample": ["scientific paper understanding is done by using Task", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "We further illustrate how Textomics can be used to advance other applications , including evaluating scientific paper embeddings and generating masked templates for scientific paper understanding .", "forward": false, "src_ids": "2022.acl-long.335_1632"}
+{"input": "generating masked templates is used for Task| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "generating masked templates", "output": "scientific paper understanding", "neg_sample": ["generating masked templates is used for Task", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "We further illustrate how Textomics can be used to advance other applications , including evaluating scientific paper embeddings and generating masked templates for scientific paper understanding .", "forward": true, "src_ids": "2022.acl-long.335_1633"}
+{"input": "genomics data is done by using OtherScientificTerm| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "genomics data", "output": "textual summaries", "neg_sample": ["genomics data is done by using OtherScientificTerm", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "Textomics serves as the first benchmark for generating textual summaries for genomics data and we envision it will be broadly applied to other biomedical and natural language processing applications .", "forward": false, "src_ids": "2022.acl-long.335_1634"}
+{"input": "textomics is used for OtherScientificTerm| context: summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually .", "entity": "textomics", "output": "textual summaries", "neg_sample": ["textomics is used for OtherScientificTerm", "summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually ."], "relation": "used for", "id": "2022.acl-long.335", "year": 2022, "rel_sent": "Textomics serves as the first benchmark for generating textual summaries for genomics data and we envision it will be broadly applied to other biomedical and natural language processing applications .", "forward": true, "src_ids": "2022.acl-long.335_1635"}
+{"input": "response generation is done by using Method| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "response generation", "output": "hetermpc", "neg_sample": ["response generation is done by using Method", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Through multi - hop updating , HeterMPC can adequately utilize the structural knowledge of conversations for response generation .", "forward": false, "src_ids": "2022.acl-long.349_1636"}
+{"input": "response generation is done by using Method| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "response generation", "output": "hetermpc", "neg_sample": ["response generation is done by using Method", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Experimental results on the Ubuntu Internet Relay Chat ( IRC ) channel benchmark show that HeterMPC outperforms various baseline models for response generation in MPCs .", "forward": false, "src_ids": "2022.acl-long.349_1637"}
+{"input": "multi - hop updating is used for Method| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "multi - hop updating", "output": "hetermpc", "neg_sample": ["multi - hop updating is used for Method", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Through multi - hop updating , HeterMPC can adequately utilize the structural knowledge of conversations for response generation .", "forward": true, "src_ids": "2022.acl-long.349_1638"}
+{"input": "response generation is done by using Method| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "response generation", "output": "heterogeneous graph - based neural network", "neg_sample": ["response generation is done by using Method", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "HeterMPC : A Heterogeneous Graph Neural Network for Response Generation in Multi - Party Conversations.", "forward": false, "src_ids": "2022.acl-long.349_1639"}
+{"input": "response generation is done by using Method| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "response generation", "output": "heterogeneous graph - based neural network", "neg_sample": ["response generation is done by using Method", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "To address these challenges , we present HeterMPC , a heterogeneous graph - based neural network for response generation in MPCs which models the semantics of utterances and interlocutors simultaneously with two types of nodes in a graph .", "forward": false, "src_ids": "2022.acl-long.349_1640"}
+{"input": "heterogeneous interactions is done by using OtherScientificTerm| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "heterogeneous interactions", "output": "meta relations", "neg_sample": ["heterogeneous interactions is done by using OtherScientificTerm", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Besides , we also design six types of meta relations with node - edge - type - dependent parameters to characterize the heterogeneous interactions within the graph .", "forward": false, "src_ids": "2022.acl-long.349_1641"}
+{"input": "meta relations is used for OtherScientificTerm| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "meta relations", "output": "heterogeneous interactions", "neg_sample": ["meta relations is used for OtherScientificTerm", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Besides , we also design six types of meta relations with node - edge - type - dependent parameters to characterize the heterogeneous interactions within the graph .", "forward": true, "src_ids": "2022.acl-long.349_1642"}
+{"input": "hetermpc is done by using Method| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "hetermpc", "output": "multi - hop updating", "neg_sample": ["hetermpc is done by using Method", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Through multi - hop updating , HeterMPC can adequately utilize the structural knowledge of conversations for response generation .", "forward": false, "src_ids": "2022.acl-long.349_1643"}
+{"input": "response generation is done by using OtherScientificTerm| context: recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated . compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e. , speaker and addressee ) and history utterances .", "entity": "response generation", "output": "structural knowledge of conversations", "neg_sample": ["response generation is done by using OtherScientificTerm", "recently , various response generation models for two - party conversations have achieved impressive improvements , but less effort has been paid to multi - party conversations ( mpcs ) which are more practical and complicated .", "compared with a two - party conversation where a dialogue context is a sequence of utterances , building a response generation model for mpcs is more challenging , since there exist complicated context structures and the generated responses heavily rely on both interlocutors ( i.e.", ", speaker and addressee ) and history utterances ."], "relation": "used for", "id": "2022.acl-long.349", "year": 2022, "rel_sent": "Through multi - hop updating , HeterMPC can adequately utilize the structural knowledge of conversations for response generation .", "forward": false, "src_ids": "2022.acl-long.349_1644"}
+{"input": "extraction emotion is done by using Method| context: over 3.96 million people use these platforms to send messages formed using texts , images , videos , audio or combinations of these to express their thoughts and feelings . text communication on social media platforms is quite overwhelming due to its enormous quantity and simplicity . the data must be processed to understand the general feeling felt by the author .", "entity": "extraction emotion", "output": "lexicon - based approach", "neg_sample": ["extraction emotion is done by using Method", "over 3.96 million people use these platforms to send messages formed using texts , images , videos , audio or combinations of these to express their thoughts and feelings .", "text communication on social media platforms is quite overwhelming due to its enormous quantity and simplicity .", "the data must be processed to understand the general feeling felt by the author ."], "relation": "used for", "id": "2022.dravidianlangtech-1.26", "year": 2022, "rel_sent": "We present a lexicon - based approach for the extraction emotion in Tamil texts .", "forward": false, "src_ids": "2022.dravidianlangtech-1.26_1645"}
+{"input": "lexicon - based approach is used for Task| context: over 3.96 million people use these platforms to send messages formed using texts , images , videos , audio or combinations of these to express their thoughts and feelings . text communication on social media platforms is quite overwhelming due to its enormous quantity and simplicity . the data must be processed to understand the general feeling felt by the author .", "entity": "lexicon - based approach", "output": "extraction emotion", "neg_sample": ["lexicon - based approach is used for Task", "over 3.96 million people use these platforms to send messages formed using texts , images , videos , audio or combinations of these to express their thoughts and feelings .", "text communication on social media platforms is quite overwhelming due to its enormous quantity and simplicity .", "the data must be processed to understand the general feeling felt by the author ."], "relation": "used for", "id": "2022.dravidianlangtech-1.26", "year": 2022, "rel_sent": "We present a lexicon - based approach for the extraction emotion in Tamil texts .", "forward": true, "src_ids": "2022.dravidianlangtech-1.26_1646"}
+{"input": "zero - shot cross - lingual transferability is used for Method| context: unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd . in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora . this makes annotation of lscd datasets very labour - consuming . the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning .", "entity": "zero - shot cross - lingual transferability", "output": "contextualized embeddings", "neg_sample": ["zero - shot cross - lingual transferability is used for Method", "unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd .", "in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora .", "this makes annotation of lscd datasets very labour - consuming .", "the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning ."], "relation": "used for", "id": "2022.lchange-1.22", "year": 2022, "rel_sent": "Then we employ zero - shot cross - lingual transferability of XLM - R to build the contextualized embeddings for examples in Spanish .", "forward": true, "src_ids": "2022.lchange-1.22_1647"}
+{"input": "xlm - r is used for Method| context: unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd . in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora . this makes annotation of lscd datasets very labour - consuming . the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning .", "entity": "xlm - r", "output": "contextualized embeddings", "neg_sample": ["xlm - r is used for Method", "unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd .", "in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora .", "this makes annotation of lscd datasets very labour - consuming .", "the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning ."], "relation": "used for", "id": "2022.lchange-1.22", "year": 2022, "rel_sent": "Then we employ zero - shot cross - lingual transferability of XLM - R to build the contextualized embeddings for examples in Spanish .", "forward": true, "src_ids": "2022.lchange-1.22_1648"}
+{"input": "gloss - based wsd system is done by using Method| context: the contextualized embeddings obtained from neural networks pre - trained as language models ( lm ) or masked language models ( mlm ) are not well suitable for solving the lexical semantic change detection ( lscd ) task because they are more sensitive to changes in word forms rather than word meaning , a property previously known as the word form bias or orthographic bias . unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd . in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora . this makes annotation of lscd datasets very labour - consuming . the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning .", "entity": "gloss - based wsd system", "output": "xlm - r mlm", "neg_sample": ["gloss - based wsd system is done by using Method", "the contextualized embeddings obtained from neural networks pre - trained as language models ( lm ) or masked language models ( mlm ) are not well suitable for solving the lexical semantic change detection ( lscd ) task because they are more sensitive to changes in word forms rather than word meaning , a property previously known as the word form bias or orthographic bias .", "unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd .", "in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora .", "this makes annotation of lscd datasets very labour - consuming .", "the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning ."], "relation": "used for", "id": "2022.lchange-1.22", "year": 2022, "rel_sent": "To solve these problems we fine - tune the XLM - R MLM as part of a gloss - based WSD system on a large WSD dataset in English .", "forward": false, "src_ids": "2022.lchange-1.22_1649"}
+{"input": "xlm - r mlm is used for Method| context: the contextualized embeddings obtained from neural networks pre - trained as language models ( lm ) or masked language models ( mlm ) are not well suitable for solving the lexical semantic change detection ( lscd ) task because they are more sensitive to changes in word forms rather than word meaning , a property previously known as the word form bias or orthographic bias . unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd . in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora . this makes annotation of lscd datasets very labour - consuming . the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning .", "entity": "xlm - r mlm", "output": "gloss - based wsd system", "neg_sample": ["xlm - r mlm is used for Method", "the contextualized embeddings obtained from neural networks pre - trained as language models ( lm ) or masked language models ( mlm ) are not well suitable for solving the lexical semantic change detection ( lscd ) task because they are more sensitive to changes in word forms rather than word meaning , a property previously known as the word form bias or orthographic bias .", "unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd .", "in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora .", "this makes annotation of lscd datasets very labour - consuming .", "the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning ."], "relation": "used for", "id": "2022.lchange-1.22", "year": 2022, "rel_sent": "To solve these problems we fine - tune the XLM - R MLM as part of a gloss - based WSD system on a large WSD dataset in English .", "forward": true, "src_ids": "2022.lchange-1.22_1650"}
+{"input": "contextualized embeddings is done by using Method| context: the contextualized embeddings obtained from neural networks pre - trained as language models ( lm ) or masked language models ( mlm ) are not well suitable for solving the lexical semantic change detection ( lscd ) task because they are more sensitive to changes in word forms rather than word meaning , a property previously known as the word form bias or orthographic bias . unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd . in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora . this makes annotation of lscd datasets very labour - consuming . the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning .", "entity": "contextualized embeddings", "output": "zero - shot cross - lingual transferability", "neg_sample": ["contextualized embeddings is done by using Method", "the contextualized embeddings obtained from neural networks pre - trained as language models ( lm ) or masked language models ( mlm ) are not well suitable for solving the lexical semantic change detection ( lscd ) task because they are more sensitive to changes in word forms rather than word meaning , a property previously known as the word form bias or orthographic bias .", "unlike many other nlp tasks , it is also not obvious how tofine - tune such models for lscd .", "in order to conclude if there are any differences between senses of a particular word in two corpora , a human annotator or a system shall analyze many examples containing this word from both corpora .", "this makes annotation of lscd datasets very labour - consuming .", "the existing lscd datasets contain up to 100 words that are labeled according to their semantic change , which is hardly enough for fine - tuning ."], "relation": "used for", "id": "2022.lchange-1.22", "year": 2022, "rel_sent": "Then we employ zero - shot cross - lingual transferability of XLM - R to build the contextualized embeddings for examples in Spanish .", "forward": false, "src_ids": "2022.lchange-1.22_1651"}
+{"input": "open - ended reasoning is done by using Method| context: images are often more significant than only the pixels to human eyes , as we can infer , associate , and reason with contextual information from other sources to establish a more complete picture . for example , in figure 1 , we can find a way to identify the news articles related to the picture through segment - wise understandings of the signs , the buildings , the crowds , and more .", "entity": "open - ended reasoning", "output": "higher - level vision - language joint models", "neg_sample": ["open - ended reasoning is done by using Method", "images are often more significant than only the pixels to human eyes , as we can infer , associate , and reason with contextual information from other sources to establish a more complete picture .", "for example , in figure 1 , we can find a way to identify the news articles related to the picture through segment - wise understandings of the signs , the buildings , the crowds , and more ."], "relation": "used for", "id": "2022.acl-long.81", "year": 2022, "rel_sent": "We show that there exists a 70 % gap between a state - of - the - art joint model and human performance , which is slightly filled by our proposed model that uses segment - wise reasoning , motivating higher - level vision - language joint models that can conduct open - ended reasoning with world knowledge .", "forward": false, "src_ids": "2022.acl-long.81_1652"}
+{"input": "higher - level vision - language joint models is used for Method| context: images are often more significant than only the pixels to human eyes , as we can infer , associate , and reason with contextual information from other sources to establish a more complete picture . for example , in figure 1 , we can find a way to identify the news articles related to the picture through segment - wise understandings of the signs , the buildings , the crowds , and more .", "entity": "higher - level vision - language joint models", "output": "open - ended reasoning", "neg_sample": ["higher - level vision - language joint models is used for Method", "images are often more significant than only the pixels to human eyes , as we can infer , associate , and reason with contextual information from other sources to establish a more complete picture .", "for example , in figure 1 , we can find a way to identify the news articles related to the picture through segment - wise understandings of the signs , the buildings , the crowds , and more ."], "relation": "used for", "id": "2022.acl-long.81", "year": 2022, "rel_sent": "We show that there exists a 70 % gap between a state - of - the - art joint model and human performance , which is slightly filled by our proposed model that uses segment - wise reasoning , motivating higher - level vision - language joint models that can conduct open - ended reasoning with world knowledge .", "forward": true, "src_ids": "2022.acl-long.81_1653"}
+{"input": "grounded summaries is used for Task| context: podcasts have shown a recent rise in popularity . summarization of podcasts is of practical benefit to both content providers and consumers . it helps people quickly decide whether they will listen to a podcast and/or reduces the cognitive load of content providers to write summaries . nevertheless , podcast summarization faces significant challenges including factual inconsistencies of summaries with respect to the inputs . the problem is exacerbated by speech disfluencies and recognition errors in transcripts of spoken language .", "entity": "grounded summaries", "output": "locating the summary and transcript segments", "neg_sample": ["grounded summaries is used for Task", "podcasts have shown a recent rise in popularity .", "summarization of podcasts is of practical benefit to both content providers and consumers .", "it helps people quickly decide whether they will listen to a podcast and/or reduces the cognitive load of content providers to write summaries .", "nevertheless , podcast summarization faces significant challenges including factual inconsistencies of summaries with respect to the inputs .", "the problem is exacerbated by speech disfluencies and recognition errors in transcripts of spoken language ."], "relation": "used for", "id": "2022.acl-long.302", "year": 2022, "rel_sent": "Grounded summaries bring clear benefits in locating the summary and transcript segments that contain inconsistent information , and hence improve summarization quality in terms of automatic and human evaluation .", "forward": true, "src_ids": "2022.acl-long.302_1654"}
+{"input": "locating the summary and transcript segments is done by using Material| context: podcasts have shown a recent rise in popularity . summarization of podcasts is of practical benefit to both content providers and consumers . it helps people quickly decide whether they will listen to a podcast and/or reduces the cognitive load of content providers to write summaries . nevertheless , podcast summarization faces significant challenges including factual inconsistencies of summaries with respect to the inputs . the problem is exacerbated by speech disfluencies and recognition errors in transcripts of spoken language .", "entity": "locating the summary and transcript segments", "output": "grounded summaries", "neg_sample": ["locating the summary and transcript segments is done by using Material", "podcasts have shown a recent rise in popularity .", "summarization of podcasts is of practical benefit to both content providers and consumers .", "it helps people quickly decide whether they will listen to a podcast and/or reduces the cognitive load of content providers to write summaries .", "nevertheless , podcast summarization faces significant challenges including factual inconsistencies of summaries with respect to the inputs .", "the problem is exacerbated by speech disfluencies and recognition errors in transcripts of spoken language ."], "relation": "used for", "id": "2022.acl-long.302", "year": 2022, "rel_sent": "Grounded summaries bring clear benefits in locating the summary and transcript segments that contain inconsistent information , and hence improve summarization quality in terms of automatic and human evaluation .", "forward": false, "src_ids": "2022.acl-long.302_1655"}
+{"input": "non - autoregressive is done by using OtherScientificTerm| context: autoregressive ( ar ) and non - autoregressive ( nar ) models have their own superiority on the performance and latency , combining them into one model may take advantage of both . current combination frameworks focus more on the integration of multiple decoding paradigms with a unified generative model , e.g. masked language model . however , the generalization can be harmful on the performance due to the gap between training objective and inference .", "entity": "non - autoregressive", "output": "dependency assumption", "neg_sample": ["non - autoregressive is done by using OtherScientificTerm", "autoregressive ( ar ) and non - autoregressive ( nar ) models have their own superiority on the performance and latency , combining them into one model may take advantage of both .", "current combination frameworks focus more on the integration of multiple decoding paradigms with a unified generative model , e.g.", "masked language model .", "however , the generalization can be harmful on the performance due to the gap between training objective and inference ."], "relation": "used for", "id": "2022.eamt-1.11", "year": 2022, "rel_sent": "The unification achieved by direction successfully preserves the original dependency assumption used in AR and NAR , retaining both generalization and performance .", "forward": false, "src_ids": "2022.eamt-1.11_1656"}
+{"input": "autoregressive is done by using OtherScientificTerm| context: autoregressive ( ar ) and non - autoregressive ( nar ) models have their own superiority on the performance and latency , combining them into one model may take advantage of both . current combination frameworks focus more on the integration of multiple decoding paradigms with a unified generative model , e.g. masked language model . however , the generalization can be harmful on the performance due to the gap between training objective and inference .", "entity": "autoregressive", "output": "dependency assumption", "neg_sample": ["autoregressive is done by using OtherScientificTerm", "autoregressive ( ar ) and non - autoregressive ( nar ) models have their own superiority on the performance and latency , combining them into one model may take advantage of both .", "current combination frameworks focus more on the integration of multiple decoding paradigms with a unified generative model , e.g.", "masked language model .", "however , the generalization can be harmful on the performance due to the gap between training objective and inference ."], "relation": "used for", "id": "2022.eamt-1.11", "year": 2022, "rel_sent": "The unification achieved by direction successfully preserves the original dependency assumption used in AR and NAR , retaining both generalization and performance .", "forward": false, "src_ids": "2022.eamt-1.11_1657"}
+{"input": "dependency assumption is used for Method| context: current combination frameworks focus more on the integration of multiple decoding paradigms with a unified generative model , e.g. masked language model . however , the generalization can be harmful on the performance due to the gap between training objective and inference .", "entity": "dependency assumption", "output": "autoregressive", "neg_sample": ["dependency assumption is used for Method", "current combination frameworks focus more on the integration of multiple decoding paradigms with a unified generative model , e.g.", "masked language model .", "however , the generalization can be harmful on the performance due to the gap between training objective and inference ."], "relation": "used for", "id": "2022.eamt-1.11", "year": 2022, "rel_sent": "The unification achieved by direction successfully preserves the original dependency assumption used in AR and NAR , retaining both generalization and performance .", "forward": true, "src_ids": "2022.eamt-1.11_1658"}
+{"input": "transformer - based misogyny detection is done by using Method| context: transformer - based natural language processing models have become the standard for hate speech detection . however , the unconscious use of these techniques for such a critical task comes with negative consequences . various works have demonstrated that hate speech classifiers are biased . these findings have prompted efforts to explain classifiers , mainly using attribution methods .", "entity": "transformer - based misogyny detection", "output": "post - hoc interpretability approaches", "neg_sample": ["transformer - based misogyny detection is done by using Method", "transformer - based natural language processing models have become the standard for hate speech detection .", "however , the unconscious use of these techniques for such a critical task comes with negative consequences .", "various works have demonstrated that hate speech classifiers are biased .", "these findings have prompted efforts to explain classifiers , mainly using attribution methods ."], "relation": "used for", "id": "2022.nlppower-1.11", "year": 2022, "rel_sent": "Benchmarking Post - Hoc Interpretability Approaches for Transformer - based Misogyny Detection.", "forward": false, "src_ids": "2022.nlppower-1.11_1659"}
+{"input": "post - hoc interpretability approaches is used for Task| context: transformer - based natural language processing models have become the standard for hate speech detection . however , the unconscious use of these techniques for such a critical task comes with negative consequences . various works have demonstrated that hate speech classifiers are biased . these findings have prompted efforts to explain classifiers , mainly using attribution methods .", "entity": "post - hoc interpretability approaches", "output": "transformer - based misogyny detection", "neg_sample": ["post - hoc interpretability approaches is used for Task", "transformer - based natural language processing models have become the standard for hate speech detection .", "however , the unconscious use of these techniques for such a critical task comes with negative consequences .", "various works have demonstrated that hate speech classifiers are biased .", "these findings have prompted efforts to explain classifiers , mainly using attribution methods ."], "relation": "used for", "id": "2022.nlppower-1.11", "year": 2022, "rel_sent": "Benchmarking Post - Hoc Interpretability Approaches for Transformer - based Misogyny Detection.", "forward": true, "src_ids": "2022.nlppower-1.11_1660"}
+{"input": "interpretability approaches is used for Task| context: however , the unconscious use of these techniques for such a critical task comes with negative consequences . various works have demonstrated that hate speech classifiers are biased . these findings have prompted efforts to explain classifiers , mainly using attribution methods .", "entity": "interpretability approaches", "output": "hate speech detection", "neg_sample": ["interpretability approaches is used for Task", "however , the unconscious use of these techniques for such a critical task comes with negative consequences .", "various works have demonstrated that hate speech classifiers are biased .", "these findings have prompted efforts to explain classifiers , mainly using attribution methods ."], "relation": "used for", "id": "2022.nlppower-1.11", "year": 2022, "rel_sent": "In this paper , we provide the first benchmark study of interpretability approaches for hate speech detection .", "forward": true, "src_ids": "2022.nlppower-1.11_1661"}
+{"input": "hate speech detection is done by using Method| context: transformer - based natural language processing models have become the standard for hate speech detection . however , the unconscious use of these techniques for such a critical task comes with negative consequences . various works have demonstrated that hate speech classifiers are biased . these findings have prompted efforts to explain classifiers , mainly using attribution methods .", "entity": "hate speech detection", "output": "interpretability approaches", "neg_sample": ["hate speech detection is done by using Method", "transformer - based natural language processing models have become the standard for hate speech detection .", "however , the unconscious use of these techniques for such a critical task comes with negative consequences .", "various works have demonstrated that hate speech classifiers are biased .", "these findings have prompted efforts to explain classifiers , mainly using attribution methods ."], "relation": "used for", "id": "2022.nlppower-1.11", "year": 2022, "rel_sent": "In this paper , we provide the first benchmark study of interpretability approaches for hate speech detection .", "forward": false, "src_ids": "2022.nlppower-1.11_1662"}
+{"input": "transformer - based misogyny classifiers is done by using Method| context: transformer - based natural language processing models have become the standard for hate speech detection . however , the unconscious use of these techniques for such a critical task comes with negative consequences . various works have demonstrated that hate speech classifiers are biased . these findings have prompted efforts to explain classifiers , mainly using attribution methods .", "entity": "transformer - based misogyny classifiers", "output": "post - hoc token attribution approaches", "neg_sample": ["transformer - based misogyny classifiers is done by using Method", "transformer - based natural language processing models have become the standard for hate speech detection .", "however , the unconscious use of these techniques for such a critical task comes with negative consequences .", "various works have demonstrated that hate speech classifiers are biased .", "these findings have prompted efforts to explain classifiers , mainly using attribution methods ."], "relation": "used for", "id": "2022.nlppower-1.11", "year": 2022, "rel_sent": "We cover four post - hoc token attribution approaches to explain the predictions of Transformer - based misogyny classifiers in English and Italian .", "forward": false, "src_ids": "2022.nlppower-1.11_1663"}
+{"input": "post - hoc token attribution approaches is used for Method| context: transformer - based natural language processing models have become the standard for hate speech detection . however , the unconscious use of these techniques for such a critical task comes with negative consequences . various works have demonstrated that hate speech classifiers are biased . these findings have prompted efforts to explain classifiers , mainly using attribution methods .", "entity": "post - hoc token attribution approaches", "output": "transformer - based misogyny classifiers", "neg_sample": ["post - hoc token attribution approaches is used for Method", "transformer - based natural language processing models have become the standard for hate speech detection .", "however , the unconscious use of these techniques for such a critical task comes with negative consequences .", "various works have demonstrated that hate speech classifiers are biased .", "these findings have prompted efforts to explain classifiers , mainly using attribution methods ."], "relation": "used for", "id": "2022.nlppower-1.11", "year": 2022, "rel_sent": "We cover four post - hoc token attribution approaches to explain the predictions of Transformer - based misogyny classifiers in English and Italian .", "forward": true, "src_ids": "2022.nlppower-1.11_1664"}
+{"input": "learn to adapt is used for Method| context: generalized zero - shot text classification aims to classify textual instances from both previously seen classes and incrementally emerging unseen classes . most existing methods generalize poorly since the learned parameters are only optimal for seen classes rather than for both classes , and the parameters keep stationary in predicting procedures .", "entity": "learn to adapt", "output": "adaptive classifier", "neg_sample": ["learn to adapt is used for Method", "generalized zero - shot text classification aims to classify textual instances from both previously seen classes and incrementally emerging unseen classes .", "most existing methods generalize poorly since the learned parameters are only optimal for seen classes rather than for both classes , and the parameters keep stationary in predicting procedures ."], "relation": "used for", "id": "2022.acl-long.39", "year": 2022, "rel_sent": "Specifically , LTA trains an adaptive classifier by using both seen and virtual unseen classes to simulate a generalized zero - shot learning ( GZSL ) scenario in accordance with the test time , and simultaneously learns to calibrate the class prototypes and sample representations to make the learned parameters adaptive to incoming unseen classes .", "forward": true, "src_ids": "2022.acl-long.39_1665"}
+{"input": "adaptive classifier is done by using Method| context: generalized zero - shot text classification aims to classify textual instances from both previously seen classes and incrementally emerging unseen classes . most existing methods generalize poorly since the learned parameters are only optimal for seen classes rather than for both classes , and the parameters keep stationary in predicting procedures .", "entity": "adaptive classifier", "output": "learn to adapt", "neg_sample": ["adaptive classifier is done by using Method", "generalized zero - shot text classification aims to classify textual instances from both previously seen classes and incrementally emerging unseen classes .", "most existing methods generalize poorly since the learned parameters are only optimal for seen classes rather than for both classes , and the parameters keep stationary in predicting procedures ."], "relation": "used for", "id": "2022.acl-long.39", "year": 2022, "rel_sent": "Specifically , LTA trains an adaptive classifier by using both seen and virtual unseen classes to simulate a generalized zero - shot learning ( GZSL ) scenario in accordance with the test time , and simultaneously learns to calibrate the class prototypes and sample representations to make the learned parameters adaptive to incoming unseen classes .", "forward": false, "src_ids": "2022.acl-long.39_1666"}
+{"input": "probing techniques is used for Method| context: given the prevalence of pre - trained contextualized representations in today 's nlp , there have been many efforts to understand what information they contain , and why they seem to be universally successful .", "entity": "probing techniques", "output": "fine - tuning", "neg_sample": ["probing techniques is used for Method", "given the prevalence of pre - trained contextualized representations in today 's nlp , there have been many efforts to understand what information they contain , and why they seem to be universally successful ."], "relation": "used for", "id": "2022.acl-long.75", "year": 2022, "rel_sent": "In this work , we study the English BERT family and use two probing techniques to analyze how fine - tuning changes the space .", "forward": true, "src_ids": "2022.acl-long.75_1667"}
+{"input": "fine - tuning is done by using Method| context: given the prevalence of pre - trained contextualized representations in today 's nlp , there have been many efforts to understand what information they contain , and why they seem to be universally successful . the most common approach to use these representations involves fine - tuning them for an end task . yet , how fine - tuning changes the underlying embedding space is less studied .", "entity": "fine - tuning", "output": "probing techniques", "neg_sample": ["fine - tuning is done by using Method", "given the prevalence of pre - trained contextualized representations in today 's nlp , there have been many efforts to understand what information they contain , and why they seem to be universally successful .", "the most common approach to use these representations involves fine - tuning them for an end task .", "yet , how fine - tuning changes the underlying embedding space is less studied ."], "relation": "used for", "id": "2022.acl-long.75", "year": 2022, "rel_sent": "In this work , we study the English BERT family and use two probing techniques to analyze how fine - tuning changes the space .", "forward": false, "src_ids": "2022.acl-long.75_1668"}
+{"input": "vae is done by using Method| context: modelling prosody variation is critical for synthesizing natural and expressive speech in end - to - end text - to - speech ( tts ) systems .", "entity": "vae", "output": "gaussian distribution", "neg_sample": ["vae is done by using Method", "modelling prosody variation is critical for synthesizing natural and expressive speech in end - to - end text - to - speech ( tts ) systems ."], "relation": "used for", "id": "2022.acl-long.30", "year": 2022, "rel_sent": "At inference time , instead of the standard Gaussian distribution used by VAE , CUC - VAE allows sampling from an utterance - specific prior distribution conditioned on cross - utterance information , which allows the prosody features generated by the TTS system to be related to the context and is more similar to how humans naturally produce prosody .", "forward": false, "src_ids": "2022.acl-long.30_1669"}
+{"input": "gaussian distribution is used for Method| context: modelling prosody variation is critical for synthesizing natural and expressive speech in end - to - end text - to - speech ( tts ) systems .", "entity": "gaussian distribution", "output": "vae", "neg_sample": ["gaussian distribution is used for Method", "modelling prosody variation is critical for synthesizing natural and expressive speech in end - to - end text - to - speech ( tts ) systems ."], "relation": "used for", "id": "2022.acl-long.30", "year": 2022, "rel_sent": "At inference time , instead of the standard Gaussian distribution used by VAE , CUC - VAE allows sampling from an utterance - specific prior distribution conditioned on cross - utterance information , which allows the prosody features generated by the TTS system to be related to the context and is more similar to how humans naturally produce prosody .", "forward": true, "src_ids": "2022.acl-long.30_1670"}
+{"input": "language models is done by using Material| context: many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for . however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time . this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize .", "entity": "language models", "output": "diagnostic dataset", "neg_sample": ["language models is done by using Material", "many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for .", "however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time .", "this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize ."], "relation": "used for", "id": "2022.tacl-1.15", "year": 2022, "rel_sent": "We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum - those trained on specific slices of temporal data , as well as those trained on a wide range of temporal data .", "forward": false, "src_ids": "2022.tacl-1.15_1671"}
+{"input": "factual knowledge is done by using Material| context: many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for . however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time . this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize .", "entity": "factual knowledge", "output": "diagnostic dataset", "neg_sample": ["factual knowledge is done by using Material", "many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for .", "however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time .", "this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize ."], "relation": "used for", "id": "2022.tacl-1.15", "year": 2022, "rel_sent": "We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum - those trained on specific slices of temporal data , as well as those trained on a wide range of temporal data .", "forward": false, "src_ids": "2022.tacl-1.15_1672"}
+{"input": "factual knowledge is done by using Method| context: many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for . this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize .", "entity": "factual knowledge", "output": "language models", "neg_sample": ["factual knowledge is done by using Method", "many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for .", "this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize ."], "relation": "used for", "id": "2022.tacl-1.15", "year": 2022, "rel_sent": "We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum - those trained on specific slices of temporal data , as well as those trained on a wide range of temporal data .", "forward": false, "src_ids": "2022.tacl-1.15_1673"}
+{"input": "diagnostic dataset is used for Method| context: many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for . this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize .", "entity": "diagnostic dataset", "output": "language models", "neg_sample": ["diagnostic dataset is used for Method", "many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for .", "this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize ."], "relation": "used for", "id": "2022.tacl-1.15", "year": 2022, "rel_sent": "We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum - those trained on specific slices of temporal data , as well as those trained on a wide range of temporal data .", "forward": true, "src_ids": "2022.tacl-1.15_1674"}
+{"input": "language models is used for OtherScientificTerm| context: many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for . however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time . this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize .", "entity": "language models", "output": "factual knowledge", "neg_sample": ["language models is used for OtherScientificTerm", "many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for .", "however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time .", "this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize ."], "relation": "used for", "id": "2022.tacl-1.15", "year": 2022, "rel_sent": "We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum - those trained on specific slices of temporal data , as well as those trained on a wide range of temporal data .", "forward": true, "src_ids": "2022.tacl-1.15_1675"}
+{"input": "diagnostic dataset is used for OtherScientificTerm| context: many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for . however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time . this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize .", "entity": "diagnostic dataset", "output": "factual knowledge", "neg_sample": ["diagnostic dataset is used for OtherScientificTerm", "many facts come with an expiration date , from the name of the president to the basketball team lebron james plays for .", "however , most language models ( lms ) are trained on snapshots of data collected at a specific moment in time .", "this can limit their utility , especially in the closed - book setting where the pretraining corpus must contain the facts the model should memorize ."], "relation": "used for", "id": "2022.tacl-1.15", "year": 2022, "rel_sent": "We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum - those trained on specific slices of temporal data , as well as those trained on a wide range of temporal data .", "forward": true, "src_ids": "2022.tacl-1.15_1676"}
+{"input": "history information enhanced methods is used for Method| context: recently , context - dependent text - to - sql semantic parsing which translates natural language into sql in an interaction process has attracted a lot of attentions . previous works leverage context dependence information either from interaction history utterances or previous predicted queries but fail in taking advantage of both of them since of the mismatch between the natural language and logic - form sql .", "entity": "history information enhanced methods", "output": "hie - sql", "neg_sample": ["history information enhanced methods is used for Method", "recently , context - dependent text - to - sql semantic parsing which translates natural language into sql in an interaction process has attracted a lot of attentions .", "previous works leverage context dependence information either from interaction history utterances or previous predicted queries but fail in taking advantage of both of them since of the mismatch between the natural language and logic - form sql ."], "relation": "used for", "id": "2022.findings-acl.236", "year": 2022, "rel_sent": "We show our history information enhanced methods improve the performance of HIE - SQL by a significant margin , which achieves new state - of - the - art results on two context - dependent text - to - SQL benchmarks , the SparC and CoSQL datasets , at the writing time .", "forward": true, "src_ids": "2022.findings-acl.236_1677"}
+{"input": "hie - sql is done by using Method| context: recently , context - dependent text - to - sql semantic parsing which translates natural language into sql in an interaction process has attracted a lot of attentions . previous works leverage context dependence information either from interaction history utterances or previous predicted queries but fail in taking advantage of both of them since of the mismatch between the natural language and logic - form sql .", "entity": "hie - sql", "output": "history information enhanced methods", "neg_sample": ["hie - sql is done by using Method", "recently , context - dependent text - to - sql semantic parsing which translates natural language into sql in an interaction process has attracted a lot of attentions .", "previous works leverage context dependence information either from interaction history utterances or previous predicted queries but fail in taking advantage of both of them since of the mismatch between the natural language and logic - form sql ."], "relation": "used for", "id": "2022.findings-acl.236", "year": 2022, "rel_sent": "We show our history information enhanced methods improve the performance of HIE - SQL by a significant margin , which achieves new state - of - the - art results on two context - dependent text - to - SQL benchmarks , the SparC and CoSQL datasets , at the writing time .", "forward": false, "src_ids": "2022.findings-acl.236_1678"}
+{"input": "similarity of frequent words is done by using OtherScientificTerm| context: here , we uncover systematic ways in which word similarities estimated by cosine over bert embeddings are understated and trace this effect to training data frequency .", "entity": "similarity of frequent words", "output": "cosine similarity", "neg_sample": ["similarity of frequent words is done by using OtherScientificTerm", "here , we uncover systematic ways in which word similarities estimated by cosine over bert embeddings are understated and trace this effect to training data frequency ."], "relation": "used for", "id": "2022.acl-short.45", "year": 2022, "rel_sent": "We find that relative to human judgements , cosine similarity underestimates the similarity of frequent words with other instances of the same word or other words across contexts , even after controlling for polysemy and other factors .", "forward": false, "src_ids": "2022.acl-short.45_1679"}
+{"input": "cosine similarity is used for OtherScientificTerm| context: cosine similarity of contextual embeddings is used in many nlp tasks ( e.g. , qa , ir , mt ) and metrics ( e.g. , bertscore ) . here , we uncover systematic ways in which word similarities estimated by cosine over bert embeddings are understated and trace this effect to training data frequency .", "entity": "cosine similarity", "output": "similarity of frequent words", "neg_sample": ["cosine similarity is used for OtherScientificTerm", "cosine similarity of contextual embeddings is used in many nlp tasks ( e.g.", ", qa , ir , mt ) and metrics ( e.g.", ", bertscore ) .", "here , we uncover systematic ways in which word similarities estimated by cosine over bert embeddings are understated and trace this effect to training data frequency ."], "relation": "used for", "id": "2022.acl-short.45", "year": 2022, "rel_sent": "We find that relative to human judgements , cosine similarity underestimates the similarity of frequent words with other instances of the same word or other words across contexts , even after controlling for polysemy and other factors .", "forward": true, "src_ids": "2022.acl-short.45_1680"}
+{"input": "syntactic associations is done by using Method| context: however , this ability has primarily been shown for constructions for which the surface strings frequently provide information about dependencies in the form of agreement patterns . for example , if a model has access to sentences with and without a noun phrase intervening between the subject and the main verb ( 1 ) , it is often able to infer the agreement dependencies from the surface string alone : ( linzen et al . , 2016 ; marvin and linzen , 2018 ; goldberg , 2019 ; gulordava et al . , 2018 ; hu et al . , 2020b ) . the surface cues are boldfaced in ( 1 ):", "entity": "syntactic associations", "output": "language models", "neg_sample": ["syntactic associations is done by using Method", "however , this ability has primarily been shown for constructions for which the surface strings frequently provide information about dependencies in the form of agreement patterns .", "for example , if a model has access to sentences with and without a noun phrase intervening between the subject and the main verb ( 1 ) , it is often able to infer the agreement dependencies from the surface string alone : ( linzen et al .", ", 2016 ; marvin and linzen , 2018 ; goldberg , 2019 ; gulordava et al .", ", 2018 ; hu et al .", ", 2020b ) .", "the surface cues are boldfaced in ( 1 ):"], "relation": "used for", "id": "2022.scil-1.18", "year": 2022, "rel_sent": "Can language models capture syntactic associations without surface cues ? A case study of reflexive anaphor licensing in English control constructions.", "forward": false, "src_ids": "2022.scil-1.18_1681"}
+{"input": "language models is used for OtherScientificTerm| context: recent studies have shown that language models ( lms ) have the ability to capture many longdistance dependencies such as filler - gap dependencies ( wilcox et al . , 2018 ) and subject - verb agreement ( linzen et al . , 2016 ) despite only learning from surface strings . however , this ability has primarily been shown for constructions for which the surface strings frequently provide information about dependencies in the form of agreement patterns . for example , if a model has access to sentences with and without a noun phrase intervening between the subject and the main verb ( 1 ) , it is often able to infer the agreement dependencies from the surface string alone : ( linzen et al . , 2016 ; marvin and linzen , 2018 ; goldberg , 2019 ; gulordava et al . , 2018 ; hu et al . , 2020b ) . the surface cues are boldfaced in ( 1 ):", "entity": "language models", "output": "syntactic associations", "neg_sample": ["language models is used for OtherScientificTerm", "recent studies have shown that language models ( lms ) have the ability to capture many longdistance dependencies such as filler - gap dependencies ( wilcox et al .", ", 2018 ) and subject - verb agreement ( linzen et al .", ", 2016 ) despite only learning from surface strings .", "however , this ability has primarily been shown for constructions for which the surface strings frequently provide information about dependencies in the form of agreement patterns .", "for example , if a model has access to sentences with and without a noun phrase intervening between the subject and the main verb ( 1 ) , it is often able to infer the agreement dependencies from the surface string alone : ( linzen et al .", ", 2016 ; marvin and linzen , 2018 ; goldberg , 2019 ; gulordava et al .", ", 2018 ; hu et al .", ", 2020b ) .", "the surface cues are boldfaced in ( 1 ):"], "relation": "used for", "id": "2022.scil-1.18", "year": 2022, "rel_sent": "Can language models capture syntactic associations without surface cues ? A case study of reflexive anaphor licensing in English control constructions.", "forward": true, "src_ids": "2022.scil-1.18_1682"}
+{"input": "scene - text is done by using Method| context: we tackle the tasks of image and text retrieval using a dual - encoder model in which images and text are encoded independently .", "entity": "scene - text", "output": "image encoder", "neg_sample": ["scene - text is done by using Method", "we tackle the tasks of image and text retrieval using a dual - encoder model in which images and text are encoded independently ."], "relation": "used for", "id": "2022.acl-srw.34", "year": 2022, "rel_sent": "This model has attracted attention as an approach that enables efficient offline inferences by connecting both vision and language in the same semantic space ; however , whether an image encoder as part of a dual - encoder model can interpret scene - text ( i.e. , the textual information in images ) is unclear . We propose pre - training methods that encourage a joint understanding of the scene - text and surrounding visual information . The experimental results demonstrate that our methods improve the retrieval performances of the dual - encoder models .", "forward": false, "src_ids": "2022.acl-srw.34_1683"}
+{"input": "image encoder is used for OtherScientificTerm| context: we tackle the tasks of image and text retrieval using a dual - encoder model in which images and text are encoded independently .", "entity": "image encoder", "output": "scene - text", "neg_sample": ["image encoder is used for OtherScientificTerm", "we tackle the tasks of image and text retrieval using a dual - encoder model in which images and text are encoded independently ."], "relation": "used for", "id": "2022.acl-srw.34", "year": 2022, "rel_sent": "This model has attracted attention as an approach that enables efficient offline inferences by connecting both vision and language in the same semantic space ; however , whether an image encoder as part of a dual - encoder model can interpret scene - text ( i.e. , the textual information in images ) is unclear . We propose pre - training methods that encourage a joint understanding of the scene - text and surrounding visual information . The experimental results demonstrate that our methods improve the retrieval performances of the dual - encoder models .", "forward": true, "src_ids": "2022.acl-srw.34_1684"}
+{"input": "kge models is done by using Method| context: knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors . these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) . kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities . for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility .", "entity": "kge models", "output": "encoder - decoder transformer model", "neg_sample": ["kge models is done by using Method", "knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors .", "these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) .", "kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities .", "for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility ."], "relation": "used for", "id": "2022.acl-long.201", "year": 2022, "rel_sent": "We show that an off - the - shelf encoder - decoder Transformer model can serve as a scalable and versatile KGE model obtaining state - of - the - art results for KG link prediction and incomplete KG question answering .", "forward": false, "src_ids": "2022.acl-long.201_1685"}
+{"input": "kg link prediction is done by using Method| context: knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors . these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) . kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities . for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility .", "entity": "kg link prediction", "output": "kge models", "neg_sample": ["kg link prediction is done by using Method", "knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors .", "these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) .", "kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities .", "for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility ."], "relation": "used for", "id": "2022.acl-long.201", "year": 2022, "rel_sent": "We show that an off - the - shelf encoder - decoder Transformer model can serve as a scalable and versatile KGE model obtaining state - of - the - art results for KG link prediction and incomplete KG question answering .", "forward": false, "src_ids": "2022.acl-long.201_1686"}
+{"input": "encoder - decoder transformer model is used for Method| context: knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors . these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) . kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities . for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility .", "entity": "encoder - decoder transformer model", "output": "kge models", "neg_sample": ["encoder - decoder transformer model is used for Method", "knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors .", "these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) .", "kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities .", "for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility ."], "relation": "used for", "id": "2022.acl-long.201", "year": 2022, "rel_sent": "We show that an off - the - shelf encoder - decoder Transformer model can serve as a scalable and versatile KGE model obtaining state - of - the - art results for KG link prediction and incomplete KG question answering .", "forward": true, "src_ids": "2022.acl-long.201_1687"}
+{"input": "kge models is used for Task| context: knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors . kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities . for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility .", "entity": "kge models", "output": "kg link prediction", "neg_sample": ["kge models is used for Task", "knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors .", "kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities .", "for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility ."], "relation": "used for", "id": "2022.acl-long.201", "year": 2022, "rel_sent": "We show that an off - the - shelf encoder - decoder Transformer model can serve as a scalable and versatile KGE model obtaining state - of - the - art results for KG link prediction and incomplete KG question answering .", "forward": true, "src_ids": "2022.acl-long.201_1688"}
+{"input": "kge methods is done by using Method| context: knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors . these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) . kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities . for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility .", "entity": "kge methods", "output": "triple scoring approach", "neg_sample": ["kge methods is done by using Method", "knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors .", "these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) .", "kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities .", "for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility ."], "relation": "used for", "id": "2022.acl-long.201", "year": 2022, "rel_sent": "We achieve this by posing KG link prediction as a sequence - to - sequence task and exchange the triple scoring approach taken by prior KGE methods with autoregressive decoding .", "forward": false, "src_ids": "2022.acl-long.201_1689"}
+{"input": "triple scoring approach is used for Method| context: knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors . these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) . kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities . for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility .", "entity": "triple scoring approach", "output": "kge methods", "neg_sample": ["triple scoring approach is used for Method", "knowledge graph embedding ( kge ) models represent each entity and relation of a knowledge graph ( kg ) with low - dimensional embedding vectors .", "these methods have recently been applied to kg link prediction and question answering over incomplete kgs ( kgqa ) .", "kges typically create an embedding for each entity in the graph , which results in large model sizes on real - world graphs with millions of entities .", "for downstream tasks these atomic entity representations often need to be integrated into a multi stage pipeline , limiting their utility ."], "relation": "used for", "id": "2022.acl-long.201", "year": 2022, "rel_sent": "We achieve this by posing KG link prediction as a sequence - to - sequence task and exchange the triple scoring approach taken by prior KGE methods with autoregressive decoding .", "forward": true, "src_ids": "2022.acl-long.201_1690"}
+{"input": "mbart is used for OtherScientificTerm| context: what can pre - trained multilingual sequence - to - sequence models like mbart contribute to translating low - resource languages ?", "entity": "mbart", "output": "domain differences", "neg_sample": ["mbart is used for OtherScientificTerm", "what can pre - trained multilingual sequence - to - sequence models like mbart contribute to translating low - resource languages ?"], "relation": "used for", "id": "2022.findings-acl.6", "year": 2022, "rel_sent": "While mBART is robust to domain differences , its translations for unseen and typologically distant languages remain below 3.0 BLEU .", "forward": true, "src_ids": "2022.findings-acl.6_1691"}
+{"input": "question answering is done by using Method| context: many e - commerce websites provide product - related question answering ( pqa ) platform where potential customers can ask questions related to a product , and other consumers can post an answer to that question based on their experience . recently , there has been a growing interest in providing automated responses to product questions .", "entity": "question answering", "output": "generative approach", "neg_sample": ["question answering is done by using Method", "many e - commerce websites provide product - related question answering ( pqa ) platform where potential customers can ask questions related to a product , and other consumers can post an answer to that question based on their experience .", "recently , there has been a growing interest in providing automated responses to product questions ."], "relation": "used for", "id": "2022.ecnlp-1.24", "year": 2022, "rel_sent": "Investigating the Generative Approach for Question Answering in E - Commerce.", "forward": false, "src_ids": "2022.ecnlp-1.24_1692"}
+{"input": "product - related question answering is done by using Method| context: many e - commerce websites provide product - related question answering ( pqa ) platform where potential customers can ask questions related to a product , and other consumers can post an answer to that question based on their experience . recently , there has been a growing interest in providing automated responses to product questions .", "entity": "product - related question answering", "output": "generative approach", "neg_sample": ["product - related question answering is done by using Method", "many e - commerce websites provide product - related question answering ( pqa ) platform where potential customers can ask questions related to a product , and other consumers can post an answer to that question based on their experience .", "recently , there has been a growing interest in providing automated responses to product questions ."], "relation": "used for", "id": "2022.ecnlp-1.24", "year": 2022, "rel_sent": "In this paper , we investigate the suitability of the generative approach for PQA .", "forward": false, "src_ids": "2022.ecnlp-1.24_1693"}
+{"input": "generative approach is used for Task| context: recently , there has been a growing interest in providing automated responses to product questions .", "entity": "generative approach", "output": "question answering", "neg_sample": ["generative approach is used for Task", "recently , there has been a growing interest in providing automated responses to product questions ."], "relation": "used for", "id": "2022.ecnlp-1.24", "year": 2022, "rel_sent": "Investigating the Generative Approach for Question Answering in E - Commerce.", "forward": true, "src_ids": "2022.ecnlp-1.24_1694"}
+{"input": "generative approach is used for Task| context: recently , there has been a growing interest in providing automated responses to product questions .", "entity": "generative approach", "output": "product - related question answering", "neg_sample": ["generative approach is used for Task", "recently , there has been a growing interest in providing automated responses to product questions ."], "relation": "used for", "id": "2022.ecnlp-1.24", "year": 2022, "rel_sent": "In this paper , we investigate the suitability of the generative approach for PQA .", "forward": true, "src_ids": "2022.ecnlp-1.24_1695"}
+{"input": "passage dropout and regularization technique is used for Task| context: multidoc2dial presents an important challenge on modeling dialogues grounded with multiple documents .", "entity": "passage dropout and regularization technique", "output": "response generation", "neg_sample": ["passage dropout and regularization technique is used for Task", "multidoc2dial presents an important challenge on modeling dialogues grounded with multiple documents ."], "relation": "used for", "id": "2022.dialdoc-1.13", "year": 2022, "rel_sent": "We also adopt a passage dropout and regularization technique to improve response generation performance .", "forward": true, "src_ids": "2022.dialdoc-1.13_1696"}
+{"input": "response generation is done by using Method| context: multidoc2dial presents an important challenge on modeling dialogues grounded with multiple documents .", "entity": "response generation", "output": "passage dropout and regularization technique", "neg_sample": ["response generation is done by using Method", "multidoc2dial presents an important challenge on modeling dialogues grounded with multiple documents ."], "relation": "used for", "id": "2022.dialdoc-1.13", "year": 2022, "rel_sent": "We also adopt a passage dropout and regularization technique to improve response generation performance .", "forward": false, "src_ids": "2022.dialdoc-1.13_1697"}
+{"input": "education applications is done by using Task| context: a clear picture of existing work can invigorate and inform how researchers conceptualise worthwhile projects .", "entity": "education applications", "output": "natural language processing", "neg_sample": ["education applications is done by using Task", "a clear picture of existing work can invigorate and inform how researchers conceptualise worthwhile projects ."], "relation": "used for", "id": "2022.acl-srw.15", "year": 2022, "rel_sent": "Scoping natural language processing in Indonesian and Malay for education applications.", "forward": false, "src_ids": "2022.acl-srw.15_1698"}
+{"input": "natural language processing is used for Task| context: indonesian and malay are underrepresented in the development of natural language processing ( nlp ) technologies and available resources are difficult tofind . a clear picture of existing work can invigorate and inform how researchers conceptualise worthwhile projects .", "entity": "natural language processing", "output": "education applications", "neg_sample": ["natural language processing is used for Task", "indonesian and malay are underrepresented in the development of natural language processing ( nlp ) technologies and available resources are difficult tofind .", "a clear picture of existing work can invigorate and inform how researchers conceptualise worthwhile projects ."], "relation": "used for", "id": "2022.acl-srw.15", "year": 2022, "rel_sent": "Scoping natural language processing in Indonesian and Malay for education applications.", "forward": true, "src_ids": "2022.acl-srw.15_1699"}
+{"input": "phoneme representations is done by using OtherScientificTerm| context: while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data .", "entity": "phoneme representations", "output": "embeddings", "neg_sample": ["phoneme representations is done by using OtherScientificTerm", "while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data ."], "relation": "used for", "id": "2022.acl-long.472", "year": 2022, "rel_sent": "In this work , we use embeddings derived from articulatory vectors rather than embeddings derived from phoneme identities to learn phoneme representations that hold across languages .", "forward": false, "src_ids": "2022.acl-long.472_1700"}
+{"input": "embeddings is used for Method| context: while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data .", "entity": "embeddings", "output": "phoneme representations", "neg_sample": ["embeddings is used for Method", "while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data ."], "relation": "used for", "id": "2022.acl-long.472", "year": 2022, "rel_sent": "In this work , we use embeddings derived from articulatory vectors rather than embeddings derived from phoneme identities to learn phoneme representations that hold across languages .", "forward": true, "src_ids": "2022.acl-long.472_1701"}
+{"input": "extractive summarization tasks is done by using OtherScientificTerm| context: it has been the norm for a long time to evaluate automated summarization tasks using the popular rouge metric . although several studies in the past have highlighted the limitations of rouge , researchers have struggled to reach a consensus on a better alternative until today . one major limitation of the traditional rouge metric is the lack of semantic understanding ( relies on direct overlap of n - grams ) .", "entity": "extractive summarization tasks", "output": "semantic - aware ground truths", "neg_sample": ["extractive summarization tasks is done by using OtherScientificTerm", "it has been the norm for a long time to evaluate automated summarization tasks using the popular rouge metric .", "although several studies in the past have highlighted the limitations of rouge , researchers have struggled to reach a consensus on a better alternative until today .", "one major limitation of the traditional rouge metric is the lack of semantic understanding ( relies on direct overlap of n - grams ) ."], "relation": "used for", "id": "2022.findings-acl.122", "year": 2022, "rel_sent": "One fundamental contribution of the paper is that it demonstrates how we can generate more reliable semantic - aware ground truths for evaluating extractive summarization tasks without any additional human intervention .", "forward": false, "src_ids": "2022.findings-acl.122_1702"}
+{"input": "semantic - aware ground truths is used for Task| context: it has been the norm for a long time to evaluate automated summarization tasks using the popular rouge metric . although several studies in the past have highlighted the limitations of rouge , researchers have struggled to reach a consensus on a better alternative until today . one major limitation of the traditional rouge metric is the lack of semantic understanding ( relies on direct overlap of n - grams ) .", "entity": "semantic - aware ground truths", "output": "extractive summarization tasks", "neg_sample": ["semantic - aware ground truths is used for Task", "it has been the norm for a long time to evaluate automated summarization tasks using the popular rouge metric .", "although several studies in the past have highlighted the limitations of rouge , researchers have struggled to reach a consensus on a better alternative until today .", "one major limitation of the traditional rouge metric is the lack of semantic understanding ( relies on direct overlap of n - grams ) ."], "relation": "used for", "id": "2022.findings-acl.122", "year": 2022, "rel_sent": "One fundamental contribution of the paper is that it demonstrates how we can generate more reliable semantic - aware ground truths for evaluating extractive summarization tasks without any additional human intervention .", "forward": true, "src_ids": "2022.findings-acl.122_1703"}
+{"input": "cascaded st is done by using Method| context: speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech . the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al . , 1998 ; waibel et al . , 1991 ) . the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al . , 2002 ) .", "entity": "cascaded st", "output": "neural mt", "neg_sample": ["cascaded st is done by using Method", "speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech .", "the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al .", ", 1998 ; waibel et al .", ", 1991 ) .", "the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al .", ", 2002 ) ."], "relation": "used for", "id": "2022.eamt-1.2", "year": 2022, "rel_sent": "We started with a study about the effects of NMT in cascaded ST , where we analyzed the translation errors of NMT and phrase - based MT ( PBMT ) for automatically transcribed input text .", "forward": false, "src_ids": "2022.eamt-1.2_1704"}
+{"input": "neural mt is used for Method| context: speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech . the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al . , 1998 ; waibel et al . , 1991 ) . the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al . , 2002 ) . direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al . , 2014 ; bahdanau et al . , 2015 ) .", "entity": "neural mt", "output": "cascaded st", "neg_sample": ["neural mt is used for Method", "speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech .", "the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al .", ", 1998 ; waibel et al .", ", 1991 ) .", "the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al .", ", 2002 ) .", "direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al .", ", 2014 ; bahdanau et al .", ", 2015 ) ."], "relation": "used for", "id": "2022.eamt-1.2", "year": 2022, "rel_sent": "We started with a study about the effects of NMT in cascaded ST , where we analyzed the translation errors of NMT and phrase - based MT ( PBMT ) for automatically transcribed input text .", "forward": true, "src_ids": "2022.eamt-1.2_1705"}
+{"input": "automatic transcripts is done by using Method| context: speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech . the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al . , 1998 ; waibel et al . , 1991 ) . the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al . , 2002 ) . direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al . , 2014 ; bahdanau et al . , 2015 ) .", "entity": "automatic transcripts", "output": "nmt system", "neg_sample": ["automatic transcripts is done by using Method", "speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech .", "the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al .", ", 1998 ; waibel et al .", ", 1991 ) .", "the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al .", ", 2002 ) .", "direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al .", ", 2014 ; bahdanau et al .", ", 2015 ) ."], "relation": "used for", "id": "2022.eamt-1.2", "year": 2022, "rel_sent": "Motivated by application needs , in a following work we studied how to use a single NMT system to translate effectively clean source text and automatic transcripts .", "forward": false, "src_ids": "2022.eamt-1.2_1706"}
+{"input": "clean source text is done by using Method| context: speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech . the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al . , 1998 ; waibel et al . , 1991 ) . the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al . , 2002 ) . direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al . , 2014 ; bahdanau et al . , 2015 ) .", "entity": "clean source text", "output": "nmt system", "neg_sample": ["clean source text is done by using Method", "speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech .", "the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al .", ", 1998 ; waibel et al .", ", 1991 ) .", "the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al .", ", 2002 ) .", "direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al .", ", 2014 ; bahdanau et al .", ", 2015 ) ."], "relation": "used for", "id": "2022.eamt-1.2", "year": 2022, "rel_sent": "Motivated by application needs , in a following work we studied how to use a single NMT system to translate effectively clean source text and automatic transcripts .", "forward": false, "src_ids": "2022.eamt-1.2_1707"}
+{"input": "nmt system is used for Material| context: speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech . the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al . , 1998 ; waibel et al . , 1991 ) . the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al . , 2002 ) . direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al . , 2014 ; bahdanau et al . , 2015 ) .", "entity": "nmt system", "output": "clean source text", "neg_sample": ["nmt system is used for Material", "speech - to - text translation , or simply speech translation ( st ) , is the task of translating automatically a spoken speech .", "the problem has classically been tackled by combining the technologies of automatic speech recognition ( asr ) and machine translation ( mt ) with different degrees of coupling ( takezawa et al .", ", 1998 ; waibel et al .", ", 1991 ) .", "the most popular approach is to cascade asr and mt systems , as it can make use of the state of the art in such mature fields ( black et al .", ", 2002 ) .", "direct speech translation ( dst ) is based on the sequence - to - sequence learning technology that allowed the spectacular advances of the field of neural mt ( nmt ) but introducing its own challenges ( sutskever et al .", ", 2014 ; bahdanau et al .", ", 2015 ) ."], "relation": "used for", "id": "2022.eamt-1.2", "year": 2022, "rel_sent": "Motivated by application needs , in a following work we studied how to use a single NMT system to translate effectively clean source text and automatic transcripts .", "forward": true, "src_ids": "2022.eamt-1.2_1708"}
+{"input": "knowledge graph embedding is done by using Method| context: while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently .", "entity": "knowledge graph embedding", "output": "hyper - parameter search", "neg_sample": ["knowledge graph embedding is done by using Method", "while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently ."], "relation": "used for", "id": "2022.acl-long.194", "year": 2022, "rel_sent": "Efficient Hyper - parameter Search for Knowledge Graph Embedding.", "forward": false, "src_ids": "2022.acl-long.194_1709"}
+{"input": "hyper - parameter search is used for Task| context: while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently .", "entity": "hyper - parameter search", "output": "knowledge graph embedding", "neg_sample": ["hyper - parameter search is used for Task", "while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently ."], "relation": "used for", "id": "2022.acl-long.194", "year": 2022, "rel_sent": "Efficient Hyper - parameter Search for Knowledge Graph Embedding.", "forward": true, "src_ids": "2022.acl-long.194_1710"}
+{"input": "hp configurations is done by using Method| context: while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently .", "entity": "hp configurations", "output": "two - stage search algorithm kgtuner", "neg_sample": ["hp configurations is done by using Method", "while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently ."], "relation": "used for", "id": "2022.acl-long.194", "year": 2022, "rel_sent": "Based on the analysis , we propose an efficient two - stage search algorithm KGTuner , which efficiently explores HP configurations on small subgraph at the first stage and transfers the top - performed configurations for fine - tuning on the large full graph at the second stage .", "forward": false, "src_ids": "2022.acl-long.194_1711"}
+{"input": "two - stage search algorithm kgtuner is used for OtherScientificTerm| context: while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently .", "entity": "two - stage search algorithm kgtuner", "output": "hp configurations", "neg_sample": ["two - stage search algorithm kgtuner is used for OtherScientificTerm", "while hyper - parameters ( hps ) are important for knowledge graph ( kg ) learning , existing methods fail to search them efficiently ."], "relation": "used for", "id": "2022.acl-long.194", "year": 2022, "rel_sent": "Based on the analysis , we propose an efficient two - stage search algorithm KGTuner , which efficiently explores HP configurations on small subgraph at the first stage and transfers the top - performed configurations for fine - tuning on the large full graph at the second stage .", "forward": true, "src_ids": "2022.acl-long.194_1712"}
+{"input": "automatic release note generation is done by using Material| context: a release note is a technical document that describes the latest changes to a software product and is crucial in open source software development . however , it still remains challenging to generate release notes automatically .", "entity": "automatic release note generation", "output": "rnsum", "neg_sample": ["automatic release note generation is done by using Material", "a release note is a technical document that describes the latest changes to a software product and is crucial in open source software development .", "however , it still remains challenging to generate release notes automatically ."], "relation": "used for", "id": "2022.acl-long.597", "year": 2022, "rel_sent": "RNSum : A Large - Scale Dataset for Automatic Release Note Generation via Commit Logs Summarization.", "forward": false, "src_ids": "2022.acl-long.597_1713"}
+{"input": "large - scale dataset is used for Task| context: a release note is a technical document that describes the latest changes to a software product and is crucial in open source software development . however , it still remains challenging to generate release notes automatically .", "entity": "large - scale dataset", "output": "automatic release note generation", "neg_sample": ["large - scale dataset is used for Task", "a release note is a technical document that describes the latest changes to a software product and is crucial in open source software development .", "however , it still remains challenging to generate release notes automatically ."], "relation": "used for", "id": "2022.acl-long.597", "year": 2022, "rel_sent": "RNSum : A Large - Scale Dataset for Automatic Release Note Generation via Commit Logs Summarization.", "forward": true, "src_ids": "2022.acl-long.597_1714"}
+{"input": "rnsum is used for Task| context: a release note is a technical document that describes the latest changes to a software product and is crucial in open source software development . however , it still remains challenging to generate release notes automatically .", "entity": "rnsum", "output": "automatic release note generation", "neg_sample": ["rnsum is used for Task", "a release note is a technical document that describes the latest changes to a software product and is crucial in open source software development .", "however , it still remains challenging to generate release notes automatically ."], "relation": "used for", "id": "2022.acl-long.597", "year": 2022, "rel_sent": "RNSum : A Large - Scale Dataset for Automatic Release Note Generation via Commit Logs Summarization.", "forward": true, "src_ids": "2022.acl-long.597_1715"}
+{"input": "human errors is done by using Method| context: one of the main challenges for cged is the lack of annotated data . to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods . however , the introduced noises are usually context - independent , which are quite different from those made by humans . in addition , generated sentences may be error - free and thus become noisy data .", "entity": "human errors", "output": "generative models", "neg_sample": ["human errors is done by using Method", "one of the main challenges for cged is the lack of annotated data .", "to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods .", "however , the introduced noises are usually context - independent , which are quite different from those made by humans .", "in addition , generated sentences may be error - free and thus become noisy data ."], "relation": "used for", "id": "2022.findings-acl.233", "year": 2022, "rel_sent": "The model - based methods utilize generative models to imitate human errors .", "forward": false, "src_ids": "2022.findings-acl.233_1716"}
+{"input": "generative models is used for OtherScientificTerm| context: one of the main challenges for cged is the lack of annotated data . to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods . however , the introduced noises are usually context - independent , which are quite different from those made by humans . in addition , generated sentences may be error - free and thus become noisy data .", "entity": "generative models", "output": "human errors", "neg_sample": ["generative models is used for OtherScientificTerm", "one of the main challenges for cged is the lack of annotated data .", "to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods .", "however , the introduced noises are usually context - independent , which are quite different from those made by humans .", "in addition , generated sentences may be error - free and thus become noisy data ."], "relation": "used for", "id": "2022.findings-acl.233", "year": 2022, "rel_sent": "The model - based methods utilize generative models to imitate human errors .", "forward": true, "src_ids": "2022.findings-acl.233_1717"}
+{"input": "semantically ambiguous sentences is done by using Method| context: one of the main challenges for cged is the lack of annotated data . to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods . however , the introduced noises are usually context - independent , which are quite different from those made by humans . in addition , generated sentences may be error - free and thus become noisy data .", "entity": "semantically ambiguous sentences", "output": "generative model", "neg_sample": ["semantically ambiguous sentences is done by using Method", "one of the main challenges for cged is the lack of annotated data .", "to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods .", "however , the introduced noises are usually context - independent , which are quite different from those made by humans .", "in addition , generated sentences may be error - free and thus become noisy data ."], "relation": "used for", "id": "2022.findings-acl.233", "year": 2022, "rel_sent": "The generative model may bring too many changes to the original sentences and generate semantically ambiguous sentences , so it is difficult to detect grammatical errors in these generated sentences .", "forward": false, "src_ids": "2022.findings-acl.233_1718"}
+{"input": "generative model is used for Material| context: one of the main challenges for cged is the lack of annotated data . to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods . however , the introduced noises are usually context - independent , which are quite different from those made by humans . in addition , generated sentences may be error - free and thus become noisy data .", "entity": "generative model", "output": "semantically ambiguous sentences", "neg_sample": ["generative model is used for Material", "one of the main challenges for cged is the lack of annotated data .", "to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods .", "however , the introduced noises are usually context - independent , which are quite different from those made by humans .", "in addition , generated sentences may be error - free and thus become noisy data ."], "relation": "used for", "id": "2022.findings-acl.233", "year": 2022, "rel_sent": "The generative model may bring too many changes to the original sentences and generate semantically ambiguous sentences , so it is difficult to detect grammatical errors in these generated sentences .", "forward": true, "src_ids": "2022.findings-acl.233_1719"}
+{"input": "generating chinese grammatical errors is done by using Method| context: one of the main challenges for cged is the lack of annotated data . to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods . however , the introduced noises are usually context - independent , which are quite different from those made by humans . in addition , generated sentences may be error - free and thus become noisy data .", "entity": "generating chinese grammatical errors", "output": "conditional non - autoregressive error generation model", "neg_sample": ["generating chinese grammatical errors is done by using Method", "one of the main challenges for cged is the lack of annotated data .", "to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods .", "however , the introduced noises are usually context - independent , which are quite different from those made by humans .", "in addition , generated sentences may be error - free and thus become noisy data ."], "relation": "used for", "id": "2022.findings-acl.233", "year": 2022, "rel_sent": "To handle these problems , we propose CNEG , a novel Conditional Non - Autoregressive Error Generation model for generating Chinese grammatical errors .", "forward": false, "src_ids": "2022.findings-acl.233_1720"}
+{"input": "conditional non - autoregressive error generation model is used for Task| context: one of the main challenges for cged is the lack of annotated data . to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods . however , the introduced noises are usually context - independent , which are quite different from those made by humans . in addition , generated sentences may be error - free and thus become noisy data .", "entity": "conditional non - autoregressive error generation model", "output": "generating chinese grammatical errors", "neg_sample": ["conditional non - autoregressive error generation model is used for Task", "one of the main challenges for cged is the lack of annotated data .", "to alleviate this problem , previous studies proposed various methods to automatically generate more training samples , which can be roughly categorized into rule - based methods and model - based methods .", "however , the introduced noises are usually context - independent , which are quite different from those made by humans .", "in addition , generated sentences may be error - free and thus become noisy data ."], "relation": "used for", "id": "2022.findings-acl.233", "year": 2022, "rel_sent": "To handle these problems , we propose CNEG , a novel Conditional Non - Autoregressive Error Generation model for generating Chinese grammatical errors .", "forward": true, "src_ids": "2022.findings-acl.233_1721"}
+{"input": "real - world pedagogical scenarios is done by using Method| context: handing in a paper or exercise and merely receiving ' bad ' or ' incorrect ' as feedback is not very helpful when the goal is to improve . unfortunately , this is currently the kind of feedback given by automatic short answer grading ( asag ) systems . one of the reasons for this is a lack of content - focused elaborated feedback datasets .", "entity": "real - world pedagogical scenarios", "output": "feedback models", "neg_sample": ["real - world pedagogical scenarios is done by using Method", "handing in a paper or exercise and merely receiving ' bad ' or ' incorrect ' as feedback is not very helpful when the goal is to improve .", "unfortunately , this is currently the kind of feedback given by automatic short answer grading ( asag ) systems .", "one of the reasons for this is a lack of content - focused elaborated feedback datasets ."], "relation": "used for", "id": "2022.acl-long.587", "year": 2022, "rel_sent": "This paper discusses the need for enhanced feedback models in real - world pedagogical scenarios , describes the dataset annotation process , gives a comprehensive analysis of SAF , and provides T5 - based baselines for future comparison .", "forward": false, "src_ids": "2022.acl-long.587_1722"}
+{"input": "feedback models is used for Material| context: handing in a paper or exercise and merely receiving ' bad ' or ' incorrect ' as feedback is not very helpful when the goal is to improve . unfortunately , this is currently the kind of feedback given by automatic short answer grading ( asag ) systems . one of the reasons for this is a lack of content - focused elaborated feedback datasets .", "entity": "feedback models", "output": "real - world pedagogical scenarios", "neg_sample": ["feedback models is used for Material", "handing in a paper or exercise and merely receiving ' bad ' or ' incorrect ' as feedback is not very helpful when the goal is to improve .", "unfortunately , this is currently the kind of feedback given by automatic short answer grading ( asag ) systems .", "one of the reasons for this is a lack of content - focused elaborated feedback datasets ."], "relation": "used for", "id": "2022.acl-long.587", "year": 2022, "rel_sent": "This paper discusses the need for enhanced feedback models in real - world pedagogical scenarios , describes the dataset annotation process , gives a comprehensive analysis of SAF , and provides T5 - based baselines for future comparison .", "forward": true, "src_ids": "2022.acl-long.587_1723"}
+{"input": "passage retrieval is done by using Method| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "passage retrieval", "output": "hyperlink - induced pre - training", "neg_sample": ["passage retrieval is done by using Method", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "Hyperlink - induced Pre - training for Passage Retrieval in Open - domain Question Answering.", "forward": false, "src_ids": "2022.acl-long.493_1724"}
+{"input": "open - domain question answering is done by using Task| context: however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "open - domain question answering", "output": "passage retrieval", "neg_sample": ["open - domain question answering is done by using Task", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "Hyperlink - induced Pre - training for Passage Retrieval in Open - domain Question Answering.", "forward": false, "src_ids": "2022.acl-long.493_1725"}
+{"input": "hyperlink - induced pre - training is used for Task| context: however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "hyperlink - induced pre - training", "output": "passage retrieval", "neg_sample": ["hyperlink - induced pre - training is used for Task", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "Hyperlink - induced Pre - training for Passage Retrieval in Open - domain Question Answering.", "forward": true, "src_ids": "2022.acl-long.493_1726"}
+{"input": "passage retrieval is used for Task| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "passage retrieval", "output": "open - domain question answering", "neg_sample": ["passage retrieval is used for Task", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "Hyperlink - induced Pre - training for Passage Retrieval in Open - domain Question Answering.", "forward": true, "src_ids": "2022.acl-long.493_1727"}
+{"input": "large - scale pre - training is done by using OtherScientificTerm| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "large - scale pre - training", "output": "hyperlink - based structures", "neg_sample": ["large - scale pre - training is done by using OtherScientificTerm", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": false, "src_ids": "2022.acl-long.493_1728"}
+{"input": "relevance signals is done by using OtherScientificTerm| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "relevance signals", "output": "hyperlink - based structures", "neg_sample": ["relevance signals is done by using OtherScientificTerm", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": false, "src_ids": "2022.acl-long.493_1729"}
+{"input": "hyperlink - based structures is used for OtherScientificTerm| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "hyperlink - based structures", "output": "relevance signals", "neg_sample": ["hyperlink - based structures is used for OtherScientificTerm", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": true, "src_ids": "2022.acl-long.493_1730"}
+{"input": "downstream passage retrieval is done by using Method| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "downstream passage retrieval", "output": "large - scale pre - training", "neg_sample": ["downstream passage retrieval is done by using Method", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": false, "src_ids": "2022.acl-long.493_1731"}
+{"input": "hyperlink - based structures is used for Method| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "hyperlink - based structures", "output": "large - scale pre - training", "neg_sample": ["hyperlink - based structures is used for Method", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": true, "src_ids": "2022.acl-long.493_1732"}
+{"input": "relevance signals is used for Method| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "relevance signals", "output": "large - scale pre - training", "neg_sample": ["relevance signals is used for Method", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": true, "src_ids": "2022.acl-long.493_1733"}
+{"input": "large - scale pre - training is used for Task| context: to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) . however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement .", "entity": "large - scale pre - training", "output": "downstream passage retrieval", "neg_sample": ["large - scale pre - training is used for Task", "to alleviate the data scarcity problem in training question answering systems , recent works propose additional intermediate pre - training for dense passage retrieval ( dpr ) .", "however , there still remains a large discrepancy between the provided upstream signals and the downstream question - passage relevance , which leads to less improvement ."], "relation": "used for", "id": "2022.acl-long.493", "year": 2022, "rel_sent": "We demonstrate that the hyperlink - based structures of dual - link and co - mention can provide effective relevance signals for large - scale pre - training that better facilitate downstream passage retrieval .", "forward": true, "src_ids": "2022.acl-long.493_1734"}
+{"input": "input - specific attention subnetworks ( ias ) is done by using Method| context: self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning . in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection .", "entity": "input - specific attention subnetworks ( ias )", "output": "input - specific attention subnetworks ( ias )", "neg_sample": ["input - specific attention subnetworks ( ias ) is done by using Method", "self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning .", "in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection ."], "relation": "used for", "id": "2022.findings-acl.4", "year": 2022, "rel_sent": "Specifically , we propose a method to construct input - specific attention subnetworks ( IAS ) from which we extract three features to discriminate between authentic and adversarial inputs .", "forward": false, "src_ids": "2022.findings-acl.4_1735"}
+{"input": "input - specific attention subnetworks ( ias ) is used for Method| context: self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning . in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection .", "entity": "input - specific attention subnetworks ( ias )", "output": "input - specific attention subnetworks ( ias )", "neg_sample": ["input - specific attention subnetworks ( ias ) is used for Method", "self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning .", "in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection ."], "relation": "used for", "id": "2022.findings-acl.4", "year": 2022, "rel_sent": "Specifically , we propose a method to construct input - specific attention subnetworks ( IAS ) from which we extract three features to discriminate between authentic and adversarial inputs .", "forward": true, "src_ids": "2022.findings-acl.4_1736"}
+{"input": "authentic and adversarial inputs is done by using OtherScientificTerm| context: self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning . in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection .", "entity": "authentic and adversarial inputs", "output": "features", "neg_sample": ["authentic and adversarial inputs is done by using OtherScientificTerm", "self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning .", "in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection ."], "relation": "used for", "id": "2022.findings-acl.4", "year": 2022, "rel_sent": "Specifically , we propose a method to construct input - specific attention subnetworks ( IAS ) from which we extract three features to discriminate between authentic and adversarial inputs .", "forward": false, "src_ids": "2022.findings-acl.4_1737"}
+{"input": "features is used for OtherScientificTerm| context: self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning . in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection .", "entity": "features", "output": "authentic and adversarial inputs", "neg_sample": ["features is used for OtherScientificTerm", "self - attention heads are characteristic of transformer models and have been well studied for interpretability and pruning .", "in this work , we demonstrate an altogether different utility of attention heads , namely for adversarial detection ."], "relation": "used for", "id": "2022.findings-acl.4", "year": 2022, "rel_sent": "Specifically , we propose a method to construct input - specific attention subnetworks ( IAS ) from which we extract three features to discriminate between authentic and adversarial inputs .", "forward": true, "src_ids": "2022.findings-acl.4_1738"}
+{"input": "prompting is done by using Method| context: prompting methods recently achieve impressive success in few - shot learning . these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels . however , such a paradigm is very inefficient for the task of slot tagging . since slot tagging samples are multiple consecutive words in a sentence , the prompting methods have to enumerate all n - grams token spans tofind all the possible slots , which greatly slows down the prediction .", "entity": "prompting", "output": "inverse paradigm", "neg_sample": ["prompting is done by using Method", "prompting methods recently achieve impressive success in few - shot learning .", "these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels .", "however , such a paradigm is very inefficient for the task of slot tagging .", "since slot tagging samples are multiple consecutive words in a sentence , the prompting methods have to enumerate all n - grams token spans tofind all the possible slots , which greatly slows down the prediction ."], "relation": "used for", "id": "2022.findings-acl.53", "year": 2022, "rel_sent": "To tackle this , we introduce an inverse paradigm for prompting .", "forward": false, "src_ids": "2022.findings-acl.53_1739"}
+{"input": "inverse paradigm is used for Method| context: these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels . however , such a paradigm is very inefficient for the task of slot tagging .", "entity": "inverse paradigm", "output": "prompting", "neg_sample": ["inverse paradigm is used for Method", "these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels .", "however , such a paradigm is very inefficient for the task of slot tagging ."], "relation": "used for", "id": "2022.findings-acl.53", "year": 2022, "rel_sent": "To tackle this , we introduce an inverse paradigm for prompting .", "forward": true, "src_ids": "2022.findings-acl.53_1740"}
+{"input": "prediction is done by using Method| context: prompting methods recently achieve impressive success in few - shot learning . these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels . however , such a paradigm is very inefficient for the task of slot tagging . since slot tagging samples are multiple consecutive words in a sentence , the prompting methods have to enumerate all n - grams token spans tofind all the possible slots , which greatly slows down the prediction .", "entity": "prediction", "output": "inverse prompting", "neg_sample": ["prediction is done by using Method", "prompting methods recently achieve impressive success in few - shot learning .", "these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels .", "however , such a paradigm is very inefficient for the task of slot tagging .", "since slot tagging samples are multiple consecutive words in a sentence , the prompting methods have to enumerate all n - grams token spans tofind all the possible slots , which greatly slows down the prediction ."], "relation": "used for", "id": "2022.findings-acl.53", "year": 2022, "rel_sent": "Such inverse prompting only requires a one - turn prediction for each slot type and greatly speeds up the prediction .", "forward": false, "src_ids": "2022.findings-acl.53_1741"}
+{"input": "inverse prompting is used for Task| context: prompting methods recently achieve impressive success in few - shot learning . these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels . however , such a paradigm is very inefficient for the task of slot tagging .", "entity": "inverse prompting", "output": "prediction", "neg_sample": ["inverse prompting is used for Task", "prompting methods recently achieve impressive success in few - shot learning .", "these methods modify input samples with prompt sentence pieces , and decode label tokens to map samples to corresponding labels .", "however , such a paradigm is very inefficient for the task of slot tagging ."], "relation": "used for", "id": "2022.findings-acl.53", "year": 2022, "rel_sent": "Such inverse prompting only requires a one - turn prediction for each slot type and greatly speeds up the prediction .", "forward": true, "src_ids": "2022.findings-acl.53_1742"}
+{"input": "pretrained models is done by using Method| context: depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities .", "entity": "pretrained models", "output": "transformers", "neg_sample": ["pretrained models is done by using Method", "depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities ."], "relation": "used for", "id": "2022.ltedi-1.26", "year": 2022, "rel_sent": "Transformers provides thousands of pretrained models to perform tasks on different modalities such as text , vision , and audio .", "forward": false, "src_ids": "2022.ltedi-1.26_1743"}
+{"input": "modalities is done by using Method| context: depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities .", "entity": "modalities", "output": "transformers", "neg_sample": ["modalities is done by using Method", "depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities ."], "relation": "used for", "id": "2022.ltedi-1.26", "year": 2022, "rel_sent": "Transformers provides thousands of pretrained models to perform tasks on different modalities such as text , vision , and audio .", "forward": false, "src_ids": "2022.ltedi-1.26_1744"}
+{"input": "transformers is used for Method| context: depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities .", "entity": "transformers", "output": "pretrained models", "neg_sample": ["transformers is used for Method", "depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities ."], "relation": "used for", "id": "2022.ltedi-1.26", "year": 2022, "rel_sent": "Transformers provides thousands of pretrained models to perform tasks on different modalities such as text , vision , and audio .", "forward": true, "src_ids": "2022.ltedi-1.26_1745"}
+{"input": "pretrained models is used for OtherScientificTerm| context: depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities .", "entity": "pretrained models", "output": "modalities", "neg_sample": ["pretrained models is used for OtherScientificTerm", "depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities ."], "relation": "used for", "id": "2022.ltedi-1.26", "year": 2022, "rel_sent": "Transformers provides thousands of pretrained models to perform tasks on different modalities such as text , vision , and audio .", "forward": true, "src_ids": "2022.ltedi-1.26_1746"}
+{"input": "transformers is used for OtherScientificTerm| context: depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities .", "entity": "transformers", "output": "modalities", "neg_sample": ["transformers is used for OtherScientificTerm", "depression is a common mental illness that involves sadness and lack of interest in all day - to - day activities ."], "relation": "used for", "id": "2022.ltedi-1.26", "year": 2022, "rel_sent": "Transformers provides thousands of pretrained models to perform tasks on different modalities such as text , vision , and audio .", "forward": true, "src_ids": "2022.ltedi-1.26_1747"}
+{"input": "automated approaches is used for OtherScientificTerm| context: we found that over 1,400 ( 14 % ) of the 10,003 training utterances may have been incorrectly labelled .", "entity": "automated approaches", "output": "label errors", "neg_sample": ["automated approaches is used for OtherScientificTerm", "we found that over 1,400 ( 14 % ) of the 10,003 training utterances may have been incorrectly labelled ."], "relation": "used for", "id": "2022.insights-1.19", "year": 2022, "rel_sent": "Motivated by our own negative results when constructing an intent classifier , we applied two automated approaches to identify potential label errors in the dataset .", "forward": true, "src_ids": "2022.insights-1.19_1748"}
+{"input": "label errors is done by using Method| context: we investigate potential label errors present in the popular banking77 dataset and the associated negative impacts on intent classification methods . we found that over 1,400 ( 14 % ) of the 10,003 training utterances may have been incorrectly labelled .", "entity": "label errors", "output": "automated approaches", "neg_sample": ["label errors is done by using Method", "we investigate potential label errors present in the popular banking77 dataset and the associated negative impacts on intent classification methods .", "we found that over 1,400 ( 14 % ) of the 10,003 training utterances may have been incorrectly labelled ."], "relation": "used for", "id": "2022.insights-1.19", "year": 2022, "rel_sent": "Motivated by our own negative results when constructing an intent classifier , we applied two automated approaches to identify potential label errors in the dataset .", "forward": false, "src_ids": "2022.insights-1.19_1749"}
+{"input": "toxic - neutral sentence pairs is done by using Material| context: we present a novel pipeline for the collection of parallel data for the detoxification task . we collect non - toxic paraphrases for over 10,000 english toxic sentences .", "entity": "toxic - neutral sentence pairs", "output": "corpus of paraphrases", "neg_sample": ["toxic - neutral sentence pairs is done by using Material", "we present a novel pipeline for the collection of parallel data for the detoxification task .", "we collect non - toxic paraphrases for over 10,000 english toxic sentences ."], "relation": "used for", "id": "2022.acl-long.469", "year": 2022, "rel_sent": "We also show that this pipeline can be used to distill a large existing corpus of paraphrases to get toxic - neutral sentence pairs .", "forward": false, "src_ids": "2022.acl-long.469_1750"}
+{"input": "corpus of paraphrases is used for OtherScientificTerm| context: we present a novel pipeline for the collection of parallel data for the detoxification task . we collect non - toxic paraphrases for over 10,000 english toxic sentences .", "entity": "corpus of paraphrases", "output": "toxic - neutral sentence pairs", "neg_sample": ["corpus of paraphrases is used for OtherScientificTerm", "we present a novel pipeline for the collection of parallel data for the detoxification task .", "we collect non - toxic paraphrases for over 10,000 english toxic sentences ."], "relation": "used for", "id": "2022.acl-long.469", "year": 2022, "rel_sent": "We also show that this pipeline can be used to distill a large existing corpus of paraphrases to get toxic - neutral sentence pairs .", "forward": true, "src_ids": "2022.acl-long.469_1751"}
+{"input": "detoxification models is done by using Material| context: we present a novel pipeline for the collection of parallel data for the detoxification task . we collect non - toxic paraphrases for over 10,000 english toxic sentences .", "entity": "detoxification models", "output": "parallel corpora", "neg_sample": ["detoxification models is done by using Material", "we present a novel pipeline for the collection of parallel data for the detoxification task .", "we collect non - toxic paraphrases for over 10,000 english toxic sentences ."], "relation": "used for", "id": "2022.acl-long.469", "year": 2022, "rel_sent": "We release two parallel corpora which can be used for the training of detoxification models .", "forward": false, "src_ids": "2022.acl-long.469_1752"}
+{"input": "parallel corpora is used for Method| context: we present a novel pipeline for the collection of parallel data for the detoxification task . we collect non - toxic paraphrases for over 10,000 english toxic sentences .", "entity": "parallel corpora", "output": "detoxification models", "neg_sample": ["parallel corpora is used for Method", "we present a novel pipeline for the collection of parallel data for the detoxification task .", "we collect non - toxic paraphrases for over 10,000 english toxic sentences ."], "relation": "used for", "id": "2022.acl-long.469", "year": 2022, "rel_sent": "We release two parallel corpora which can be used for the training of detoxification models .", "forward": true, "src_ids": "2022.acl-long.469_1753"}
+{"input": "model training is done by using OtherScientificTerm| context: the secret behind the success of this model is in the preprocessing step where all words are transformed to their universal language representation via the international phonetic alphabet ( ipa ) .", "entity": "model training", "output": "information theoretic", "neg_sample": ["model training is done by using OtherScientificTerm", "the secret behind the success of this model is in the preprocessing step where all words are transformed to their universal language representation via the international phonetic alphabet ( ipa ) ."], "relation": "used for", "id": "2022.cmcl-1.12", "year": 2022, "rel_sent": "Various feature types were extracted covering basic frequencies , n - grams , information theoretic , and psycholinguistically - motivated predictors for model training .", "forward": false, "src_ids": "2022.cmcl-1.12_1754"}
+{"input": "psycholinguistically - motivated predictors is used for Task| context: the secret behind the success of this model is in the preprocessing step where all words are transformed to their universal language representation via the international phonetic alphabet ( ipa ) .", "entity": "psycholinguistically - motivated predictors", "output": "model training", "neg_sample": ["psycholinguistically - motivated predictors is used for Task", "the secret behind the success of this model is in the preprocessing step where all words are transformed to their universal language representation via the international phonetic alphabet ( ipa ) ."], "relation": "used for", "id": "2022.cmcl-1.12", "year": 2022, "rel_sent": "Various feature types were extracted covering basic frequencies , n - grams , information theoretic , and psycholinguistically - motivated predictors for model training .", "forward": true, "src_ids": "2022.cmcl-1.12_1755"}
+{"input": "information theoretic is used for Task| context: the secret behind the success of this model is in the preprocessing step where all words are transformed to their universal language representation via the international phonetic alphabet ( ipa ) .", "entity": "information theoretic", "output": "model training", "neg_sample": ["information theoretic is used for Task", "the secret behind the success of this model is in the preprocessing step where all words are transformed to their universal language representation via the international phonetic alphabet ( ipa ) ."], "relation": "used for", "id": "2022.cmcl-1.12", "year": 2022, "rel_sent": "Various feature types were extracted covering basic frequencies , n - grams , information theoretic , and psycholinguistically - motivated predictors for model training .", "forward": true, "src_ids": "2022.cmcl-1.12_1756"}
+{"input": "surface - level noise is used for OtherScientificTerm| context: cross - lingual transfer between a high - resource language and its dialects or closely related language varieties should be facilitated by their similarity . however , current approaches that operate in the embedding space do not take surface similarity into account . this work presents a simple yet effective strategy to improve cross - lingual transfer between closely related varieties .", "entity": "surface - level noise", "output": "transfer between language varieties", "neg_sample": ["surface - level noise is used for OtherScientificTerm", "cross - lingual transfer between a high - resource language and its dialects or closely related language varieties should be facilitated by their similarity .", "however , current approaches that operate in the embedding space do not take surface similarity into account .", "this work presents a simple yet effective strategy to improve cross - lingual transfer between closely related varieties ."], "relation": "used for", "id": "2022.findings-acl.321", "year": 2022, "rel_sent": "Our work provides evidence for the usefulness of simple surface - level noise in improving transfer between language varieties .", "forward": true, "src_ids": "2022.findings-acl.321_1757"}
+{"input": "transfer between language varieties is done by using OtherScientificTerm| context: cross - lingual transfer between a high - resource language and its dialects or closely related language varieties should be facilitated by their similarity . however , current approaches that operate in the embedding space do not take surface similarity into account . this work presents a simple yet effective strategy to improve cross - lingual transfer between closely related varieties .", "entity": "transfer between language varieties", "output": "surface - level noise", "neg_sample": ["transfer between language varieties is done by using OtherScientificTerm", "cross - lingual transfer between a high - resource language and its dialects or closely related language varieties should be facilitated by their similarity .", "however , current approaches that operate in the embedding space do not take surface similarity into account .", "this work presents a simple yet effective strategy to improve cross - lingual transfer between closely related varieties ."], "relation": "used for", "id": "2022.findings-acl.321", "year": 2022, "rel_sent": "Our work provides evidence for the usefulness of simple surface - level noise in improving transfer between language varieties .", "forward": false, "src_ids": "2022.findings-acl.321_1758"}
+{"input": "text - based games is done by using Method| context: text - based games provide an interactive way to study natural language processing . while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world .", "entity": "text - based games", "output": "question - guided reinforcement learning", "neg_sample": ["text - based games is done by using Method", "text - based games provide an interactive way to study natural language processing .", "while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world ."], "relation": "used for", "id": "2022.acl-long.41", "year": 2022, "rel_sent": "Perceiving the World : Question - guided Reinforcement Learning for Text - based Games.", "forward": false, "src_ids": "2022.acl-long.41_1759"}
+{"input": "question - guided reinforcement learning is used for Task| context: while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world .", "entity": "question - guided reinforcement learning", "output": "text - based games", "neg_sample": ["question - guided reinforcement learning is used for Task", "while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world ."], "relation": "used for", "id": "2022.acl-long.41", "year": 2022, "rel_sent": "Perceiving the World : Question - guided Reinforcement Learning for Text - based Games.", "forward": true, "src_ids": "2022.acl-long.41_1760"}
+{"input": "language learning is done by using Method| context: text - based games provide an interactive way to study natural language processing . while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world .", "entity": "language learning", "output": "two - phase training framework", "neg_sample": ["language learning is done by using Method", "text - based games provide an interactive way to study natural language processing .", "while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world ."], "relation": "used for", "id": "2022.acl-long.41", "year": 2022, "rel_sent": "We then propose a two - phase training framework to decouple language learning from reinforcement learning , which further improves the sample efficiency .", "forward": false, "src_ids": "2022.acl-long.41_1761"}
+{"input": "reinforcement learning is done by using Method| context: text - based games provide an interactive way to study natural language processing . while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world .", "entity": "reinforcement learning", "output": "two - phase training framework", "neg_sample": ["reinforcement learning is done by using Method", "text - based games provide an interactive way to study natural language processing .", "while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world ."], "relation": "used for", "id": "2022.acl-long.41", "year": 2022, "rel_sent": "We then propose a two - phase training framework to decouple language learning from reinforcement learning , which further improves the sample efficiency .", "forward": false, "src_ids": "2022.acl-long.41_1762"}
+{"input": "two - phase training framework is used for Task| context: text - based games provide an interactive way to study natural language processing . while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world .", "entity": "two - phase training framework", "output": "language learning", "neg_sample": ["two - phase training framework is used for Task", "text - based games provide an interactive way to study natural language processing .", "while deep reinforcement learning has shown effectiveness in developing the game playing agent , the low sample efficiency and the large action space remain to be the two major challenges that hinder the drl from being applied in the real world ."], "relation": "used for", "id": "2022.acl-long.41", "year": 2022, "rel_sent": "We then propose a two - phase training framework to decouple language learning from reinforcement learning , which further improves the sample efficiency .", "forward": true, "src_ids": "2022.acl-long.41_1763"}
+{"input": "two - phase training framework is used for Method| context: text - based games provide an interactive way to study natural language processing .", "entity": "two - phase training framework", "output": "reinforcement learning", "neg_sample": ["two - phase training framework is used for Method", "text - based games provide an interactive way to study natural language processing ."], "relation": "used for", "id": "2022.acl-long.41", "year": 2022, "rel_sent": "We then propose a two - phase training framework to decouple language learning from reinforcement learning , which further improves the sample efficiency .", "forward": true, "src_ids": "2022.acl-long.41_1764"}
+{"input": "generating code is done by using Method| context: considering the seq2seq architecture of yin and neubig ( 2018 ) for natural language to code translation , we identify four key components of importance : grammatical constraints , lexical preprocessing , input representations , and copy mechanisms .", "entity": "generating code", "output": "lexical and grammatical processing", "neg_sample": ["generating code is done by using Method", "considering the seq2seq architecture of yin and neubig ( 2018 ) for natural language to code translation , we identify four key components of importance : grammatical constraints , lexical preprocessing , input representations , and copy mechanisms ."], "relation": "used for", "id": "2022.findings-acl.173", "year": 2022, "rel_sent": "The impact of lexical and grammatical processing on generating code from natural language.", "forward": false, "src_ids": "2022.findings-acl.173_1765"}
+{"input": "lexical and grammatical processing is used for Task| context: considering the seq2seq architecture of yin and neubig ( 2018 ) for natural language to code translation , we identify four key components of importance : grammatical constraints , lexical preprocessing , input representations , and copy mechanisms .", "entity": "lexical and grammatical processing", "output": "generating code", "neg_sample": ["lexical and grammatical processing is used for Task", "considering the seq2seq architecture of yin and neubig ( 2018 ) for natural language to code translation , we identify four key components of importance : grammatical constraints , lexical preprocessing , input representations , and copy mechanisms ."], "relation": "used for", "id": "2022.findings-acl.173", "year": 2022, "rel_sent": "The impact of lexical and grammatical processing on generating code from natural language.", "forward": true, "src_ids": "2022.findings-acl.173_1766"}
+{"input": "semantic biases is used for Task| context: over the last few decades , multiple efforts have been undertaken to investigate incorrect translations caused by the polysemous nature of words . within this body of research , some studies have posited that models pick up semantic biases existing in the training data , thus producing translation errors .", "entity": "semantic biases", "output": "machine translation", "neg_sample": ["semantic biases is used for Task", "over the last few decades , multiple efforts have been undertaken to investigate incorrect translations caused by the polysemous nature of words .", "within this body of research , some studies have posited that models pick up semantic biases existing in the training data , thus producing translation errors ."], "relation": "used for", "id": "2022.acl-long.298", "year": 2022, "rel_sent": "In this paper , we present DiBiMT , the first entirely manually - curated evaluation benchmark which enables an extensive study of semantic biases in Machine Translation of nominal and verbal words in five different language combinations , namely , English and one or other of the following languages : Chinese , German , Italian , Russian and Spanish .", "forward": true, "src_ids": "2022.acl-long.298_1767"}
+{"input": "machine translation is done by using OtherScientificTerm| context: lexical ambiguity poses one of the greatest challenges in the field of machine translation . over the last few decades , multiple efforts have been undertaken to investigate incorrect translations caused by the polysemous nature of words .", "entity": "machine translation", "output": "semantic biases", "neg_sample": ["machine translation is done by using OtherScientificTerm", "lexical ambiguity poses one of the greatest challenges in the field of machine translation .", "over the last few decades , multiple efforts have been undertaken to investigate incorrect translations caused by the polysemous nature of words ."], "relation": "used for", "id": "2022.acl-long.298", "year": 2022, "rel_sent": "In this paper , we present DiBiMT , the first entirely manually - curated evaluation benchmark which enables an extensive study of semantic biases in Machine Translation of nominal and verbal words in five different language combinations , namely , English and one or other of the following languages : Chinese , German , Italian , Russian and Spanish .", "forward": false, "src_ids": "2022.acl-long.298_1768"}
+{"input": "ai tasks is done by using OtherScientificTerm| context: several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized . at issue here are not just individual systems and datasets , but also the ai tasks themselves .", "entity": "ai tasks", "output": "ethics sheets", "neg_sample": ["ai tasks is done by using OtherScientificTerm", "several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized .", "at issue here are not just individual systems and datasets , but also the ai tasks themselves ."], "relation": "used for", "id": "2022.acl-long.573", "year": 2022, "rel_sent": "Ethics Sheets for AI Tasks.", "forward": false, "src_ids": "2022.acl-long.573_1769"}
+{"input": "ethical considerations is done by using OtherScientificTerm| context: several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized . at issue here are not just individual systems and datasets , but also the ai tasks themselves .", "entity": "ethical considerations", "output": "ethics sheets", "neg_sample": ["ethical considerations is done by using OtherScientificTerm", "several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized .", "at issue here are not just individual systems and datasets , but also the ai tasks themselves ."], "relation": "used for", "id": "2022.acl-long.573", "year": 2022, "rel_sent": "Ethics sheets are a mechanism to engage with and document ethical considerations before building datasets and systems .", "forward": false, "src_ids": "2022.acl-long.573_1770"}
+{"input": "ethics sheets is used for Task| context: several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized .", "entity": "ethics sheets", "output": "ai tasks", "neg_sample": ["ethics sheets is used for Task", "several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized ."], "relation": "used for", "id": "2022.acl-long.573", "year": 2022, "rel_sent": "Ethics Sheets for AI Tasks.", "forward": true, "src_ids": "2022.acl-long.573_1771"}
+{"input": "ethics sheets is used for OtherScientificTerm| context: several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized . at issue here are not just individual systems and datasets , but also the ai tasks themselves .", "entity": "ethics sheets", "output": "ethical considerations", "neg_sample": ["ethics sheets is used for OtherScientificTerm", "several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized .", "at issue here are not just individual systems and datasets , but also the ai tasks themselves ."], "relation": "used for", "id": "2022.acl-long.573", "year": 2022, "rel_sent": "Ethics sheets are a mechanism to engage with and document ethical considerations before building datasets and systems .", "forward": true, "src_ids": "2022.acl-long.573_1772"}
+{"input": "task - oriented dialog is done by using Method| context: recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) . these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "task - oriented dialog", "output": "domain specialization", "neg_sample": ["task - oriented dialog is done by using Method", "recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) .", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.72", "year": 2022, "rel_sent": "DS - TOD : Efficient Domain Specialization for Task - Oriented Dialog.", "forward": false, "src_ids": "2022.findings-acl.72_1773"}
+{"input": "domain specialization is used for Task| context: these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "domain specialization", "output": "task - oriented dialog", "neg_sample": ["domain specialization is used for Task", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.72", "year": 2022, "rel_sent": "DS - TOD : Efficient Domain Specialization for Task - Oriented Dialog.", "forward": true, "src_ids": "2022.findings-acl.72_1774"}
+{"input": "pretrained language models ( plms ) is used for Task| context: these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "pretrained language models ( plms )", "output": "task - oriented dialog", "neg_sample": ["pretrained language models ( plms ) is used for Task", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.72", "year": 2022, "rel_sent": "In this work , we investigate the effects of domain specialization of pretrained language models ( PLMs ) for TOD .", "forward": true, "src_ids": "2022.findings-acl.72_1775"}
+{"input": "task - oriented dialog is done by using Method| context: recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) . these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "task - oriented dialog", "output": "pretrained language models ( plms )", "neg_sample": ["task - oriented dialog is done by using Method", "recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) .", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.72", "year": 2022, "rel_sent": "In this work , we investigate the effects of domain specialization of pretrained language models ( PLMs ) for TOD .", "forward": false, "src_ids": "2022.findings-acl.72_1776"}
+{"input": "domain - specific pretraining is done by using OtherScientificTerm| context: recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) . these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "domain - specific pretraining", "output": "domaincc and domainreddit - resources", "neg_sample": ["domain - specific pretraining is done by using OtherScientificTerm", "recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) .", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.72", "year": 2022, "rel_sent": "Within our DS - TOD framework , we first automatically extract salient domain - specific terms , and then use them to construct DomainCC and DomainReddit - resources that we leverage for domain - specific pretraining , based on ( i ) masked language modeling ( MLM ) and ( ii ) response selection ( RS ) objectives , respectively .", "forward": false, "src_ids": "2022.findings-acl.72_1777"}
+{"input": "domaincc and domainreddit - resources is used for Task| context: recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) . these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "domaincc and domainreddit - resources", "output": "domain - specific pretraining", "neg_sample": ["domaincc and domainreddit - resources is used for Task", "recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) .", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.72", "year": 2022, "rel_sent": "Within our DS - TOD framework , we first automatically extract salient domain - specific terms , and then use them to construct DomainCC and DomainReddit - resources that we leverage for domain - specific pretraining , based on ( i ) masked language modeling ( MLM ) and ( ii ) response selection ( RS ) objectives , respectively .", "forward": true, "src_ids": "2022.findings-acl.72_1778"}
+{"input": "mention representations is done by using OtherScientificTerm| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "mention representations", "output": "sibling mentions", "neg_sample": ["mention representations is done by using OtherScientificTerm", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.147", "year": 2022, "rel_sent": "To this end , we propose to exploit sibling mentions for enhancing the mention representations . Specifically , we present two different metrics for sibling selection and employ an attentive graph neural network to aggregate information from sibling mentions .", "forward": false, "src_ids": "2022.acl-long.147_1779"}
+{"input": "hard mentions is done by using OtherScientificTerm| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "hard mentions", "output": "sibling mentions", "neg_sample": ["hard mentions is done by using OtherScientificTerm", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.147", "year": 2022, "rel_sent": "Moreover , our experiments indeed prove the superiority of sibling mentions in helping clarify the types for hard mentions .", "forward": false, "src_ids": "2022.acl-long.147_1780"}
+{"input": "fine - grained entity typing is done by using Method| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "fine - grained entity typing", "output": "scalable graph inference", "neg_sample": ["fine - grained entity typing is done by using Method", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.147", "year": 2022, "rel_sent": "Learning from Sibling Mentions with Scalable Graph Inference in Fine - Grained Entity Typing.", "forward": false, "src_ids": "2022.acl-long.147_1781"}
+{"input": "scalable graph inference is used for Task| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "scalable graph inference", "output": "fine - grained entity typing", "neg_sample": ["scalable graph inference is used for Task", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.147", "year": 2022, "rel_sent": "Learning from Sibling Mentions with Scalable Graph Inference in Fine - Grained Entity Typing.", "forward": true, "src_ids": "2022.acl-long.147_1782"}
+{"input": "sibling mentions is used for Method| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "sibling mentions", "output": "mention representations", "neg_sample": ["sibling mentions is used for Method", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.147", "year": 2022, "rel_sent": "To this end , we propose to exploit sibling mentions for enhancing the mention representations . Specifically , we present two different metrics for sibling selection and employ an attentive graph neural network to aggregate information from sibling mentions .", "forward": true, "src_ids": "2022.acl-long.147_1783"}
+{"input": "psycholinguistic research is used for Method| context: in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs . as a result , the verb is the primary determinant of the meaning of a clause . in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs . two decades of psycholinguistic research have produced substantial empirical evidence in favor of the construction view .", "entity": "psycholinguistic research", "output": "transformer - based language models ( lms )", "neg_sample": ["psycholinguistic research is used for Method", "in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs .", "as a result , the verb is the primary determinant of the meaning of a clause .", "in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs .", "two decades of psycholinguistic research have produced substantial empirical evidence in favor of the construction view ."], "relation": "used for", "id": "2022.acl-long.512", "year": 2022, "rel_sent": "Here we adapt several psycholinguistic studies to probe for the existence of argument structure constructions ( ASCs ) in Transformer - based language models ( LMs ) .", "forward": true, "src_ids": "2022.acl-long.512_1784"}
+{"input": "argument structure constructions ( ascs ) is done by using Task| context: in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs . as a result , the verb is the primary determinant of the meaning of a clause . in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs .", "entity": "argument structure constructions ( ascs )", "output": "psycholinguistic research", "neg_sample": ["argument structure constructions ( ascs ) is done by using Task", "in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs .", "as a result , the verb is the primary determinant of the meaning of a clause .", "in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs ."], "relation": "used for", "id": "2022.acl-long.512", "year": 2022, "rel_sent": "Here we adapt several psycholinguistic studies to probe for the existence of argument structure constructions ( ASCs ) in Transformer - based language models ( LMs ) .", "forward": false, "src_ids": "2022.acl-long.512_1785"}
+{"input": "transformer - based language models ( lms ) is done by using Task| context: in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs . as a result , the verb is the primary determinant of the meaning of a clause . in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs .", "entity": "transformer - based language models ( lms )", "output": "psycholinguistic research", "neg_sample": ["transformer - based language models ( lms ) is done by using Task", "in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs .", "as a result , the verb is the primary determinant of the meaning of a clause .", "in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs ."], "relation": "used for", "id": "2022.acl-long.512", "year": 2022, "rel_sent": "Here we adapt several psycholinguistic studies to probe for the existence of argument structure constructions ( ASCs ) in Transformer - based language models ( LMs ) .", "forward": false, "src_ids": "2022.acl-long.512_1786"}
+{"input": "psycholinguistic research is used for OtherScientificTerm| context: in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs . as a result , the verb is the primary determinant of the meaning of a clause . in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs . two decades of psycholinguistic research have produced substantial empirical evidence in favor of the construction view .", "entity": "psycholinguistic research", "output": "argument structure constructions ( ascs )", "neg_sample": ["psycholinguistic research is used for OtherScientificTerm", "in lexicalist linguistic theories , argument structure is assumed to be predictable from the meaning of verbs .", "as a result , the verb is the primary determinant of the meaning of a clause .", "in contrast , construction grammarians propose that argument structure is encoded in constructions ( or form - meaning pairs ) that are distinct from verbs .", "two decades of psycholinguistic research have produced substantial empirical evidence in favor of the construction view ."], "relation": "used for", "id": "2022.acl-long.512", "year": 2022, "rel_sent": "Here we adapt several psycholinguistic studies to probe for the existence of argument structure constructions ( ASCs ) in Transformer - based language models ( LMs ) .", "forward": true, "src_ids": "2022.acl-long.512_1787"}
+{"input": "oxmc is done by using Method| context: the extreme multi - label classification ( xmc ) task aims at tagging content with a subset of labels from an extremely large label set . the label vocabulary is typically defined in advance by domain experts and assumed to capture all necessary tags . however in real world scenarios this label set , although large , is often incomplete and experts frequently need to refine it .", "entity": "oxmc", "output": "fine - tuned seq2seq model", "neg_sample": ["oxmc is done by using Method", "the extreme multi - label classification ( xmc ) task aims at tagging content with a subset of labels from an extremely large label set .", "the label vocabulary is typically defined in advance by domain experts and assumed to capture all necessary tags .", "however in real world scenarios this label set , although large , is often incomplete and experts frequently need to refine it ."], "relation": "used for", "id": "2022.findings-acl.123", "year": 2022, "rel_sent": "We propose GROOV , a fine - tuned seq2seq model for OXMC that generates the set of labels as a flat sequence and is trained using a novel loss independent of predicted label order .", "forward": false, "src_ids": "2022.findings-acl.123_1788"}
+{"input": "fine - tuned seq2seq model is used for Method| context: the extreme multi - label classification ( xmc ) task aims at tagging content with a subset of labels from an extremely large label set . the label vocabulary is typically defined in advance by domain experts and assumed to capture all necessary tags . however in real world scenarios this label set , although large , is often incomplete and experts frequently need to refine it .", "entity": "fine - tuned seq2seq model", "output": "oxmc", "neg_sample": ["fine - tuned seq2seq model is used for Method", "the extreme multi - label classification ( xmc ) task aims at tagging content with a subset of labels from an extremely large label set .", "the label vocabulary is typically defined in advance by domain experts and assumed to capture all necessary tags .", "however in real world scenarios this label set , although large , is often incomplete and experts frequently need to refine it ."], "relation": "used for", "id": "2022.findings-acl.123", "year": 2022, "rel_sent": "We propose GROOV , a fine - tuned seq2seq model for OXMC that generates the set of labels as a flat sequence and is trained using a novel loss independent of predicted label order .", "forward": true, "src_ids": "2022.findings-acl.123_1789"}
+{"input": "generating faithful plot events is done by using Method| context: we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps . the dataset provides a challenging testbed for abstractive summarization for several reasons . plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript . these details must be found and integrated toform the succinct plot descriptions in the recaps . also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief . this information is rarely contained in recaps .", "entity": "generating faithful plot events", "output": "non - oracle models", "neg_sample": ["generating faithful plot events is done by using Method", "we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps .", "the dataset provides a challenging testbed for abstractive summarization for several reasons .", "plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript .", "these details must be found and integrated toform the succinct plot descriptions in the recaps .", "also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief .", "this information is rarely contained in recaps ."], "relation": "used for", "id": "2022.acl-long.589", "year": 2022, "rel_sent": "Human evaluation and qualitative analysis reveal that our non - oracle models are competitive with their oracle counterparts in terms of generating faithful plot events and can benefit from better content selectors .", "forward": false, "src_ids": "2022.acl-long.589_1790"}
+{"input": "generating faithful plot events is done by using Generic| context: we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps . the dataset provides a challenging testbed for abstractive summarization for several reasons . plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript . these details must be found and integrated toform the succinct plot descriptions in the recaps . also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief . this information is rarely contained in recaps .", "entity": "generating faithful plot events", "output": "oracle counterparts", "neg_sample": ["generating faithful plot events is done by using Generic", "we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps .", "the dataset provides a challenging testbed for abstractive summarization for several reasons .", "plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript .", "these details must be found and integrated toform the succinct plot descriptions in the recaps .", "also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief .", "this information is rarely contained in recaps ."], "relation": "used for", "id": "2022.acl-long.589", "year": 2022, "rel_sent": "Human evaluation and qualitative analysis reveal that our non - oracle models are competitive with their oracle counterparts in terms of generating faithful plot events and can benefit from better content selectors .", "forward": false, "src_ids": "2022.acl-long.589_1791"}
+{"input": "oracle counterparts is used for Task| context: we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps . the dataset provides a challenging testbed for abstractive summarization for several reasons . plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript . these details must be found and integrated toform the succinct plot descriptions in the recaps . also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief . this information is rarely contained in recaps .", "entity": "oracle counterparts", "output": "generating faithful plot events", "neg_sample": ["oracle counterparts is used for Task", "we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps .", "the dataset provides a challenging testbed for abstractive summarization for several reasons .", "plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript .", "these details must be found and integrated toform the succinct plot descriptions in the recaps .", "also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief .", "this information is rarely contained in recaps ."], "relation": "used for", "id": "2022.acl-long.589", "year": 2022, "rel_sent": "Human evaluation and qualitative analysis reveal that our non - oracle models are competitive with their oracle counterparts in terms of generating faithful plot events and can benefit from better content selectors .", "forward": true, "src_ids": "2022.acl-long.589_1792"}
+{"input": "non - oracle models is used for Task| context: we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps . the dataset provides a challenging testbed for abstractive summarization for several reasons . plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript . these details must be found and integrated toform the succinct plot descriptions in the recaps . also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief . this information is rarely contained in recaps .", "entity": "non - oracle models", "output": "generating faithful plot events", "neg_sample": ["non - oracle models is used for Task", "we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps .", "the dataset provides a challenging testbed for abstractive summarization for several reasons .", "plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript .", "these details must be found and integrated toform the succinct plot descriptions in the recaps .", "also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief .", "this information is rarely contained in recaps ."], "relation": "used for", "id": "2022.acl-long.589", "year": 2022, "rel_sent": "Human evaluation and qualitative analysis reveal that our non - oracle models are competitive with their oracle counterparts in terms of generating faithful plot events and can benefit from better content selectors .", "forward": true, "src_ids": "2022.acl-long.589_1793"}
+{"input": "unfaithful facts is done by using Method| context: we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps . the dataset provides a challenging testbed for abstractive summarization for several reasons . plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript . these details must be found and integrated toform the succinct plot descriptions in the recaps . also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief . this information is rarely contained in recaps .", "entity": "unfaithful facts", "output": "oracle and non - oracle models", "neg_sample": ["unfaithful facts is done by using Method", "we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps .", "the dataset provides a challenging testbed for abstractive summarization for several reasons .", "plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript .", "these details must be found and integrated toform the succinct plot descriptions in the recaps .", "also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief .", "this information is rarely contained in recaps ."], "relation": "used for", "id": "2022.acl-long.589", "year": 2022, "rel_sent": "Both oracle and non - oracle models generate unfaithful facts , suggesting future research directions .", "forward": false, "src_ids": "2022.acl-long.589_1794"}
+{"input": "oracle and non - oracle models is used for OtherScientificTerm| context: we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps . the dataset provides a challenging testbed for abstractive summarization for several reasons . plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript . these details must be found and integrated toform the succinct plot descriptions in the recaps . also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief . this information is rarely contained in recaps .", "entity": "oracle and non - oracle models", "output": "unfaithful facts", "neg_sample": ["oracle and non - oracle models is used for OtherScientificTerm", "we introduce summscreen , a summarization dataset comprised of pairs of tv series transcripts and human written recaps .", "the dataset provides a challenging testbed for abstractive summarization for several reasons .", "plot details are often expressed indirectly in character dialogues and may be scattered across the entirety of the transcript .", "these details must be found and integrated toform the succinct plot descriptions in the recaps .", "also , tv scripts contain content that does not directly pertain to the central plot but rather serves to develop characters or provide comic relief .", "this information is rarely contained in recaps ."], "relation": "used for", "id": "2022.acl-long.589", "year": 2022, "rel_sent": "Both oracle and non - oracle models generate unfaithful facts , suggesting future research directions .", "forward": true, "src_ids": "2022.acl-long.589_1795"}
+{"input": "dependency modeling objective is used for OtherScientificTerm| context: various models have been proposed to incorporate knowledge of syntactic structures into neural language models . however , previous works have relied heavily on elaborate components for a specific language model , usually recurrent neural network ( rnn ) , which makes themselves unwieldy in practice tofit into other neural language models , such as transformer and gpt-2 .", "entity": "dependency modeling objective", "output": "probability distribution", "neg_sample": ["dependency modeling objective is used for OtherScientificTerm", "various models have been proposed to incorporate knowledge of syntactic structures into neural language models .", "however , previous works have relied heavily on elaborate components for a specific language model , usually recurrent neural network ( rnn ) , which makes themselves unwieldy in practice tofit into other neural language models , such as transformer and gpt-2 ."], "relation": "used for", "id": "2022.acl-long.535", "year": 2022, "rel_sent": "In detail , we first train neural language models with a novel dependency modeling objective to learn the probability distribution of future dependent tokens given context .", "forward": true, "src_ids": "2022.acl-long.535_1796"}
+{"input": "neural language models is used for OtherScientificTerm| context: various models have been proposed to incorporate knowledge of syntactic structures into neural language models . however , previous works have relied heavily on elaborate components for a specific language model , usually recurrent neural network ( rnn ) , which makes themselves unwieldy in practice tofit into other neural language models , such as transformer and gpt-2 .", "entity": "neural language models", "output": "probability distribution", "neg_sample": ["neural language models is used for OtherScientificTerm", "various models have been proposed to incorporate knowledge of syntactic structures into neural language models .", "however , previous works have relied heavily on elaborate components for a specific language model , usually recurrent neural network ( rnn ) , which makes themselves unwieldy in practice tofit into other neural language models , such as transformer and gpt-2 ."], "relation": "used for", "id": "2022.acl-long.535", "year": 2022, "rel_sent": "In detail , we first train neural language models with a novel dependency modeling objective to learn the probability distribution of future dependent tokens given context .", "forward": true, "src_ids": "2022.acl-long.535_1797"}
+{"input": "probability distribution is done by using OtherScientificTerm| context: various models have been proposed to incorporate knowledge of syntactic structures into neural language models . however , previous works have relied heavily on elaborate components for a specific language model , usually recurrent neural network ( rnn ) , which makes themselves unwieldy in practice tofit into other neural language models , such as transformer and gpt-2 .", "entity": "probability distribution", "output": "dependency modeling objective", "neg_sample": ["probability distribution is done by using OtherScientificTerm", "various models have been proposed to incorporate knowledge of syntactic structures into neural language models .", "however , previous works have relied heavily on elaborate components for a specific language model , usually recurrent neural network ( rnn ) , which makes themselves unwieldy in practice tofit into other neural language models , such as transformer and gpt-2 ."], "relation": "used for", "id": "2022.acl-long.535", "year": 2022, "rel_sent": "In detail , we first train neural language models with a novel dependency modeling objective to learn the probability distribution of future dependent tokens given context .", "forward": false, "src_ids": "2022.acl-long.535_1798"}
+{"input": "interactive weakly - supervised learning is done by using Method| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "interactive weakly - supervised learning", "output": "prompt - based rule discovery", "neg_sample": ["interactive weakly - supervised learning is done by using Method", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "Prompt - Based Rule Discovery and Boosting for Interactive Weakly - Supervised Learning.", "forward": false, "src_ids": "2022.acl-long.55_1799"}
+{"input": "large - error instances is done by using Method| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "large - error instances", "output": "boosting", "neg_sample": ["large - error instances is done by using Method", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "It uses boosting to identify large - error instances and discovers candidate rules from them by prompting pre - trained LMs with rule templates .", "forward": false, "src_ids": "2022.acl-long.55_1800"}
+{"input": "boosting is used for Task| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "boosting", "output": "interactive weakly - supervised learning", "neg_sample": ["boosting is used for Task", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "Prompt - Based Rule Discovery and Boosting for Interactive Weakly - Supervised Learning.", "forward": true, "src_ids": "2022.acl-long.55_1801"}
+{"input": "prompt - based rule discovery is used for Task| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "prompt - based rule discovery", "output": "interactive weakly - supervised learning", "neg_sample": ["prompt - based rule discovery is used for Task", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "Prompt - Based Rule Discovery and Boosting for Interactive Weakly - Supervised Learning.", "forward": true, "src_ids": "2022.acl-long.55_1802"}
+{"input": "boosting is used for OtherScientificTerm| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "boosting", "output": "large - error instances", "neg_sample": ["boosting is used for OtherScientificTerm", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "It uses boosting to identify large - error instances and discovers candidate rules from them by prompting pre - trained LMs with rule templates .", "forward": true, "src_ids": "2022.acl-long.55_1803"}
+{"input": "complementary weak labels is done by using OtherScientificTerm| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "complementary weak labels", "output": "rules", "neg_sample": ["complementary weak labels is done by using OtherScientificTerm", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "The candidate rules are judged by human experts , and the accepted rules are used to generate complementary weak labels and strengthen the current model .", "forward": false, "src_ids": "2022.acl-long.55_1804"}
+{"input": "rules is used for OtherScientificTerm| context: weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult .", "entity": "rules", "output": "complementary weak labels", "neg_sample": ["rules is used for OtherScientificTerm", "weakly - supervised learning ( wsl ) has shown promising results in addressing label scarcity on many nlp tasks , but manually designing a comprehensive , high - quality labeling rule set is tedious and difficult ."], "relation": "used for", "id": "2022.acl-long.55", "year": 2022, "rel_sent": "The candidate rules are judged by human experts , and the accepted rules are used to generate complementary weak labels and strengthen the current model .", "forward": true, "src_ids": "2022.acl-long.55_1805"}
+{"input": "abstractive long - input summarization is done by using Method| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "abstractive long - input summarization", "output": "dynamic latent extraction approach", "neg_sample": ["abstractive long - input summarization is done by using Method", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "In this paper , we present DYLE , a novel dynamic latent extraction approach for abstractive long - input summarization .", "forward": false, "src_ids": "2022.acl-long.118_1806"}
+{"input": "dynamic latent extraction approach is used for Task| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "dynamic latent extraction approach", "output": "abstractive long - input summarization", "neg_sample": ["dynamic latent extraction approach is used for Task", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "In this paper , we present DYLE , a novel dynamic latent extraction approach for abstractive long - input summarization .", "forward": true, "src_ids": "2022.acl-long.118_1807"}
+{"input": "decoding is done by using OtherScientificTerm| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "decoding", "output": "dynamic snippet - level attention weights", "neg_sample": ["decoding is done by using OtherScientificTerm", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "DYLE jointly trains an extractor and a generator and treats the extracted text snippets as the latent variable , allowing dynamic snippet - level attention weights during decoding .", "forward": false, "src_ids": "2022.acl-long.118_1808"}
+{"input": "dynamic snippet - level attention weights is used for Task| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "dynamic snippet - level attention weights", "output": "decoding", "neg_sample": ["dynamic snippet - level attention weights is used for Task", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "DYLE jointly trains an extractor and a generator and treats the extracted text snippets as the latent variable , allowing dynamic snippet - level attention weights during decoding .", "forward": true, "src_ids": "2022.acl-long.118_1809"}
+{"input": "oracle extraction is done by using Method| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "oracle extraction", "output": "heuristics", "neg_sample": ["oracle extraction is done by using Method", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "To provide adequate supervision , we propose simple yet effective heuristics for oracle extraction as well as a consistency loss term , which encourages the extractor to approximate the averaged dynamic weights predicted by the generator .", "forward": false, "src_ids": "2022.acl-long.118_1810"}
+{"input": "heuristics is used for Task| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "heuristics", "output": "oracle extraction", "neg_sample": ["heuristics is used for Task", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "To provide adequate supervision , we propose simple yet effective heuristics for oracle extraction as well as a consistency loss term , which encourages the extractor to approximate the averaged dynamic weights predicted by the generator .", "forward": true, "src_ids": "2022.acl-long.118_1811"}
+{"input": "generation process is done by using OtherScientificTerm| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "generation process", "output": "dynamic weights", "neg_sample": ["generation process is done by using OtherScientificTerm", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "Further analysis shows that the proposed dynamic weights provide interpretability of our generation process .", "forward": false, "src_ids": "2022.acl-long.118_1812"}
+{"input": "dynamic weights is used for Task| context: transformer - based models have achieved state - of - the - art performance on short - input summarization . however , they still struggle with summarizing longer text .", "entity": "dynamic weights", "output": "generation process", "neg_sample": ["dynamic weights is used for Task", "transformer - based models have achieved state - of - the - art performance on short - input summarization .", "however , they still struggle with summarizing longer text ."], "relation": "used for", "id": "2022.acl-long.118", "year": 2022, "rel_sent": "Further analysis shows that the proposed dynamic weights provide interpretability of our generation process .", "forward": true, "src_ids": "2022.acl-long.118_1813"}
+{"input": "identification of training data artifacts is done by using Method| context: training the deep neural networks that dominate nlp requires large datasets . these are often collected automatically or via crowdsourcing , and may exhibit systematic biases or annotation artifacts . by the latter we mean spurious correlations between inputs and outputs that do not represent a generally held causal relationship between features and classes ; models that exploit such correlations may appear to perform a given task well , but fail on out of sample data .", "entity": "identification of training data artifacts", "output": "attribution methods", "neg_sample": ["identification of training data artifacts is done by using Method", "training the deep neural networks that dominate nlp requires large datasets .", "these are often collected automatically or via crowdsourcing , and may exhibit systematic biases or annotation artifacts .", "by the latter we mean spurious correlations between inputs and outputs that do not represent a generally held causal relationship between features and classes ; models that exploit such correlations may appear to perform a given task well , but fail on out of sample data ."], "relation": "used for", "id": "2022.findings-acl.153", "year": 2022, "rel_sent": "In this paper , we evaluate use of different attribution methods for aiding identification of training data artifacts .", "forward": false, "src_ids": "2022.findings-acl.153_1814"}
+{"input": "attribution methods is used for Task| context: training the deep neural networks that dominate nlp requires large datasets . these are often collected automatically or via crowdsourcing , and may exhibit systematic biases or annotation artifacts . by the latter we mean spurious correlations between inputs and outputs that do not represent a generally held causal relationship between features and classes ; models that exploit such correlations may appear to perform a given task well , but fail on out of sample data .", "entity": "attribution methods", "output": "identification of training data artifacts", "neg_sample": ["attribution methods is used for Task", "training the deep neural networks that dominate nlp requires large datasets .", "these are often collected automatically or via crowdsourcing , and may exhibit systematic biases or annotation artifacts .", "by the latter we mean spurious correlations between inputs and outputs that do not represent a generally held causal relationship between features and classes ; models that exploit such correlations may appear to perform a given task well , but fail on out of sample data ."], "relation": "used for", "id": "2022.findings-acl.153", "year": 2022, "rel_sent": "In this paper , we evaluate use of different attribution methods for aiding identification of training data artifacts .", "forward": true, "src_ids": "2022.findings-acl.153_1815"}
+{"input": "discriminative token representation is done by using OtherScientificTerm| context: we propose a novel framework to conduct field extraction from forms with unlabeled data .", "entity": "discriminative token representation", "output": "supervisory signal", "neg_sample": ["discriminative token representation is done by using OtherScientificTerm", "we propose a novel framework to conduct field extraction from forms with unlabeled data ."], "relation": "used for", "id": "2022.spanlp-1.4", "year": 2022, "rel_sent": "Using the supervisory signal from the pseudo - labels , we extract a discriminative token representation from a transformer - based model by modeling the interaction between text in the form .", "forward": false, "src_ids": "2022.spanlp-1.4_1816"}
+{"input": "supervisory signal is used for Method| context: we propose a novel framework to conduct field extraction from forms with unlabeled data .", "entity": "supervisory signal", "output": "discriminative token representation", "neg_sample": ["supervisory signal is used for Method", "we propose a novel framework to conduct field extraction from forms with unlabeled data ."], "relation": "used for", "id": "2022.spanlp-1.4", "year": 2022, "rel_sent": "Using the supervisory signal from the pseudo - labels , we extract a discriminative token representation from a transformer - based model by modeling the interaction between text in the form .", "forward": true, "src_ids": "2022.spanlp-1.4_1817"}
+{"input": "mining noisy pseudo - labels is done by using Method| context: we propose a novel framework to conduct field extraction from forms with unlabeled data .", "entity": "mining noisy pseudo - labels", "output": "rule - based method", "neg_sample": ["mining noisy pseudo - labels is done by using Method", "we propose a novel framework to conduct field extraction from forms with unlabeled data ."], "relation": "used for", "id": "2022.spanlp-1.4", "year": 2022, "rel_sent": "To bootstrap the training process , we develop a rule - based method for mining noisy pseudo - labels from unlabeled forms .", "forward": false, "src_ids": "2022.spanlp-1.4_1818"}
+{"input": "rule - based method is used for Task| context: we propose a novel framework to conduct field extraction from forms with unlabeled data .", "entity": "rule - based method", "output": "mining noisy pseudo - labels", "neg_sample": ["rule - based method is used for Task", "we propose a novel framework to conduct field extraction from forms with unlabeled data ."], "relation": "used for", "id": "2022.spanlp-1.4", "year": 2022, "rel_sent": "To bootstrap the training process , we develop a rule - based method for mining noisy pseudo - labels from unlabeled forms .", "forward": true, "src_ids": "2022.spanlp-1.4_1819"}
+{"input": "hope speech detection is done by using Method| context: language should be accommodating of equality and diversity as a fundamental aspect of communication . the language of internet users has a big impact on peer users all over the world . on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages . people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail . such motivational remarks are hope speech remarks . simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors .", "entity": "hope speech detection", "output": "ensemble model", "neg_sample": ["hope speech detection is done by using Method", "language should be accommodating of equality and diversity as a fundamental aspect of communication .", "the language of internet users has a big impact on peer users all over the world .", "on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages .", "people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail .", "such motivational remarks are hope speech remarks .", "simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors ."], "relation": "used for", "id": "2022.ltedi-1.31", "year": 2022, "rel_sent": "SOA_NLP@LT - EDI - ACL2022 : An Ensemble Model for Hope Speech Detection from YouTube Comments.", "forward": false, "src_ids": "2022.ltedi-1.31_1820"}
+{"input": "ensemble model is used for Task| context: language should be accommodating of equality and diversity as a fundamental aspect of communication . the language of internet users has a big impact on peer users all over the world . on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages . people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail . such motivational remarks are hope speech remarks . simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors .", "entity": "ensemble model", "output": "hope speech detection", "neg_sample": ["ensemble model is used for Task", "language should be accommodating of equality and diversity as a fundamental aspect of communication .", "the language of internet users has a big impact on peer users all over the world .", "on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages .", "people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail .", "such motivational remarks are hope speech remarks .", "simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors ."], "relation": "used for", "id": "2022.ltedi-1.31", "year": 2022, "rel_sent": "SOA_NLP@LT - EDI - ACL2022 : An Ensemble Model for Hope Speech Detection from YouTube Comments.", "forward": true, "src_ids": "2022.ltedi-1.31_1821"}
+{"input": "classifiers is done by using OtherScientificTerm| context: language should be accommodating of equality and diversity as a fundamental aspect of communication . the language of internet users has a big impact on peer users all over the world . on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages . people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail . such motivational remarks are hope speech remarks . simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors .", "entity": "classifiers", "output": "features", "neg_sample": ["classifiers is done by using OtherScientificTerm", "language should be accommodating of equality and diversity as a fundamental aspect of communication .", "the language of internet users has a big impact on peer users all over the world .", "on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages .", "people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail .", "such motivational remarks are hope speech remarks .", "simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors ."], "relation": "used for", "id": "2022.ltedi-1.31", "year": 2022, "rel_sent": "Extensive testing was carried out to discover the best features for the aforementioned classifiers .", "forward": false, "src_ids": "2022.ltedi-1.31_1822"}
+{"input": "features is used for Method| context: language should be accommodating of equality and diversity as a fundamental aspect of communication . the language of internet users has a big impact on peer users all over the world . on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages . people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail . such motivational remarks are hope speech remarks . simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors .", "entity": "features", "output": "classifiers", "neg_sample": ["features is used for Method", "language should be accommodating of equality and diversity as a fundamental aspect of communication .", "the language of internet users has a big impact on peer users all over the world .", "on virtual platforms such as facebook , twitter , and youtube , people express their opinions in different languages .", "people respect others ' accomplishments , pray for their well - being , and cheer them on when they fail .", "such motivational remarks are hope speech remarks .", "simultaneously , a group of users encourages discrimination against women , people of color , people with disabilities , and other minorities based on gender , race , sexual orientation , and other factors ."], "relation": "used for", "id": "2022.ltedi-1.31", "year": 2022, "rel_sent": "Extensive testing was carried out to discover the best features for the aforementioned classifiers .", "forward": true, "src_ids": "2022.ltedi-1.31_1823"}
+{"input": "natural language grammar is done by using Method| context: using the notion of polarity as a case study , we show that this is not always the most adequate set - up .", "entity": "natural language grammar", "output": "language models", "neg_sample": ["natural language grammar is done by using Method", "using the notion of polarity as a case study , we show that this is not always the most adequate set - up ."], "relation": "used for", "id": "2022.acl-long.455", "year": 2022, "rel_sent": "Establishing this allows us to more adequately evaluate the performance of language models and also to use language models to discover new insights into natural language grammar beyond existing linguistic theories .", "forward": false, "src_ids": "2022.acl-long.455_1824"}
+{"input": "language models is used for Method| context: representation of linguistic phenomena in computational language models is typically assessed against the predictions of existing linguistic theories of these phenomena . using the notion of polarity as a case study , we show that this is not always the most adequate set - up .", "entity": "language models", "output": "natural language grammar", "neg_sample": ["language models is used for Method", "representation of linguistic phenomena in computational language models is typically assessed against the predictions of existing linguistic theories of these phenomena .", "using the notion of polarity as a case study , we show that this is not always the most adequate set - up ."], "relation": "used for", "id": "2022.acl-long.455", "year": 2022, "rel_sent": "Establishing this allows us to more adequately evaluate the performance of language models and also to use language models to discover new insights into natural language grammar beyond existing linguistic theories .", "forward": true, "src_ids": "2022.acl-long.455_1825"}
+{"input": "task oriented dialogue systems is done by using Method| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "task oriented dialogue systems", "output": "logical reasoning", "neg_sample": ["task oriented dialogue systems is done by using Method", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "Logical Reasoning for Task Oriented Dialogue Systems.", "forward": false, "src_ids": "2022.ecnlp-1.10_1826"}
+{"input": "transformer based models is used for Method| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "transformer based models", "output": "logical reasoning", "neg_sample": ["transformer based models is used for Method", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "We show that the transformer based model can perform logical reasoning to answer questions when the dialogue context contains all the required information , otherwise it is able to extract appropriate constraints to pass to downstream components ( e.g.", "forward": true, "src_ids": "2022.ecnlp-1.10_1827"}
+{"input": "transformer based models is used for Method| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "transformer based models", "output": "logical reasoning", "neg_sample": ["transformer based models is used for Method", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "We observe that transformer based models such as UnifiedQA - T5 can be fine - tuned to perform logical reasoning ( such as numerical and categorical attributes ' comparison ) over attributes seen at training time ( e.g. , accuracy of 90%+ for comparison of smaller than kmax=5 values over heldout test dataset ) .", "forward": true, "src_ids": "2022.ecnlp-1.10_1828"}
+{"input": "logical reasoning is used for Task| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "logical reasoning", "output": "task oriented dialogue systems", "neg_sample": ["logical reasoning is used for Task", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "Logical Reasoning for Task Oriented Dialogue Systems.", "forward": true, "src_ids": "2022.ecnlp-1.10_1829"}
+{"input": "logical relations is done by using Method| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "logical relations", "output": "synthetic data generation mechanism", "neg_sample": ["logical relations is done by using Method", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "In this work , we propose a novel method tofine - tune pretrained transformer models such as Roberta and T5 , to reason over a set of facts in a given dialogue context . Our method includes a synthetic data generation mechanism which helps the model learn logical relations , such as comparison between list of numerical values , inverse relations ( and negation ) , inclusion and exclusion for categorical attributes , and application of a combination of attributes over both numerical and categorical values , and spoken form for numerical values , without need for additional training data .", "forward": false, "src_ids": "2022.ecnlp-1.10_1830"}
+{"input": "synthetic data generation mechanism is used for OtherScientificTerm| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "synthetic data generation mechanism", "output": "logical relations", "neg_sample": ["synthetic data generation mechanism is used for OtherScientificTerm", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "In this work , we propose a novel method tofine - tune pretrained transformer models such as Roberta and T5 , to reason over a set of facts in a given dialogue context . Our method includes a synthetic data generation mechanism which helps the model learn logical relations , such as comparison between list of numerical values , inverse relations ( and negation ) , inclusion and exclusion for categorical attributes , and application of a combination of attributes over both numerical and categorical values , and spoken form for numerical values , without need for additional training data .", "forward": true, "src_ids": "2022.ecnlp-1.10_1831"}
+{"input": "numerical values is done by using OtherScientificTerm| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "numerical values", "output": "spoken form", "neg_sample": ["numerical values is done by using OtherScientificTerm", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "In this work , we propose a novel method tofine - tune pretrained transformer models such as Roberta and T5 , to reason over a set of facts in a given dialogue context . Our method includes a synthetic data generation mechanism which helps the model learn logical relations , such as comparison between list of numerical values , inverse relations ( and negation ) , inclusion and exclusion for categorical attributes , and application of a combination of attributes over both numerical and categorical values , and spoken form for numerical values , without need for additional training data .", "forward": false, "src_ids": "2022.ecnlp-1.10_1832"}
+{"input": "spoken form is used for OtherScientificTerm| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "spoken form", "output": "numerical values", "neg_sample": ["spoken form is used for OtherScientificTerm", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "In this work , we propose a novel method tofine - tune pretrained transformer models such as Roberta and T5 , to reason over a set of facts in a given dialogue context . Our method includes a synthetic data generation mechanism which helps the model learn logical relations , such as comparison between list of numerical values , inverse relations ( and negation ) , inclusion and exclusion for categorical attributes , and application of a combination of attributes over both numerical and categorical values , and spoken form for numerical values , without need for additional training data .", "forward": true, "src_ids": "2022.ecnlp-1.10_1833"}
+{"input": "logical reasoning is done by using Method| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "logical reasoning", "output": "transformer based models", "neg_sample": ["logical reasoning is done by using Method", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "We show that the transformer based model can perform logical reasoning to answer questions when the dialogue context contains all the required information , otherwise it is able to extract appropriate constraints to pass to downstream components ( e.g.", "forward": false, "src_ids": "2022.ecnlp-1.10_1834"}
+{"input": "logical reasoning is done by using Method| context: in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates . however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules .", "entity": "logical reasoning", "output": "transformer based models", "neg_sample": ["logical reasoning is done by using Method", "in recent years , large pretrained models have been used in dialogue systems to improve successful task completion rates .", "however , lack of reasoning capabilities of dialogue platforms make it difficult to provide relevant and fluent responses , unless the designers of a conversational experience spend a considerable amount of time implementing these capabilities in external rule based modules ."], "relation": "used for", "id": "2022.ecnlp-1.10", "year": 2022, "rel_sent": "We observe that transformer based models such as UnifiedQA - T5 can be fine - tuned to perform logical reasoning ( such as numerical and categorical attributes ' comparison ) over attributes seen at training time ( e.g. , accuracy of 90%+ for comparison of smaller than kmax=5 values over heldout test dataset ) .", "forward": false, "src_ids": "2022.ecnlp-1.10_1835"}
+{"input": "natural language processing is done by using Method| context: the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear .", "entity": "natural language processing", "output": "data augmentations", "neg_sample": ["natural language processing is done by using Method", "the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear ."], "relation": "used for", "id": "2022.insights-1.12", "year": 2022, "rel_sent": "On the Impact of Data Augmentation on Downstream Performance in Natural Language Processing.", "forward": false, "src_ids": "2022.insights-1.12_1836"}
+{"input": "nlp tasks is done by using Method| context: the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear .", "entity": "nlp tasks", "output": "data augmentations", "neg_sample": ["nlp tasks is done by using Method", "the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear ."], "relation": "used for", "id": "2022.insights-1.12", "year": 2022, "rel_sent": "This all alludes to the difficulty of data augmentations for NLP tasks and we are inclined to believe that static data augmentations are not broadly applicable given these properties .", "forward": false, "src_ids": "2022.insights-1.12_1837"}
+{"input": "data augmentations is used for Task| context: with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models . while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited . the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear .", "entity": "data augmentations", "output": "natural language processing", "neg_sample": ["data augmentations is used for Task", "with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models .", "while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited .", "the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear ."], "relation": "used for", "id": "2022.insights-1.12", "year": 2022, "rel_sent": "On the Impact of Data Augmentation on Downstream Performance in Natural Language Processing.", "forward": true, "src_ids": "2022.insights-1.12_1838"}
+{"input": "pre - trained language models is done by using Method| context: with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models . while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited .", "entity": "pre - trained language models", "output": "data augmentation methods", "neg_sample": ["pre - trained language models is done by using Method", "with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models .", "while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited ."], "relation": "used for", "id": "2022.insights-1.12", "year": 2022, "rel_sent": "In this paper , we look to tackle this by evaluating the impact of 12 data augmentation methods on multiple datasets when finetuning pre - trained language models .", "forward": false, "src_ids": "2022.insights-1.12_1839"}
+{"input": "data augmentation methods is used for Method| context: with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models . while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited . the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear .", "entity": "data augmentation methods", "output": "pre - trained language models", "neg_sample": ["data augmentation methods is used for Method", "with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models .", "while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited .", "the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear ."], "relation": "used for", "id": "2022.insights-1.12", "year": 2022, "rel_sent": "In this paper , we look to tackle this by evaluating the impact of 12 data augmentation methods on multiple datasets when finetuning pre - trained language models .", "forward": true, "src_ids": "2022.insights-1.12_1840"}
+{"input": "data augmentations is used for Task| context: with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models . while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited . the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear .", "entity": "data augmentations", "output": "nlp tasks", "neg_sample": ["data augmentations is used for Task", "with in the broader scope of machine learning , data augmentation is a common strategy to improve generalization and robustness of machine learning models .", "while data augmentation has been widely used within computer vision , its use in the nlp has been been comparably rather limited .", "the reason for this is that within nlp , the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner , and effective data augmentation methods are unclear ."], "relation": "used for", "id": "2022.insights-1.12", "year": 2022, "rel_sent": "This all alludes to the difficulty of data augmentations for NLP tasks and we are inclined to believe that static data augmentations are not broadly applicable given these properties .", "forward": true, "src_ids": "2022.insights-1.12_1841"}
+{"input": "robustness methods is used for Task| context: our benchmarks cover four jurisdictions ( european council , usa , switzerland , and china ) , five languages ( english , german , french , italian and chinese ) and fairness across five attributes ( gender , age , region , language , and legal area ) .", "entity": "robustness methods", "output": "legal nlp", "neg_sample": ["robustness methods is used for Task", "our benchmarks cover four jurisdictions ( european council , usa , switzerland , and china ) , five languages ( english , german , french , italian and chinese ) and fairness across five attributes ( gender , age , region , language , and legal area ) ."], "relation": "used for", "id": "2022.acl-long.301", "year": 2022, "rel_sent": "Furthermore , we provide a quantitative and qualitative analysis of our results , highlighting open challenges in the development of robustness methods in legal NLP .", "forward": true, "src_ids": "2022.acl-long.301_1842"}
+{"input": "legal nlp is done by using Method| context: our benchmarks cover four jurisdictions ( european council , usa , switzerland , and china ) , five languages ( english , german , french , italian and chinese ) and fairness across five attributes ( gender , age , region , language , and legal area ) .", "entity": "legal nlp", "output": "robustness methods", "neg_sample": ["legal nlp is done by using Method", "our benchmarks cover four jurisdictions ( european council , usa , switzerland , and china ) , five languages ( english , german , french , italian and chinese ) and fairness across five attributes ( gender , age , region , language , and legal area ) ."], "relation": "used for", "id": "2022.acl-long.301", "year": 2022, "rel_sent": "Furthermore , we provide a quantitative and qualitative analysis of our results , highlighting open challenges in the development of robustness methods in legal NLP .", "forward": false, "src_ids": "2022.acl-long.301_1843"}
+{"input": "hierarchical importance scores is done by using Method| context: interpreting nlp models is fundamental for their development as it can shed light on hidden properties and unexpected behaviors . however , while transformer architectures exploit contextual information to enhance their predictive capabilities , most of the available methods to explain such predictions only provide importance scores at the word level .", "entity": "hierarchical importance scores", "output": "grammarshap", "neg_sample": ["hierarchical importance scores is done by using Method", "interpreting nlp models is fundamental for their development as it can shed light on hidden properties and unexpected behaviors .", "however , while transformer architectures exploit contextual information to enhance their predictive capabilities , most of the available methods to explain such predictions only provide importance scores at the word level ."], "relation": "used for", "id": "2022.lnls-1.2", "year": 2022, "rel_sent": "We extend the SHAP framework by proposing GrammarSHAP - a model - agnostic explainer leveraging the sentence 's constituency parsing to generate hierarchical importance scores .", "forward": false, "src_ids": "2022.lnls-1.2_1844"}
+{"input": "grammarshap is used for OtherScientificTerm| context: interpreting nlp models is fundamental for their development as it can shed light on hidden properties and unexpected behaviors . however , while transformer architectures exploit contextual information to enhance their predictive capabilities , most of the available methods to explain such predictions only provide importance scores at the word level .", "entity": "grammarshap", "output": "hierarchical importance scores", "neg_sample": ["grammarshap is used for OtherScientificTerm", "interpreting nlp models is fundamental for their development as it can shed light on hidden properties and unexpected behaviors .", "however , while transformer architectures exploit contextual information to enhance their predictive capabilities , most of the available methods to explain such predictions only provide importance scores at the word level ."], "relation": "used for", "id": "2022.lnls-1.2", "year": 2022, "rel_sent": "We extend the SHAP framework by proposing GrammarSHAP - a model - agnostic explainer leveraging the sentence 's constituency parsing to generate hierarchical importance scores .", "forward": true, "src_ids": "2022.lnls-1.2_1845"}
+{"input": "model - agnostic explainer is used for OtherScientificTerm| context: interpreting nlp models is fundamental for their development as it can shed light on hidden properties and unexpected behaviors . however , while transformer architectures exploit contextual information to enhance their predictive capabilities , most of the available methods to explain such predictions only provide importance scores at the word level .", "entity": "model - agnostic explainer", "output": "hierarchical importance scores", "neg_sample": ["model - agnostic explainer is used for OtherScientificTerm", "interpreting nlp models is fundamental for their development as it can shed light on hidden properties and unexpected behaviors .", "however , while transformer architectures exploit contextual information to enhance their predictive capabilities , most of the available methods to explain such predictions only provide importance scores at the word level ."], "relation": "used for", "id": "2022.lnls-1.2", "year": 2022, "rel_sent": "We extend the SHAP framework by proposing GrammarSHAP - a model - agnostic explainer leveraging the sentence 's constituency parsing to generate hierarchical importance scores .", "forward": true, "src_ids": "2022.lnls-1.2_1846"}
+{"input": "nlu test suites is done by using Task| context: we study few - shot debugging of transformer based natural language understanding models , using recently popularized test suites to not just diagnose but correct a problem .", "entity": "nlu test suites", "output": "fast few - shot debugging", "neg_sample": ["nlu test suites is done by using Task", "we study few - shot debugging of transformer based natural language understanding models , using recently popularized test suites to not just diagnose but correct a problem ."], "relation": "used for", "id": "2022.deelio-1.8", "year": 2022, "rel_sent": "Fast Few - shot Debugging for NLU Test Suites.", "forward": false, "src_ids": "2022.deelio-1.8_1847"}
+{"input": "fast few - shot debugging is used for Material| context: we study few - shot debugging of transformer based natural language understanding models , using recently popularized test suites to not just diagnose but correct a problem .", "entity": "fast few - shot debugging", "output": "nlu test suites", "neg_sample": ["fast few - shot debugging is used for Material", "we study few - shot debugging of transformer based natural language understanding models , using recently popularized test suites to not just diagnose but correct a problem ."], "relation": "used for", "id": "2022.deelio-1.8", "year": 2022, "rel_sent": "Fast Few - shot Debugging for NLU Test Suites.", "forward": true, "src_ids": "2022.deelio-1.8_1848"}
+{"input": "multimodal transformer - based models is done by using Method| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "multimodal transformer - based models", "output": "adapter - based approaches", "neg_sample": ["multimodal transformer - based models is done by using Method", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.196", "year": 2022, "rel_sent": "We further propose new adapter - based approaches to adapt multimodal transformer - based models to become multilingual , and - vice versa - multilingual models to become multimodal .", "forward": false, "src_ids": "2022.findings-acl.196_1849"}
+{"input": "adapter - based approaches is used for Method| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "adapter - based approaches", "output": "multimodal transformer - based models", "neg_sample": ["adapter - based approaches is used for Method", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.196", "year": 2022, "rel_sent": "We further propose new adapter - based approaches to adapt multimodal transformer - based models to become multilingual , and - vice versa - multilingual models to become multimodal .", "forward": true, "src_ids": "2022.findings-acl.196_1850"}
+{"input": "machine - generated translations is done by using Method| context: while state - of - the - art qe models have been shown to achieve good results , they over - rely on features that do not have a causal impact on the quality of a translation . in particular , there appears to be a partial input bias , i.e. , a tendency to assign high - quality scores to translations that are fluent and grammatically correct , even though they do not preserve the meaning of the source .", "entity": "machine - generated translations", "output": "predictive models", "neg_sample": ["machine - generated translations is done by using Method", "while state - of - the - art qe models have been shown to achieve good results , they over - rely on features that do not have a causal impact on the quality of a translation .", "in particular , there appears to be a partial input bias , i.e.", ", a tendency to assign high - quality scores to translations that are fluent and grammatically correct , even though they do not preserve the meaning of the source ."], "relation": "used for", "id": "2022.acl-long.104", "year": 2022, "rel_sent": "Machine Translation Quality Estimation ( QE ) aims to build predictive models to assess the quality of machine - generated translations in the absence of reference translations .", "forward": false, "src_ids": "2022.acl-long.104_1851"}
+{"input": "predictive models is used for Material| context: while state - of - the - art qe models have been shown to achieve good results , they over - rely on features that do not have a causal impact on the quality of a translation . in particular , there appears to be a partial input bias , i.e. , a tendency to assign high - quality scores to translations that are fluent and grammatically correct , even though they do not preserve the meaning of the source .", "entity": "predictive models", "output": "machine - generated translations", "neg_sample": ["predictive models is used for Material", "while state - of - the - art qe models have been shown to achieve good results , they over - rely on features that do not have a causal impact on the quality of a translation .", "in particular , there appears to be a partial input bias , i.e.", ", a tendency to assign high - quality scores to translations that are fluent and grammatically correct , even though they do not preserve the meaning of the source ."], "relation": "used for", "id": "2022.acl-long.104", "year": 2022, "rel_sent": "Machine Translation Quality Estimation ( QE ) aims to build predictive models to assess the quality of machine - generated translations in the absence of reference translations .", "forward": true, "src_ids": "2022.acl-long.104_1852"}
+{"input": "unified analysis is done by using Method| context: the central idea is that movement , albeit unbounded , boils down to local mother - daughter dependencies on a specific substructure called a tree tier . this re - veals interesting parallels between syntax and phonology , but it also looks very different from the standard view of movement .", "entity": "unified analysis", "output": "tsl-2", "neg_sample": ["unified analysis is done by using Method", "the central idea is that movement , albeit unbounded , boils down to local mother - daughter dependencies on a specific substructure called a tree tier .", "this re - veals interesting parallels between syntax and phonology , but it also looks very different from the standard view of movement ."], "relation": "used for", "id": "2022.scil-1.15", "year": 2022, "rel_sent": "I argue that this is indeed the case because TSL-2 furnishes a unified analysis of a variety of phenomena : multiple wh - movement , expletive constructions , the that -trace effect and the anti - that -trace effect , islands , and wh - agreement .", "forward": false, "src_ids": "2022.scil-1.15_1853"}
+{"input": "tsl-2 is used for Method| context: earlier work has shown that movement , which forms the backbone of minimalist syntax , belongs in the subregular class of tsl-2 dependencies over trees . the central idea is that movement , albeit unbounded , boils down to local mother - daughter dependencies on a specific substructure called a tree tier . this re - veals interesting parallels between syntax and phonology , but it also looks very different from the standard view of movement . one may wonder , then , whether the tsl-2 characterization is linguistically natural .", "entity": "tsl-2", "output": "unified analysis", "neg_sample": ["tsl-2 is used for Method", "earlier work has shown that movement , which forms the backbone of minimalist syntax , belongs in the subregular class of tsl-2 dependencies over trees .", "the central idea is that movement , albeit unbounded , boils down to local mother - daughter dependencies on a specific substructure called a tree tier .", "this re - veals interesting parallels between syntax and phonology , but it also looks very different from the standard view of movement .", "one may wonder , then , whether the tsl-2 characterization is linguistically natural ."], "relation": "used for", "id": "2022.scil-1.15", "year": 2022, "rel_sent": "I argue that this is indeed the case because TSL-2 furnishes a unified analysis of a variety of phenomena : multiple wh - movement , expletive constructions , the that -trace effect and the anti - that -trace effect , islands , and wh - agreement .", "forward": true, "src_ids": "2022.scil-1.15_1854"}
+{"input": "unsupervised parallel text mining is done by using Task| context: this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts . we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "unsupervised parallel text mining", "output": "bilingual alignment", "neg_sample": ["unsupervised parallel text mining is done by using Task", "this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts .", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "Bilingual alignment transfers to multilingual alignment for unsupervised parallel text mining.", "forward": false, "src_ids": "2022.acl-long.595_1855"}
+{"input": "multilingual alignment is used for Task| context: this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts . we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "multilingual alignment", "output": "unsupervised parallel text mining", "neg_sample": ["multilingual alignment is used for Task", "this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts .", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "Bilingual alignment transfers to multilingual alignment for unsupervised parallel text mining.", "forward": true, "src_ids": "2022.acl-long.595_1856"}
+{"input": "bilingual alignment is used for Task| context: this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts . we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "bilingual alignment", "output": "unsupervised parallel text mining", "neg_sample": ["bilingual alignment is used for Task", "this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts .", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "Bilingual alignment transfers to multilingual alignment for unsupervised parallel text mining.", "forward": true, "src_ids": "2022.acl-long.595_1857"}
+{"input": "unsupervised bitext mining is done by using Method| context: this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts . we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "unsupervised bitext mining", "output": "universal sentence encoders", "neg_sample": ["unsupervised bitext mining is done by using Method", "this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts .", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "The experiments evaluate the models as universal sentence encoders on the task of unsupervised bitext mining on two datasets , where the unsupervised model reaches the state of the art of unsupervised retrieval , and the alternative single - pair supervised model approaches the performance of multilingually supervised models .", "forward": false, "src_ids": "2022.acl-long.595_1858"}
+{"input": "universal sentence encoders is used for Task| context: this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts . we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "universal sentence encoders", "output": "unsupervised bitext mining", "neg_sample": ["universal sentence encoders is used for Task", "this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts .", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "The experiments evaluate the models as universal sentence encoders on the task of unsupervised bitext mining on two datasets , where the unsupervised model reaches the state of the art of unsupervised retrieval , and the alternative single - pair supervised model approaches the performance of multilingually supervised models .", "forward": true, "src_ids": "2022.acl-long.595_1859"}
+{"input": "sentence representations is done by using Method| context: this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts . we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "sentence representations", "output": "bilingual training techniques", "neg_sample": ["sentence representations is done by using Method", "this work presents methods for learning cross - lingual sentence representations using paired or unpaired bilingual texts .", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "The results suggest that bilingual training techniques as proposed can be applied to get sentence representations with multilingual alignment .", "forward": false, "src_ids": "2022.acl-long.595_1860"}
+{"input": "bilingual training techniques is used for Method| context: we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations .", "entity": "bilingual training techniques", "output": "sentence representations", "neg_sample": ["bilingual training techniques is used for Method", "we hypothesize that the cross - lingual alignment strategy is transferable , and therefore a model trained to align only two languages can encode multilingually more aligned representations ."], "relation": "used for", "id": "2022.acl-long.595", "year": 2022, "rel_sent": "The results suggest that bilingual training techniques as proposed can be applied to get sentence representations with multilingual alignment .", "forward": true, "src_ids": "2022.acl-long.595_1861"}
+{"input": "causal relations is done by using Generic| context: understanding causal narratives communicated in clinical notes can help make strides towards personalized healthcare . extracted causal information from clinical notes can be combined with structured ehr data such as patients ' demographics , diagnoses , and medications .", "entity": "causal relations", "output": "baseline scores", "neg_sample": ["causal relations is done by using Generic", "understanding causal narratives communicated in clinical notes can help make strides towards personalized healthcare .", "extracted causal information from clinical notes can be combined with structured ehr data such as patients ' demographics , diagnoses , and medications ."], "relation": "used for", "id": "2022.findings-acl.63", "year": 2022, "rel_sent": "In this work , we propose annotation guidelines , develop an annotated corpus and provide baseline scores to identify types and direction of causal relations between a pair of biomedical concepts in clinical notes ; communicated implicitly or explicitly , identified either in a single sentence or across multiple sentences .", "forward": false, "src_ids": "2022.findings-acl.63_1862"}
+{"input": "baseline scores is used for OtherScientificTerm| context: understanding causal narratives communicated in clinical notes can help make strides towards personalized healthcare . extracted causal information from clinical notes can be combined with structured ehr data such as patients ' demographics , diagnoses , and medications .", "entity": "baseline scores", "output": "causal relations", "neg_sample": ["baseline scores is used for OtherScientificTerm", "understanding causal narratives communicated in clinical notes can help make strides towards personalized healthcare .", "extracted causal information from clinical notes can be combined with structured ehr data such as patients ' demographics , diagnoses , and medications ."], "relation": "used for", "id": "2022.findings-acl.63", "year": 2022, "rel_sent": "In this work , we propose annotation guidelines , develop an annotated corpus and provide baseline scores to identify types and direction of causal relations between a pair of biomedical concepts in clinical notes ; communicated implicitly or explicitly , identified either in a single sentence or across multiple sentences .", "forward": true, "src_ids": "2022.findings-acl.63_1863"}
+{"input": "answer prediction is done by using Method| context: current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages .", "entity": "answer prediction", "output": "kg - fid", "neg_sample": ["answer prediction is done by using Method", "current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages ."], "relation": "used for", "id": "2022.acl-long.340", "year": 2022, "rel_sent": "Our experiments on common ODQA benchmark datasets ( Natural Questions and TriviaQA ) demonstrate that KG - FiD can achieve comparable or better performance in answer prediction than FiD , with less than 40 % of the computation cost .", "forward": false, "src_ids": "2022.acl-long.340_1864"}
+{"input": "fusion - in - decoder is done by using OtherScientificTerm| context: current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages .", "entity": "fusion - in - decoder", "output": "knowledge graph", "neg_sample": ["fusion - in - decoder is done by using OtherScientificTerm", "current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages ."], "relation": "used for", "id": "2022.acl-long.340", "year": 2022, "rel_sent": "KG - FiD : Infusing Knowledge Graph in Fusion - in - Decoder for Open - Domain Question Answering.", "forward": false, "src_ids": "2022.acl-long.340_1865"}
+{"input": "open - domain question answering is done by using Method| context: current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages .", "entity": "open - domain question answering", "output": "fusion - in - decoder", "neg_sample": ["open - domain question answering is done by using Method", "current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages ."], "relation": "used for", "id": "2022.acl-long.340", "year": 2022, "rel_sent": "KG - FiD : Infusing Knowledge Graph in Fusion - in - Decoder for Open - Domain Question Answering.", "forward": false, "src_ids": "2022.acl-long.340_1866"}
+{"input": "knowledge graph is used for Method| context: current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages .", "entity": "knowledge graph", "output": "fusion - in - decoder", "neg_sample": ["knowledge graph is used for Method", "current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages ."], "relation": "used for", "id": "2022.acl-long.340", "year": 2022, "rel_sent": "KG - FiD : Infusing Knowledge Graph in Fusion - in - Decoder for Open - Domain Question Answering.", "forward": true, "src_ids": "2022.acl-long.340_1867"}
+{"input": "kg - fid is used for Task| context: current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages .", "entity": "kg - fid", "output": "answer prediction", "neg_sample": ["kg - fid is used for Task", "current open - domain question answering ( odqa ) models typically include a retrieving module and a reading module , where the retriever selects potentially relevant passages from open - source documents for a given question , and the reader produces an answer based on the retrieved passages ."], "relation": "used for", "id": "2022.acl-long.340", "year": 2022, "rel_sent": "Our experiments on common ODQA benchmark datasets ( Natural Questions and TriviaQA ) demonstrate that KG - FiD can achieve comparable or better performance in answer prediction than FiD , with less than 40 % of the computation cost .", "forward": true, "src_ids": "2022.acl-long.340_1868"}
+{"input": "crossword puzzle is done by using Method| context: solving crossword puzzles requires diverse reasoning capabilities , access to a vast amount of knowledge about language and the world , and the ability to satisfy the constraints imposed by the structure of the puzzle .", "entity": "crossword puzzle", "output": "non - parametric constraint satisfaction baseline", "neg_sample": ["crossword puzzle is done by using Method", "solving crossword puzzles requires diverse reasoning capabilities , access to a vast amount of knowledge about language and the world , and the ability to satisfy the constraints imposed by the structure of the puzzle ."], "relation": "used for", "id": "2022.acl-long.189", "year": 2022, "rel_sent": "We also introduce a non - parametric constraint satisfaction baseline for solving the entire crossword puzzle .", "forward": false, "src_ids": "2022.acl-long.189_1869"}
+{"input": "document classifiers is done by using Method| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics . class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "document classifiers", "output": "optimization algorithms", "neg_sample": ["document classifiers is done by using Method", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics .", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Instead of simply resampling uniformly to hedge our bets , we focus on the underlying optimization algorithms used to train such document classifiers and evaluate several group - robust optimization algorithms , initially proposed to mitigate group - level disparities .", "forward": false, "src_ids": "2022.findings-acl.192_1870"}
+{"input": "optimization algorithms is used for Method| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics . class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "optimization algorithms", "output": "document classifiers", "neg_sample": ["optimization algorithms is used for Method", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics .", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Instead of simply resampling uniformly to hedge our bets , we focus on the underlying optimization algorithms used to train such document classifiers and evaluate several group - robust optimization algorithms , initially proposed to mitigate group - level disparities .", "forward": true, "src_ids": "2022.findings-acl.192_1871"}
+{"input": "group - level disparities is done by using Method| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics . class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "group - level disparities", "output": "group - robust optimization algorithms", "neg_sample": ["group - level disparities is done by using Method", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics .", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Instead of simply resampling uniformly to hedge our bets , we focus on the underlying optimization algorithms used to train such document classifiers and evaluate several group - robust optimization algorithms , initially proposed to mitigate group - level disparities .", "forward": false, "src_ids": "2022.findings-acl.192_1872"}
+{"input": "group - robust optimization algorithms is used for OtherScientificTerm| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics . class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "group - robust optimization algorithms", "output": "group - level disparities", "neg_sample": ["group - robust optimization algorithms is used for OtherScientificTerm", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics .", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Instead of simply resampling uniformly to hedge our bets , we focus on the underlying optimization algorithms used to train such document classifiers and evaluate several group - robust optimization algorithms , initially proposed to mitigate group - level disparities .", "forward": true, "src_ids": "2022.findings-acl.192_1873"}
+{"input": "sampling - based approaches is used for OtherScientificTerm| context: class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "sampling - based approaches", "output": "concept drift", "neg_sample": ["sampling - based approaches is used for OtherScientificTerm", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": true, "src_ids": "2022.findings-acl.192_1874"}
+{"input": "invariant risk minimization is used for OtherScientificTerm| context: class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "invariant risk minimization", "output": "concept drift", "neg_sample": ["invariant risk minimization is used for OtherScientificTerm", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": true, "src_ids": "2022.findings-acl.192_1875"}
+{"input": "spectral decoupling is used for OtherScientificTerm| context: class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "spectral decoupling", "output": "concept drift", "neg_sample": ["spectral decoupling is used for OtherScientificTerm", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": true, "src_ids": "2022.findings-acl.192_1876"}
+{"input": "class imbalance is done by using Method| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics . class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "class imbalance", "output": "invariant risk minimization", "neg_sample": ["class imbalance is done by using Method", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics .", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": false, "src_ids": "2022.findings-acl.192_1877"}
+{"input": "concept drift is done by using Method| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics . class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?", "entity": "concept drift", "output": "invariant risk minimization", "neg_sample": ["concept drift is done by using Method", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics .", "class imbalance and drift can sometimes be mitigated by resampling the training data to simulate ( or compensate for ) a known target distribution , but what if the target distribution is determined by unknown future events ?"], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": false, "src_ids": "2022.findings-acl.192_1878"}
+{"input": "sampling - based approaches is used for OtherScientificTerm| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics .", "entity": "sampling - based approaches", "output": "class imbalance", "neg_sample": ["sampling - based approaches is used for OtherScientificTerm", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics ."], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": true, "src_ids": "2022.findings-acl.192_1879"}
+{"input": "invariant risk minimization is used for OtherScientificTerm| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics .", "entity": "invariant risk minimization", "output": "class imbalance", "neg_sample": ["invariant risk minimization is used for OtherScientificTerm", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics ."], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": true, "src_ids": "2022.findings-acl.192_1880"}
+{"input": "spectral decoupling is used for OtherScientificTerm| context: in document classification for , e.g. , legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g. , policy changes , conflicts , or pandemics .", "entity": "spectral decoupling", "output": "class imbalance", "neg_sample": ["spectral decoupling is used for OtherScientificTerm", "in document classification for , e.g.", ", legal and biomedical text , we often deal with hundreds of classes , including very infrequent ones , as well as temporal concept drift caused by the influence of real world events , e.g.", ", policy changes , conflicts , or pandemics ."], "relation": "used for", "id": "2022.findings-acl.192", "year": 2022, "rel_sent": "Reframing group - robust algorithms as adaptation algorithms under concept drift , we find that Invariant Risk Minimization and Spectral Decoupling outperform sampling - based approaches to class imbalance and concept drift , and lead to much better performance on minority classes .", "forward": true, "src_ids": "2022.findings-acl.192_1881"}
+{"input": "explanation generation is done by using Method| context: generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation . previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details .", "entity": "explanation generation", "output": "multi - scale distribution deep variational autoencoder", "neg_sample": ["explanation generation is done by using Method", "generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation .", "previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details ."], "relation": "used for", "id": "2022.findings-acl.7", "year": 2022, "rel_sent": "Multi - Scale Distribution Deep Variational Autoencoder for Explanation Generation.", "forward": false, "src_ids": "2022.findings-acl.7_1882"}
+{"input": "multi - scale distribution deep variational autoencoder is used for Task| context: generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation . previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details .", "entity": "multi - scale distribution deep variational autoencoder", "output": "explanation generation", "neg_sample": ["multi - scale distribution deep variational autoencoder is used for Task", "generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation .", "previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details ."], "relation": "used for", "id": "2022.findings-acl.7", "year": 2022, "rel_sent": "Multi - Scale Distribution Deep Variational Autoencoder for Explanation Generation.", "forward": true, "src_ids": "2022.findings-acl.7_1883"}
+{"input": "noise is done by using Method| context: generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation . previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details .", "entity": "noise", "output": "prior network", "neg_sample": ["noise is done by using Method", "generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation .", "previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details ."], "relation": "used for", "id": "2022.findings-acl.7", "year": 2022, "rel_sent": "To resolve this problem , we present Multi - Scale Distribution Deep Variational Autoencoders ( MVAE).These are deep hierarchical VAEs with a prior network that eliminates noise while retaining meaningful signals in the input , coupled with a recognition network serving as the source of information to guide the learning of the prior network .", "forward": false, "src_ids": "2022.findings-acl.7_1884"}
+{"input": "prior network is used for OtherScientificTerm| context: generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation . previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details .", "entity": "prior network", "output": "noise", "neg_sample": ["prior network is used for OtherScientificTerm", "generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation .", "previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details ."], "relation": "used for", "id": "2022.findings-acl.7", "year": 2022, "rel_sent": "To resolve this problem , we present Multi - Scale Distribution Deep Variational Autoencoders ( MVAE).These are deep hierarchical VAEs with a prior network that eliminates noise while retaining meaningful signals in the input , coupled with a recognition network serving as the source of information to guide the learning of the prior network .", "forward": true, "src_ids": "2022.findings-acl.7_1885"}
+{"input": "learning is done by using OtherScientificTerm| context: generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation . previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details .", "entity": "learning", "output": "multi kl divergences", "neg_sample": ["learning is done by using OtherScientificTerm", "generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation .", "previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details ."], "relation": "used for", "id": "2022.findings-acl.7", "year": 2022, "rel_sent": "Further , the Multi - scale distribution Learning Framework ( MLF ) along with a Target Tracking Kullback - Leibler divergence ( TKL ) mechanism are proposed to employ multi KL divergences at different scales for more effective learning .", "forward": false, "src_ids": "2022.findings-acl.7_1886"}
+{"input": "multi kl divergences is used for Task| context: generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation . previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details .", "entity": "multi kl divergences", "output": "learning", "neg_sample": ["multi kl divergences is used for Task", "generating explanations for recommender systems is essential for improving their transparency , as users often wish to understand the reason for receiving a specified recommendation .", "previous methods mainly focus on improving the generation quality , but often produce generic explanations that fail to incorporate user and item specific details ."], "relation": "used for", "id": "2022.findings-acl.7", "year": 2022, "rel_sent": "Further , the Multi - scale distribution Learning Framework ( MLF ) along with a Target Tracking Kullback - Leibler divergence ( TKL ) mechanism are proposed to employ multi KL divergences at different scales for more effective learning .", "forward": true, "src_ids": "2022.findings-acl.7_1887"}
+{"input": "zero - shot scientific fact checking is done by using Task| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "zero - shot scientific fact checking", "output": "generating scientific claims", "neg_sample": ["zero - shot scientific fact checking is done by using Task", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "Generating Scientific Claims for Zero - Shot Scientific Fact Checking.", "forward": false, "src_ids": "2022.acl-long.175_1888"}
+{"input": "atomic and verifiable claims is done by using Task| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "atomic and verifiable claims", "output": "generating scientific claims", "neg_sample": ["atomic and verifiable claims is done by using Task", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "To address this challenge , we propose scientific claim generation , the task of generating one or more atomic and verifiable claims from scientific sentences , and demonstrate its usefulness in zero - shot fact checking for biomedical claims .", "forward": false, "src_ids": "2022.acl-long.175_1889"}
+{"input": "generating scientific claims is used for Task| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "generating scientific claims", "output": "zero - shot scientific fact checking", "neg_sample": ["generating scientific claims is used for Task", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "Generating Scientific Claims for Zero - Shot Scientific Fact Checking.", "forward": true, "src_ids": "2022.acl-long.175_1890"}
+{"input": "generating scientific claims is used for OtherScientificTerm| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "generating scientific claims", "output": "atomic and verifiable claims", "neg_sample": ["generating scientific claims is used for OtherScientificTerm", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "To address this challenge , we propose scientific claim generation , the task of generating one or more atomic and verifiable claims from scientific sentences , and demonstrate its usefulness in zero - shot fact checking for biomedical claims .", "forward": true, "src_ids": "2022.acl-long.175_1891"}
+{"input": "biomedical claims is done by using Task| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "biomedical claims", "output": "zero - shot fact checking", "neg_sample": ["biomedical claims is done by using Task", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "To address this challenge , we propose scientific claim generation , the task of generating one or more atomic and verifiable claims from scientific sentences , and demonstrate its usefulness in zero - shot fact checking for biomedical claims .", "forward": false, "src_ids": "2022.acl-long.175_1892"}
+{"input": "zero - shot fact checking is used for Material| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "zero - shot fact checking", "output": "biomedical claims", "neg_sample": ["zero - shot fact checking is used for Material", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "To address this challenge , we propose scientific claim generation , the task of generating one or more atomic and verifiable claims from scientific sentences , and demonstrate its usefulness in zero - shot fact checking for biomedical claims .", "forward": true, "src_ids": "2022.acl-long.175_1893"}
+{"input": "unsupervised entity - centric method is used for Material| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "unsupervised entity - centric method", "output": "biomedical claims", "neg_sample": ["unsupervised entity - centric method is used for Material", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "Additionally , we adapt an existing unsupervised entity - centric method of claim generation to biomedical claims , which we call CLAIMGEN - ENTITY .", "forward": true, "src_ids": "2022.acl-long.175_1894"}
+{"input": "generating claim negations is done by using Method| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "generating claim negations", "output": "kbin", "neg_sample": ["generating claim negations is done by using Method", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "We propose CLAIMGEN - BART , a new supervised method for generating claims supported by the literature , as well as KBIN , a novel method for generating claim negations .", "forward": false, "src_ids": "2022.acl-long.175_1895"}
+{"input": "kbin is used for Task| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "kbin", "output": "generating claim negations", "neg_sample": ["kbin is used for Task", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "We propose CLAIMGEN - BART , a new supervised method for generating claims supported by the literature , as well as KBIN , a novel method for generating claim negations .", "forward": true, "src_ids": "2022.acl-long.175_1896"}
+{"input": "claim generation is done by using Method| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "claim generation", "output": "unsupervised entity - centric method", "neg_sample": ["claim generation is done by using Method", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "Additionally , we adapt an existing unsupervised entity - centric method of claim generation to biomedical claims , which we call CLAIMGEN - ENTITY .", "forward": false, "src_ids": "2022.acl-long.175_1897"}
+{"input": "biomedical claims is done by using Method| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "biomedical claims", "output": "unsupervised entity - centric method", "neg_sample": ["biomedical claims is done by using Method", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "Additionally , we adapt an existing unsupervised entity - centric method of claim generation to biomedical claims , which we call CLAIMGEN - ENTITY .", "forward": false, "src_ids": "2022.acl-long.175_1898"}
+{"input": "unsupervised entity - centric method is used for Task| context: automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise .", "entity": "unsupervised entity - centric method", "output": "claim generation", "neg_sample": ["unsupervised entity - centric method is used for Task", "automated scientific fact checking is difficult due to the complexity of scientific language and a lack of significant amounts of training data , as annotation requires domain expertise ."], "relation": "used for", "id": "2022.acl-long.175", "year": 2022, "rel_sent": "Additionally , we adapt an existing unsupervised entity - centric method of claim generation to biomedical claims , which we call CLAIMGEN - ENTITY .", "forward": true, "src_ids": "2022.acl-long.175_1899"}
+{"input": "document - level evaluation is done by using OtherScientificTerm| context: this paper presents the results of the dela project .", "entity": "document - level evaluation", "output": "context span", "neg_sample": ["document - level evaluation is done by using OtherScientificTerm", "this paper presents the results of the dela project ."], "relation": "used for", "id": "2022.eamt-1.50", "year": 2022, "rel_sent": "We describe the testing of context span for document - level evaluation , construction of a document - level corpus , and context position , as well as the latest developments of the project when looking at human and automatic evaluation metrics for document - level evaluation .", "forward": false, "src_ids": "2022.eamt-1.50_1900"}
+{"input": "context span is used for Task| context: this paper presents the results of the dela project .", "entity": "context span", "output": "document - level evaluation", "neg_sample": ["context span is used for Task", "this paper presents the results of the dela project ."], "relation": "used for", "id": "2022.eamt-1.50", "year": 2022, "rel_sent": "We describe the testing of context span for document - level evaluation , construction of a document - level corpus , and context position , as well as the latest developments of the project when looking at human and automatic evaluation metrics for document - level evaluation .", "forward": true, "src_ids": "2022.eamt-1.50_1901"}
+{"input": "binary bragging classification is done by using Method| context: bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself . it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "binary bragging classification", "output": "transformer - based models", "neg_sample": ["binary bragging classification is done by using Method", "bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself .", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.273", "year": 2022, "rel_sent": "We empirically evaluate different transformer - based models injected with linguistic information in ( a ) binary bragging classification , i.e. , if tweets contain bragging statements or not ; and ( b ) multi - class bragging type prediction including not bragging .", "forward": false, "src_ids": "2022.acl-long.273_1902"}
+{"input": "multi - class bragging type prediction is done by using Method| context: bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself . it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "multi - class bragging type prediction", "output": "transformer - based models", "neg_sample": ["multi - class bragging type prediction is done by using Method", "bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself .", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.273", "year": 2022, "rel_sent": "We empirically evaluate different transformer - based models injected with linguistic information in ( a ) binary bragging classification , i.e. , if tweets contain bragging statements or not ; and ( b ) multi - class bragging type prediction including not bragging .", "forward": false, "src_ids": "2022.acl-long.273_1903"}
+{"input": "transformer - based models is used for Task| context: bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself . it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "transformer - based models", "output": "binary bragging classification", "neg_sample": ["transformer - based models is used for Task", "bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself .", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.273", "year": 2022, "rel_sent": "We empirically evaluate different transformer - based models injected with linguistic information in ( a ) binary bragging classification , i.e. , if tweets contain bragging statements or not ; and ( b ) multi - class bragging type prediction including not bragging .", "forward": true, "src_ids": "2022.acl-long.273_1904"}
+{"input": "grapheme to phoneme conversion is done by using Method| context: grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields . most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages .", "entity": "grapheme to phoneme conversion", "output": "zero - shot learning", "neg_sample": ["grapheme to phoneme conversion is done by using Method", "grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields .", "most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages ."], "relation": "used for", "id": "2022.findings-acl.166", "year": 2022, "rel_sent": "Zero - shot Learning for Grapheme to Phoneme Conversion with Language Ensemble.", "forward": false, "src_ids": "2022.findings-acl.166_1905"}
+{"input": "approximate g2p models is done by using Method| context: grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields . most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages .", "entity": "approximate g2p models", "output": "zero - shot learning", "neg_sample": ["approximate g2p models is done by using Method", "grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields .", "most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages ."], "relation": "used for", "id": "2022.findings-acl.166", "year": 2022, "rel_sent": "This work attempts to apply zero - shot learning to approximate G2P models for all low - resource and endangered languages in Glottolog ( about 8k languages ) .", "forward": false, "src_ids": "2022.findings-acl.166_1906"}
+{"input": "zero - shot learning is used for Task| context: grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields . most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages .", "entity": "zero - shot learning", "output": "grapheme to phoneme conversion", "neg_sample": ["zero - shot learning is used for Task", "grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields .", "most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages ."], "relation": "used for", "id": "2022.findings-acl.166", "year": 2022, "rel_sent": "Zero - shot Learning for Grapheme to Phoneme Conversion with Language Ensemble.", "forward": true, "src_ids": "2022.findings-acl.166_1907"}
+{"input": "low - resource and endangered languages is done by using Method| context: grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields . most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages .", "entity": "low - resource and endangered languages", "output": "approximate g2p models", "neg_sample": ["low - resource and endangered languages is done by using Method", "grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields .", "most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages ."], "relation": "used for", "id": "2022.findings-acl.166", "year": 2022, "rel_sent": "This work attempts to apply zero - shot learning to approximate G2P models for all low - resource and endangered languages in Glottolog ( about 8k languages ) .", "forward": false, "src_ids": "2022.findings-acl.166_1908"}
+{"input": "zero - shot learning is used for Method| context: grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields . most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages .", "entity": "zero - shot learning", "output": "approximate g2p models", "neg_sample": ["zero - shot learning is used for Method", "grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields .", "most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages ."], "relation": "used for", "id": "2022.findings-acl.166", "year": 2022, "rel_sent": "This work attempts to apply zero - shot learning to approximate G2P models for all low - resource and endangered languages in Glottolog ( about 8k languages ) .", "forward": true, "src_ids": "2022.findings-acl.166_1909"}
+{"input": "approximate g2p models is used for Material| context: grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields . most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages .", "entity": "approximate g2p models", "output": "low - resource and endangered languages", "neg_sample": ["approximate g2p models is used for Material", "grapheme - to - phoneme ( g2p ) has many applications in nlp and speech fields .", "most existing work focuses heavily on languages with abundant training datasets , which limits the scope of target languages to less than 100 languages ."], "relation": "used for", "id": "2022.findings-acl.166", "year": 2022, "rel_sent": "This work attempts to apply zero - shot learning to approximate G2P models for all low - resource and endangered languages in Glottolog ( about 8k languages ) .", "forward": true, "src_ids": "2022.findings-acl.166_1910"}
+{"input": "automatic icd coding is done by using Method| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets . unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd .", "entity": "automatic icd coding", "output": "multiple synonyms matching network", "neg_sample": ["automatic icd coding is done by using Method", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets .", "unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd ."], "relation": "used for", "id": "2022.acl-short.91", "year": 2022, "rel_sent": "Code Synonyms Do Matter : Multiple Synonyms Matching Network for Automatic ICD Coding.", "forward": false, "src_ids": "2022.acl-short.91_1911"}
+{"input": "code representation learning is done by using Method| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets . unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd .", "entity": "code representation learning", "output": "multiple synonyms matching network", "neg_sample": ["code representation learning is done by using Method", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets .", "unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd ."], "relation": "used for", "id": "2022.acl-short.91", "year": 2022, "rel_sent": "Then , we propose a multiple synonyms matching network to leverage synonyms for better code representation learning , and finally help the code classification .", "forward": false, "src_ids": "2022.acl-short.91_1912"}
+{"input": "code classification is done by using Method| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets . unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd .", "entity": "code classification", "output": "multiple synonyms matching network", "neg_sample": ["code classification is done by using Method", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets .", "unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd ."], "relation": "used for", "id": "2022.acl-short.91", "year": 2022, "rel_sent": "Then , we propose a multiple synonyms matching network to leverage synonyms for better code representation learning , and finally help the code classification .", "forward": false, "src_ids": "2022.acl-short.91_1913"}
+{"input": "multiple synonyms matching network is used for Task| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets . unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd .", "entity": "multiple synonyms matching network", "output": "code representation learning", "neg_sample": ["multiple synonyms matching network is used for Task", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets .", "unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd ."], "relation": "used for", "id": "2022.acl-short.91", "year": 2022, "rel_sent": "Then , we propose a multiple synonyms matching network to leverage synonyms for better code representation learning , and finally help the code classification .", "forward": true, "src_ids": "2022.acl-short.91_1914"}
+{"input": "synonyms is used for Task| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets . unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd .", "entity": "synonyms", "output": "code representation learning", "neg_sample": ["synonyms is used for Task", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets .", "unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd ."], "relation": "used for", "id": "2022.acl-short.91", "year": 2022, "rel_sent": "Then , we propose a multiple synonyms matching network to leverage synonyms for better code representation learning , and finally help the code classification .", "forward": true, "src_ids": "2022.acl-short.91_1915"}
+{"input": "multiple synonyms matching network is used for Method| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets . unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd .", "entity": "multiple synonyms matching network", "output": "code classification", "neg_sample": ["multiple synonyms matching network is used for Method", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs).existing methods usually apply label attention with code representations to match related text snippets .", "unlike these works that model the label with the code hierarchy or description , we argue that the code synonyms can provide more comprehensive knowledge based on the observation that the code expressions in emrs vary from their descriptions in icd ."], "relation": "used for", "id": "2022.acl-short.91", "year": 2022, "rel_sent": "Then , we propose a multiple synonyms matching network to leverage synonyms for better code representation learning , and finally help the code classification .", "forward": true, "src_ids": "2022.acl-short.91_1916"}
+{"input": "model interpretability is done by using OtherScientificTerm| context: knowledge graphs are often used to store common sense information that is useful for various tasks . however , the extraction of contextually - relevant knowledge is an unsolved problem , and current approaches are relatively simple .", "entity": "model interpretability", "output": "graph connectivity", "neg_sample": ["model interpretability is done by using OtherScientificTerm", "knowledge graphs are often used to store common sense information that is useful for various tasks .", "however , the extraction of contextually - relevant knowledge is an unsolved problem , and current approaches are relatively simple ."], "relation": "used for", "id": "2022.csrr-1.1", "year": 2022, "rel_sent": "Graph connectivity is important for model interpretability , as paths are frequently used as explanations for the reasoning that connects question and answer .", "forward": false, "src_ids": "2022.csrr-1.1_1917"}
+{"input": "graph connectivity is used for Metric| context: knowledge graphs are often used to store common sense information that is useful for various tasks . however , the extraction of contextually - relevant knowledge is an unsolved problem , and current approaches are relatively simple .", "entity": "graph connectivity", "output": "model interpretability", "neg_sample": ["graph connectivity is used for Metric", "knowledge graphs are often used to store common sense information that is useful for various tasks .", "however , the extraction of contextually - relevant knowledge is an unsolved problem , and current approaches are relatively simple ."], "relation": "used for", "id": "2022.csrr-1.1", "year": 2022, "rel_sent": "Graph connectivity is important for model interpretability , as paths are frequently used as explanations for the reasoning that connects question and answer .", "forward": true, "src_ids": "2022.csrr-1.1_1918"}
+{"input": "multi - source and heterogeneous knowledge is done by using Method| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "multi - source and heterogeneous knowledge", "output": "knowledge graph embedding ( kge ) toolkit", "neg_sample": ["multi - source and heterogeneous knowledge is done by using Method", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "In this paper , we propose CogKGE , a knowledge graph embedding ( KGE ) toolkit , which aims to represent multi - source and heterogeneous knowledge .", "forward": false, "src_ids": "2022.acl-demo.16_1919"}
+{"input": "knowledge graph embedding ( kge ) toolkit is used for OtherScientificTerm| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "knowledge graph embedding ( kge ) toolkit", "output": "multi - source and heterogeneous knowledge", "neg_sample": ["knowledge graph embedding ( kge ) toolkit is used for OtherScientificTerm", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "In this paper , we propose CogKGE , a knowledge graph embedding ( KGE ) toolkit , which aims to represent multi - source and heterogeneous knowledge .", "forward": true, "src_ids": "2022.acl-demo.16_1920"}
+{"input": "cogkge is used for OtherScientificTerm| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "cogkge", "output": "multi - source and heterogeneous knowledge", "neg_sample": ["cogkge is used for OtherScientificTerm", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "In this paper , we propose CogKGE , a knowledge graph embedding ( KGE ) toolkit , which aims to represent multi - source and heterogeneous knowledge .", "forward": true, "src_ids": "2022.acl-demo.16_1921"}
+{"input": "cogkge is used for OtherScientificTerm| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "cogkge", "output": "embeddings", "neg_sample": ["cogkge is used for OtherScientificTerm", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "For heterogeneous knowledge , besides structured triple facts , CogKGE leverages additional unstructured information , such as text descriptions , node types and temporal information , to enhance the meaning of embeddings .", "forward": true, "src_ids": "2022.acl-demo.16_1922"}
+{"input": "kge tasks is done by using Method| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "kge tasks", "output": "unified programming framework", "neg_sample": ["kge tasks is done by using Method", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "Designing CogKGE aims to provide a unified programming framework for KGE tasks and a series of knowledge representations for downstream tasks .", "forward": false, "src_ids": "2022.acl-demo.16_1923"}
+{"input": "cogkge is used for Method| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "cogkge", "output": "unified programming framework", "neg_sample": ["cogkge is used for Method", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "Designing CogKGE aims to provide a unified programming framework for KGE tasks and a series of knowledge representations for downstream tasks .", "forward": true, "src_ids": "2022.acl-demo.16_1924"}
+{"input": "unified programming framework is used for Task| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "unified programming framework", "output": "kge tasks", "neg_sample": ["unified programming framework is used for Task", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "Designing CogKGE aims to provide a unified programming framework for KGE tasks and a series of knowledge representations for downstream tasks .", "forward": true, "src_ids": "2022.acl-demo.16_1925"}
+{"input": "cogkge is used for Task| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "cogkge", "output": "kge tasks", "neg_sample": ["cogkge is used for Task", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "Designing CogKGE aims to provide a unified programming framework for KGE tasks and a series of knowledge representations for downstream tasks .", "forward": true, "src_ids": "2022.acl-demo.16_1926"}
+{"input": "downstream tasks is done by using Method| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "downstream tasks", "output": "knowledge representations", "neg_sample": ["downstream tasks is done by using Method", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "Designing CogKGE aims to provide a unified programming framework for KGE tasks and a series of knowledge representations for downstream tasks .", "forward": false, "src_ids": "2022.acl-demo.16_1927"}
+{"input": "knowledge representations is used for Task| context: for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge .", "entity": "knowledge representations", "output": "downstream tasks", "neg_sample": ["knowledge representations is used for Task", "for multi - source knowledge , unlike existing methods that mainly focus on entity - centric knowledge , cogkge also supports the representations of event - centric , commonsense and linguistic knowledge ."], "relation": "used for", "id": "2022.acl-demo.16", "year": 2022, "rel_sent": "Designing CogKGE aims to provide a unified programming framework for KGE tasks and a series of knowledge representations for downstream tasks .", "forward": true, "src_ids": "2022.acl-demo.16_1928"}
+{"input": "interpreting figurative language types is done by using Method| context: figurative language is ubiquitous in english . yet , the vast majority of nlp research focuses on literal language . existing text representations by design rely on compositionality , while figurative language is often non- compositional .", "entity": "interpreting figurative language types", "output": "human strategies", "neg_sample": ["interpreting figurative language types is done by using Method", "figurative language is ubiquitous in english .", "yet , the vast majority of nlp research focuses on literal language .", "existing text representations by design rely on compositionality , while figurative language is often non- compositional ."], "relation": "used for", "id": "2022.tacl-1.34", "year": 2022, "rel_sent": "We additionally propose knowledge - enhanced models , adopting human strategies for interpreting figurative language types : inferring meaning from the context and relying on the constituent words ' literal meanings .", "forward": false, "src_ids": "2022.tacl-1.34_1929"}
+{"input": "human strategies is used for Task| context: figurative language is ubiquitous in english . yet , the vast majority of nlp research focuses on literal language . existing text representations by design rely on compositionality , while figurative language is often non- compositional .", "entity": "human strategies", "output": "interpreting figurative language types", "neg_sample": ["human strategies is used for Task", "figurative language is ubiquitous in english .", "yet , the vast majority of nlp research focuses on literal language .", "existing text representations by design rely on compositionality , while figurative language is often non- compositional ."], "relation": "used for", "id": "2022.tacl-1.34", "year": 2022, "rel_sent": "We additionally propose knowledge - enhanced models , adopting human strategies for interpreting figurative language types : inferring meaning from the context and relying on the constituent words ' literal meanings .", "forward": true, "src_ids": "2022.tacl-1.34_1930"}
+{"input": "children 's story books is done by using Task| context: existing question answering ( qa ) techniques are created mainly to answer questions asked by humans . but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities .", "entity": "children 's story books", "output": "question - answer pair generation", "neg_sample": ["children 's story books is done by using Task", "existing question answering ( qa ) techniques are created mainly to answer questions asked by humans .", "but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities ."], "relation": "used for", "id": "2022.acl-long.54", "year": 2022, "rel_sent": "It is AI 's Turn to Ask Humans a Question : Question - Answer Pair Generation for Children 's Story Books.", "forward": false, "src_ids": "2022.acl-long.54_1931"}
+{"input": "question - answer pair generation is used for Material| context: existing question answering ( qa ) techniques are created mainly to answer questions asked by humans . but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities .", "entity": "question - answer pair generation", "output": "children 's story books", "neg_sample": ["question - answer pair generation is used for Material", "existing question answering ( qa ) techniques are created mainly to answer questions asked by humans .", "but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities ."], "relation": "used for", "id": "2022.acl-long.54", "year": 2022, "rel_sent": "It is AI 's Turn to Ask Humans a Question : Question - Answer Pair Generation for Children 's Story Books.", "forward": true, "src_ids": "2022.acl-long.54_1932"}
+{"input": "interactive story - telling application is done by using Method| context: existing question answering ( qa ) techniques are created mainly to answer questions asked by humans . but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities .", "entity": "interactive story - telling application", "output": "qag system", "neg_sample": ["interactive story - telling application is done by using Method", "existing question answering ( qa ) techniques are created mainly to answer questions asked by humans .", "but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities ."], "relation": "used for", "id": "2022.acl-long.54", "year": 2022, "rel_sent": "On top of our QAG system , we also start to build an interactive story - telling application for the future real - world deployment in this educational scenario .", "forward": false, "src_ids": "2022.acl-long.54_1933"}
+{"input": "educational scenario is done by using Task| context: existing question answering ( qa ) techniques are created mainly to answer questions asked by humans . but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities .", "entity": "educational scenario", "output": "interactive story - telling application", "neg_sample": ["educational scenario is done by using Task", "existing question answering ( qa ) techniques are created mainly to answer questions asked by humans .", "but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities ."], "relation": "used for", "id": "2022.acl-long.54", "year": 2022, "rel_sent": "On top of our QAG system , we also start to build an interactive story - telling application for the future real - world deployment in this educational scenario .", "forward": false, "src_ids": "2022.acl-long.54_1934"}
+{"input": "qag system is used for Task| context: existing question answering ( qa ) techniques are created mainly to answer questions asked by humans . but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities .", "entity": "qag system", "output": "interactive story - telling application", "neg_sample": ["qag system is used for Task", "existing question answering ( qa ) techniques are created mainly to answer questions asked by humans .", "but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities ."], "relation": "used for", "id": "2022.acl-long.54", "year": 2022, "rel_sent": "On top of our QAG system , we also start to build an interactive story - telling application for the future real - world deployment in this educational scenario .", "forward": true, "src_ids": "2022.acl-long.54_1935"}
+{"input": "interactive story - telling application is used for Material| context: existing question answering ( qa ) techniques are created mainly to answer questions asked by humans . but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities .", "entity": "interactive story - telling application", "output": "educational scenario", "neg_sample": ["interactive story - telling application is used for Material", "existing question answering ( qa ) techniques are created mainly to answer questions asked by humans .", "but in educational applications , teachers often need to decide what questions they should ask , in order to help students to improve their narrative understanding capabilities ."], "relation": "used for", "id": "2022.acl-long.54", "year": 2022, "rel_sent": "On top of our QAG system , we also start to build an interactive story - telling application for the future real - world deployment in this educational scenario .", "forward": true, "src_ids": "2022.acl-long.54_1936"}
+{"input": "nlp is done by using Method| context: although deep neural networks have achieved state - of - the - art performance in various machine learning tasks , adversarial examples , constructed by adding small non - random perturbations to correctly classified inputs , successfully fool highly expressive deep classifiers into incorrect predictions . approaches to adversarial attacks in natural language tasks have boomed in the last five years using character - level , word - level , phrase - level , or sentence - level textual perturbations . while there is some work in nlp on defending against such attacks through proactive methods , like adversarial training , there is to our knowledge no effective general reactive approaches to defence via detection of textual adversarial examples such as is found in the image processing literature .", "entity": "nlp", "output": "reactive methods", "neg_sample": ["nlp is done by using Method", "although deep neural networks have achieved state - of - the - art performance in various machine learning tasks , adversarial examples , constructed by adding small non - random perturbations to correctly classified inputs , successfully fool highly expressive deep classifiers into incorrect predictions .", "approaches to adversarial attacks in natural language tasks have boomed in the last five years using character - level , word - level , phrase - level , or sentence - level textual perturbations .", "while there is some work in nlp on defending against such attacks through proactive methods , like adversarial training , there is to our knowledge no effective general reactive approaches to defence via detection of textual adversarial examples such as is found in the image processing literature ."], "relation": "used for", "id": "2022.repl4nlp-1.9", "year": 2022, "rel_sent": "In this paper , we propose two new reactive methods for NLP tofill this gap , which unlike the few limited application baselines from NLP are based entirely on distribution characteristics of learned representations ' : ' we adapt one from the image processing literature ( Local Intrinsic Dimensionality ( LID ) ) , and propose a novel one ( MultiDistance Representation Ensemble Method ( MDRE ) ) .", "forward": false, "src_ids": "2022.repl4nlp-1.9_1937"}
+{"input": "reactive methods is used for Task| context: although deep neural networks have achieved state - of - the - art performance in various machine learning tasks , adversarial examples , constructed by adding small non - random perturbations to correctly classified inputs , successfully fool highly expressive deep classifiers into incorrect predictions . approaches to adversarial attacks in natural language tasks have boomed in the last five years using character - level , word - level , phrase - level , or sentence - level textual perturbations .", "entity": "reactive methods", "output": "nlp", "neg_sample": ["reactive methods is used for Task", "although deep neural networks have achieved state - of - the - art performance in various machine learning tasks , adversarial examples , constructed by adding small non - random perturbations to correctly classified inputs , successfully fool highly expressive deep classifiers into incorrect predictions .", "approaches to adversarial attacks in natural language tasks have boomed in the last five years using character - level , word - level , phrase - level , or sentence - level textual perturbations ."], "relation": "used for", "id": "2022.repl4nlp-1.9", "year": 2022, "rel_sent": "In this paper , we propose two new reactive methods for NLP tofill this gap , which unlike the few limited application baselines from NLP are based entirely on distribution characteristics of learned representations ' : ' we adapt one from the image processing literature ( Local Intrinsic Dimensionality ( LID ) ) , and propose a novel one ( MultiDistance Representation Ensemble Method ( MDRE ) ) .", "forward": true, "src_ids": "2022.repl4nlp-1.9_1938"}
+{"input": "modalities is done by using OtherScientificTerm| context: recent research has made impressive progress in large - scale multimodal pre - training . in the context of the rapid growth of model size , it is necessary to seek efficient and flexible methods other than finetuning .", "entity": "modalities", "output": "prompt vectors", "neg_sample": ["modalities is done by using OtherScientificTerm", "recent research has made impressive progress in large - scale multimodal pre - training .", "in the context of the rapid growth of model size , it is necessary to seek efficient and flexible methods other than finetuning ."], "relation": "used for", "id": "2022.findings-acl.234", "year": 2022, "rel_sent": "In this paper , we propose to use prompt vectors to align the modalities .", "forward": false, "src_ids": "2022.findings-acl.234_1939"}
+{"input": "prompt vectors is used for OtherScientificTerm| context: recent research has made impressive progress in large - scale multimodal pre - training . in the context of the rapid growth of model size , it is necessary to seek efficient and flexible methods other than finetuning .", "entity": "prompt vectors", "output": "modalities", "neg_sample": ["prompt vectors is used for OtherScientificTerm", "recent research has made impressive progress in large - scale multimodal pre - training .", "in the context of the rapid growth of model size , it is necessary to seek efficient and flexible methods other than finetuning ."], "relation": "used for", "id": "2022.findings-acl.234", "year": 2022, "rel_sent": "In this paper , we propose to use prompt vectors to align the modalities .", "forward": true, "src_ids": "2022.findings-acl.234_1940"}
+{"input": "multilingual customer support is done by using Method| context: we describe the multilingual customer solution by language i / o in this paper . with data security and confidentiality ensured by the iso 27001 certification , global corporations can turn monolingual customer support agents into efficient multilingual brand ambassadors in less than 24 hours . our solution supports more than 100 languages .", "entity": "multilingual customer support", "output": "language i / o solution", "neg_sample": ["multilingual customer support is done by using Method", "we describe the multilingual customer solution by language i / o in this paper .", "with data security and confidentiality ensured by the iso 27001 certification , global corporations can turn monolingual customer support agents into efficient multilingual brand ambassadors in less than 24 hours .", "our solution supports more than 100 languages ."], "relation": "used for", "id": "2022.eamt-1.32", "year": 2022, "rel_sent": "Language I / O Solution for Multilingual Customer Support.", "forward": false, "src_ids": "2022.eamt-1.32_1941"}
+{"input": "language i / o solution is used for Task| context: we describe the multilingual customer solution by language i / o in this paper . with data security and confidentiality ensured by the iso 27001 certification , global corporations can turn monolingual customer support agents into efficient multilingual brand ambassadors in less than 24 hours . our solution supports more than 100 languages .", "entity": "language i / o solution", "output": "multilingual customer support", "neg_sample": ["language i / o solution is used for Task", "we describe the multilingual customer solution by language i / o in this paper .", "with data security and confidentiality ensured by the iso 27001 certification , global corporations can turn monolingual customer support agents into efficient multilingual brand ambassadors in less than 24 hours .", "our solution supports more than 100 languages ."], "relation": "used for", "id": "2022.eamt-1.32", "year": 2022, "rel_sent": "Language I / O Solution for Multilingual Customer Support.", "forward": true, "src_ids": "2022.eamt-1.32_1942"}
+{"input": "equality is done by using Task| context: in recent years social media has become one of the major forums for expressing human views and emotions . with the help of smartphones and high - speed internet , anyone can express their views on social media . however , this can also lead to the spread of hatred and violence in society . therefore it is necessary to build a method tofind and support helpful social media content .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "in recent years social media has become one of the major forums for expressing human views and emotions .", "with the help of smartphones and high - speed internet , anyone can express their views on social media .", "however , this can also lead to the spread of hatred and violence in society .", "therefore it is necessary to build a method tofind and support helpful social media content ."], "relation": "used for", "id": "2022.ltedi-1.22", "year": 2022, "rel_sent": "SSN_ARMM@ LT - EDI -ACL2022 : Hope Speech Detection for Equality , Diversity , and Inclusion Using ALBERT model.", "forward": false, "src_ids": "2022.ltedi-1.22_1943"}
+{"input": "diversity is done by using Task| context: in recent years social media has become one of the major forums for expressing human views and emotions . with the help of smartphones and high - speed internet , anyone can express their views on social media . however , this can also lead to the spread of hatred and violence in society . therefore it is necessary to build a method tofind and support helpful social media content .", "entity": "diversity", "output": "hope speech detection", "neg_sample": ["diversity is done by using Task", "in recent years social media has become one of the major forums for expressing human views and emotions .", "with the help of smartphones and high - speed internet , anyone can express their views on social media .", "however , this can also lead to the spread of hatred and violence in society .", "therefore it is necessary to build a method tofind and support helpful social media content ."], "relation": "used for", "id": "2022.ltedi-1.22", "year": 2022, "rel_sent": "SSN_ARMM@ LT - EDI -ACL2022 : Hope Speech Detection for Equality , Diversity , and Inclusion Using ALBERT model.", "forward": false, "src_ids": "2022.ltedi-1.22_1944"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: in recent years social media has become one of the major forums for expressing human views and emotions . with the help of smartphones and high - speed internet , anyone can express their views on social media . however , this can also lead to the spread of hatred and violence in society . therefore it is necessary to build a method tofind and support helpful social media content .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "in recent years social media has become one of the major forums for expressing human views and emotions .", "with the help of smartphones and high - speed internet , anyone can express their views on social media .", "however , this can also lead to the spread of hatred and violence in society .", "therefore it is necessary to build a method tofind and support helpful social media content ."], "relation": "used for", "id": "2022.ltedi-1.22", "year": 2022, "rel_sent": "SSN_ARMM@ LT - EDI -ACL2022 : Hope Speech Detection for Equality , Diversity , and Inclusion Using ALBERT model.", "forward": true, "src_ids": "2022.ltedi-1.22_1945"}
+{"input": "hope speech is done by using Method| context: in recent years social media has become one of the major forums for expressing human views and emotions . with the help of smartphones and high - speed internet , anyone can express their views on social media . however , this can also lead to the spread of hatred and violence in society . therefore it is necessary to build a method tofind and support helpful social media content .", "entity": "hope speech", "output": "natural language processing approach", "neg_sample": ["hope speech is done by using Method", "in recent years social media has become one of the major forums for expressing human views and emotions .", "with the help of smartphones and high - speed internet , anyone can express their views on social media .", "however , this can also lead to the spread of hatred and violence in society .", "therefore it is necessary to build a method tofind and support helpful social media content ."], "relation": "used for", "id": "2022.ltedi-1.22", "year": 2022, "rel_sent": "In this paper , we studied Natural Language Processing approach for detecting Hope speech in a given sentence .", "forward": false, "src_ids": "2022.ltedi-1.22_1946"}
+{"input": "natural language processing approach is used for OtherScientificTerm| context: in recent years social media has become one of the major forums for expressing human views and emotions . with the help of smartphones and high - speed internet , anyone can express their views on social media . however , this can also lead to the spread of hatred and violence in society . therefore it is necessary to build a method tofind and support helpful social media content .", "entity": "natural language processing approach", "output": "hope speech", "neg_sample": ["natural language processing approach is used for OtherScientificTerm", "in recent years social media has become one of the major forums for expressing human views and emotions .", "with the help of smartphones and high - speed internet , anyone can express their views on social media .", "however , this can also lead to the spread of hatred and violence in society .", "therefore it is necessary to build a method tofind and support helpful social media content ."], "relation": "used for", "id": "2022.ltedi-1.22", "year": 2022, "rel_sent": "In this paper , we studied Natural Language Processing approach for detecting Hope speech in a given sentence .", "forward": true, "src_ids": "2022.ltedi-1.22_1947"}
+{"input": "reading tasks is done by using OtherScientificTerm| context: attention describes cognitive processes that are important to many human phenomena including reading . the term is also used to describe the way in which transformer neural networks perform natural language processing .", "entity": "reading tasks", "output": "overt human attention", "neg_sample": ["reading tasks is done by using OtherScientificTerm", "attention describes cognitive processes that are important to many human phenomena including reading .", "the term is also used to describe the way in which transformer neural networks perform natural language processing ."], "relation": "used for", "id": "2022.cmcl-1.9", "year": 2022, "rel_sent": "While attention appears to be very different under these two contexts , this paper presents an analysis of the correlations between transformer attention and overt human attention during reading tasks .", "forward": false, "src_ids": "2022.cmcl-1.9_1948"}
+{"input": "overt human attention is used for Task| context: attention describes cognitive processes that are important to many human phenomena including reading . the term is also used to describe the way in which transformer neural networks perform natural language processing .", "entity": "overt human attention", "output": "reading tasks", "neg_sample": ["overt human attention is used for Task", "attention describes cognitive processes that are important to many human phenomena including reading .", "the term is also used to describe the way in which transformer neural networks perform natural language processing ."], "relation": "used for", "id": "2022.cmcl-1.9", "year": 2022, "rel_sent": "While attention appears to be very different under these two contexts , this paper presents an analysis of the correlations between transformer attention and overt human attention during reading tasks .", "forward": true, "src_ids": "2022.cmcl-1.9_1949"}
+{"input": "deep - learning sequence labelling model is done by using Method| context: biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining . most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets .", "entity": "deep - learning sequence labelling model", "output": "distant supervision", "neg_sample": ["deep - learning sequence labelling model is done by using Method", "biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining .", "most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets ."], "relation": "used for", "id": "2022.bionlp-1.15", "year": 2022, "rel_sent": "In this paper , we introduce a novel annotation schema to capture complex entities , and explore the effects of distant supervision on our deep - learning sequence labelling model .", "forward": false, "src_ids": "2022.bionlp-1.15_1950"}
+{"input": "complex entities is done by using Method| context: biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining . most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets .", "entity": "complex entities", "output": "annotation schema", "neg_sample": ["complex entities is done by using Method", "biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining .", "most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets ."], "relation": "used for", "id": "2022.bionlp-1.15", "year": 2022, "rel_sent": "In this paper , we introduce a novel annotation schema to capture complex entities , and explore the effects of distant supervision on our deep - learning sequence labelling model .", "forward": false, "src_ids": "2022.bionlp-1.15_1951"}
+{"input": "annotation schema is used for OtherScientificTerm| context: biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining . most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets .", "entity": "annotation schema", "output": "complex entities", "neg_sample": ["annotation schema is used for OtherScientificTerm", "biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining .", "most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets ."], "relation": "used for", "id": "2022.bionlp-1.15", "year": 2022, "rel_sent": "In this paper , we introduce a novel annotation schema to capture complex entities , and explore the effects of distant supervision on our deep - learning sequence labelling model .", "forward": true, "src_ids": "2022.bionlp-1.15_1952"}
+{"input": "distant supervision is used for Method| context: biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining . most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets .", "entity": "distant supervision", "output": "deep - learning sequence labelling model", "neg_sample": ["distant supervision is used for Method", "biomedical named entity recognition ( bmner ) is one of the most important tasks in the field of biomedical text mining .", "most work sofar on this task has not focused on identification of discontinuous and overlapping entities , even though they are present in significant fractions in real - life biomedical datasets ."], "relation": "used for", "id": "2022.bionlp-1.15", "year": 2022, "rel_sent": "In this paper , we introduce a novel annotation schema to capture complex entities , and explore the effects of distant supervision on our deep - learning sequence labelling model .", "forward": true, "src_ids": "2022.bionlp-1.15_1953"}
+{"input": "cross - lingual ner is done by using Method| context: cross - lingual named entity recognition task is one of the critical problems for evaluating the potential transfer learning techniques on low resource languages . knowledge distillation using pre - trained multilingual language models between source and target languages have shown their superiority in transfer . however , existing cross - lingual distillation models merely consider the potential transferability between two identical single tasks across both domains . other possible auxiliary tasks to improve the learning performance have not been fully investigated .", "entity": "cross - lingual ner", "output": "similarity metric model", "neg_sample": ["cross - lingual ner is done by using Method", "cross - lingual named entity recognition task is one of the critical problems for evaluating the potential transfer learning techniques on low resource languages .", "knowledge distillation using pre - trained multilingual language models between source and target languages have shown their superiority in transfer .", "however , existing cross - lingual distillation models merely consider the potential transferability between two identical single tasks across both domains .", "other possible auxiliary tasks to improve the learning performance have not been fully investigated ."], "relation": "used for", "id": "2022.acl-long.14", "year": 2022, "rel_sent": "In this study , based on the knowledge distillation framework and multi - task learning , we introduce the similarity metric model as an auxiliary task to improve the cross - lingual NER performance on the target domain .", "forward": false, "src_ids": "2022.acl-long.14_1954"}
+{"input": "similarity metric model is used for Task| context: cross - lingual named entity recognition task is one of the critical problems for evaluating the potential transfer learning techniques on low resource languages . knowledge distillation using pre - trained multilingual language models between source and target languages have shown their superiority in transfer . however , existing cross - lingual distillation models merely consider the potential transferability between two identical single tasks across both domains . other possible auxiliary tasks to improve the learning performance have not been fully investigated .", "entity": "similarity metric model", "output": "cross - lingual ner", "neg_sample": ["similarity metric model is used for Task", "cross - lingual named entity recognition task is one of the critical problems for evaluating the potential transfer learning techniques on low resource languages .", "knowledge distillation using pre - trained multilingual language models between source and target languages have shown their superiority in transfer .", "however , existing cross - lingual distillation models merely consider the potential transferability between two identical single tasks across both domains .", "other possible auxiliary tasks to improve the learning performance have not been fully investigated ."], "relation": "used for", "id": "2022.acl-long.14", "year": 2022, "rel_sent": "In this study , based on the knowledge distillation framework and multi - task learning , we introduce the similarity metric model as an auxiliary task to improve the cross - lingual NER performance on the target domain .", "forward": true, "src_ids": "2022.acl-long.14_1955"}
+{"input": "conversational emotion recognition is done by using Method| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "conversational emotion recognition", "output": "emotion capsule based model", "neg_sample": ["conversational emotion recognition is done by using Method", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "EmoCaps : Emotion Capsule based Model for Conversational Emotion Recognition.", "forward": false, "src_ids": "2022.findings-acl.126_1956"}
+{"input": "emotion capsule based model is used for Task| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "emotion capsule based model", "output": "conversational emotion recognition", "neg_sample": ["emotion capsule based model is used for Task", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "EmoCaps : Emotion Capsule based Model for Conversational Emotion Recognition.", "forward": true, "src_ids": "2022.findings-acl.126_1957"}
+{"input": "multi - modal emotion vectors is done by using OtherScientificTerm| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "multi - modal emotion vectors", "output": "emoformer", "neg_sample": ["multi - modal emotion vectors is done by using OtherScientificTerm", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "In order to extract multi - modal information and the emotional tendency of the utterance effectively , we propose a new structure named Emoformer to extract multi - modal emotion vectors from different modalities and fuse them with sentence vector to be an emotion capsule .", "forward": false, "src_ids": "2022.findings-acl.126_1958"}
+{"input": "emoformer is used for OtherScientificTerm| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "emoformer", "output": "multi - modal emotion vectors", "neg_sample": ["emoformer is used for OtherScientificTerm", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "In order to extract multi - modal information and the emotional tendency of the utterance effectively , we propose a new structure named Emoformer to extract multi - modal emotion vectors from different modalities and fuse them with sentence vector to be an emotion capsule .", "forward": true, "src_ids": "2022.findings-acl.126_1959"}
+{"input": "emotion capsule is done by using OtherScientificTerm| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "emotion capsule", "output": "sentence vector", "neg_sample": ["emotion capsule is done by using OtherScientificTerm", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "In order to extract multi - modal information and the emotional tendency of the utterance effectively , we propose a new structure named Emoformer to extract multi - modal emotion vectors from different modalities and fuse them with sentence vector to be an emotion capsule .", "forward": false, "src_ids": "2022.findings-acl.126_1960"}
+{"input": "sentence vector is used for OtherScientificTerm| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "sentence vector", "output": "emotion capsule", "neg_sample": ["sentence vector is used for OtherScientificTerm", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "In order to extract multi - modal information and the emotional tendency of the utterance effectively , we propose a new structure named Emoformer to extract multi - modal emotion vectors from different modalities and fuse them with sentence vector to be an emotion capsule .", "forward": true, "src_ids": "2022.findings-acl.126_1961"}
+{"input": "emotion vectors is done by using Method| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "emotion vectors", "output": "end - to - end erc model", "neg_sample": ["emotion vectors is done by using Method", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "Furthermore , we design an end - to - end ERC model called EmoCaps , which extracts emotion vectors through the Emoformer structure and obtain the emotion classification results from a context analysis model .", "forward": false, "src_ids": "2022.findings-acl.126_1962"}
+{"input": "end - to - end erc model is used for OtherScientificTerm| context: recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency .", "entity": "end - to - end erc model", "output": "emotion vectors", "neg_sample": ["end - to - end erc model is used for OtherScientificTerm", "recent works in erc focus on context modeling but ignore the representation of contextual emotional tendency ."], "relation": "used for", "id": "2022.findings-acl.126", "year": 2022, "rel_sent": "Furthermore , we design an end - to - end ERC model called EmoCaps , which extracts emotion vectors through the Emoformer structure and obtain the emotion classification results from a context analysis model .", "forward": true, "src_ids": "2022.findings-acl.126_1963"}
+{"input": "scrambled domains is done by using Method| context: there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training . we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling .", "entity": "scrambled domains", "output": "bert", "neg_sample": ["scrambled domains is done by using Method", "there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training .", "we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling ."], "relation": "used for", "id": "2022.repl4nlp-1.11", "year": 2022, "rel_sent": "Among these models , we find that only BERT shows high rates of transfer into our scrambled domains , and for classification but not sequence labeling tasks .", "forward": false, "src_ids": "2022.repl4nlp-1.11_1964"}
+{"input": "classification but not sequence labeling tasks is done by using Method| context: there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training . we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling .", "entity": "classification but not sequence labeling tasks", "output": "bert", "neg_sample": ["classification but not sequence labeling tasks is done by using Method", "there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training .", "we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling ."], "relation": "used for", "id": "2022.repl4nlp-1.11", "year": 2022, "rel_sent": "Among these models , we find that only BERT shows high rates of transfer into our scrambled domains , and for classification but not sequence labeling tasks .", "forward": false, "src_ids": "2022.repl4nlp-1.11_1965"}
+{"input": "bert is used for Material| context: there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training . we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling .", "entity": "bert", "output": "scrambled domains", "neg_sample": ["bert is used for Material", "there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training .", "we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling ."], "relation": "used for", "id": "2022.repl4nlp-1.11", "year": 2022, "rel_sent": "Among these models , we find that only BERT shows high rates of transfer into our scrambled domains , and for classification but not sequence labeling tasks .", "forward": true, "src_ids": "2022.repl4nlp-1.11_1966"}
+{"input": "bert is used for Task| context: there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training . we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling .", "entity": "bert", "output": "classification but not sequence labeling tasks", "neg_sample": ["bert is used for Task", "there is growing evidence that pretrained language models improve task - specific fine - tuning even where the task examples are radically different from those seen in training .", "we study an extreme case of transfer learning by providing a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling ."], "relation": "used for", "id": "2022.repl4nlp-1.11", "year": 2022, "rel_sent": "Among these models , we find that only BERT shows high rates of transfer into our scrambled domains , and for classification but not sequence labeling tasks .", "forward": true, "src_ids": "2022.repl4nlp-1.11_1967"}
+{"input": "emotions is done by using Method| context: we participated in track 2 for predicting emotion at the essay level .", "entity": "emotions", "output": "ensemble approach", "neg_sample": ["emotions is done by using Method", "we participated in track 2 for predicting emotion at the essay level ."], "relation": "used for", "id": "2022.wassa-1.30", "year": 2022, "rel_sent": "An Ensemble Approach to Detect Emotions at an Essay Level.", "forward": false, "src_ids": "2022.wassa-1.30_1968"}
+{"input": "ensemble approach is used for OtherScientificTerm| context: we participated in track 2 for predicting emotion at the essay level .", "entity": "ensemble approach", "output": "emotions", "neg_sample": ["ensemble approach is used for OtherScientificTerm", "we participated in track 2 for predicting emotion at the essay level ."], "relation": "used for", "id": "2022.wassa-1.30", "year": 2022, "rel_sent": "An Ensemble Approach to Detect Emotions at an Essay Level.", "forward": true, "src_ids": "2022.wassa-1.30_1969"}
+{"input": "computational approaches is used for Task| context: we participated in track 2 for predicting emotion at the essay level .", "entity": "computational approaches", "output": "sentiment & social media analysis", "neg_sample": ["computational approaches is used for Task", "we participated in track 2 for predicting emotion at the essay level ."], "relation": "used for", "id": "2022.wassa-1.30", "year": 2022, "rel_sent": "This paper describes our system ( IREL , reffered as himanshu.1007 on Codalab ) for Shared Task on Empathy Detection , Emotion Classification , and Personality Detection at 12th Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis at ACL 2022 .", "forward": true, "src_ids": "2022.wassa-1.30_1970"}
+{"input": "sentiment & social media analysis is done by using Method| context: we participated in track 2 for predicting emotion at the essay level .", "entity": "sentiment & social media analysis", "output": "computational approaches", "neg_sample": ["sentiment & social media analysis is done by using Method", "we participated in track 2 for predicting emotion at the essay level ."], "relation": "used for", "id": "2022.wassa-1.30", "year": 2022, "rel_sent": "This paper describes our system ( IREL , reffered as himanshu.1007 on Codalab ) for Shared Task on Empathy Detection , Emotion Classification , and Personality Detection at 12th Workshop on Computational Approaches to Subjectivity , Sentiment & Social Media Analysis at ACL 2022 .", "forward": false, "src_ids": "2022.wassa-1.30_1971"}
+{"input": "sign language learners is done by using Material| context: a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words . searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language . however , this is not true for sign language , where gestural elements preclude this type of easy lookup .", "entity": "sign language learners", "output": "automatic gloss dictionary", "neg_sample": ["sign language learners is done by using Material", "a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words .", "searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language .", "however , this is not true for sign language , where gestural elements preclude this type of easy lookup ."], "relation": "used for", "id": "2022.acl-demo.8", "year": 2022, "rel_sent": "Automatic Gloss Dictionary for Sign Language Learners.", "forward": false, "src_ids": "2022.acl-demo.8_1972"}
+{"input": "automatic gloss dictionary is used for OtherScientificTerm| context: a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words . searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language . however , this is not true for sign language , where gestural elements preclude this type of easy lookup .", "entity": "automatic gloss dictionary", "output": "sign language learners", "neg_sample": ["automatic gloss dictionary is used for OtherScientificTerm", "a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words .", "searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language .", "however , this is not true for sign language , where gestural elements preclude this type of easy lookup ."], "relation": "used for", "id": "2022.acl-demo.8", "year": 2022, "rel_sent": "Automatic Gloss Dictionary for Sign Language Learners.", "forward": true, "src_ids": "2022.acl-demo.8_1973"}
+{"input": "sign language learning is done by using Method| context: a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words . searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language . however , this is not true for sign language , where gestural elements preclude this type of easy lookup .", "entity": "sign language learning", "output": "glossfinder", "neg_sample": ["sign language learning is done by using Method", "a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words .", "searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language .", "however , this is not true for sign language , where gestural elements preclude this type of easy lookup ."], "relation": "used for", "id": "2022.acl-demo.8", "year": 2022, "rel_sent": "This implies that GlossFinder can lower the barrier in sign language learning by addressing the common problem of sign finding and make it accessible to the wider community .", "forward": false, "src_ids": "2022.acl-demo.8_1974"}
+{"input": "language learners is done by using Generic| context: a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words . searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language . however , this is not true for sign language , where gestural elements preclude this type of easy lookup .", "entity": "language learners", "output": "online tool", "neg_sample": ["language learners is done by using Generic", "a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words .", "searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language .", "however , this is not true for sign language , where gestural elements preclude this type of easy lookup ."], "relation": "used for", "id": "2022.acl-demo.8", "year": 2022, "rel_sent": "This paper introduces GlossFinder , an online tool supporting 2 , 000 signs to assist language learners in determining the meaning of given signs .", "forward": false, "src_ids": "2022.acl-demo.8_1975"}
+{"input": "online tool is used for OtherScientificTerm| context: a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words . searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language . however , this is not true for sign language , where gestural elements preclude this type of easy lookup .", "entity": "online tool", "output": "language learners", "neg_sample": ["online tool is used for OtherScientificTerm", "a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words .", "searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language .", "however , this is not true for sign language , where gestural elements preclude this type of easy lookup ."], "relation": "used for", "id": "2022.acl-demo.8", "year": 2022, "rel_sent": "This paper introduces GlossFinder , an online tool supporting 2 , 000 signs to assist language learners in determining the meaning of given signs .", "forward": true, "src_ids": "2022.acl-demo.8_1976"}
+{"input": "glossfinder is used for Task| context: a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words . searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language . however , this is not true for sign language , where gestural elements preclude this type of easy lookup .", "entity": "glossfinder", "output": "sign language learning", "neg_sample": ["glossfinder is used for Task", "a multi - language dictionary is a fundamental tool for language learning , allowing the learner to look up unfamiliar words .", "searching an unrecognized word in the dictionary does not usually require deep knowledge of the target language .", "however , this is not true for sign language , where gestural elements preclude this type of easy lookup ."], "relation": "used for", "id": "2022.acl-demo.8", "year": 2022, "rel_sent": "This implies that GlossFinder can lower the barrier in sign language learning by addressing the common problem of sign finding and make it accessible to the wider community .", "forward": true, "src_ids": "2022.acl-demo.8_1977"}
+{"input": "multilingual open information extraction is done by using Method| context: progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages .", "entity": "multilingual open information extraction", "output": "alignment - augmented consistent translation", "neg_sample": ["multilingual open information extraction is done by using Method", "progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages ."], "relation": "used for", "id": "2022.acl-long.179", "year": 2022, "rel_sent": "Alignment - Augmented Consistent Translation for Multilingual Open Information Extraction.", "forward": false, "src_ids": "2022.acl-long.179_1978"}
+{"input": "alignment - augmented consistent translation is used for Task| context: progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages .", "entity": "alignment - augmented consistent translation", "output": "multilingual open information extraction", "neg_sample": ["alignment - augmented consistent translation is used for Task", "progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages ."], "relation": "used for", "id": "2022.acl-long.179", "year": 2022, "rel_sent": "Alignment - Augmented Consistent Translation for Multilingual Open Information Extraction.", "forward": true, "src_ids": "2022.acl-long.179_1979"}
+{"input": "openie systems is done by using Material| context: progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages .", "entity": "openie systems", "output": "english text", "neg_sample": ["openie systems is done by using Material", "progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages ."], "relation": "used for", "id": "2022.acl-long.179", "year": 2022, "rel_sent": "In this paper , we explore techniques to automatically convert English text for training OpenIE systems in other languages .", "forward": false, "src_ids": "2022.acl-long.179_1980"}
+{"input": "english text is used for Method| context: progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages .", "entity": "english text", "output": "openie systems", "neg_sample": ["english text is used for Method", "progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages ."], "relation": "used for", "id": "2022.acl-long.179", "year": 2022, "rel_sent": "In this paper , we explore techniques to automatically convert English text for training OpenIE systems in other languages .", "forward": true, "src_ids": "2022.acl-long.179_1981"}
+{"input": "two - stage generative openie model is done by using Method| context: progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages .", "entity": "two - stage generative openie model", "output": "alignment - augmented constrained translation", "neg_sample": ["two - stage generative openie model is done by using Method", "progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages ."], "relation": "used for", "id": "2022.acl-long.179", "year": 2022, "rel_sent": "Using the data generated with AACTrans , we train a novel two - stage generative OpenIE model , which we call Gen2OIE , that outputs for each sentence : 1 ) relations in the first stage and 2 ) all extractions containing the relation in the second stage .", "forward": false, "src_ids": "2022.acl-long.179_1982"}
+{"input": "alignment - augmented constrained translation is used for Method| context: progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages .", "entity": "alignment - augmented constrained translation", "output": "two - stage generative openie model", "neg_sample": ["alignment - augmented constrained translation is used for Method", "progress with supervised open information extraction ( openie ) has been primarily limited to english due to the scarcity of training data in other languages ."], "relation": "used for", "id": "2022.acl-long.179", "year": 2022, "rel_sent": "Using the data generated with AACTrans , we train a novel two - stage generative OpenIE model , which we call Gen2OIE , that outputs for each sentence : 1 ) relations in the first stage and 2 ) all extractions containing the relation in the second stage .", "forward": true, "src_ids": "2022.acl-long.179_1983"}
+{"input": "implicit context aggregation is done by using Method| context: class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms .", "entity": "implicit context aggregation", "output": "class - based prediction", "neg_sample": ["implicit context aggregation is done by using Method", "class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms ."], "relation": "used for", "id": "2022.acl-long.96", "year": 2022, "rel_sent": "We hypothesize that class - based prediction leads to an implicit context aggregation for similar words and thus can improve generalization for rare words .", "forward": false, "src_ids": "2022.acl-long.96_1984"}
+{"input": "similar words is done by using Task| context: class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms .", "entity": "similar words", "output": "implicit context aggregation", "neg_sample": ["similar words is done by using Task", "class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms ."], "relation": "used for", "id": "2022.acl-long.96", "year": 2022, "rel_sent": "We hypothesize that class - based prediction leads to an implicit context aggregation for similar words and thus can improve generalization for rare words .", "forward": false, "src_ids": "2022.acl-long.96_1985"}
+{"input": "class - based prediction is used for Task| context: class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms .", "entity": "class - based prediction", "output": "implicit context aggregation", "neg_sample": ["class - based prediction is used for Task", "class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms ."], "relation": "used for", "id": "2022.acl-long.96", "year": 2022, "rel_sent": "We hypothesize that class - based prediction leads to an implicit context aggregation for similar words and thus can improve generalization for rare words .", "forward": true, "src_ids": "2022.acl-long.96_1986"}
+{"input": "implicit context aggregation is used for OtherScientificTerm| context: class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms .", "entity": "implicit context aggregation", "output": "similar words", "neg_sample": ["implicit context aggregation is used for OtherScientificTerm", "class - based language models ( lms ) have been long devised to address context sparsity in n - gram lms ."], "relation": "used for", "id": "2022.acl-long.96", "year": 2022, "rel_sent": "We hypothesize that class - based prediction leads to an implicit context aggregation for similar words and thus can improve generalization for rare words .", "forward": true, "src_ids": "2022.acl-long.96_1987"}
+{"input": "news recommendation is done by using Method| context: personalized news recommendation is an essential technique to help users find interested news . accurately matching user 's interests and candidate news is the key to news recommendation . most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest . however , user interest is usually diverse and may not be adequately modeled by a single user embedding .", "entity": "news recommendation", "output": "multi - interest matching network", "neg_sample": ["news recommendation is done by using Method", "personalized news recommendation is an essential technique to help users find interested news .", "accurately matching user 's interests and candidate news is the key to news recommendation .", "most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest .", "however , user interest is usually diverse and may not be adequately modeled by a single user embedding ."], "relation": "used for", "id": "2022.findings-acl.29", "year": 2022, "rel_sent": "MINER : Multi - Interest Matching Network for News Recommendation.", "forward": false, "src_ids": "2022.findings-acl.29_1988"}
+{"input": "multi - interest matching network is used for Task| context: most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest . however , user interest is usually diverse and may not be adequately modeled by a single user embedding .", "entity": "multi - interest matching network", "output": "news recommendation", "neg_sample": ["multi - interest matching network is used for Task", "most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest .", "however , user interest is usually diverse and may not be adequately modeled by a single user embedding ."], "relation": "used for", "id": "2022.findings-acl.29", "year": 2022, "rel_sent": "MINER : Multi - Interest Matching Network for News Recommendation.", "forward": true, "src_ids": "2022.findings-acl.29_1989"}
+{"input": "multiple interest vectors is done by using Method| context: personalized news recommendation is an essential technique to help users find interested news . accurately matching user 's interests and candidate news is the key to news recommendation . most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest . however , user interest is usually diverse and may not be adequately modeled by a single user embedding .", "entity": "multiple interest vectors", "output": "poly attention scheme", "neg_sample": ["multiple interest vectors is done by using Method", "personalized news recommendation is an essential technique to help users find interested news .", "accurately matching user 's interests and candidate news is the key to news recommendation .", "most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest .", "however , user interest is usually diverse and may not be adequately modeled by a single user embedding ."], "relation": "used for", "id": "2022.findings-acl.29", "year": 2022, "rel_sent": "In this paper , we propose a poly attention scheme to learn multiple interest vectors for each user , which encodes the different aspects of user interest .", "forward": false, "src_ids": "2022.findings-acl.29_1990"}
+{"input": "poly attention scheme is used for OtherScientificTerm| context: personalized news recommendation is an essential technique to help users find interested news . accurately matching user 's interests and candidate news is the key to news recommendation . most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest . however , user interest is usually diverse and may not be adequately modeled by a single user embedding .", "entity": "poly attention scheme", "output": "multiple interest vectors", "neg_sample": ["poly attention scheme is used for OtherScientificTerm", "personalized news recommendation is an essential technique to help users find interested news .", "accurately matching user 's interests and candidate news is the key to news recommendation .", "most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest .", "however , user interest is usually diverse and may not be adequately modeled by a single user embedding ."], "relation": "used for", "id": "2022.findings-acl.29", "year": 2022, "rel_sent": "In this paper , we propose a poly attention scheme to learn multiple interest vectors for each user , which encodes the different aspects of user interest .", "forward": true, "src_ids": "2022.findings-acl.29_1991"}
+{"input": "interests vectors is done by using Method| context: personalized news recommendation is an essential technique to help users find interested news . accurately matching user 's interests and candidate news is the key to news recommendation . most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest . however , user interest is usually diverse and may not be adequately modeled by a single user embedding .", "entity": "interests vectors", "output": "disagreement regularization", "neg_sample": ["interests vectors is done by using Method", "personalized news recommendation is an essential technique to help users find interested news .", "accurately matching user 's interests and candidate news is the key to news recommendation .", "most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest .", "however , user interest is usually diverse and may not be adequately modeled by a single user embedding ."], "relation": "used for", "id": "2022.findings-acl.29", "year": 2022, "rel_sent": "We further propose a disagreement regularization to make the learned interests vectors more diverse .", "forward": false, "src_ids": "2022.findings-acl.29_1992"}
+{"input": "disagreement regularization is used for OtherScientificTerm| context: personalized news recommendation is an essential technique to help users find interested news . accurately matching user 's interests and candidate news is the key to news recommendation . most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest . however , user interest is usually diverse and may not be adequately modeled by a single user embedding .", "entity": "disagreement regularization", "output": "interests vectors", "neg_sample": ["disagreement regularization is used for OtherScientificTerm", "personalized news recommendation is an essential technique to help users find interested news .", "accurately matching user 's interests and candidate news is the key to news recommendation .", "most existing methods learn a single user embedding from user 's historical behaviors to represent the reading interest .", "however , user interest is usually diverse and may not be adequately modeled by a single user embedding ."], "relation": "used for", "id": "2022.findings-acl.29", "year": 2022, "rel_sent": "We further propose a disagreement regularization to make the learned interests vectors more diverse .", "forward": true, "src_ids": "2022.findings-acl.29_1993"}
+{"input": "vision and language navigation is done by using Method| context: in this paper , we investigate the problem of vision and language navigation . to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important .", "entity": "vision and language navigation", "output": "explicit object relation alignment", "neg_sample": ["vision and language navigation is done by using Method", "in this paper , we investigate the problem of vision and language navigation .", "to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important ."], "relation": "used for", "id": "2022.acl-srw.24", "year": 2022, "rel_sent": "Explicit Object Relation Alignment for Vision and Language Navigation.", "forward": false, "src_ids": "2022.acl-srw.24_1994"}
+{"input": "explicit object relation alignment is used for Task| context: to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important .", "entity": "explicit object relation alignment", "output": "vision and language navigation", "neg_sample": ["explicit object relation alignment is used for Task", "to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important ."], "relation": "used for", "id": "2022.acl-srw.24", "year": 2022, "rel_sent": "Explicit Object Relation Alignment for Vision and Language Navigation.", "forward": true, "src_ids": "2022.acl-srw.24_1995"}
+{"input": "spatial information is done by using Method| context: in this paper , we investigate the problem of vision and language navigation . to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important .", "entity": "spatial information", "output": "neural agent", "neg_sample": ["spatial information is done by using Method", "in this paper , we investigate the problem of vision and language navigation .", "to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important ."], "relation": "used for", "id": "2022.acl-srw.24", "year": 2022, "rel_sent": "We propose a neural agent named Explicit Object Relation Alignment Agent ( EXOR),to explicitly align the spatial information in both instruction and the visual environment , including landmarks and spatial relationships between the agent and landmarks . Empirically , our proposed method surpasses the baseline by a large margin on the R2R dataset .", "forward": false, "src_ids": "2022.acl-srw.24_1996"}
+{"input": "neural agent is used for OtherScientificTerm| context: in this paper , we investigate the problem of vision and language navigation . to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important .", "entity": "neural agent", "output": "spatial information", "neg_sample": ["neural agent is used for OtherScientificTerm", "in this paper , we investigate the problem of vision and language navigation .", "to solve this problem , grounding the landmarks and spatial relations in the textual instructions into visual modality is important ."], "relation": "used for", "id": "2022.acl-srw.24", "year": 2022, "rel_sent": "We propose a neural agent named Explicit Object Relation Alignment Agent ( EXOR),to explicitly align the spatial information in both instruction and the visual environment , including landmarks and spatial relationships between the agent and landmarks . Empirically , our proposed method surpasses the baseline by a large margin on the R2R dataset .", "forward": true, "src_ids": "2022.acl-srw.24_1997"}
+{"input": "isometric - speech language translation is done by using Method| context: this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) . there are numerous state - of - art models for translation problems . however , these models lack any length constraint to produce short or long outputs from the source text .", "entity": "isometric - speech language translation", "output": "hierarchical multi - task learning framework", "neg_sample": ["isometric - speech language translation is done by using Method", "this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) .", "there are numerous state - of - art models for translation problems .", "however , these models lack any length constraint to produce short or long outputs from the source text ."], "relation": "used for", "id": "2022.iwslt-1.35", "year": 2022, "rel_sent": "Hierarchical Multi - task learning framework for Isometric - Speech Language Translation.", "forward": false, "src_ids": "2022.iwslt-1.35_1998"}
+{"input": "hierarchical multi - task learning framework is used for Task| context: this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) . there are numerous state - of - art models for translation problems . however , these models lack any length constraint to produce short or long outputs from the source text .", "entity": "hierarchical multi - task learning framework", "output": "isometric - speech language translation", "neg_sample": ["hierarchical multi - task learning framework is used for Task", "this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) .", "there are numerous state - of - art models for translation problems .", "however , these models lack any length constraint to produce short or long outputs from the source text ."], "relation": "used for", "id": "2022.iwslt-1.35", "year": 2022, "rel_sent": "Hierarchical Multi - task learning framework for Isometric - Speech Language Translation.", "forward": true, "src_ids": "2022.iwslt-1.35_1999"}
+{"input": "hierarchical approach is used for Task| context: this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) . there are numerous state - of - art models for translation problems . however , these models lack any length constraint to produce short or long outputs from the source text .", "entity": "hierarchical approach", "output": "isometric translation", "neg_sample": ["hierarchical approach is used for Task", "this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) .", "there are numerous state - of - art models for translation problems .", "however , these models lack any length constraint to produce short or long outputs from the source text ."], "relation": "used for", "id": "2022.iwslt-1.35", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical approach to generate isometric translation on MUST - C dataset , we achieve a BERTscore of 0.85 , a length ratio of 1.087 , a BLEU score of 42.3 , and a length range of 51.03 % .", "forward": true, "src_ids": "2022.iwslt-1.35_2000"}
+{"input": "isometric translation is done by using Method| context: this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) . there are numerous state - of - art models for translation problems . however , these models lack any length constraint to produce short or long outputs from the source text .", "entity": "isometric translation", "output": "hierarchical approach", "neg_sample": ["isometric translation is done by using Method", "this paper presents our submission for the shared task on isometric neural machine translation in international conference on spoken language translation ( iwslt ) .", "there are numerous state - of - art models for translation problems .", "however , these models lack any length constraint to produce short or long outputs from the source text ."], "relation": "used for", "id": "2022.iwslt-1.35", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical approach to generate isometric translation on MUST - C dataset , we achieve a BERTscore of 0.85 , a length ratio of 1.087 , a BLEU score of 42.3 , and a length range of 51.03 % .", "forward": false, "src_ids": "2022.iwslt-1.35_2001"}
+{"input": "translation industry is done by using Method| context: with the increase in machine translation ( mt ) quality over the latest years , it has now become a common practice to integrate mt in the workflow of language service providers ( lsps ) and other actors in the translation industry . with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions . the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows .", "entity": "translation industry", "output": "quality estimation and quality evaluation tool", "neg_sample": ["translation industry is done by using Method", "with the increase in machine translation ( mt ) quality over the latest years , it has now become a common practice to integrate mt in the workflow of language service providers ( lsps ) and other actors in the translation industry .", "with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions .", "the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows ."], "relation": "used for", "id": "2022.eamt-1.43", "year": 2022, "rel_sent": "A Quality Estimation and Quality Evaluation Tool for the Translation Industry.", "forward": false, "src_ids": "2022.eamt-1.43_2002"}
+{"input": "quality estimation and quality evaluation tool is used for Task| context: with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions . the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows .", "entity": "quality estimation and quality evaluation tool", "output": "translation industry", "neg_sample": ["quality estimation and quality evaluation tool is used for Task", "with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions .", "the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows ."], "relation": "used for", "id": "2022.eamt-1.43", "year": 2022, "rel_sent": "A Quality Estimation and Quality Evaluation Tool for the Translation Industry.", "forward": true, "src_ids": "2022.eamt-1.43_2003"}
+{"input": "mt quality assessment ( mtqa ) tool is used for Task| context: with the increase in machine translation ( mt ) quality over the latest years , it has now become a common practice to integrate mt in the workflow of language service providers ( lsps ) and other actors in the translation industry . with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions . the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows .", "entity": "mt quality assessment ( mtqa ) tool", "output": "quality assessment of mt engines", "neg_sample": ["mt quality assessment ( mtqa ) tool is used for Task", "with the increase in machine translation ( mt ) quality over the latest years , it has now become a common practice to integrate mt in the workflow of language service providers ( lsps ) and other actors in the translation industry .", "with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions .", "the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows ."], "relation": "used for", "id": "2022.eamt-1.43", "year": 2022, "rel_sent": "The overall objective of the project presented in this paper is to develop an MT quality assessment ( MTQA ) tool that simplifies the quality assessment of MT engines , combining quality evaluation and quality estimation on document- and sentence- level .", "forward": true, "src_ids": "2022.eamt-1.43_2004"}
+{"input": "quality assessment of mt engines is done by using Method| context: with the increase in machine translation ( mt ) quality over the latest years , it has now become a common practice to integrate mt in the workflow of language service providers ( lsps ) and other actors in the translation industry . with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions . the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows .", "entity": "quality assessment of mt engines", "output": "mt quality assessment ( mtqa ) tool", "neg_sample": ["quality assessment of mt engines is done by using Method", "with the increase in machine translation ( mt ) quality over the latest years , it has now become a common practice to integrate mt in the workflow of language service providers ( lsps ) and other actors in the translation industry .", "with mt having a direct impact on the translation workflow , it is important not only to use high - quality mt systems , but also to understand the quality dimension so that the humans involved in the translation workflow can make informed decisions .", "the evaluation and monitoring of mt output quality has become one of the essential aspects of language technology management in lsps ' workflows ."], "relation": "used for", "id": "2022.eamt-1.43", "year": 2022, "rel_sent": "The overall objective of the project presented in this paper is to develop an MT quality assessment ( MTQA ) tool that simplifies the quality assessment of MT engines , combining quality evaluation and quality estimation on document- and sentence- level .", "forward": false, "src_ids": "2022.eamt-1.43_2005"}
+{"input": "health outcome generation is done by using Method| context: probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases . to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves . we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task .", "entity": "health outcome generation", "output": "position - based prompting", "neg_sample": ["health outcome generation is done by using Method", "probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases .", "to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves .", "we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task ."], "relation": "used for", "id": "2022.bionlp-1.3", "year": 2022, "rel_sent": "Position - based Prompting for Health Outcome Generation.", "forward": false, "src_ids": "2022.bionlp-1.3_2006"}
+{"input": "position - based prompting is used for Task| context: probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases . to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves . we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task .", "entity": "position - based prompting", "output": "health outcome generation", "neg_sample": ["position - based prompting is used for Task", "probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases .", "to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves .", "we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task ."], "relation": "used for", "id": "2022.bionlp-1.3", "year": 2022, "rel_sent": "Position - based Prompting for Health Outcome Generation.", "forward": true, "src_ids": "2022.bionlp-1.3_2007"}
+{"input": "positional information is done by using Method| context: probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases . to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves . we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task .", "entity": "positional information", "output": "position - attention mechanism", "neg_sample": ["positional information is done by using Method", "probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases .", "to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves .", "we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task ."], "relation": "used for", "id": "2022.bionlp-1.3", "year": 2022, "rel_sent": "To alleviate this constraint , we propose using a position - attention mechanism to capture positional information of each word in a prompt relative to the mask to be filled , hence avoiding the need to re - construct prompts when the prompts ' linguistic pattern changes .", "forward": false, "src_ids": "2022.bionlp-1.3_2008"}
+{"input": "position - attention mechanism is used for OtherScientificTerm| context: probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases . to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves . we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task .", "entity": "position - attention mechanism", "output": "positional information", "neg_sample": ["position - attention mechanism is used for OtherScientificTerm", "probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases .", "to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves .", "we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task ."], "relation": "used for", "id": "2022.bionlp-1.3", "year": 2022, "rel_sent": "To alleviate this constraint , we propose using a position - attention mechanism to capture positional information of each word in a prompt relative to the mask to be filled , hence avoiding the need to re - construct prompts when the prompts ' linguistic pattern changes .", "forward": true, "src_ids": "2022.bionlp-1.3_2009"}
+{"input": "masked tokens is done by using Method| context: probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases . to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves . we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task .", "entity": "masked tokens", "output": "plms representation", "neg_sample": ["masked tokens is done by using Method", "probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases .", "to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves .", "we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task ."], "relation": "used for", "id": "2022.bionlp-1.3", "year": 2022, "rel_sent": "More so , using various biomedical PLMs , our approach consistently outperforms a baseline in which the default PLMs representation is used to predict masked tokens .", "forward": false, "src_ids": "2022.bionlp-1.3_2010"}
+{"input": "plms representation is used for OtherScientificTerm| context: probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases . to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves . we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task .", "entity": "plms representation", "output": "masked tokens", "neg_sample": ["plms representation is used for OtherScientificTerm", "probing factual knowledge in pre - trained language models ( plms ) using prompts has indirectly implied that language models ( lms ) can be treated as knowledge bases .", "to this end , this phenomenon has been effective , especially when these lms are fine - tuned towards not just data , but also to the style or linguistic pattern of the prompts themselves .", "we observe that satisfying a particular linguistic pattern in prompts is an unsustainable , time - consuming constraint in the probing task , especially because they are often manually designed and the range of possible prompt template patterns can vary depending on the prompting task ."], "relation": "used for", "id": "2022.bionlp-1.3", "year": 2022, "rel_sent": "More so , using various biomedical PLMs , our approach consistently outperforms a baseline in which the default PLMs representation is used to predict masked tokens .", "forward": true, "src_ids": "2022.bionlp-1.3_2011"}
+{"input": "feature system is done by using OtherScientificTerm| context: this paper uses model theory to analyze the formal properties of three phonological feature systems : privative , full binary , and underspecified .", "entity": "feature system", "output": "logical negation", "neg_sample": ["feature system is done by using OtherScientificTerm", "this paper uses model theory to analyze the formal properties of three phonological feature systems : privative , full binary , and underspecified ."], "relation": "used for", "id": "2022.scil-1.1", "year": 2022, "rel_sent": "By systematically manipulating the choice of logical language and representational primitives , it is shown that logical negation effectively converts any feature system into a full binary one .", "forward": false, "src_ids": "2022.scil-1.1_2012"}
+{"input": "positive reframing is done by using OtherScientificTerm| context: with a sentiment reversal comes also a reversal in meaning .", "entity": "positive reframing", "output": "meaning preservation", "neg_sample": ["positive reframing is done by using OtherScientificTerm", "with a sentiment reversal comes also a reversal in meaning ."], "relation": "used for", "id": "2022.acl-long.257", "year": 2022, "rel_sent": "Our insistence on meaning preservation makes positive reframing a challenging and semantically rich task .", "forward": false, "src_ids": "2022.acl-long.257_2013"}
+{"input": "positive reframing is done by using OtherScientificTerm| context: with a sentiment reversal comes also a reversal in meaning .", "entity": "positive reframing", "output": "positive psychology frames", "neg_sample": ["positive reframing is done by using OtherScientificTerm", "with a sentiment reversal comes also a reversal in meaning ."], "relation": "used for", "id": "2022.acl-long.257", "year": 2022, "rel_sent": "Tofacilitate rapid progress , we introduce a large - scale benchmark , Positive Psychology Frames , with 8,349 sentence pairs and 12,755 structured annotations to explain positive reframing in terms of six theoretically - motivated reframing strategies .", "forward": false, "src_ids": "2022.acl-long.257_2014"}
+{"input": "meaning preservation is used for Task| context: with a sentiment reversal comes also a reversal in meaning .", "entity": "meaning preservation", "output": "positive reframing", "neg_sample": ["meaning preservation is used for Task", "with a sentiment reversal comes also a reversal in meaning ."], "relation": "used for", "id": "2022.acl-long.257", "year": 2022, "rel_sent": "Our insistence on meaning preservation makes positive reframing a challenging and semantically rich task .", "forward": true, "src_ids": "2022.acl-long.257_2015"}
+{"input": "positive psychology frames is used for Task| context: with a sentiment reversal comes also a reversal in meaning .", "entity": "positive psychology frames", "output": "positive reframing", "neg_sample": ["positive psychology frames is used for Task", "with a sentiment reversal comes also a reversal in meaning ."], "relation": "used for", "id": "2022.acl-long.257", "year": 2022, "rel_sent": "Tofacilitate rapid progress , we introduce a large - scale benchmark , Positive Psychology Frames , with 8,349 sentence pairs and 12,755 structured annotations to explain positive reframing in terms of six theoretically - motivated reframing strategies .", "forward": true, "src_ids": "2022.acl-long.257_2016"}
+{"input": "resolving overfitting is done by using Method| context: conventional wisdom in pruning transformer - based language models is that pruning reduces the model expressiveness and thus is more likely to underfit rather than overfit . however , under the trending pretrain - and - finetune paradigm , we postulate a counter - traditional hypothesis , that is : pruning increases the risk of overfitting when performed at the fine - tuning phase .", "entity": "resolving overfitting", "output": "sparse progressive distillation", "neg_sample": ["resolving overfitting is done by using Method", "conventional wisdom in pruning transformer - based language models is that pruning reduces the model expressiveness and thus is more likely to underfit rather than overfit .", "however , under the trending pretrain - and - finetune paradigm , we postulate a counter - traditional hypothesis , that is : pruning increases the risk of overfitting when performed at the fine - tuning phase ."], "relation": "used for", "id": "2022.acl-long.16", "year": 2022, "rel_sent": "Sparse Progressive Distillation : Resolving Overfitting under Pretrain - and - Finetune Paradigm.", "forward": false, "src_ids": "2022.acl-long.16_2017"}
+{"input": "sparse progressive distillation is used for Task| context: conventional wisdom in pruning transformer - based language models is that pruning reduces the model expressiveness and thus is more likely to underfit rather than overfit . however , under the trending pretrain - and - finetune paradigm , we postulate a counter - traditional hypothesis , that is : pruning increases the risk of overfitting when performed at the fine - tuning phase .", "entity": "sparse progressive distillation", "output": "resolving overfitting", "neg_sample": ["sparse progressive distillation is used for Task", "conventional wisdom in pruning transformer - based language models is that pruning reduces the model expressiveness and thus is more likely to underfit rather than overfit .", "however , under the trending pretrain - and - finetune paradigm , we postulate a counter - traditional hypothesis , that is : pruning increases the risk of overfitting when performed at the fine - tuning phase ."], "relation": "used for", "id": "2022.acl-long.16", "year": 2022, "rel_sent": "Sparse Progressive Distillation : Resolving Overfitting under Pretrain - and - Finetune Paradigm.", "forward": true, "src_ids": "2022.acl-long.16_2018"}
+{"input": "document re - ranking inference is done by using Method| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "document re - ranking inference", "output": "ed2lm", "neg_sample": ["document re - ranking inference is done by using Method", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "ED2LM : Encoder - Decoder to Language Model for Faster Document Re - ranking Inference.", "forward": false, "src_ids": "2022.findings-acl.295_2019"}
+{"input": "ed2lm is used for Task| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "ed2lm", "output": "document re - ranking inference", "neg_sample": ["ed2lm is used for Task", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "ED2LM : Encoder - Decoder to Language Model for Faster Document Re - ranking Inference.", "forward": true, "src_ids": "2022.findings-acl.295_2020"}
+{"input": "training and inference paradigm is used for Task| context: to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "training and inference paradigm", "output": "re - ranking", "neg_sample": ["training and inference paradigm is used for Task", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "This paper proposes a new training and inference paradigm for re - ranking .", "forward": true, "src_ids": "2022.findings-acl.295_2021"}
+{"input": "re - ranking is done by using Method| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "re - ranking", "output": "training and inference paradigm", "neg_sample": ["re - ranking is done by using Method", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "This paper proposes a new training and inference paradigm for re - ranking .", "forward": false, "src_ids": "2022.findings-acl.295_2022"}
+{"input": "inference is done by using Method| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "inference", "output": "decoder - only language model", "neg_sample": ["inference is done by using Method", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "Subsequently , we show that this encoder - decoder architecture can be decomposed into a decoder - only language model during inference .", "forward": false, "src_ids": "2022.findings-acl.295_2023"}
+{"input": "decoder - only language model is used for Task| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "entity": "decoder - only language model", "output": "inference", "neg_sample": ["decoder - only language model is used for Task", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "Subsequently , we show that this encoder - decoder architecture can be decomposed into a decoder - only language model during inference .", "forward": true, "src_ids": "2022.findings-acl.295_2024"}
+{"input": "static encoder embeddings is used for Task| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "entity": "static encoder embeddings", "output": "inference", "neg_sample": ["static encoder embeddings is used for Task", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "This results in significant inference time speedups since the decoder - only architecture only needs to learn to interpret static encoder embeddings during inference .", "forward": true, "src_ids": "2022.findings-acl.295_2025"}
+{"input": "static encoder embeddings is done by using Method| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "static encoder embeddings", "output": "decoder - only architecture", "neg_sample": ["static encoder embeddings is done by using Method", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "This results in significant inference time speedups since the decoder - only architecture only needs to learn to interpret static encoder embeddings during inference .", "forward": false, "src_ids": "2022.findings-acl.295_2026"}
+{"input": "inference is done by using OtherScientificTerm| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "inference", "output": "static encoder embeddings", "neg_sample": ["inference is done by using OtherScientificTerm", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "This results in significant inference time speedups since the decoder - only architecture only needs to learn to interpret static encoder embeddings during inference .", "forward": false, "src_ids": "2022.findings-acl.295_2027"}
+{"input": "decoder - only architecture is used for OtherScientificTerm| context: state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking . to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach . these paradigms , however , are not without flaws , i.e. , running the model on all query - document pairs at inference - time incurs a significant computational cost .", "entity": "decoder - only architecture", "output": "static encoder embeddings", "neg_sample": ["decoder - only architecture is used for OtherScientificTerm", "state - of - the - art neural models typically encode document - query pairs using cross - attention for re - ranking .", "to this end , models generally utilize an encoder - only ( like bert ) paradigm or an encoder - decoder ( like t5 ) approach .", "these paradigms , however , are not without flaws , i.e.", ", running the model on all query - document pairs at inference - time incurs a significant computational cost ."], "relation": "used for", "id": "2022.findings-acl.295", "year": 2022, "rel_sent": "This results in significant inference time speedups since the decoder - only architecture only needs to learn to interpret static encoder embeddings during inference .", "forward": true, "src_ids": "2022.findings-acl.295_2028"}
+{"input": "contextual representations is used for Method| context: human languages are full of metaphorical expressions . metaphors help people understand the world by connecting new concepts and domains to more familiar ones .", "entity": "contextual representations", "output": "pre - trained language models", "neg_sample": ["contextual representations is used for Method", "human languages are full of metaphorical expressions .", "metaphors help people understand the world by connecting new concepts and domains to more familiar ones ."], "relation": "used for", "id": "2022.acl-long.144", "year": 2022, "rel_sent": "Our extensive experiments suggest that contextual representations in PLMs do encode metaphorical knowledge , and mostly in their middle layers .", "forward": true, "src_ids": "2022.acl-long.144_2029"}
+{"input": "contextual representations is used for OtherScientificTerm| context: human languages are full of metaphorical expressions . metaphors help people understand the world by connecting new concepts and domains to more familiar ones .", "entity": "contextual representations", "output": "metaphorical knowledge", "neg_sample": ["contextual representations is used for OtherScientificTerm", "human languages are full of metaphorical expressions .", "metaphors help people understand the world by connecting new concepts and domains to more familiar ones ."], "relation": "used for", "id": "2022.acl-long.144", "year": 2022, "rel_sent": "Our extensive experiments suggest that contextual representations in PLMs do encode metaphorical knowledge , and mostly in their middle layers .", "forward": true, "src_ids": "2022.acl-long.144_2030"}
+{"input": "metaphorical knowledge is done by using Method| context: human languages are full of metaphorical expressions . metaphors help people understand the world by connecting new concepts and domains to more familiar ones . large pre - trained language models ( plms ) are therefore assumed to encode metaphorical knowledge useful for nlp systems .", "entity": "metaphorical knowledge", "output": "contextual representations", "neg_sample": ["metaphorical knowledge is done by using Method", "human languages are full of metaphorical expressions .", "metaphors help people understand the world by connecting new concepts and domains to more familiar ones .", "large pre - trained language models ( plms ) are therefore assumed to encode metaphorical knowledge useful for nlp systems ."], "relation": "used for", "id": "2022.acl-long.144", "year": 2022, "rel_sent": "Our extensive experiments suggest that contextual representations in PLMs do encode metaphorical knowledge , and mostly in their middle layers .", "forward": false, "src_ids": "2022.acl-long.144_2031"}
+{"input": "pre - trained language models is done by using Method| context: human languages are full of metaphorical expressions . metaphors help people understand the world by connecting new concepts and domains to more familiar ones . large pre - trained language models ( plms ) are therefore assumed to encode metaphorical knowledge useful for nlp systems .", "entity": "pre - trained language models", "output": "contextual representations", "neg_sample": ["pre - trained language models is done by using Method", "human languages are full of metaphorical expressions .", "metaphors help people understand the world by connecting new concepts and domains to more familiar ones .", "large pre - trained language models ( plms ) are therefore assumed to encode metaphorical knowledge useful for nlp systems ."], "relation": "used for", "id": "2022.acl-long.144", "year": 2022, "rel_sent": "Our extensive experiments suggest that contextual representations in PLMs do encode metaphorical knowledge , and mostly in their middle layers .", "forward": false, "src_ids": "2022.acl-long.144_2032"}
+{"input": "lexical data is done by using Material| context: the universal knowledge core ( ukc ) is a large multilingual lexical database with a focus on language diversity and covering over two thousand languages .", "entity": "lexical data", "output": "ukc livelanguage catalogue", "neg_sample": ["lexical data is done by using Material", "the universal knowledge core ( ukc ) is a large multilingual lexical database with a focus on language diversity and covering over two thousand languages ."], "relation": "used for", "id": "2022.acl-demo.15", "year": 2022, "rel_sent": "The UKC LiveLanguage Catalogue , in turn , provides access to the underlying lexical data in a computer - processable form , ready to be reused in cross - lingual applications .", "forward": false, "src_ids": "2022.acl-demo.15_2033"}
+{"input": "cross - lingual applications is done by using Material| context: the universal knowledge core ( ukc ) is a large multilingual lexical database with a focus on language diversity and covering over two thousand languages .", "entity": "cross - lingual applications", "output": "ukc livelanguage catalogue", "neg_sample": ["cross - lingual applications is done by using Material", "the universal knowledge core ( ukc ) is a large multilingual lexical database with a focus on language diversity and covering over two thousand languages ."], "relation": "used for", "id": "2022.acl-demo.15", "year": 2022, "rel_sent": "The UKC LiveLanguage Catalogue , in turn , provides access to the underlying lexical data in a computer - processable form , ready to be reused in cross - lingual applications .", "forward": false, "src_ids": "2022.acl-demo.15_2034"}
+{"input": "ukc livelanguage catalogue is used for Task| context: the universal knowledge core ( ukc ) is a large multilingual lexical database with a focus on language diversity and covering over two thousand languages .", "entity": "ukc livelanguage catalogue", "output": "cross - lingual applications", "neg_sample": ["ukc livelanguage catalogue is used for Task", "the universal knowledge core ( ukc ) is a large multilingual lexical database with a focus on language diversity and covering over two thousand languages ."], "relation": "used for", "id": "2022.acl-demo.15", "year": 2022, "rel_sent": "The UKC LiveLanguage Catalogue , in turn , provides access to the underlying lexical data in a computer - processable form , ready to be reused in cross - lingual applications .", "forward": true, "src_ids": "2022.acl-demo.15_2035"}
+{"input": "generalization is done by using Method| context: adversarial robustness has attracted much attention recently , and the mainstream solution is adversarial training . however , the tradition of generating adversarial perturbations for each input embedding ( in the settings of nlp ) scales up the training computational complexity by the number of gradient steps it takes to obtain the adversarial samples .", "entity": "generalization", "output": "flooding method", "neg_sample": ["generalization is done by using Method", "adversarial robustness has attracted much attention recently , and the mainstream solution is adversarial training .", "however , the tradition of generating adversarial perturbations for each input embedding ( in the settings of nlp ) scales up the training computational complexity by the number of gradient steps it takes to obtain the adversarial samples ."], "relation": "used for", "id": "2022.acl-long.386", "year": 2022, "rel_sent": "To address this problem , we leverage Flooding method which primarily aims at better generalization and we find promising in defending adversarial attacks .", "forward": false, "src_ids": "2022.acl-long.386_2036"}
+{"input": "flooding method is used for Task| context: adversarial robustness has attracted much attention recently , and the mainstream solution is adversarial training . however , the tradition of generating adversarial perturbations for each input embedding ( in the settings of nlp ) scales up the training computational complexity by the number of gradient steps it takes to obtain the adversarial samples .", "entity": "flooding method", "output": "generalization", "neg_sample": ["flooding method is used for Task", "adversarial robustness has attracted much attention recently , and the mainstream solution is adversarial training .", "however , the tradition of generating adversarial perturbations for each input embedding ( in the settings of nlp ) scales up the training computational complexity by the number of gradient steps it takes to obtain the adversarial samples ."], "relation": "used for", "id": "2022.acl-long.386", "year": 2022, "rel_sent": "To address this problem , we leverage Flooding method which primarily aims at better generalization and we find promising in defending adversarial attacks .", "forward": true, "src_ids": "2022.acl-long.386_2037"}
+{"input": "cross - lingual transfer is used for Task| context: this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training .", "entity": "cross - lingual transfer", "output": "zero - shot neural machine translation", "neg_sample": ["cross - lingual transfer is used for Task", "this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training ."], "relation": "used for", "id": "2022.acl-long.12", "year": 2022, "rel_sent": "Towards Making the Most of Cross - Lingual Transfer for Zero - Shot Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.12_2038"}
+{"input": "unsupervised tasks is done by using Method| context: this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training .", "entity": "unsupervised tasks", "output": "sixt+", "neg_sample": ["unsupervised tasks is done by using Method", "this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training ."], "relation": "used for", "id": "2022.acl-long.12", "year": 2022, "rel_sent": "Additionally , SixT+ offers a set of model parameters that can be further fine - tuned to other unsupervised tasks .", "forward": false, "src_ids": "2022.acl-long.12_2039"}
+{"input": "unsupervised tasks is done by using OtherScientificTerm| context: this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training .", "entity": "unsupervised tasks", "output": "model parameters", "neg_sample": ["unsupervised tasks is done by using OtherScientificTerm", "this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training ."], "relation": "used for", "id": "2022.acl-long.12", "year": 2022, "rel_sent": "Additionally , SixT+ offers a set of model parameters that can be further fine - tuned to other unsupervised tasks .", "forward": false, "src_ids": "2022.acl-long.12_2040"}
+{"input": "sixt+ is used for Task| context: this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training .", "entity": "sixt+", "output": "unsupervised tasks", "neg_sample": ["sixt+ is used for Task", "this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training ."], "relation": "used for", "id": "2022.acl-long.12", "year": 2022, "rel_sent": "Additionally , SixT+ offers a set of model parameters that can be further fine - tuned to other unsupervised tasks .", "forward": true, "src_ids": "2022.acl-long.12_2041"}
+{"input": "model parameters is used for Task| context: this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training .", "entity": "model parameters", "output": "unsupervised tasks", "neg_sample": ["model parameters is used for Task", "this paper demonstrates that multilingual pretraining and multilingual fine - tuning are both critical for facilitating cross - lingual transfer in zero - shot translation , where the neural machine translation ( nmt ) model is tested on source languages unseen during supervised training ."], "relation": "used for", "id": "2022.acl-long.12", "year": 2022, "rel_sent": "Additionally , SixT+ offers a set of model parameters that can be further fine - tuned to other unsupervised tasks .", "forward": true, "src_ids": "2022.acl-long.12_2042"}
+{"input": "identifying abusive content is done by using Method| context: identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times . the major driving force behind this is the widespread use of social media websites . further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics . as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages .", "entity": "identifying abusive content", "output": "transformer - based approach", "neg_sample": ["identifying abusive content is done by using Method", "identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times .", "the major driving force behind this is the widespread use of social media websites .", "further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics .", "as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.16", "year": 2022, "rel_sent": "IIITDWD@TamilNLP - ACL2022 : Transformer - based approach to classify abusive content in Dravidian Code - mixed text.", "forward": false, "src_ids": "2022.dravidianlangtech-1.16_2043"}
+{"input": "transformer - based approach is used for Task| context: the major driving force behind this is the widespread use of social media websites . as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages .", "entity": "transformer - based approach", "output": "identifying abusive content", "neg_sample": ["transformer - based approach is used for Task", "the major driving force behind this is the widespread use of social media websites .", "as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.16", "year": 2022, "rel_sent": "IIITDWD@TamilNLP - ACL2022 : Transformer - based approach to classify abusive content in Dravidian Code - mixed text.", "forward": true, "src_ids": "2022.dravidianlangtech-1.16_2044"}
+{"input": "feature extraction is done by using OtherScientificTerm| context: identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times . the major driving force behind this is the widespread use of social media websites . further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics . as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages .", "entity": "feature extraction", "output": "indic - bert", "neg_sample": ["feature extraction is done by using OtherScientificTerm", "identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times .", "the major driving force behind this is the widespread use of social media websites .", "further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics .", "as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.16", "year": 2022, "rel_sent": "In our proposed approach , we used a pre - trained transformer model such as Indic - bert for feature extraction , and on top of that , SVM classifier is used for stance detection .", "forward": false, "src_ids": "2022.dravidianlangtech-1.16_2045"}
+{"input": "indic - bert is used for Task| context: identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times . the major driving force behind this is the widespread use of social media websites . further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics . as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages .", "entity": "indic - bert", "output": "feature extraction", "neg_sample": ["indic - bert is used for Task", "identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times .", "the major driving force behind this is the widespread use of social media websites .", "further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics .", "as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.16", "year": 2022, "rel_sent": "In our proposed approach , we used a pre - trained transformer model such as Indic - bert for feature extraction , and on top of that , SVM classifier is used for stance detection .", "forward": true, "src_ids": "2022.dravidianlangtech-1.16_2046"}
+{"input": "stance detection is done by using Method| context: identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times . the major driving force behind this is the widespread use of social media websites . further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics . as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages .", "entity": "stance detection", "output": "svm classifier", "neg_sample": ["stance detection is done by using Method", "identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times .", "the major driving force behind this is the widespread use of social media websites .", "further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics .", "as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.16", "year": 2022, "rel_sent": "In our proposed approach , we used a pre - trained transformer model such as Indic - bert for feature extraction , and on top of that , SVM classifier is used for stance detection .", "forward": false, "src_ids": "2022.dravidianlangtech-1.16_2047"}
+{"input": "svm classifier is used for Task| context: identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times . the major driving force behind this is the widespread use of social media websites . further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics . as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages .", "entity": "svm classifier", "output": "stance detection", "neg_sample": ["svm classifier is used for Task", "identifying abusive content or hate speech in social media text has raised the research community 's interest in recent times .", "the major driving force behind this is the widespread use of social media websites .", "further , it also leads to identifying abusive content in low - resource regional languages , which is an important research problem in computational linguistics .", "as part of acl-2022 , organizers of dravidianlangtech@acl 2022 have released a shared task on abusive category identification in tamil and tamil - english code - mixed text to encourage further research on offensive content identification in low - resource indic languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.16", "year": 2022, "rel_sent": "In our proposed approach , we used a pre - trained transformer model such as Indic - bert for feature extraction , and on top of that , SVM classifier is used for stance detection .", "forward": true, "src_ids": "2022.dravidianlangtech-1.16_2048"}
+{"input": "writing suggestions is done by using Method| context: while developing a story , novices and published writers alike have had to look outside themselves for inspiration . language models have recently been able to generate text fluently , producing new stochastic narratives upon request . however , effectively integrating such capabilities with human cognitive faculties and creative processes remains challenging .", "entity": "writing suggestions", "output": "multimodal writing support interface", "neg_sample": ["writing suggestions is done by using Method", "while developing a story , novices and published writers alike have had to look outside themselves for inspiration .", "language models have recently been able to generate text fluently , producing new stochastic narratives upon request .", "however , effectively integrating such capabilities with human cognitive faculties and creative processes remains challenging ."], "relation": "used for", "id": "2022.in2writing-1.3", "year": 2022, "rel_sent": "We propose to investigate this integration with a multimodal writing support interface that offers writing suggestions textually , visually , and aurally .", "forward": false, "src_ids": "2022.in2writing-1.3_2049"}
+{"input": "multimodal writing support interface is used for OtherScientificTerm| context: while developing a story , novices and published writers alike have had to look outside themselves for inspiration . language models have recently been able to generate text fluently , producing new stochastic narratives upon request . however , effectively integrating such capabilities with human cognitive faculties and creative processes remains challenging .", "entity": "multimodal writing support interface", "output": "writing suggestions", "neg_sample": ["multimodal writing support interface is used for OtherScientificTerm", "while developing a story , novices and published writers alike have had to look outside themselves for inspiration .", "language models have recently been able to generate text fluently , producing new stochastic narratives upon request .", "however , effectively integrating such capabilities with human cognitive faculties and creative processes remains challenging ."], "relation": "used for", "id": "2022.in2writing-1.3", "year": 2022, "rel_sent": "We propose to investigate this integration with a multimodal writing support interface that offers writing suggestions textually , visually , and aurally .", "forward": true, "src_ids": "2022.in2writing-1.3_2050"}
+{"input": "tl methods is used for Task| context: transfer learning ( tl ) in natural language processing ( nlp ) has seen a surge of interest in recent years , as pre - trained models have shown an impressive ability to transfer to novel tasks . three main strategies have emerged for making use of multiple supervised datasets during fine - tuning : training on an intermediate task before training on the target task ( stilts ) , using multi - task learning ( mtl ) to train jointly on a supplementary task and the target task ( pairwise mtl ) , or simply using mtl to train jointly on all available datasets ( mtl - all ) .", "entity": "tl methods", "output": "nlp tasks", "neg_sample": ["tl methods is used for Task", "transfer learning ( tl ) in natural language processing ( nlp ) has seen a surge of interest in recent years , as pre - trained models have shown an impressive ability to transfer to novel tasks .", "three main strategies have emerged for making use of multiple supervised datasets during fine - tuning : training on an intermediate task before training on the target task ( stilts ) , using multi - task learning ( mtl ) to train jointly on a supplementary task and the target task ( pairwise mtl ) , or simply using mtl to train jointly on all available datasets ( mtl - all ) ."], "relation": "used for", "id": "2022.acl-short.30", "year": 2022, "rel_sent": "We hope this study will aid others as they choose between TL methods for NLP tasks .", "forward": true, "src_ids": "2022.acl-short.30_2051"}
+{"input": "nlp tasks is done by using Method| context: transfer learning ( tl ) in natural language processing ( nlp ) has seen a surge of interest in recent years , as pre - trained models have shown an impressive ability to transfer to novel tasks . three main strategies have emerged for making use of multiple supervised datasets during fine - tuning : training on an intermediate task before training on the target task ( stilts ) , using multi - task learning ( mtl ) to train jointly on a supplementary task and the target task ( pairwise mtl ) , or simply using mtl to train jointly on all available datasets ( mtl - all ) .", "entity": "nlp tasks", "output": "tl methods", "neg_sample": ["nlp tasks is done by using Method", "transfer learning ( tl ) in natural language processing ( nlp ) has seen a surge of interest in recent years , as pre - trained models have shown an impressive ability to transfer to novel tasks .", "three main strategies have emerged for making use of multiple supervised datasets during fine - tuning : training on an intermediate task before training on the target task ( stilts ) , using multi - task learning ( mtl ) to train jointly on a supplementary task and the target task ( pairwise mtl ) , or simply using mtl to train jointly on all available datasets ( mtl - all ) ."], "relation": "used for", "id": "2022.acl-short.30", "year": 2022, "rel_sent": "We hope this study will aid others as they choose between TL methods for NLP tasks .", "forward": false, "src_ids": "2022.acl-short.30_2052"}
+{"input": "chinese bert is done by using Method| context: for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "chinese bert", "output": "whole word masking", "neg_sample": ["chinese bert is done by using Method", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "' Is Whole Word Masking Always Better for Chinese BERT ? ' : Probing on Chinese Grammatical Error Correction.", "forward": false, "src_ids": "2022.findings-acl.1_2053"}
+{"input": "context understanding ability is done by using Method| context: for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "context understanding ability", "output": "whole word masking", "neg_sample": ["context understanding ability is done by using Method", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "Such difference motivates us to investigate whether WWM leads to better context understanding ability for Chinese BERT .", "forward": false, "src_ids": "2022.findings-acl.1_2054"}
+{"input": "chinese bert is done by using Method| context: for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "chinese bert", "output": "whole word masking", "neg_sample": ["chinese bert is done by using Method", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "Such difference motivates us to investigate whether WWM leads to better context understanding ability for Chinese BERT .", "forward": false, "src_ids": "2022.findings-acl.1_2055"}
+{"input": "whole word masking is used for Task| context: whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model . for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "whole word masking", "output": "chinese bert", "neg_sample": ["whole word masking is used for Task", "whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model .", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "' Is Whole Word Masking Always Better for Chinese BERT ? ' : Probing on Chinese Grammatical Error Correction.", "forward": true, "src_ids": "2022.findings-acl.1_2056"}
+{"input": "whole word masking is used for Task| context: whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model . for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "whole word masking", "output": "chinese bert", "neg_sample": ["whole word masking is used for Task", "whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model .", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "Such difference motivates us to investigate whether WWM leads to better context understanding ability for Chinese BERT .", "forward": true, "src_ids": "2022.findings-acl.1_2057"}
+{"input": "grammatical error correction is done by using Task| context: whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model . for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "grammatical error correction", "output": "probing tasks", "neg_sample": ["grammatical error correction is done by using Task", "whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model .", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "To achieve this , we introduce two probing tasks related to grammatical error correction and ask pretrained models to revise or insert tokens in a masked language modeling manner .", "forward": false, "src_ids": "2022.findings-acl.1_2058"}
+{"input": "probing tasks is used for Task| context: whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model . for the chinese language , however , there is no subword because each token is an atomic character . the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters .", "entity": "probing tasks", "output": "grammatical error correction", "neg_sample": ["probing tasks is used for Task", "whole word masking ( wwm ) , which masks all subwords corresponding to a word at once , makes a better english bert model .", "for the chinese language , however , there is no subword because each token is an atomic character .", "the meaning of a word in chinese is different in that a word is a compositional unit consisting of multiple characters ."], "relation": "used for", "id": "2022.findings-acl.1", "year": 2022, "rel_sent": "To achieve this , we introduce two probing tasks related to grammatical error correction and ask pretrained models to revise or insert tokens in a masked language modeling manner .", "forward": true, "src_ids": "2022.findings-acl.1_2059"}
+{"input": "covariate drift is done by using Method| context: in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) . covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it .", "entity": "covariate drift", "output": "distributionally robust finetuning bert", "neg_sample": ["covariate drift is done by using Method", "in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) .", "covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it ."], "relation": "used for", "id": "2022.acl-long.139", "year": 2022, "rel_sent": "Distributionally Robust Finetuning BERT for Covariate Drift in Spoken Language Understanding.", "forward": false, "src_ids": "2022.acl-long.139_2060"}
+{"input": "covariate drift is done by using OtherScientificTerm| context: in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) . covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it .", "entity": "covariate drift", "output": "natural variations in data", "neg_sample": ["covariate drift is done by using OtherScientificTerm", "in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) .", "covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it ."], "relation": "used for", "id": "2022.acl-long.139", "year": 2022, "rel_sent": "To study this we propose a method that exploits natural variations in data to create a covariate drift in SLU datasets .", "forward": false, "src_ids": "2022.acl-long.139_2061"}
+{"input": "finetuning bert - based models is done by using Method| context: in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) . covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it .", "entity": "finetuning bert - based models", "output": "distributionally robust optimization ( dro )", "neg_sample": ["finetuning bert - based models is done by using Method", "in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) .", "covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it ."], "relation": "used for", "id": "2022.acl-long.139", "year": 2022, "rel_sent": "To mitigate the performance loss , we investigate distributionally robust optimization ( DRO ) for finetuning BERT - based models .", "forward": false, "src_ids": "2022.acl-long.139_2062"}
+{"input": "distributionally robust optimization ( dro ) is used for Task| context: in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) . covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it .", "entity": "distributionally robust optimization ( dro )", "output": "finetuning bert - based models", "neg_sample": ["distributionally robust optimization ( dro ) is used for Task", "in this study , we investigate robustness against covariate drift in spoken language understanding ( slu ) .", "covariate drift can occur in sluwhen there is a drift between training and testing regarding what users request or how they request it ."], "relation": "used for", "id": "2022.acl-long.139", "year": 2022, "rel_sent": "To mitigate the performance loss , we investigate distributionally robust optimization ( DRO ) for finetuning BERT - based models .", "forward": true, "src_ids": "2022.acl-long.139_2063"}
+{"input": "toxic span identification is done by using Method| context: toxic span identification in tamil is a shared task that focuses on identifying harmful content , contributing to offensiveness .", "entity": "toxic span identification", "output": "transformer - based models", "neg_sample": ["toxic span identification is done by using Method", "toxic span identification in tamil is a shared task that focuses on identifying harmful content , contributing to offensiveness ."], "relation": "used for", "id": "2022.dravidianlangtech-1.12", "year": 2022, "rel_sent": "NITK - IT_NLP@TamilNLP - ACL2022 : Transformer based model for Toxic Span Identification in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.12_2064"}
+{"input": "out - of - domain intent classification is done by using Method| context: the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems . previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features . then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples .", "entity": "out - of - domain intent classification", "output": "knn - contrastive learning", "neg_sample": ["out - of - domain intent classification is done by using Method", "the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems .", "previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features .", "then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples ."], "relation": "used for", "id": "2022.acl-long.352", "year": 2022, "rel_sent": "KNN - Contrastive Learning for Out - of - Domain Intent Classification.", "forward": false, "src_ids": "2022.acl-long.352_2065"}
+{"input": "knn - contrastive learning is used for Task| context: the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems . previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features . then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples .", "entity": "knn - contrastive learning", "output": "out - of - domain intent classification", "neg_sample": ["knn - contrastive learning is used for Task", "the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems .", "previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features .", "then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples ."], "relation": "used for", "id": "2022.acl-long.352", "year": 2022, "rel_sent": "KNN - Contrastive Learning for Out - of - Domain Intent Classification.", "forward": true, "src_ids": "2022.acl-long.352_2066"}
+{"input": "ood detection is done by using OtherScientificTerm| context: the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems . then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples .", "entity": "ood detection", "output": "discriminative semantic features", "neg_sample": ["ood detection is done by using OtherScientificTerm", "the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems .", "then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples ."], "relation": "used for", "id": "2022.acl-long.352", "year": 2022, "rel_sent": "Our approach utilizes k - nearest neighbors ( KNN ) of IND intents to learn discriminative semantic features that are more conducive to OOD detection . Notably , the density - based novelty detection algorithm is so well - grounded in the essence of our method that it is reasonable to use it as the OOD detection algorithm without making any requirements for the feature distribution . Extensive experiments on four public datasets show that our approach can not only enhance the OOD detection performance substantially but also improve the IND intent classification while requiring no restrictions on feature distribution .", "forward": false, "src_ids": "2022.acl-long.352_2067"}
+{"input": "discriminative semantic features is used for Task| context: the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems . previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features . then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples .", "entity": "discriminative semantic features", "output": "ood detection", "neg_sample": ["discriminative semantic features is used for Task", "the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems .", "previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features .", "then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples ."], "relation": "used for", "id": "2022.acl-long.352", "year": 2022, "rel_sent": "Our approach utilizes k - nearest neighbors ( KNN ) of IND intents to learn discriminative semantic features that are more conducive to OOD detection . Notably , the density - based novelty detection algorithm is so well - grounded in the essence of our method that it is reasonable to use it as the OOD detection algorithm without making any requirements for the feature distribution . Extensive experiments on four public datasets show that our approach can not only enhance the OOD detection performance substantially but also improve the IND intent classification while requiring no restrictions on feature distribution .", "forward": true, "src_ids": "2022.acl-long.352_2068"}
+{"input": "blank filling pretraining is done by using Method| context: there have been various types of pretraining architectures including autoencoding models ( e.g. , bert ) , autoregressive models ( e.g. , gpt ) , and encoder - decoder models ( e.g. , t5 ) . however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation .", "entity": "blank filling pretraining", "output": "general language model", "neg_sample": ["blank filling pretraining is done by using Method", "there have been various types of pretraining architectures including autoencoding models ( e.g.", ", bert ) , autoregressive models ( e.g.", ", gpt ) , and encoder - decoder models ( e.g.", ", t5 ) .", "however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation ."], "relation": "used for", "id": "2022.acl-long.26", "year": 2022, "rel_sent": "GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans , which results in performance gains over BERT and T5 on NLU tasks .", "forward": false, "src_ids": "2022.acl-long.26_2069"}
+{"input": "nlu tasks is done by using Method| context: however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation .", "entity": "nlu tasks", "output": "bert", "neg_sample": ["nlu tasks is done by using Method", "however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation ."], "relation": "used for", "id": "2022.acl-long.26", "year": 2022, "rel_sent": "GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans , which results in performance gains over BERT and T5 on NLU tasks .", "forward": false, "src_ids": "2022.acl-long.26_2070"}
+{"input": "general language model is used for Method| context: there have been various types of pretraining architectures including autoencoding models ( e.g. , bert ) , autoregressive models ( e.g. , gpt ) , and encoder - decoder models ( e.g. , t5 ) . however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation .", "entity": "general language model", "output": "blank filling pretraining", "neg_sample": ["general language model is used for Method", "there have been various types of pretraining architectures including autoencoding models ( e.g.", ", bert ) , autoregressive models ( e.g.", ", gpt ) , and encoder - decoder models ( e.g.", ", t5 ) .", "however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation ."], "relation": "used for", "id": "2022.acl-long.26", "year": 2022, "rel_sent": "GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans , which results in performance gains over BERT and T5 on NLU tasks .", "forward": true, "src_ids": "2022.acl-long.26_2071"}
+{"input": "bert is used for Task| context: there have been various types of pretraining architectures including autoencoding models ( e.g. , bert ) , autoregressive models ( e.g. , gpt ) , and encoder - decoder models ( e.g. , t5 ) . however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation .", "entity": "bert", "output": "nlu tasks", "neg_sample": ["bert is used for Task", "there have been various types of pretraining architectures including autoencoding models ( e.g.", ", bert ) , autoregressive models ( e.g.", ", gpt ) , and encoder - decoder models ( e.g.", ", t5 ) .", "however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation ."], "relation": "used for", "id": "2022.acl-long.26", "year": 2022, "rel_sent": "GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans , which results in performance gains over BERT and T5 on NLU tasks .", "forward": true, "src_ids": "2022.acl-long.26_2072"}
+{"input": "downstream tasks is done by using Method| context: there have been various types of pretraining architectures including autoencoding models ( e.g. , bert ) , autoregressive models ( e.g. , gpt ) , and encoder - decoder models ( e.g. , t5 ) . however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation .", "entity": "downstream tasks", "output": "pretrained model", "neg_sample": ["downstream tasks is done by using Method", "there have been various types of pretraining architectures including autoencoding models ( e.g.", ", bert ) , autoregressive models ( e.g.", ", gpt ) , and encoder - decoder models ( e.g.", ", t5 ) .", "however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation ."], "relation": "used for", "id": "2022.acl-long.26", "year": 2022, "rel_sent": "On a wide range of tasks across NLU , conditional and unconditional generation , GLM outperforms BERT , T5 , and GPT given the same model sizes and data , and achieves the best performance from a single pretrained model with 1.25 times parameters of BERT Large , demonstrating its generalizability to different downstream tasks .", "forward": false, "src_ids": "2022.acl-long.26_2073"}
+{"input": "pretrained model is used for Task| context: there have been various types of pretraining architectures including autoencoding models ( e.g. , bert ) , autoregressive models ( e.g. , gpt ) , and encoder - decoder models ( e.g. , t5 ) . however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation .", "entity": "pretrained model", "output": "downstream tasks", "neg_sample": ["pretrained model is used for Task", "there have been various types of pretraining architectures including autoencoding models ( e.g.", ", bert ) , autoregressive models ( e.g.", ", gpt ) , and encoder - decoder models ( e.g.", ", t5 ) .", "however , none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding ( nlu ) , unconditional generation , and conditional generation ."], "relation": "used for", "id": "2022.acl-long.26", "year": 2022, "rel_sent": "On a wide range of tasks across NLU , conditional and unconditional generation , GLM outperforms BERT , T5 , and GPT given the same model sizes and data , and achieves the best performance from a single pretrained model with 1.25 times parameters of BERT Large , demonstrating its generalizability to different downstream tasks .", "forward": true, "src_ids": "2022.acl-long.26_2074"}
+{"input": "multiparallel word alignment is done by using Method| context: after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation . generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel . here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together .", "entity": "multiparallel word alignment", "output": "graph neural networks ( gnns )", "neg_sample": ["multiparallel word alignment is done by using Method", "after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation .", "generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel .", "here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together ."], "relation": "used for", "id": "2022.findings-acl.108", "year": 2022, "rel_sent": "Graph Neural Networks for Multiparallel Word Alignment.", "forward": false, "src_ids": "2022.findings-acl.108_2075"}
+{"input": "graph structure is done by using Method| context: after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation . generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel . here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together .", "entity": "graph structure", "output": "graph neural networks ( gnns )", "neg_sample": ["graph structure is done by using Method", "after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation .", "generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel .", "here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together ."], "relation": "used for", "id": "2022.findings-acl.108", "year": 2022, "rel_sent": "Next , we use graph neural networks ( GNNs ) to exploit the graph structure .", "forward": false, "src_ids": "2022.findings-acl.108_2076"}
+{"input": "graph neural networks ( gnns ) is used for Task| context: after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation . generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel . here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together .", "entity": "graph neural networks ( gnns )", "output": "multiparallel word alignment", "neg_sample": ["graph neural networks ( gnns ) is used for Task", "after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation .", "generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel .", "here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together ."], "relation": "used for", "id": "2022.findings-acl.108", "year": 2022, "rel_sent": "Graph Neural Networks for Multiparallel Word Alignment.", "forward": true, "src_ids": "2022.findings-acl.108_2077"}
+{"input": "community detection algorithms is used for Task| context: after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation . generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel . here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together .", "entity": "community detection algorithms", "output": "multiparallel word alignment", "neg_sample": ["community detection algorithms is used for Task", "after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation .", "generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel .", "here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together ."], "relation": "used for", "id": "2022.findings-acl.108", "year": 2022, "rel_sent": "We show that community detection algorithms can provide valuable information for multiparallel word alignment .", "forward": true, "src_ids": "2022.findings-acl.108_2078"}
+{"input": "multiparallel word alignment is done by using Method| context: after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation . generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel . here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together .", "entity": "multiparallel word alignment", "output": "community detection algorithms", "neg_sample": ["multiparallel word alignment is done by using Method", "after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation .", "generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel .", "here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together ."], "relation": "used for", "id": "2022.findings-acl.108", "year": 2022, "rel_sent": "We show that community detection algorithms can provide valuable information for multiparallel word alignment .", "forward": false, "src_ids": "2022.findings-acl.108_2079"}
+{"input": "graph neural networks ( gnns ) is used for OtherScientificTerm| context: after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation . generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel . here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together .", "entity": "graph neural networks ( gnns )", "output": "graph structure", "neg_sample": ["graph neural networks ( gnns ) is used for OtherScientificTerm", "after a period of decrease , interest in word alignments is increasing again for their usefulness in domains such as typological research , cross - lingual annotation projection and machine translation .", "generally , alignment algorithms only use bitext and do not make use of the fact that many parallel corpora are multiparallel .", "here , we compute high - quality word alignments between multiple language pairs by considering all language pairs together ."], "relation": "used for", "id": "2022.findings-acl.108", "year": 2022, "rel_sent": "Next , we use graph neural networks ( GNNs ) to exploit the graph structure .", "forward": true, "src_ids": "2022.findings-acl.108_2080"}
+{"input": "oie tasks is done by using OtherScientificTerm| context: different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements .", "entity": "oie tasks", "output": "oia graph", "neg_sample": ["oie tasks is done by using OtherScientificTerm", "different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements ."], "relation": "used for", "id": "2022.acl-long.430", "year": 2022, "rel_sent": "OIE@OIA follows the methodology of Open Information eXpression ( OIX ): parsing a sentence to an Open Information Annotation ( OIA ) Graph and then adapting the OIA graph to different OIE tasks with simple rules .", "forward": false, "src_ids": "2022.acl-long.430_2081"}
+{"input": "learning algorithm is used for OtherScientificTerm| context: different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements .", "entity": "learning algorithm", "output": "oia graph", "neg_sample": ["learning algorithm is used for OtherScientificTerm", "different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements ."], "relation": "used for", "id": "2022.acl-long.430", "year": 2022, "rel_sent": "As the core of our OIE@OIA system , we implement an end - to - end OIA generator by annotating a dataset ( we make it open available ) and designing an efficient learning algorithm for the complex OIA graph .", "forward": true, "src_ids": "2022.acl-long.430_2082"}
+{"input": "oia graph is used for Task| context: different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements .", "entity": "oia graph", "output": "oie tasks", "neg_sample": ["oia graph is used for Task", "different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements ."], "relation": "used for", "id": "2022.acl-long.430", "year": 2022, "rel_sent": "OIE@OIA follows the methodology of Open Information eXpression ( OIX ): parsing a sentence to an Open Information Annotation ( OIA ) Graph and then adapting the OIA graph to different OIE tasks with simple rules .", "forward": true, "src_ids": "2022.acl-long.430_2083"}
+{"input": "oie@oia system is used for Task| context: different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements .", "entity": "oie@oia system", "output": "oie tasks", "neg_sample": ["oie@oia system is used for Task", "different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements ."], "relation": "used for", "id": "2022.acl-long.430", "year": 2022, "rel_sent": "We easily adapt the OIE@OIA system to accomplish three popular OIE tasks .", "forward": true, "src_ids": "2022.acl-long.430_2084"}
+{"input": "oie tasks is done by using Method| context: different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements .", "entity": "oie tasks", "output": "oie@oia system", "neg_sample": ["oie tasks is done by using Method", "different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements ."], "relation": "used for", "id": "2022.acl-long.430", "year": 2022, "rel_sent": "We easily adapt the OIE@OIA system to accomplish three popular OIE tasks .", "forward": false, "src_ids": "2022.acl-long.430_2085"}
+{"input": "oia graph is done by using Method| context: different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements .", "entity": "oia graph", "output": "learning algorithm", "neg_sample": ["oia graph is done by using Method", "different open information extraction ( oie ) tasks require different types of information , so the oie field requires strong adaptability of oie algorithms to meet different task requirements ."], "relation": "used for", "id": "2022.acl-long.430", "year": 2022, "rel_sent": "As the core of our OIE@OIA system , we implement an end - to - end OIA generator by annotating a dataset ( we make it open available ) and designing an efficient learning algorithm for the complex OIA graph .", "forward": false, "src_ids": "2022.acl-long.430_2086"}
+{"input": "cross - modal retrieval is done by using Method| context: current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image . while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "cross - modal retrieval", "output": "cooperative and joint approaches", "neg_sample": ["cross - modal retrieval is done by using Method", "current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image .", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "Retrieve Fast , Rerank Smart : Cooperative and Joint Approaches for Improved Cross - Modal Retrieval.", "forward": false, "src_ids": "2022.tacl-1.29_2087"}
+{"input": "cooperative and joint approaches is used for Task| context: while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "cooperative and joint approaches", "output": "cross - modal retrieval", "neg_sample": ["cooperative and joint approaches is used for Task", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "Retrieve Fast , Rerank Smart : Cooperative and Joint Approaches for Improved Cross - Modal Retrieval.", "forward": true, "src_ids": "2022.tacl-1.29_2088"}
+{"input": "pretrained text - image multi - modal model is done by using Method| context: current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image . while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "pretrained text - image multi - modal model", "output": "fine - tuning framework", "neg_sample": ["pretrained text - image multi - modal model is done by using Method", "current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image .", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "To address these crucial gaps towards both improved and efficient cross- modal retrieval , we propose a novel fine - tuning framework that turns any pretrained text - image multi - modal model into an efficient retrieval model .", "forward": false, "src_ids": "2022.tacl-1.29_2089"}
+{"input": "retrieval model is done by using Method| context: current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image . while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "retrieval model", "output": "fine - tuning framework", "neg_sample": ["retrieval model is done by using Method", "current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image .", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "To address these crucial gaps towards both improved and efficient cross- modal retrieval , we propose a novel fine - tuning framework that turns any pretrained text - image multi - modal model into an efficient retrieval model .", "forward": false, "src_ids": "2022.tacl-1.29_2090"}
+{"input": "fine - tuning framework is used for Method| context: current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image . while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "fine - tuning framework", "output": "pretrained text - image multi - modal model", "neg_sample": ["fine - tuning framework is used for Method", "current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image .", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "To address these crucial gaps towards both improved and efficient cross- modal retrieval , we propose a novel fine - tuning framework that turns any pretrained text - image multi - modal model into an efficient retrieval model .", "forward": true, "src_ids": "2022.tacl-1.29_2091"}
+{"input": "ranking is done by using Method| context: current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image . while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "ranking", "output": "cross - encoder component", "neg_sample": ["ranking is done by using Method", "current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image .", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "The framework is based on a cooperative retrieve - and - rerank approach that combines : 1 ) twin networks ( i.e. , a bi - encoder ) to separately encode all items of a corpus , enabling efficient initial retrieval , and 2 ) a cross - encoder component for a more nuanced ( i.e. , smarter ) ranking of the retrieved small set of items .", "forward": false, "src_ids": "2022.tacl-1.29_2092"}
+{"input": "cross - encoder component is used for Task| context: current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image . while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications .", "entity": "cross - encoder component", "output": "ranking", "neg_sample": ["cross - encoder component is used for Task", "current state - of - the - art approaches to cross- modal retrieval process text and visual input jointly , relying on transformer - based architectures with cross - attention mechanisms that attend over all words and objects in an image .", "while offering unmatched retrieval performance , such models : 1 ) are typically pretrained from scratch and thus less scalable , 2 ) suffer from huge retrieval latency and inefficiency issues , which makes them impractical in realistic applications ."], "relation": "used for", "id": "2022.tacl-1.29", "year": 2022, "rel_sent": "The framework is based on a cooperative retrieve - and - rerank approach that combines : 1 ) twin networks ( i.e. , a bi - encoder ) to separately encode all items of a corpus , enabling efficient initial retrieval , and 2 ) a cross - encoder component for a more nuanced ( i.e. , smarter ) ranking of the retrieved small set of items .", "forward": true, "src_ids": "2022.tacl-1.29_2093"}
+{"input": "complex question answering over knowledge base complex question answering over knowledge base is done by using Method| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "complex question answering over knowledge base complex question answering over knowledge base", "output": "kqa pro", "neg_sample": ["complex question answering over knowledge base complex question answering over knowledge base is done by using Method", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "Experimental results show that state - of - the - art KBQA methods can not achieve promising results on KQA Pro as on current datasets , which suggests that KQA Pro is challenging and Complex KBQA requires further research efforts .", "forward": false, "src_ids": "2022.acl-long.422_2094"}
+{"input": "reasoning skills is done by using Method| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "reasoning skills", "output": "kqa pro", "neg_sample": ["reasoning skills is done by using Method", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "We also treat KQA Pro as a diagnostic dataset for testing multiple reasoning skills , conduct a thorough evaluation of existing models and discuss further directions for Complex KBQA .", "forward": false, "src_ids": "2022.acl-long.422_2095"}
+{"input": "complex question answering is done by using Method| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "complex question answering", "output": "compositional programs", "neg_sample": ["complex question answering is done by using Method", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "KQA Pro : A Dataset with Explicit Compositional Programs for Complex Question Answering over Knowledge Base.", "forward": false, "src_ids": "2022.acl-long.422_2096"}
+{"input": "compositional programs is used for Task| context: existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "compositional programs", "output": "complex question answering", "neg_sample": ["compositional programs is used for Task", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "KQA Pro : A Dataset with Explicit Compositional Programs for Complex Question Answering over Knowledge Base.", "forward": true, "src_ids": "2022.acl-long.422_2097"}
+{"input": "kqa pro is used for Task| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "entity": "kqa pro", "output": "complex question answering over knowledge base complex question answering over knowledge base", "neg_sample": ["kqa pro is used for Task", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "Experimental results show that state - of - the - art KBQA methods can not achieve promising results on KQA Pro as on current datasets , which suggests that KQA Pro is challenging and Complex KBQA requires further research efforts .", "forward": true, "src_ids": "2022.acl-long.422_2098"}
+{"input": "reasoning process of complex questions is done by using Method| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "reasoning process of complex questions", "output": "compositional and interpretable programming language kopl", "neg_sample": ["reasoning process of complex questions is done by using Method", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "We introduce a compositional and interpretable programming language KoPL to represent the reasoning process of complex questions .", "forward": false, "src_ids": "2022.acl-long.422_2099"}
+{"input": "compositional and interpretable programming language kopl is used for Task| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "compositional and interpretable programming language kopl", "output": "reasoning process of complex questions", "neg_sample": ["compositional and interpretable programming language kopl is used for Task", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "We introduce a compositional and interpretable programming language KoPL to represent the reasoning process of complex questions .", "forward": true, "src_ids": "2022.acl-long.422_2100"}
+{"input": "reasoning skills is done by using Material| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "reasoning skills", "output": "diagnostic dataset", "neg_sample": ["reasoning skills is done by using Material", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "We also treat KQA Pro as a diagnostic dataset for testing multiple reasoning skills , conduct a thorough evaluation of existing models and discuss further directions for Complex KBQA .", "forward": false, "src_ids": "2022.acl-long.422_2101"}
+{"input": "diagnostic dataset is used for OtherScientificTerm| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "diagnostic dataset", "output": "reasoning skills", "neg_sample": ["diagnostic dataset is used for OtherScientificTerm", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "We also treat KQA Pro as a diagnostic dataset for testing multiple reasoning skills , conduct a thorough evaluation of existing models and discuss further directions for Complex KBQA .", "forward": true, "src_ids": "2022.acl-long.422_2102"}
+{"input": "kqa pro is used for OtherScientificTerm| context: complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc . existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale .", "entity": "kqa pro", "output": "reasoning skills", "neg_sample": ["kqa pro is used for OtherScientificTerm", "complex question answering over knowledge base ( complex kbqa ) is challenging because it requires various compositional reasoning capabilities , such as multi - hop inference , attribute comparison , set operation , etc .", "existing benchmarks have some shortcomings that limit the development of complex kbqa : 1 ) they only provide qa pairs without explicit reasoning processes ; 2 ) questions are poor in diversity or scale ."], "relation": "used for", "id": "2022.acl-long.422", "year": 2022, "rel_sent": "We also treat KQA Pro as a diagnostic dataset for testing multiple reasoning skills , conduct a thorough evaluation of existing models and discuss further directions for Complex KBQA .", "forward": true, "src_ids": "2022.acl-long.422_2103"}
+{"input": "knowledge graph extraction is done by using Method| context: acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators . adoption of unsupervised techniques for automated annotation have thus become popular . however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases . while such resources are abundant for general domains , they are scarce for specialised technical domains .", "entity": "knowledge graph extraction", "output": "quickgraph", "neg_sample": ["knowledge graph extraction is done by using Method", "acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators .", "adoption of unsupervised techniques for automated annotation have thus become popular .", "however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases .", "while such resources are abundant for general domains , they are scarce for specialised technical domains ."], "relation": "used for", "id": "2022.acl-demo.27", "year": 2022, "rel_sent": "To tackle this challenge , we present QuickGraph , the first collaborative MT - IE annotation tool built with indirect weak supervision and clustering to maximise annotator productivity . QuickGraph 's main contribution is a set of novel features that enable knowledge graph extraction through rapid and consistent complex multi - task entity and relation annotation .", "forward": false, "src_ids": "2022.acl-demo.27_2104"}
+{"input": "knowledge graph extraction is done by using Method| context: acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators . adoption of unsupervised techniques for automated annotation have thus become popular . however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases . while such resources are abundant for general domains , they are scarce for specialised technical domains .", "entity": "knowledge graph extraction", "output": "rapid annotation tool", "neg_sample": ["knowledge graph extraction is done by using Method", "acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators .", "adoption of unsupervised techniques for automated annotation have thus become popular .", "however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases .", "while such resources are abundant for general domains , they are scarce for specialised technical domains ."], "relation": "used for", "id": "2022.acl-demo.27", "year": 2022, "rel_sent": "QuickGraph : A Rapid Annotation Tool for Knowledge Graph Extraction from Technical Text.", "forward": false, "src_ids": "2022.acl-demo.27_2105"}
+{"input": "rapid annotation tool is used for Task| context: acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators . adoption of unsupervised techniques for automated annotation have thus become popular . however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases . while such resources are abundant for general domains , they are scarce for specialised technical domains .", "entity": "rapid annotation tool", "output": "knowledge graph extraction", "neg_sample": ["rapid annotation tool is used for Task", "acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators .", "adoption of unsupervised techniques for automated annotation have thus become popular .", "however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases .", "while such resources are abundant for general domains , they are scarce for specialised technical domains ."], "relation": "used for", "id": "2022.acl-demo.27", "year": 2022, "rel_sent": "QuickGraph : A Rapid Annotation Tool for Knowledge Graph Extraction from Technical Text.", "forward": true, "src_ids": "2022.acl-demo.27_2106"}
+{"input": "features is used for Task| context: acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators . adoption of unsupervised techniques for automated annotation have thus become popular . however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases . while such resources are abundant for general domains , they are scarce for specialised technical domains .", "entity": "features", "output": "knowledge graph extraction", "neg_sample": ["features is used for Task", "acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators .", "adoption of unsupervised techniques for automated annotation have thus become popular .", "however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases .", "while such resources are abundant for general domains , they are scarce for specialised technical domains ."], "relation": "used for", "id": "2022.acl-demo.27", "year": 2022, "rel_sent": "To tackle this challenge , we present QuickGraph , the first collaborative MT - IE annotation tool built with indirect weak supervision and clustering to maximise annotator productivity . QuickGraph 's main contribution is a set of novel features that enable knowledge graph extraction through rapid and consistent complex multi - task entity and relation annotation .", "forward": true, "src_ids": "2022.acl-demo.27_2107"}
+{"input": "quickgraph is used for Task| context: acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators . adoption of unsupervised techniques for automated annotation have thus become popular . however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases . while such resources are abundant for general domains , they are scarce for specialised technical domains .", "entity": "quickgraph", "output": "knowledge graph extraction", "neg_sample": ["quickgraph is used for Task", "acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators .", "adoption of unsupervised techniques for automated annotation have thus become popular .", "however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases .", "while such resources are abundant for general domains , they are scarce for specialised technical domains ."], "relation": "used for", "id": "2022.acl-demo.27", "year": 2022, "rel_sent": "To tackle this challenge , we present QuickGraph , the first collaborative MT - IE annotation tool built with indirect weak supervision and clustering to maximise annotator productivity . QuickGraph 's main contribution is a set of novel features that enable knowledge graph extraction through rapid and consistent complex multi - task entity and relation annotation .", "forward": true, "src_ids": "2022.acl-demo.27_2108"}
+{"input": "knowledge graph extraction is done by using OtherScientificTerm| context: acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators . adoption of unsupervised techniques for automated annotation have thus become popular . however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases . while such resources are abundant for general domains , they are scarce for specialised technical domains .", "entity": "knowledge graph extraction", "output": "features", "neg_sample": ["knowledge graph extraction is done by using OtherScientificTerm", "acquiring high - quality annotated corpora for complex multi - task information extraction ( mt - ie ) is an arduous and costly process for human - annotators .", "adoption of unsupervised techniques for automated annotation have thus become popular .", "however , these techniques rely heavily on dictionaries , gazetteers , and knowledge bases .", "while such resources are abundant for general domains , they are scarce for specialised technical domains ."], "relation": "used for", "id": "2022.acl-demo.27", "year": 2022, "rel_sent": "To tackle this challenge , we present QuickGraph , the first collaborative MT - IE annotation tool built with indirect weak supervision and clustering to maximise annotator productivity . QuickGraph 's main contribution is a set of novel features that enable knowledge graph extraction through rapid and consistent complex multi - task entity and relation annotation .", "forward": false, "src_ids": "2022.acl-demo.27_2109"}
+{"input": "anomaly detector is used for Metric| context: latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "anomaly detector", "output": "robustness", "neg_sample": ["anomaly detector is used for Metric", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "We question the validity of the current evaluation of robustness of PrLMs based on these non - natural adversarial samples and propose an anomaly detector to evaluate the robustness of PrLMs with more natural adversarial samples .", "forward": true, "src_ids": "2022.findings-acl.73_2110"}
+{"input": "anomaly detector is used for Metric| context: latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "anomaly detector", "output": "robustness", "neg_sample": ["anomaly detector is used for Metric", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "( 2 ) We apply the anomaly detector to a defense framework to enhance the robustness of PrLMs .", "forward": true, "src_ids": "2022.findings-acl.73_2111"}
+{"input": "anomaly detector is used for Method| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "anomaly detector", "output": "defense frameworks", "neg_sample": ["anomaly detector is used for Method", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "( 2 ) We apply the anomaly detector to a defense framework to enhance the robustness of PrLMs .", "forward": true, "src_ids": "2022.findings-acl.73_2112"}
+{"input": "robustness is done by using Method| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "robustness", "output": "anomaly detector", "neg_sample": ["robustness is done by using Method", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "We question the validity of the current evaluation of robustness of PrLMs based on these non - natural adversarial samples and propose an anomaly detector to evaluate the robustness of PrLMs with more natural adversarial samples .", "forward": false, "src_ids": "2022.findings-acl.73_2113"}
+{"input": "defense frameworks is done by using Method| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "defense frameworks", "output": "anomaly detector", "neg_sample": ["defense frameworks is done by using Method", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "( 2 ) We apply the anomaly detector to a defense framework to enhance the robustness of PrLMs .", "forward": false, "src_ids": "2022.findings-acl.73_2114"}
+{"input": "pre - trained language models ( prlms ) is done by using Method| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "pre - trained language models ( prlms )", "output": "anomaly detector", "neg_sample": ["pre - trained language models ( prlms ) is done by using Method", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "( 2 ) We apply the anomaly detector to a defense framework to enhance the robustness of PrLMs .", "forward": false, "src_ids": "2022.findings-acl.73_2115"}
+{"input": "robustness is done by using Method| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "robustness", "output": "anomaly detector", "neg_sample": ["robustness is done by using Method", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.73", "year": 2022, "rel_sent": "( 2 ) We apply the anomaly detector to a defense framework to enhance the robustness of PrLMs .", "forward": false, "src_ids": "2022.findings-acl.73_2116"}
+{"input": "product question answering is done by using Task| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "product question answering", "output": "semipqa", "neg_sample": ["product question answering is done by using Task", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "semiPQA : A Study on Product Question Answering over Semi - structured Data.", "forward": false, "src_ids": "2022.ecnlp-1.14_2117"}
+{"input": "product question answering is done by using Task| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "product question answering", "output": "semipqa", "neg_sample": ["product question answering is done by using Task", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "To this end , we present semiPQA : a dataset to benchmark PQA over semi - structured data .", "forward": false, "src_ids": "2022.ecnlp-1.14_2118"}
+{"input": "semipqa is used for Task| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "semipqa", "output": "product question answering", "neg_sample": ["semipqa is used for Task", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "semiPQA : A Study on Product Question Answering over Semi - structured Data.", "forward": true, "src_ids": "2022.ecnlp-1.14_2119"}
+{"input": "semi - structured answer sources is used for Task| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "semi - structured answer sources", "output": "product question answering", "neg_sample": ["semi - structured answer sources is used for Task", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "Tofill in this blank , here we study how to effectively incorporate semi - structured answer sources for PQA and focus on presenting answers in a natural , fluent sentence .", "forward": true, "src_ids": "2022.ecnlp-1.14_2120"}
+{"input": "semipqa is used for Task| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "semipqa", "output": "product question answering", "neg_sample": ["semipqa is used for Task", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "To this end , we present semiPQA : a dataset to benchmark PQA over semi - structured data .", "forward": true, "src_ids": "2022.ecnlp-1.14_2121"}
+{"input": "semi - structured data is used for Task| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "semi - structured data", "output": "product question answering", "neg_sample": ["semi - structured data is used for Task", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "We provide baseline results and a deep analysis on the successes and challenges of leveraging semi - structured data for PQA .", "forward": true, "src_ids": "2022.ecnlp-1.14_2122"}
+{"input": "product question answering is done by using Material| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . however , little attention has been paid to them .", "entity": "product question answering", "output": "semi - structured data", "neg_sample": ["product question answering is done by using Material", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "We provide baseline results and a deep analysis on the successes and challenges of leveraging semi - structured data for PQA .", "forward": false, "src_ids": "2022.ecnlp-1.14_2123"}
+{"input": "product question answering is done by using Material| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "product question answering", "output": "semi - structured answer sources", "neg_sample": ["product question answering is done by using Material", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "Tofill in this blank , here we study how to effectively incorporate semi - structured answer sources for PQA and focus on presenting answers in a natural , fluent sentence .", "forward": false, "src_ids": "2022.ecnlp-1.14_2124"}
+{"input": "seen attribute types is done by using Method| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "seen attribute types", "output": "neural models", "neg_sample": ["seen attribute types is done by using Method", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "In general , state - of - the - art neural models can perform remarkably well when dealing with seen attribute types .", "forward": false, "src_ids": "2022.ecnlp-1.14_2125"}
+{"input": "neural models is used for OtherScientificTerm| context: current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas . apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g. , key - value pairs , lists , tables , json and xml files , etc . these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases . however , little attention has been paid to them .", "entity": "neural models", "output": "seen attribute types", "neg_sample": ["neural models is used for OtherScientificTerm", "current research mainly focuses on finding answers from either unstructured text , like product descriptions and user reviews , or structured knowledge bases with pre - defined schemas .", "apart from the above two sources , a lot of product information is represented in a semi - structured way , e.g.", ", key - value pairs , lists , tables , json and xml files , etc .", "these semi - structured data can be a valuable answer source since they are better organized than free text , while being easier to construct than structured knowledge bases .", "however , little attention has been paid to them ."], "relation": "used for", "id": "2022.ecnlp-1.14", "year": 2022, "rel_sent": "In general , state - of - the - art neural models can perform remarkably well when dealing with seen attribute types .", "forward": true, "src_ids": "2022.ecnlp-1.14_2126"}
+{"input": "out - of - domain explicitly abusive utterances is done by using Method| context: robustness of machine learning models on ever - changing real - world data is critical , especially for applications affecting human well - being such as content moderation . new kinds of abusive language continually emerge in online discussions in response to current events ( e.g. , covid-19 ) , and the deployed abuse detection systems should be updated regularly to remain accurate .", "entity": "out - of - domain explicitly abusive utterances", "output": "general abusive language classifiers", "neg_sample": ["out - of - domain explicitly abusive utterances is done by using Method", "robustness of machine learning models on ever - changing real - world data is critical , especially for applications affecting human well - being such as content moderation .", "new kinds of abusive language continually emerge in online discussions in response to current events ( e.g.", ", covid-19 ) , and the deployed abuse detection systems should be updated regularly to remain accurate ."], "relation": "used for", "id": "2022.acl-long.378", "year": 2022, "rel_sent": "In this paper , we show that general abusive language classifiers tend to be fairly reliable in detecting out - of - domain explicitly abusive utterances but fail to detect new types of more subtle , implicit abuse .", "forward": false, "src_ids": "2022.acl-long.378_2127"}
+{"input": "general abusive language classifiers is used for OtherScientificTerm| context: robustness of machine learning models on ever - changing real - world data is critical , especially for applications affecting human well - being such as content moderation . new kinds of abusive language continually emerge in online discussions in response to current events ( e.g. , covid-19 ) , and the deployed abuse detection systems should be updated regularly to remain accurate .", "entity": "general abusive language classifiers", "output": "out - of - domain explicitly abusive utterances", "neg_sample": ["general abusive language classifiers is used for OtherScientificTerm", "robustness of machine learning models on ever - changing real - world data is critical , especially for applications affecting human well - being such as content moderation .", "new kinds of abusive language continually emerge in online discussions in response to current events ( e.g.", ", covid-19 ) , and the deployed abuse detection systems should be updated regularly to remain accurate ."], "relation": "used for", "id": "2022.acl-long.378", "year": 2022, "rel_sent": "In this paper , we show that general abusive language classifiers tend to be fairly reliable in detecting out - of - domain explicitly abusive utterances but fail to detect new types of more subtle , implicit abuse .", "forward": true, "src_ids": "2022.acl-long.378_2128"}
+{"input": "fake news detection is done by using Method| context: fake news detection is crucial for preventing the dissemination of misinformation on social media . to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies . however , these methods neglect the information in the external news environment where a fake news post is created and disseminated . the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread .", "entity": "fake news detection", "output": "news environment perception", "neg_sample": ["fake news detection is done by using Method", "fake news detection is crucial for preventing the dissemination of misinformation on social media .", "to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies .", "however , these methods neglect the information in the external news environment where a fake news post is created and disseminated .", "the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread ."], "relation": "used for", "id": "2022.acl-long.311", "year": 2022, "rel_sent": "Zoom Out and Observe : News Environment Perception for Fake News Detection.", "forward": false, "src_ids": "2022.acl-long.311_2129"}
+{"input": "news environment perception is used for Task| context: to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies . however , these methods neglect the information in the external news environment where a fake news post is created and disseminated . the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread .", "entity": "news environment perception", "output": "fake news detection", "neg_sample": ["news environment perception is used for Task", "to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies .", "however , these methods neglect the information in the external news environment where a fake news post is created and disseminated .", "the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread ."], "relation": "used for", "id": "2022.acl-long.311", "year": 2022, "rel_sent": "Zoom Out and Observe : News Environment Perception for Fake News Detection.", "forward": true, "src_ids": "2022.acl-long.311_2130"}
+{"input": "fake news detectors is done by using Method| context: fake news detection is crucial for preventing the dissemination of misinformation on social media . to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies . however , these methods neglect the information in the external news environment where a fake news post is created and disseminated . the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread .", "entity": "fake news detectors", "output": "news environment perception framework", "neg_sample": ["fake news detectors is done by using Method", "fake news detection is crucial for preventing the dissemination of misinformation on social media .", "to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies .", "however , these methods neglect the information in the external news environment where a fake news post is created and disseminated .", "the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread ."], "relation": "used for", "id": "2022.acl-long.311", "year": 2022, "rel_sent": "Experiments on our newly built datasets show that the NEP can efficiently improve the performance of basic fake news detectors .", "forward": false, "src_ids": "2022.acl-long.311_2131"}
+{"input": "news environment perception framework is used for Method| context: fake news detection is crucial for preventing the dissemination of misinformation on social media . to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies . however , these methods neglect the information in the external news environment where a fake news post is created and disseminated . the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread .", "entity": "news environment perception framework", "output": "fake news detectors", "neg_sample": ["news environment perception framework is used for Method", "fake news detection is crucial for preventing the dissemination of misinformation on social media .", "to differentiate fake news from real ones , existing methods observe the language patterns of the news post and ' zoom in ' to verify its content with knowledge sources or check its readers ' replies .", "however , these methods neglect the information in the external news environment where a fake news post is created and disseminated .", "the news environment represents recent mainstream media opinion and public attention , which is an important inspiration of fake news fabrication because fake news is often designed to ride the wave of popular events and catch public attention with unexpected novel content for greater exposure and spread ."], "relation": "used for", "id": "2022.acl-long.311", "year": 2022, "rel_sent": "Experiments on our newly built datasets show that the NEP can efficiently improve the performance of basic fake news detectors .", "forward": true, "src_ids": "2022.acl-long.311_2132"}
+{"input": "tabular reasoning is done by using Method| context: natural language inference on tabular data is a challenging task . existing approaches lack the world and common sense knowledge required to perform at a human level . while massive amounts of kg data exist , approaches to integrate them with deep learning models to enhance tabular reasoning are uncommon .", "entity": "tabular reasoning", "output": "external knowledge enhanced transformer bilstm model", "neg_sample": ["tabular reasoning is done by using Method", "natural language inference on tabular data is a challenging task .", "existing approaches lack the world and common sense knowledge required to perform at a human level .", "while massive amounts of kg data exist , approaches to integrate them with deep learning models to enhance tabular reasoning are uncommon ."], "relation": "used for", "id": "2022.deelio-1.7", "year": 2022, "rel_sent": "Trans - KBLSTM : An External Knowledge Enhanced Transformer BiLSTM Model for Tabular Reasoning.", "forward": false, "src_ids": "2022.deelio-1.7_2133"}
+{"input": "external knowledge enhanced transformer bilstm model is used for Task| context: natural language inference on tabular data is a challenging task . existing approaches lack the world and common sense knowledge required to perform at a human level .", "entity": "external knowledge enhanced transformer bilstm model", "output": "tabular reasoning", "neg_sample": ["external knowledge enhanced transformer bilstm model is used for Task", "natural language inference on tabular data is a challenging task .", "existing approaches lack the world and common sense knowledge required to perform at a human level ."], "relation": "used for", "id": "2022.deelio-1.7", "year": 2022, "rel_sent": "Trans - KBLSTM : An External Knowledge Enhanced Transformer BiLSTM Model for Tabular Reasoning.", "forward": true, "src_ids": "2022.deelio-1.7_2134"}
+{"input": "procedural multimodal machine comprehension is done by using Method| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step . comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) .", "entity": "procedural multimodal machine comprehension", "output": "temporal - modal entity graph ( tmeg )", "neg_sample": ["procedural multimodal machine comprehension is done by using Method", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "Modeling Temporal - Modal Entity Graph for Procedural Multimodal Machine Comprehension.", "forward": false, "src_ids": "2022.acl-long.84_2135"}
+{"input": "temporal - modal entity graph ( tmeg ) is used for Task| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "entity": "temporal - modal entity graph ( tmeg )", "output": "procedural multimodal machine comprehension", "neg_sample": ["temporal - modal entity graph ( tmeg ) is used for Task", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "Modeling Temporal - Modal Entity Graph for Procedural Multimodal Machine Comprehension.", "forward": true, "src_ids": "2022.acl-long.84_2136"}
+{"input": "textual and visual entities is done by using OtherScientificTerm| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step . comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) .", "entity": "textual and visual entities", "output": "graph structure", "neg_sample": ["textual and visual entities is done by using OtherScientificTerm", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "Specifically , graph structure is formulated to capture textual and visual entities and trace their temporal - modal evolution .", "forward": false, "src_ids": "2022.acl-long.84_2137"}
+{"input": "temporal - modal evolution is done by using OtherScientificTerm| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step . comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) .", "entity": "temporal - modal evolution", "output": "graph structure", "neg_sample": ["temporal - modal evolution is done by using OtherScientificTerm", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "Specifically , graph structure is formulated to capture textual and visual entities and trace their temporal - modal evolution .", "forward": false, "src_ids": "2022.acl-long.84_2138"}
+{"input": "graph structure is used for OtherScientificTerm| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step . comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) .", "entity": "graph structure", "output": "textual and visual entities", "neg_sample": ["graph structure is used for OtherScientificTerm", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "Specifically , graph structure is formulated to capture textual and visual entities and trace their temporal - modal evolution .", "forward": true, "src_ids": "2022.acl-long.84_2139"}
+{"input": "graph encoding and reasoning is done by using Method| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step . comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) .", "entity": "graph encoding and reasoning", "output": "graph aggregation module", "neg_sample": ["graph encoding and reasoning is done by using Method", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "In addition , a graph aggregation module is introduced to conduct graph encoding and reasoning .", "forward": false, "src_ids": "2022.acl-long.84_2140"}
+{"input": "graph aggregation module is used for Task| context: procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step . comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) .", "entity": "graph aggregation module", "output": "graph encoding and reasoning", "neg_sample": ["graph aggregation module is used for Task", "procedural multimodal documents ( pmds ) organize textual instructions and corresponding images step by step .", "comprehending pmds and inducing their representations for the downstream reasoning tasks is designated as procedural multimodal machine comprehension ( m3c ) ."], "relation": "used for", "id": "2022.acl-long.84", "year": 2022, "rel_sent": "In addition , a graph aggregation module is introduced to conduct graph encoding and reasoning .", "forward": true, "src_ids": "2022.acl-long.84_2141"}
+{"input": "word meaning is done by using OtherScientificTerm| context: natural language processing models learn word representations based on the distributional hypothesis , which asserts that word context ( e.g. , co - occurrence ) correlates with meaning .", "entity": "word meaning", "output": "n - grams", "neg_sample": ["word meaning is done by using OtherScientificTerm", "natural language processing models learn word representations based on the distributional hypothesis , which asserts that word context ( e.g.", ", co - occurrence ) correlates with meaning ."], "relation": "used for", "id": "2022.acl-long.492", "year": 2022, "rel_sent": "We propose that n - grams composed of random character sequences , or garble , provide a novel context for studying word meaning both within and beyond extant language .", "forward": false, "src_ids": "2022.acl-long.492_2142"}
+{"input": "n - grams is used for OtherScientificTerm| context: natural language processing models learn word representations based on the distributional hypothesis , which asserts that word context ( e.g. , co - occurrence ) correlates with meaning . in particular , randomly generated character n - grams lack meaning but contain primitive information based on the distribution of characters they contain .", "entity": "n - grams", "output": "word meaning", "neg_sample": ["n - grams is used for OtherScientificTerm", "natural language processing models learn word representations based on the distributional hypothesis , which asserts that word context ( e.g.", ", co - occurrence ) correlates with meaning .", "in particular , randomly generated character n - grams lack meaning but contain primitive information based on the distribution of characters they contain ."], "relation": "used for", "id": "2022.acl-long.492", "year": 2022, "rel_sent": "We propose that n - grams composed of random character sequences , or garble , provide a novel context for studying word meaning both within and beyond extant language .", "forward": true, "src_ids": "2022.acl-long.492_2143"}
+{"input": "counseling dialogues is done by using Task| context: in this paper , we study the effect of commonsense and domain knowledge while generating responses in counseling conversations using retrieval and generative methods for knowledge integration .", "entity": "counseling dialogues", "output": "knowledge enhanced reflection generation", "neg_sample": ["counseling dialogues is done by using Task", "in this paper , we study the effect of commonsense and domain knowledge while generating responses in counseling conversations using retrieval and generative methods for knowledge integration ."], "relation": "used for", "id": "2022.acl-long.221", "year": 2022, "rel_sent": "Knowledge Enhanced Reflection Generation for Counseling Dialogues.", "forward": false, "src_ids": "2022.acl-long.221_2144"}
+{"input": "knowledge enhanced reflection generation is used for Task| context: in this paper , we study the effect of commonsense and domain knowledge while generating responses in counseling conversations using retrieval and generative methods for knowledge integration .", "entity": "knowledge enhanced reflection generation", "output": "counseling dialogues", "neg_sample": ["knowledge enhanced reflection generation is used for Task", "in this paper , we study the effect of commonsense and domain knowledge while generating responses in counseling conversations using retrieval and generative methods for knowledge integration ."], "relation": "used for", "id": "2022.acl-long.221", "year": 2022, "rel_sent": "Knowledge Enhanced Reflection Generation for Counseling Dialogues.", "forward": true, "src_ids": "2022.acl-long.221_2145"}
+{"input": "comet versions is done by using Method| context: in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed . these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g. scoring multiple systems or minimum bayes risk decoding ) .", "entity": "comet versions", "output": "optimization techniques", "neg_sample": ["comet versions is done by using Method", "in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed .", "these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g.", "scoring multiple systems or minimum bayes risk decoding ) ."], "relation": "used for", "id": "2022.eamt-1.9", "year": 2022, "rel_sent": "In this paper , we explore optimization techniques , pruning , and knowledge distillation to create more compact and faster COMET versions .", "forward": false, "src_ids": "2022.eamt-1.9_2146"}
+{"input": "knowledge distillation is used for Generic| context: in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed . these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g. scoring multiple systems or minimum bayes risk decoding ) .", "entity": "knowledge distillation", "output": "comet versions", "neg_sample": ["knowledge distillation is used for Generic", "in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed .", "these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g.", "scoring multiple systems or minimum bayes risk decoding ) ."], "relation": "used for", "id": "2022.eamt-1.9", "year": 2022, "rel_sent": "In this paper , we explore optimization techniques , pruning , and knowledge distillation to create more compact and faster COMET versions .", "forward": true, "src_ids": "2022.eamt-1.9_2147"}
+{"input": "optimization techniques is used for Generic| context: in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed . these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g. scoring multiple systems or minimum bayes risk decoding ) .", "entity": "optimization techniques", "output": "comet versions", "neg_sample": ["optimization techniques is used for Generic", "in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed .", "these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g.", "scoring multiple systems or minimum bayes risk decoding ) ."], "relation": "used for", "id": "2022.eamt-1.9", "year": 2022, "rel_sent": "In this paper , we explore optimization techniques , pruning , and knowledge distillation to create more compact and faster COMET versions .", "forward": true, "src_ids": "2022.eamt-1.9_2148"}
+{"input": "pruning is used for Generic| context: in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed . these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g. scoring multiple systems or minimum bayes risk decoding ) .", "entity": "pruning", "output": "comet versions", "neg_sample": ["pruning is used for Generic", "in recent years , several neural fine - tuned machine translation evaluation metrics such as comet and bleurt have been proposed .", "these metrics achieve much higher correlations with human judgments than lexical overlap metrics at the cost of computational efficiency and simplicity , limiting their applications to scenarios in which one has to score thousands of translation hypothesis ( e.g.", "scoring multiple systems or minimum bayes risk decoding ) ."], "relation": "used for", "id": "2022.eamt-1.9", "year": 2022, "rel_sent": "In this paper , we explore optimization techniques , pruning , and knowledge distillation to create more compact and faster COMET versions .", "forward": true, "src_ids": "2022.eamt-1.9_2149"}
+{"input": "cross - lingual word embeddings is done by using OtherScientificTerm| context: as high - quality malay language resources are still a scarcity , cross lingual word embeddings make it possible for richer english resources to be leveraged for downstream malay text classification tasks .", "entity": "cross - lingual word embeddings", "output": "malay word coverage", "neg_sample": ["cross - lingual word embeddings is done by using OtherScientificTerm", "as high - quality malay language resources are still a scarcity , cross lingual word embeddings make it possible for richer english resources to be leveraged for downstream malay text classification tasks ."], "relation": "used for", "id": "2022.acl-srw.16", "year": 2022, "rel_sent": "We also examined the effect of Malay word coverage on the quality of cross - lingual word embeddings .", "forward": false, "src_ids": "2022.acl-srw.16_2150"}
+{"input": "malay translations is done by using Method| context: as high - quality malay language resources are still a scarcity , cross lingual word embeddings make it possible for richer english resources to be leveraged for downstream malay text classification tasks .", "entity": "malay translations", "output": "english - malay embeddings alignment approach", "neg_sample": ["malay translations is done by using Method", "as high - quality malay language resources are still a scarcity , cross lingual word embeddings make it possible for richer english resources to be leveraged for downstream malay text classification tasks ."], "relation": "used for", "id": "2022.acl-srw.16", "year": 2022, "rel_sent": "As the English and Malay monolingual embeddings are pre - trained on informal language corpora , our proposed English - Malay embeddings alignment approach is also able to map non - standard Malay translations in the English nearest neighbours .", "forward": false, "src_ids": "2022.acl-srw.16_2151"}
+{"input": "english - malay embeddings alignment approach is used for Material| context: as high - quality malay language resources are still a scarcity , cross lingual word embeddings make it possible for richer english resources to be leveraged for downstream malay text classification tasks .", "entity": "english - malay embeddings alignment approach", "output": "malay translations", "neg_sample": ["english - malay embeddings alignment approach is used for Material", "as high - quality malay language resources are still a scarcity , cross lingual word embeddings make it possible for richer english resources to be leveraged for downstream malay text classification tasks ."], "relation": "used for", "id": "2022.acl-srw.16", "year": 2022, "rel_sent": "As the English and Malay monolingual embeddings are pre - trained on informal language corpora , our proposed English - Malay embeddings alignment approach is also able to map non - standard Malay translations in the English nearest neighbours .", "forward": true, "src_ids": "2022.acl-srw.16_2152"}
+{"input": "stock markets is done by using Method| context: with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert .", "entity": "stock markets", "output": "language models", "neg_sample": ["stock markets is done by using Method", "with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert ."], "relation": "used for", "id": "2022.acl-short.12", "year": 2022, "rel_sent": "For example , the language models are overall more positive towards the stock market , but there are significant differences in preferences between a pair of industry sectors , or even within a sector .", "forward": false, "src_ids": "2022.acl-short.12_2153"}
+{"input": "language models is used for Material| context: pretrained language models such as bert have achieved remarkable success in several nlp tasks . with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert .", "entity": "language models", "output": "stock markets", "neg_sample": ["language models is used for Material", "pretrained language models such as bert have achieved remarkable success in several nlp tasks .", "with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert ."], "relation": "used for", "id": "2022.acl-short.12", "year": 2022, "rel_sent": "For example , the language models are overall more positive towards the stock market , but there are significant differences in preferences between a pair of industry sectors , or even within a sector .", "forward": true, "src_ids": "2022.acl-short.12_2154"}
+{"input": "financial decision making systems is done by using Method| context: pretrained language models such as bert have achieved remarkable success in several nlp tasks . with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert .", "entity": "financial decision making systems", "output": "nlp models", "neg_sample": ["financial decision making systems is done by using Method", "pretrained language models such as bert have achieved remarkable success in several nlp tasks .", "with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert ."], "relation": "used for", "id": "2022.acl-short.12", "year": 2022, "rel_sent": "Given the prevalence of NLP models in financial decision making systems , this work raises the awareness of their potential implicit preferences in the stock markets .", "forward": false, "src_ids": "2022.acl-short.12_2155"}
+{"input": "nlp models is used for Task| context: pretrained language models such as bert have achieved remarkable success in several nlp tasks . with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert .", "entity": "nlp models", "output": "financial decision making systems", "neg_sample": ["nlp models is used for Task", "pretrained language models such as bert have achieved remarkable success in several nlp tasks .", "with the wide adoption of bert in real - world applications , researchers begin to investigate the implicit biases encoded in the bert ."], "relation": "used for", "id": "2022.acl-short.12", "year": 2022, "rel_sent": "Given the prevalence of NLP models in financial decision making systems , this work raises the awareness of their potential implicit preferences in the stock markets .", "forward": true, "src_ids": "2022.acl-short.12_2156"}
+{"input": "task knowledge is done by using Method| context: task - oriented personal assistants enable people to interact with a host of devices and services using natural language . one of the challenges of making neural dialogue systems available to more users is the lack of training data for all but a few languages .", "entity": "task knowledge", "output": "fine - tuned language models", "neg_sample": ["task knowledge is done by using Method", "task - oriented personal assistants enable people to interact with a host of devices and services using natural language .", "one of the challenges of making neural dialogue systems available to more users is the lack of training data for all but a few languages ."], "relation": "used for", "id": "2022.findings-acl.319", "year": 2022, "rel_sent": "A detailed qualitative error analysis of the best methods shows that our fine - tuned language models can zero - shot transfer the task knowledge better than anticipated .", "forward": false, "src_ids": "2022.findings-acl.319_2157"}
+{"input": "fine - tuned language models is used for OtherScientificTerm| context: task - oriented personal assistants enable people to interact with a host of devices and services using natural language . one of the challenges of making neural dialogue systems available to more users is the lack of training data for all but a few languages .", "entity": "fine - tuned language models", "output": "task knowledge", "neg_sample": ["fine - tuned language models is used for OtherScientificTerm", "task - oriented personal assistants enable people to interact with a host of devices and services using natural language .", "one of the challenges of making neural dialogue systems available to more users is the lack of training data for all but a few languages ."], "relation": "used for", "id": "2022.findings-acl.319", "year": 2022, "rel_sent": "A detailed qualitative error analysis of the best methods shows that our fine - tuned language models can zero - shot transfer the task knowledge better than anticipated .", "forward": true, "src_ids": "2022.findings-acl.319_2158"}
+{"input": "speech translation is done by using Method| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "speech translation", "output": "pretrained speech encoders", "neg_sample": ["speech translation is done by using Method", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "Pretrained Speech Encoders and Efficient Fine - tuning Methods for Speech Translation : UPC at IWSLT 2022.", "forward": false, "src_ids": "2022.iwslt-1.23_2159"}
+{"input": "fine - tuning methods is used for Task| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "fine - tuning methods", "output": "speech translation", "neg_sample": ["fine - tuning methods is used for Task", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "Pretrained Speech Encoders and Efficient Fine - tuning Methods for Speech Translation : UPC at IWSLT 2022.", "forward": true, "src_ids": "2022.iwslt-1.23_2160"}
+{"input": "pretrained speech encoders is used for Task| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "pretrained speech encoders", "output": "speech translation", "neg_sample": ["pretrained speech encoders is used for Task", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "Pretrained Speech Encoders and Efficient Fine - tuning Methods for Speech Translation : UPC at IWSLT 2022.", "forward": true, "src_ids": "2022.iwslt-1.23_2161"}
+{"input": "non - trainable layers is done by using Method| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "non - trainable layers", "output": "adapter modules", "neg_sample": ["non - trainable layers is done by using Method", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "We use an efficient fine - tuning technique that trains only specific layers of our system , and explore the use of adapter modules for the non - trainable layers .", "forward": false, "src_ids": "2022.iwslt-1.23_2162"}
+{"input": "adapter modules is used for OtherScientificTerm| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "adapter modules", "output": "non - trainable layers", "neg_sample": ["adapter modules is used for OtherScientificTerm", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "We use an efficient fine - tuning technique that trains only specific layers of our system , and explore the use of adapter modules for the non - trainable layers .", "forward": true, "src_ids": "2022.iwslt-1.23_2163"}
+{"input": "decoder ( mbart ) is done by using Method| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "decoder ( mbart )", "output": "machine translation model", "neg_sample": ["decoder ( mbart ) is done by using Method", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "We further investigate the suitability of different speech encoders ( wav2vec 2.0 , HuBERT ) for our models and the impact of knowledge distillation from the Machine Translation model that we use for the decoder ( mBART ) .", "forward": false, "src_ids": "2022.iwslt-1.23_2164"}
+{"input": "machine translation model is used for Method| context: the offline task involves translating english speech to german , japanese and chinese text .", "entity": "machine translation model", "output": "decoder ( mbart )", "neg_sample": ["machine translation model is used for Method", "the offline task involves translating english speech to german , japanese and chinese text ."], "relation": "used for", "id": "2022.iwslt-1.23", "year": 2022, "rel_sent": "We further investigate the suitability of different speech encoders ( wav2vec 2.0 , HuBERT ) for our models and the impact of knowledge distillation from the Machine Translation model that we use for the decoder ( mBART ) .", "forward": true, "src_ids": "2022.iwslt-1.23_2165"}
+{"input": "spatial commonsense is done by using Material| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "spatial commonsense", "output": "images", "neg_sample": ["spatial commonsense is done by using Material", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "Starting from the observation that images are more likely to exhibit spatial commonsense than texts , we explore whether models with visual signals learn more spatial commonsense than text - based PLMs .", "forward": false, "src_ids": "2022.acl-long.168_2166"}
+{"input": "spatial knowledge is done by using Method| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "spatial knowledge", "output": "image synthesis models", "neg_sample": ["spatial knowledge is done by using Method", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "We propose a spatial commonsense benchmark that focuses on the relative scales of objects , and the positional relationship between people and objects under different actions . We probe PLMs and models with visual signals , including vision - language pretrained models and image synthesis models , on this benchmark , and find that image synthesis models are more capable of learning accurate and consistent spatial knowledge than other models .", "forward": false, "src_ids": "2022.acl-long.168_2167"}
+{"input": "natural language understanding tasks is done by using Method| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "natural language understanding tasks", "output": "image synthesis models", "neg_sample": ["natural language understanding tasks is done by using Method", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "The spatial knowledge from image synthesis models also helps in natural language understanding tasks that require spatial commonsense .", "forward": false, "src_ids": "2022.acl-long.168_2168"}
+{"input": "natural language understanding tasks is done by using OtherScientificTerm| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "natural language understanding tasks", "output": "spatial knowledge", "neg_sample": ["natural language understanding tasks is done by using OtherScientificTerm", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "The spatial knowledge from image synthesis models also helps in natural language understanding tasks that require spatial commonsense .", "forward": false, "src_ids": "2022.acl-long.168_2169"}
+{"input": "image synthesis models is used for OtherScientificTerm| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "image synthesis models", "output": "spatial knowledge", "neg_sample": ["image synthesis models is used for OtherScientificTerm", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "We propose a spatial commonsense benchmark that focuses on the relative scales of objects , and the positional relationship between people and objects under different actions . We probe PLMs and models with visual signals , including vision - language pretrained models and image synthesis models , on this benchmark , and find that image synthesis models are more capable of learning accurate and consistent spatial knowledge than other models .", "forward": true, "src_ids": "2022.acl-long.168_2170"}
+{"input": "spatial knowledge is used for Task| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "spatial knowledge", "output": "natural language understanding tasks", "neg_sample": ["spatial knowledge is used for Task", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "The spatial knowledge from image synthesis models also helps in natural language understanding tasks that require spatial commonsense .", "forward": true, "src_ids": "2022.acl-long.168_2171"}
+{"input": "image synthesis models is used for Task| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "image synthesis models", "output": "natural language understanding tasks", "neg_sample": ["image synthesis models is used for Task", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.168", "year": 2022, "rel_sent": "The spatial knowledge from image synthesis models also helps in natural language understanding tasks that require spatial commonsense .", "forward": true, "src_ids": "2022.acl-long.168_2172"}
+{"input": "keyword lists of classes is done by using Method| context: input saliency methods have recently become a popular tool for explaining predictions of deep learning models in nlp . nevertheless , there has been little work investigating methods for aggregating prediction - level explanations to the class level , nor has a framework for evaluating such class explanations been established .", "entity": "keyword lists of classes", "output": "stable attribution class explanation method ( sacx )", "neg_sample": ["keyword lists of classes is done by using Method", "input saliency methods have recently become a popular tool for explaining predictions of deep learning models in nlp .", "nevertheless , there has been little work investigating methods for aggregating prediction - level explanations to the class level , nor has a framework for evaluating such class explanations been established ."], "relation": "used for", "id": "2022.findings-acl.85", "year": 2022, "rel_sent": "We explore explanations based on XLM - R and the Integrated Gradients input attribution method , and propose 1 ) the Stable Attribution Class Explanation method ( SACX ) to extract keyword lists of classes in text classification tasks , and 2 ) a framework for the systematic evaluation of the keyword lists .", "forward": false, "src_ids": "2022.findings-acl.85_2173"}
+{"input": "stable attribution class explanation method ( sacx ) is used for OtherScientificTerm| context: input saliency methods have recently become a popular tool for explaining predictions of deep learning models in nlp . nevertheless , there has been little work investigating methods for aggregating prediction - level explanations to the class level , nor has a framework for evaluating such class explanations been established .", "entity": "stable attribution class explanation method ( sacx )", "output": "keyword lists of classes", "neg_sample": ["stable attribution class explanation method ( sacx ) is used for OtherScientificTerm", "input saliency methods have recently become a popular tool for explaining predictions of deep learning models in nlp .", "nevertheless , there has been little work investigating methods for aggregating prediction - level explanations to the class level , nor has a framework for evaluating such class explanations been established ."], "relation": "used for", "id": "2022.findings-acl.85", "year": 2022, "rel_sent": "We explore explanations based on XLM - R and the Integrated Gradients input attribution method , and propose 1 ) the Stable Attribution Class Explanation method ( SACX ) to extract keyword lists of classes in text classification tasks , and 2 ) a framework for the systematic evaluation of the keyword lists .", "forward": true, "src_ids": "2022.findings-acl.85_2174"}
+{"input": "conversational goals is done by using Method| context: a limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses , primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge . one way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response .", "entity": "conversational goals", "output": "knowledge - augmentation", "neg_sample": ["conversational goals is done by using Method", "a limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses , primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge .", "one way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response ."], "relation": "used for", "id": "2022.acl-long.224", "year": 2022, "rel_sent": "We further show that knowledge - augmentation promotes success in achieving conversational goals in both experimental settings .", "forward": false, "src_ids": "2022.acl-long.224_2175"}
+{"input": "knowledge - augmentation is used for OtherScientificTerm| context: a limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses , primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge . one way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response .", "entity": "knowledge - augmentation", "output": "conversational goals", "neg_sample": ["knowledge - augmentation is used for OtherScientificTerm", "a limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses , primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge .", "one way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response ."], "relation": "used for", "id": "2022.acl-long.224", "year": 2022, "rel_sent": "We further show that knowledge - augmentation promotes success in achieving conversational goals in both experimental settings .", "forward": true, "src_ids": "2022.acl-long.224_2176"}
+{"input": "transformer architecture is used for Task| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "transformer architecture", "output": "fast long - range sequence modeling", "neg_sample": ["transformer architecture is used for Task", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "Therefore , in this paper , we design an efficient Transformer architecture , named Fourier Sparse Attention for Transformer ( FSAT ) , for fast long - range sequence modeling .", "forward": true, "src_ids": "2022.acl-long.19_2177"}
+{"input": "fast long - range sequence modeling is done by using Method| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "fast long - range sequence modeling", "output": "transformer architecture", "neg_sample": ["fast long - range sequence modeling is done by using Method", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "Therefore , in this paper , we design an efficient Transformer architecture , named Fourier Sparse Attention for Transformer ( FSAT ) , for fast long - range sequence modeling .", "forward": false, "src_ids": "2022.acl-long.19_2178"}
+{"input": "gradient truncation is used for Method| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "gradient truncation", "output": "fourier sparse attention for transformer ( fsat )", "neg_sample": ["gradient truncation is used for Method", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "By reparameterization and gradient truncation , FSAT successfully learned the index of dominant elements .", "forward": true, "src_ids": "2022.acl-long.19_2179"}
+{"input": "reparameterization is used for Method| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "reparameterization", "output": "fourier sparse attention for transformer ( fsat )", "neg_sample": ["reparameterization is used for Method", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "By reparameterization and gradient truncation , FSAT successfully learned the index of dominant elements .", "forward": true, "src_ids": "2022.acl-long.19_2180"}
+{"input": "attention matrix is done by using Method| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "attention matrix", "output": "sparse attention matrix estimation module", "neg_sample": ["attention matrix is done by using Method", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "( 2 ) A sparse attention matrix estimation module , which predicts dominant elements of an attention matrix based on the output of the previous hidden state cross module .", "forward": false, "src_ids": "2022.acl-long.19_2181"}
+{"input": "sparse attention matrix estimation module is used for OtherScientificTerm| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "entity": "sparse attention matrix estimation module", "output": "attention matrix", "neg_sample": ["sparse attention matrix estimation module is used for OtherScientificTerm", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "( 2 ) A sparse attention matrix estimation module , which predicts dominant elements of an attention matrix based on the output of the previous hidden state cross module .", "forward": true, "src_ids": "2022.acl-long.19_2182"}
+{"input": "fourier sparse attention for transformer ( fsat ) is done by using Method| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "fourier sparse attention for transformer ( fsat )", "output": "reparameterization", "neg_sample": ["fourier sparse attention for transformer ( fsat ) is done by using Method", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "By reparameterization and gradient truncation , FSAT successfully learned the index of dominant elements .", "forward": false, "src_ids": "2022.acl-long.19_2183"}
+{"input": "long - sequence tasks is done by using OtherScientificTerm| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "long - sequence tasks", "output": "multi - head attention", "neg_sample": ["long - sequence tasks is done by using OtherScientificTerm", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "Extensive experiments ( natural language , vision , and math ) show that FSAT remarkably outperforms the standard multi - head attention and its variants in various long - sequence tasks with low computational costs , and achieves new state - of - the - art results on the Long Range Arena benchmark .", "forward": false, "src_ids": "2022.acl-long.19_2184"}
+{"input": "multi - head attention is used for Task| context: self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage . due to the sparsity of the attention matrix , much computation is redundant .", "entity": "multi - head attention", "output": "long - sequence tasks", "neg_sample": ["multi - head attention is used for Task", "self - attention mechanism has been shown to be an effective approach for capturing global context dependencies in sequence modeling , but it suffers from quadratic complexity in time and memory usage .", "due to the sparsity of the attention matrix , much computation is redundant ."], "relation": "used for", "id": "2022.acl-long.19", "year": 2022, "rel_sent": "Extensive experiments ( natural language , vision , and math ) show that FSAT remarkably outperforms the standard multi - head attention and its variants in various long - sequence tasks with low computational costs , and achieves new state - of - the - art results on the Long Range Arena benchmark .", "forward": true, "src_ids": "2022.acl-long.19_2185"}
+{"input": "black - box adversarial attacks is done by using Method| context: even though several methods have proposed to defend textual neural network ( nn ) models against black - box adversarial attacks , they often defend against a specific text perturbation strategy and/or require re - training the models from scratch . this leads to a lack of generalization in practice and redundant computation . in particular , the state - of - the - art transformer models ( e.g. , bert , roberta ) require great time and computation resources .", "entity": "black - box adversarial attacks", "output": "textual neural networks", "neg_sample": ["black - box adversarial attacks is done by using Method", "even though several methods have proposed to defend textual neural network ( nn ) models against black - box adversarial attacks , they often defend against a specific text perturbation strategy and/or require re - training the models from scratch .", "this leads to a lack of generalization in practice and redundant computation .", "in particular , the state - of - the - art transformer models ( e.g.", ", bert , roberta ) require great time and computation resources ."], "relation": "used for", "id": "2022.acl-long.459", "year": 2022, "rel_sent": "SHIELD : Defending Textual Neural Networks against Multiple Black - Box Adversarial Attacks with Stochastic Multi - Expert Patcher.", "forward": false, "src_ids": "2022.acl-long.459_2186"}
+{"input": "textual neural networks is used for Task| context: this leads to a lack of generalization in practice and redundant computation . in particular , the state - of - the - art transformer models ( e.g. , bert , roberta ) require great time and computation resources .", "entity": "textual neural networks", "output": "black - box adversarial attacks", "neg_sample": ["textual neural networks is used for Task", "this leads to a lack of generalization in practice and redundant computation .", "in particular , the state - of - the - art transformer models ( e.g.", ", bert , roberta ) require great time and computation resources ."], "relation": "used for", "id": "2022.acl-long.459", "year": 2022, "rel_sent": "SHIELD : Defending Textual Neural Networks against Multiple Black - Box Adversarial Attacks with Stochastic Multi - Expert Patcher.", "forward": true, "src_ids": "2022.acl-long.459_2187"}
+{"input": "adversarial perturbations is done by using Method| context: even though several methods have proposed to defend textual neural network ( nn ) models against black - box adversarial attacks , they often defend against a specific text perturbation strategy and/or require re - training the models from scratch . this leads to a lack of generalization in practice and redundant computation . in particular , the state - of - the - art transformer models ( e.g. , bert , roberta ) require great time and computation resources .", "entity": "adversarial perturbations", "output": "iterative search mechanisms", "neg_sample": ["adversarial perturbations is done by using Method", "even though several methods have proposed to defend textual neural network ( nn ) models against black - box adversarial attacks , they often defend against a specific text perturbation strategy and/or require re - training the models from scratch .", "this leads to a lack of generalization in practice and redundant computation .", "in particular , the state - of - the - art transformer models ( e.g.", ", bert , roberta ) require great time and computation resources ."], "relation": "used for", "id": "2022.acl-long.459", "year": 2022, "rel_sent": "Considering that most of current black - box attacks rely on iterative search mechanisms to optimize their adversarial perturbations , SHIELD confuses the attackers by automatically utilizing different weighted ensembles of predictors depending on the input .", "forward": false, "src_ids": "2022.acl-long.459_2188"}
+{"input": "iterative search mechanisms is used for OtherScientificTerm| context: even though several methods have proposed to defend textual neural network ( nn ) models against black - box adversarial attacks , they often defend against a specific text perturbation strategy and/or require re - training the models from scratch . this leads to a lack of generalization in practice and redundant computation . in particular , the state - of - the - art transformer models ( e.g. , bert , roberta ) require great time and computation resources .", "entity": "iterative search mechanisms", "output": "adversarial perturbations", "neg_sample": ["iterative search mechanisms is used for OtherScientificTerm", "even though several methods have proposed to defend textual neural network ( nn ) models against black - box adversarial attacks , they often defend against a specific text perturbation strategy and/or require re - training the models from scratch .", "this leads to a lack of generalization in practice and redundant computation .", "in particular , the state - of - the - art transformer models ( e.g.", ", bert , roberta ) require great time and computation resources ."], "relation": "used for", "id": "2022.acl-long.459", "year": 2022, "rel_sent": "Considering that most of current black - box attacks rely on iterative search mechanisms to optimize their adversarial perturbations , SHIELD confuses the attackers by automatically utilizing different weighted ensembles of predictors depending on the input .", "forward": true, "src_ids": "2022.acl-long.459_2189"}
+{"input": "synthetic benchmark is used for Task| context: training giant models from scratch for each complex task is resource- and data - inefficient .", "entity": "synthetic benchmark", "output": "complex reasoning tasks", "neg_sample": ["synthetic benchmark is used for Task", "training giant models from scratch for each complex task is resource- and data - inefficient ."], "relation": "used for", "id": "2022.findings-acl.142", "year": 2022, "rel_sent": "We design a synthetic benchmark , CommaQA , with three complex reasoning tasks ( explicit , implicit , numeric ) designed to be solved by communicating with existing QA agents .", "forward": true, "src_ids": "2022.findings-acl.142_2190"}
+{"input": "complex reasoning tasks is done by using Material| context: training giant models from scratch for each complex task is resource- and data - inefficient .", "entity": "complex reasoning tasks", "output": "synthetic benchmark", "neg_sample": ["complex reasoning tasks is done by using Material", "training giant models from scratch for each complex task is resource- and data - inefficient ."], "relation": "used for", "id": "2022.findings-acl.142", "year": 2022, "rel_sent": "We design a synthetic benchmark , CommaQA , with three complex reasoning tasks ( explicit , implicit , numeric ) designed to be solved by communicating with existing QA agents .", "forward": false, "src_ids": "2022.findings-acl.142_2191"}
+{"input": "post - hoc explanation faithfulness is done by using OtherScientificTerm| context: recent work in natural language processing has focused on developing approaches that extract faithful explanations , either via identifying the most important tokens in the input ( i.e. post - hoc explanations ) or by designing inherently faithful models that first select the most important tokens and then use them to predict the correct label ( i.e. select - then - predict models ) . currently , these approaches are largely evaluated on in - domain settings . yet , little is known about how post - hoc explanations and inherently faithful models perform in out - of - domain settings .", "entity": "post - hoc explanation faithfulness", "output": "random baseline", "neg_sample": ["post - hoc explanation faithfulness is done by using OtherScientificTerm", "recent work in natural language processing has focused on developing approaches that extract faithful explanations , either via identifying the most important tokens in the input ( i.e.", "post - hoc explanations ) or by designing inherently faithful models that first select the most important tokens and then use them to predict the correct label ( i.e.", "select - then - predict models ) .", "currently , these approaches are largely evaluated on in - domain settings .", "yet , little is known about how post - hoc explanations and inherently faithful models perform in out - of - domain settings ."], "relation": "used for", "id": "2022.acl-long.477", "year": 2022, "rel_sent": "We find this misleading and suggest using a random baseline as a yardstick for evaluating post - hoc explanation faithfulness .", "forward": false, "src_ids": "2022.acl-long.477_2192"}
+{"input": "random baseline is used for OtherScientificTerm| context: recent work in natural language processing has focused on developing approaches that extract faithful explanations , either via identifying the most important tokens in the input ( i.e. post - hoc explanations ) or by designing inherently faithful models that first select the most important tokens and then use them to predict the correct label ( i.e. select - then - predict models ) . currently , these approaches are largely evaluated on in - domain settings . yet , little is known about how post - hoc explanations and inherently faithful models perform in out - of - domain settings .", "entity": "random baseline", "output": "post - hoc explanation faithfulness", "neg_sample": ["random baseline is used for OtherScientificTerm", "recent work in natural language processing has focused on developing approaches that extract faithful explanations , either via identifying the most important tokens in the input ( i.e.", "post - hoc explanations ) or by designing inherently faithful models that first select the most important tokens and then use them to predict the correct label ( i.e.", "select - then - predict models ) .", "currently , these approaches are largely evaluated on in - domain settings .", "yet , little is known about how post - hoc explanations and inherently faithful models perform in out - of - domain settings ."], "relation": "used for", "id": "2022.acl-long.477", "year": 2022, "rel_sent": "We find this misleading and suggest using a random baseline as a yardstick for evaluating post - hoc explanation faithfulness .", "forward": true, "src_ids": "2022.acl-long.477_2193"}
+{"input": "human judgment is used for Task| context: distinct is a widely used automatic metric for evaluating diversity in language generation tasks . however , we observed that the original approach to calculating distinct scores has evident biases that tend to assign higher penalties to longer sequences .", "entity": "human judgment", "output": "evaluating response diversity", "neg_sample": ["human judgment is used for Task", "distinct is a widely used automatic metric for evaluating diversity in language generation tasks .", "however , we observed that the original approach to calculating distinct scores has evident biases that tend to assign higher penalties to longer sequences ."], "relation": "used for", "id": "2022.acl-short.86", "year": 2022, "rel_sent": "Our experiments show that our proposed metric , Expectation - Adjusted Distinct ( EAD ) , correlates better with human judgment in evaluating response diversity .", "forward": true, "src_ids": "2022.acl-short.86_2194"}
+{"input": "evaluating response diversity is done by using OtherScientificTerm| context: distinct is a widely used automatic metric for evaluating diversity in language generation tasks . however , we observed that the original approach to calculating distinct scores has evident biases that tend to assign higher penalties to longer sequences .", "entity": "evaluating response diversity", "output": "human judgment", "neg_sample": ["evaluating response diversity is done by using OtherScientificTerm", "distinct is a widely used automatic metric for evaluating diversity in language generation tasks .", "however , we observed that the original approach to calculating distinct scores has evident biases that tend to assign higher penalties to longer sequences ."], "relation": "used for", "id": "2022.acl-short.86", "year": 2022, "rel_sent": "Our experiments show that our proposed metric , Expectation - Adjusted Distinct ( EAD ) , correlates better with human judgment in evaluating response diversity .", "forward": false, "src_ids": "2022.acl-short.86_2195"}
+{"input": "snare is done by using Method| context: natural language applied to natural 2d images describes a fundamentally 3d world .", "entity": "snare", "output": "voxel - informed language grounder", "neg_sample": ["snare is done by using Method", "natural language applied to natural 2d images describes a fundamentally 3d world ."], "relation": "used for", "id": "2022.acl-short.7", "year": 2022, "rel_sent": "We show that VLG significantly improves grounding accuracy on SNARE , an object reference game task . At the time of writing , VLG holds the top place on the SNARE leaderboard , achieving SOTA results with a 2.0 % absolute improvement .", "forward": false, "src_ids": "2022.acl-short.7_2196"}
+{"input": "voxel - informed language grounder is used for Task| context: natural language applied to natural 2d images describes a fundamentally 3d world .", "entity": "voxel - informed language grounder", "output": "snare", "neg_sample": ["voxel - informed language grounder is used for Task", "natural language applied to natural 2d images describes a fundamentally 3d world ."], "relation": "used for", "id": "2022.acl-short.7", "year": 2022, "rel_sent": "We show that VLG significantly improves grounding accuracy on SNARE , an object reference game task . At the time of writing , VLG holds the top place on the SNARE leaderboard , achieving SOTA results with a 2.0 % absolute improvement .", "forward": true, "src_ids": "2022.acl-short.7_2197"}
+{"input": "linguistic knowledge is done by using Method| context: recent progress in large pretrained language models ( lms ) has led to a growth of analyses examining what kinds of linguistic knowledge are encoded by these models . due to computational constraints , existing analyses are mostly conducted on publicly - released lm checkpoints , which makes it difficult to study how various factors during training affect the models ' acquisition of linguistic knowledge .", "entity": "linguistic knowledge", "output": "transformer lms", "neg_sample": ["linguistic knowledge is done by using Method", "recent progress in large pretrained language models ( lms ) has led to a growth of analyses examining what kinds of linguistic knowledge are encoded by these models .", "due to computational constraints , existing analyses are mostly conducted on publicly - released lm checkpoints , which makes it difficult to study how various factors during training affect the models ' acquisition of linguistic knowledge ."], "relation": "used for", "id": "2022.insights-1.6", "year": 2022, "rel_sent": "We hope our work offers useful insights for future research into designing Transformer LMs that more effectively learn linguistic knowledge .", "forward": false, "src_ids": "2022.insights-1.6_2198"}
+{"input": "endangered language documentation is done by using Method| context: languages are classified as low - resource when they lack the quantity of data necessary for training statistical and machine learning tools and models . causes of resource scarcity vary but can include poor access to technology for developing these resources , a relatively small population of speakers , or a lack of urgency for collecting such resources in bilingual populations where the second language is high - resource . as a result , the languages described as low - resource in the literature are as different as finnish on the one hand , with millions of speakers using it in every imaginable domain , and seneca , with only a small - handful of fluent speakers using the language primarily in a restricted domain . while issues stemming from the lack of resources necessary to train models unite this disparate group of languages , many other issues cut across the divide between widely - spoken low - resource languages and endangered languages .", "entity": "endangered language documentation", "output": "language technology", "neg_sample": ["endangered language documentation is done by using Method", "languages are classified as low - resource when they lack the quantity of data necessary for training statistical and machine learning tools and models .", "causes of resource scarcity vary but can include poor access to technology for developing these resources , a relatively small population of speakers , or a lack of urgency for collecting such resources in bilingual populations where the second language is high - resource .", "as a result , the languages described as low - resource in the literature are as different as finnish on the one hand , with millions of speakers using it in every imaginable domain , and seneca , with only a small - handful of fluent speakers using the language primarily in a restricted domain .", "while issues stemming from the lack of resources necessary to train models unite this disparate group of languages , many other issues cut across the divide between widely - spoken low - resource languages and endangered languages ."], "relation": "used for", "id": "2022.acl-long.272", "year": 2022, "rel_sent": "In this position paper , we discuss the unique technological , cultural , practical , and ethical challenges that researchers and indigenous speech community members face when working together to develop language technology to support endangered language documentation and revitalization .", "forward": false, "src_ids": "2022.acl-long.272_2199"}
+{"input": "revitalization is done by using Method| context: languages are classified as low - resource when they lack the quantity of data necessary for training statistical and machine learning tools and models . causes of resource scarcity vary but can include poor access to technology for developing these resources , a relatively small population of speakers , or a lack of urgency for collecting such resources in bilingual populations where the second language is high - resource . as a result , the languages described as low - resource in the literature are as different as finnish on the one hand , with millions of speakers using it in every imaginable domain , and seneca , with only a small - handful of fluent speakers using the language primarily in a restricted domain . while issues stemming from the lack of resources necessary to train models unite this disparate group of languages , many other issues cut across the divide between widely - spoken low - resource languages and endangered languages .", "entity": "revitalization", "output": "language technology", "neg_sample": ["revitalization is done by using Method", "languages are classified as low - resource when they lack the quantity of data necessary for training statistical and machine learning tools and models .", "causes of resource scarcity vary but can include poor access to technology for developing these resources , a relatively small population of speakers , or a lack of urgency for collecting such resources in bilingual populations where the second language is high - resource .", "as a result , the languages described as low - resource in the literature are as different as finnish on the one hand , with millions of speakers using it in every imaginable domain , and seneca , with only a small - handful of fluent speakers using the language primarily in a restricted domain .", "while issues stemming from the lack of resources necessary to train models unite this disparate group of languages , many other issues cut across the divide between widely - spoken low - resource languages and endangered languages ."], "relation": "used for", "id": "2022.acl-long.272", "year": 2022, "rel_sent": "In this position paper , we discuss the unique technological , cultural , practical , and ethical challenges that researchers and indigenous speech community members face when working together to develop language technology to support endangered language documentation and revitalization .", "forward": false, "src_ids": "2022.acl-long.272_2200"}
+{"input": "language technology is used for Task| context: languages are classified as low - resource when they lack the quantity of data necessary for training statistical and machine learning tools and models . causes of resource scarcity vary but can include poor access to technology for developing these resources , a relatively small population of speakers , or a lack of urgency for collecting such resources in bilingual populations where the second language is high - resource . as a result , the languages described as low - resource in the literature are as different as finnish on the one hand , with millions of speakers using it in every imaginable domain , and seneca , with only a small - handful of fluent speakers using the language primarily in a restricted domain . while issues stemming from the lack of resources necessary to train models unite this disparate group of languages , many other issues cut across the divide between widely - spoken low - resource languages and endangered languages .", "entity": "language technology", "output": "endangered language documentation", "neg_sample": ["language technology is used for Task", "languages are classified as low - resource when they lack the quantity of data necessary for training statistical and machine learning tools and models .", "causes of resource scarcity vary but can include poor access to technology for developing these resources , a relatively small population of speakers , or a lack of urgency for collecting such resources in bilingual populations where the second language is high - resource .", "as a result , the languages described as low - resource in the literature are as different as finnish on the one hand , with millions of speakers using it in every imaginable domain , and seneca , with only a small - handful of fluent speakers using the language primarily in a restricted domain .", "while issues stemming from the lack of resources necessary to train models unite this disparate group of languages , many other issues cut across the divide between widely - spoken low - resource languages and endangered languages ."], "relation": "used for", "id": "2022.acl-long.272", "year": 2022, "rel_sent": "In this position paper , we discuss the unique technological , cultural , practical , and ethical challenges that researchers and indigenous speech community members face when working together to develop language technology to support endangered language documentation and revitalization .", "forward": true, "src_ids": "2022.acl-long.272_2201"}
+{"input": "telugu language is done by using Material| context: named entity recognition ( ner ) is a successful and well - researched problem in english due to the availability of resources . the transformer models , specifically the masked - language models ( mlm ) , have shown remarkable performance in ner during recent times . with growing data in different online platforms , there is a need for ner in other languages too . ner remains to be underexplored in indian languages due to the lack of resources and tools .", "entity": "telugu language", "output": "annotated ner datasets", "neg_sample": ["telugu language is done by using Material", "named entity recognition ( ner ) is a successful and well - researched problem in english due to the availability of resources .", "the transformer models , specifically the masked - language models ( mlm ) , have shown remarkable performance in ner during recent times .", "with growing data in different online platforms , there is a need for ner in other languages too .", "ner remains to be underexplored in indian languages due to the lack of resources and tools ."], "relation": "used for", "id": "2022.acl-srw.20", "year": 2022, "rel_sent": "Our contributions in this paper include ( i ) Two annotated NER datasets for the Telugu language in multiple domains : Newswire Dataset ( ND ) and Medical Dataset ( MD ) , and we combined ND and MD toform Combined Dataset ( CD ) ( ii ) Comparison of the finetuned Telugu pretrained transformer models ( BERT - Te , RoBERTa - Te , and ELECTRA - Te ) with other baseline models ( CRF , LSTM - CRF , and BiLSTM - CRF ) ( iii ) Further investigation of the performance of Telugu pretrained transformer models against the multilingual models mBERT , XLM - R , and IndicBERT .", "forward": false, "src_ids": "2022.acl-srw.20_2202"}
+{"input": "annotated ner datasets is used for Material| context: named entity recognition ( ner ) is a successful and well - researched problem in english due to the availability of resources . the transformer models , specifically the masked - language models ( mlm ) , have shown remarkable performance in ner during recent times . with growing data in different online platforms , there is a need for ner in other languages too . ner remains to be underexplored in indian languages due to the lack of resources and tools .", "entity": "annotated ner datasets", "output": "telugu language", "neg_sample": ["annotated ner datasets is used for Material", "named entity recognition ( ner ) is a successful and well - researched problem in english due to the availability of resources .", "the transformer models , specifically the masked - language models ( mlm ) , have shown remarkable performance in ner during recent times .", "with growing data in different online platforms , there is a need for ner in other languages too .", "ner remains to be underexplored in indian languages due to the lack of resources and tools ."], "relation": "used for", "id": "2022.acl-srw.20", "year": 2022, "rel_sent": "Our contributions in this paper include ( i ) Two annotated NER datasets for the Telugu language in multiple domains : Newswire Dataset ( ND ) and Medical Dataset ( MD ) , and we combined ND and MD toform Combined Dataset ( CD ) ( ii ) Comparison of the finetuned Telugu pretrained transformer models ( BERT - Te , RoBERTa - Te , and ELECTRA - Te ) with other baseline models ( CRF , LSTM - CRF , and BiLSTM - CRF ) ( iii ) Further investigation of the performance of Telugu pretrained transformer models against the multilingual models mBERT , XLM - R , and IndicBERT .", "forward": true, "src_ids": "2022.acl-srw.20_2203"}
+{"input": "named entity recognition is done by using Method| context: named entity recognition ( ner ) is a successful and well - researched problem in english due to the availability of resources . the transformer models , specifically the masked - language models ( mlm ) , have shown remarkable performance in ner during recent times . with growing data in different online platforms , there is a need for ner in other languages too . ner remains to be underexplored in indian languages due to the lack of resources and tools .", "entity": "named entity recognition", "output": "pretrained multilingual and baseline models", "neg_sample": ["named entity recognition is done by using Method", "named entity recognition ( ner ) is a successful and well - researched problem in english due to the availability of resources .", "the transformer models , specifically the masked - language models ( mlm ) , have shown remarkable performance in ner during recent times .", "with growing data in different online platforms , there is a need for ner in other languages too .", "ner remains to be underexplored in indian languages due to the lack of resources and tools ."], "relation": "used for", "id": "2022.acl-srw.20", "year": 2022, "rel_sent": "We find that pretrained Telugu language models ( BERT - Te and RoBERTa ) outperform the existing pretrained multilingual and baseline models in NER .", "forward": false, "src_ids": "2022.acl-srw.20_2204"}
+{"input": "text editing suggestions is done by using Method| context: revision is an essential part of the human writing process . it tends to be strategic , adaptive , and , more importantly , iterative in nature . despite the success of large language models on text revision tasks , they are limited to non - iterative , one - shot revisions . examining and evaluating the capability of large language models for making continuous revisions and collaborating with human writers is a critical step towards building effective writing assistants .", "entity": "text editing suggestions", "output": "text revision model", "neg_sample": ["text editing suggestions is done by using Method", "revision is an essential part of the human writing process .", "it tends to be strategic , adaptive , and , more importantly , iterative in nature .", "despite the success of large language models on text revision tasks , they are limited to non - iterative , one - shot revisions .", "examining and evaluating the capability of large language models for making continuous revisions and collaborating with human writers is a critical step towards building effective writing assistants ."], "relation": "used for", "id": "2022.in2writing-1.14", "year": 2022, "rel_sent": "In R3 , a text revision model provides text editing suggestions for human writers , who can accept or reject the suggested edits .", "forward": false, "src_ids": "2022.in2writing-1.14_2205"}
+{"input": "text revision model is used for OtherScientificTerm| context: revision is an essential part of the human writing process . it tends to be strategic , adaptive , and , more importantly , iterative in nature . despite the success of large language models on text revision tasks , they are limited to non - iterative , one - shot revisions . examining and evaluating the capability of large language models for making continuous revisions and collaborating with human writers is a critical step towards building effective writing assistants .", "entity": "text revision model", "output": "text editing suggestions", "neg_sample": ["text revision model is used for OtherScientificTerm", "revision is an essential part of the human writing process .", "it tends to be strategic , adaptive , and , more importantly , iterative in nature .", "despite the success of large language models on text revision tasks , they are limited to non - iterative , one - shot revisions .", "examining and evaluating the capability of large language models for making continuous revisions and collaborating with human writers is a critical step towards building effective writing assistants ."], "relation": "used for", "id": "2022.in2writing-1.14", "year": 2022, "rel_sent": "In R3 , a text revision model provides text editing suggestions for human writers , who can accept or reject the suggested edits .", "forward": true, "src_ids": "2022.in2writing-1.14_2206"}
+{"input": "text simplification is done by using Task| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "text simplification", "output": "predicting sentence deletions", "neg_sample": ["text simplification is done by using Task", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "Predicting Sentence Deletions for Text Simplification Using a Functional Discourse Structure.", "forward": false, "src_ids": "2022.acl-short.28_2207"}
+{"input": "news genre - specific functional discourse structure is used for Task| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "news genre - specific functional discourse structure", "output": "predicting sentence deletions", "neg_sample": ["news genre - specific functional discourse structure is used for Task", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "In this work , we focus on sentence deletions for text simplification and use a news genre - specific functional discourse structure , which categorizes sentences based on their contents and their function roles in telling a news story , for predicting sentence deletion .", "forward": true, "src_ids": "2022.acl-short.28_2208"}
+{"input": "neural net model is used for Task| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "neural net model", "output": "predicting sentence deletions", "neg_sample": ["neural net model is used for Task", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "We incorporate sentence categories into a neural net model in two ways for predicting sentence deletions , either as additional features or by jointly predicting sentence deletions and sentence categories .", "forward": true, "src_ids": "2022.acl-short.28_2209"}
+{"input": "sentence categories is used for Task| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "sentence categories", "output": "predicting sentence deletions", "neg_sample": ["sentence categories is used for Task", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "We incorporate sentence categories into a neural net model in two ways for predicting sentence deletions , either as additional features or by jointly predicting sentence deletions and sentence categories .", "forward": true, "src_ids": "2022.acl-short.28_2210"}
+{"input": "text simplification is done by using Task| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "text simplification", "output": "sentence deletions", "neg_sample": ["text simplification is done by using Task", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "In this work , we focus on sentence deletions for text simplification and use a news genre - specific functional discourse structure , which categorizes sentences based on their contents and their function roles in telling a news story , for predicting sentence deletion .", "forward": false, "src_ids": "2022.acl-short.28_2211"}
+{"input": "predicting sentence deletions is done by using OtherScientificTerm| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "predicting sentence deletions", "output": "news genre - specific functional discourse structure", "neg_sample": ["predicting sentence deletions is done by using OtherScientificTerm", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "In this work , we focus on sentence deletions for text simplification and use a news genre - specific functional discourse structure , which categorizes sentences based on their contents and their function roles in telling a news story , for predicting sentence deletion .", "forward": false, "src_ids": "2022.acl-short.28_2212"}
+{"input": "predicting sentence deletions is done by using OtherScientificTerm| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "predicting sentence deletions", "output": "sentence categories", "neg_sample": ["predicting sentence deletions is done by using OtherScientificTerm", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "We incorporate sentence categories into a neural net model in two ways for predicting sentence deletions , either as additional features or by jointly predicting sentence deletions and sentence categories .", "forward": false, "src_ids": "2022.acl-short.28_2213"}
+{"input": "predicting sentence deletions is done by using Method| context: document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity .", "entity": "predicting sentence deletions", "output": "neural net model", "neg_sample": ["predicting sentence deletions is done by using Method", "document - level text simplification often deletes some sentences besides performing lexical , grammatical or structural simplification to reduce text complexity ."], "relation": "used for", "id": "2022.acl-short.28", "year": 2022, "rel_sent": "We incorporate sentence categories into a neural net model in two ways for predicting sentence deletions , either as additional features or by jointly predicting sentence deletions and sentence categories .", "forward": false, "src_ids": "2022.acl-short.28_2214"}
+{"input": "event representations is done by using Method| context: representations of events described in text are important for various tasks .", "entity": "event representations", "output": "swcc", "neg_sample": ["event representations is done by using Method", "representations of events described in text are important for various tasks ."], "relation": "used for", "id": "2022.acl-long.216", "year": 2022, "rel_sent": "SWCC learns event representations by making better use of co - occurrence information of events .", "forward": false, "src_ids": "2022.acl-long.216_2215"}
+{"input": "event representation learning is done by using Method| context: representations of events described in text are important for various tasks .", "entity": "event representation learning", "output": "simultaneous weakly supervised contrastive learning and clustering framework", "neg_sample": ["event representation learning is done by using Method", "representations of events described in text are important for various tasks ."], "relation": "used for", "id": "2022.acl-long.216", "year": 2022, "rel_sent": "In this work , we present SWCC : a Simultaneous Weakly supervised Contrastive learning and Clustering framework for event representation learning .", "forward": false, "src_ids": "2022.acl-long.216_2216"}
+{"input": "simultaneous weakly supervised contrastive learning and clustering framework is used for Method| context: representations of events described in text are important for various tasks .", "entity": "simultaneous weakly supervised contrastive learning and clustering framework", "output": "event representation learning", "neg_sample": ["simultaneous weakly supervised contrastive learning and clustering framework is used for Method", "representations of events described in text are important for various tasks ."], "relation": "used for", "id": "2022.acl-long.216", "year": 2022, "rel_sent": "In this work , we present SWCC : a Simultaneous Weakly supervised Contrastive learning and Clustering framework for event representation learning .", "forward": true, "src_ids": "2022.acl-long.216_2217"}
+{"input": "swcc is used for Method| context: representations of events described in text are important for various tasks .", "entity": "swcc", "output": "event representations", "neg_sample": ["swcc is used for Method", "representations of events described in text are important for various tasks ."], "relation": "used for", "id": "2022.acl-long.216", "year": 2022, "rel_sent": "SWCC learns event representations by making better use of co - occurrence information of events .", "forward": true, "src_ids": "2022.acl-long.216_2218"}
+{"input": "cad is used for OtherScientificTerm| context: recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "cad", "output": "spurious correlations", "neg_sample": ["cad is used for OtherScientificTerm", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "To explain this discrepancy , through a toy theoretical example and empirical analysis on two crowdsourced CAD datasets , we show that : ( a ) while features perturbed in CAD are indeed robust features , it may prevent the model from learning unperturbed robust features ; and ( b ) CAD may exacerbate existing spurious correlations in the data .", "forward": true, "src_ids": "2022.acl-long.256_2219"}
+{"input": "ood generalization is done by using Method| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "entity": "ood generalization", "output": "cad", "neg_sample": ["ood generalization is done by using Method", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "However , empirical results using CAD during training for OOD generalization have been mixed .", "forward": false, "src_ids": "2022.acl-long.256_2220"}
+{"input": "training is done by using Method| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "entity": "training", "output": "cad", "neg_sample": ["training is done by using Method", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "However , empirical results using CAD during training for OOD generalization have been mixed .", "forward": false, "src_ids": "2022.acl-long.256_2221"}
+{"input": "spurious correlations is done by using Method| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "entity": "spurious correlations", "output": "cad", "neg_sample": ["spurious correlations is done by using Method", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "To explain this discrepancy , through a toy theoretical example and empirical analysis on two crowdsourced CAD datasets , we show that : ( a ) while features perturbed in CAD are indeed robust features , it may prevent the model from learning unperturbed robust features ; and ( b ) CAD may exacerbate existing spurious correlations in the data .", "forward": false, "src_ids": "2022.acl-long.256_2222"}
+{"input": "ood generalization is done by using Method| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "entity": "ood generalization", "output": "cad", "neg_sample": ["ood generalization is done by using Method", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "Our results thus show that the lack of perturbation diversity limits CAD 's effectiveness on OOD generalization , calling for innovative crowdsourcing procedures to elicit diverse perturbation of examples .", "forward": false, "src_ids": "2022.acl-long.256_2223"}
+{"input": "ood generalization is done by using Task| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data . recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "ood generalization", "output": "training", "neg_sample": ["ood generalization is done by using Task", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "However , empirical results using CAD during training for OOD generalization have been mixed .", "forward": false, "src_ids": "2022.acl-long.256_2224"}
+{"input": "cad is used for Task| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data . recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "cad", "output": "training", "neg_sample": ["cad is used for Task", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "However , empirical results using CAD during training for OOD generalization have been mixed .", "forward": true, "src_ids": "2022.acl-long.256_2225"}
+{"input": "training is used for Task| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data . recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "training", "output": "ood generalization", "neg_sample": ["training is used for Task", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "However , empirical results using CAD during training for OOD generalization have been mixed .", "forward": true, "src_ids": "2022.acl-long.256_2226"}
+{"input": "cad is used for Task| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data . recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "cad", "output": "ood generalization", "neg_sample": ["cad is used for Task", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "Our results thus show that the lack of perturbation diversity limits CAD 's effectiveness on OOD generalization , calling for innovative crowdsourcing procedures to elicit diverse perturbation of examples .", "forward": true, "src_ids": "2022.acl-long.256_2227"}
+{"input": "diverse perturbation of examples is done by using Method| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data . recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "diverse perturbation of examples", "output": "crowdsourcing procedures", "neg_sample": ["diverse perturbation of examples is done by using Method", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "Our results thus show that the lack of perturbation diversity limits CAD 's effectiveness on OOD generalization , calling for innovative crowdsourcing procedures to elicit diverse perturbation of examples .", "forward": false, "src_ids": "2022.acl-long.256_2228"}
+{"input": "crowdsourcing procedures is used for OtherScientificTerm| context: while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data . recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift .", "entity": "crowdsourcing procedures", "output": "diverse perturbation of examples", "neg_sample": ["crowdsourcing procedures is used for OtherScientificTerm", "while pretrained language models achieve excellent performance on natural language understanding benchmarks , they tend to rely on spurious correlations and generalize poorly to out - of - distribution ( ood ) data .", "recent work has explored using counterfactually - augmented data ( cad)-data generated by minimally perturbing examples toflip the ground - truth label - to identify robust features that are invariant under distribution shift ."], "relation": "used for", "id": "2022.acl-long.256", "year": 2022, "rel_sent": "Our results thus show that the lack of perturbation diversity limits CAD 's effectiveness on OOD generalization , calling for innovative crowdsourcing procedures to elicit diverse perturbation of examples .", "forward": true, "src_ids": "2022.acl-long.256_2229"}
+{"input": "syntactic information is used for Task| context: among different types of contextual information , the auto - generated syntactic information ( namely , word dependencies ) has shown its effectiveness for the task . however , most existing studies require modifications to the existing baseline architectures ( e.g. , adding new components , such as gcn , on the top of an encoder ) to leverage the syntactic information .", "entity": "syntactic information", "output": "relation extraction", "neg_sample": ["syntactic information is used for Task", "among different types of contextual information , the auto - generated syntactic information ( namely , word dependencies ) has shown its effectiveness for the task .", "however , most existing studies require modifications to the existing baseline architectures ( e.g.", ", adding new components , such as gcn , on the top of an encoder ) to leverage the syntactic information ."], "relation": "used for", "id": "2022.findings-acl.147", "year": 2022, "rel_sent": "To offer an alternative solution , we propose to leverage syntactic information to improve RE by training a syntax - induced encoder on auto - parsed data through dependency masking .", "forward": true, "src_ids": "2022.findings-acl.147_2230"}
+{"input": "relation extraction is done by using OtherScientificTerm| context: relation extraction ( re ) is an important natural language processing task that predicts the relation between two given entities , where a good understanding of the contextual information is essential to achieve an outstanding model performance .", "entity": "relation extraction", "output": "syntactic information", "neg_sample": ["relation extraction is done by using OtherScientificTerm", "relation extraction ( re ) is an important natural language processing task that predicts the relation between two given entities , where a good understanding of the contextual information is essential to achieve an outstanding model performance ."], "relation": "used for", "id": "2022.findings-acl.147", "year": 2022, "rel_sent": "To offer an alternative solution , we propose to leverage syntactic information to improve RE by training a syntax - induced encoder on auto - parsed data through dependency masking .", "forward": false, "src_ids": "2022.findings-acl.147_2231"}
+{"input": "syntax - induced encoder is done by using OtherScientificTerm| context: relation extraction ( re ) is an important natural language processing task that predicts the relation between two given entities , where a good understanding of the contextual information is essential to achieve an outstanding model performance .", "entity": "syntax - induced encoder", "output": "syntactic information", "neg_sample": ["syntax - induced encoder is done by using OtherScientificTerm", "relation extraction ( re ) is an important natural language processing task that predicts the relation between two given entities , where a good understanding of the contextual information is essential to achieve an outstanding model performance ."], "relation": "used for", "id": "2022.findings-acl.147", "year": 2022, "rel_sent": "To offer an alternative solution , we propose to leverage syntactic information to improve RE by training a syntax - induced encoder on auto - parsed data through dependency masking .", "forward": false, "src_ids": "2022.findings-acl.147_2232"}
+{"input": "syntactic information is used for Method| context: relation extraction ( re ) is an important natural language processing task that predicts the relation between two given entities , where a good understanding of the contextual information is essential to achieve an outstanding model performance . among different types of contextual information , the auto - generated syntactic information ( namely , word dependencies ) has shown its effectiveness for the task . however , most existing studies require modifications to the existing baseline architectures ( e.g. , adding new components , such as gcn , on the top of an encoder ) to leverage the syntactic information .", "entity": "syntactic information", "output": "syntax - induced encoder", "neg_sample": ["syntactic information is used for Method", "relation extraction ( re ) is an important natural language processing task that predicts the relation between two given entities , where a good understanding of the contextual information is essential to achieve an outstanding model performance .", "among different types of contextual information , the auto - generated syntactic information ( namely , word dependencies ) has shown its effectiveness for the task .", "however , most existing studies require modifications to the existing baseline architectures ( e.g.", ", adding new components , such as gcn , on the top of an encoder ) to leverage the syntactic information ."], "relation": "used for", "id": "2022.findings-acl.147", "year": 2022, "rel_sent": "To offer an alternative solution , we propose to leverage syntactic information to improve RE by training a syntax - induced encoder on auto - parsed data through dependency masking .", "forward": true, "src_ids": "2022.findings-acl.147_2233"}
+{"input": "automatic translation of tv dialogue is done by using OtherScientificTerm| context: unlike english , morphologically rich languages can reveal characteristics of speakers or their conversational partners , such as gender and number , via pronouns , morphological endings of words and syntax . when translating from english to such languages , a machine translation model needs to opt for a certain interpretation of textual context , which may lead to serious translation errors if extra - textual information is unavailable .", "entity": "automatic translation of tv dialogue", "output": "external metadata", "neg_sample": ["automatic translation of tv dialogue is done by using OtherScientificTerm", "unlike english , morphologically rich languages can reveal characteristics of speakers or their conversational partners , such as gender and number , via pronouns , morphological endings of words and syntax .", "when translating from english to such languages , a machine translation model needs to opt for a certain interpretation of textual context , which may lead to serious translation errors if extra - textual information is unavailable ."], "relation": "used for", "id": "2022.eamt-1.15", "year": 2022, "rel_sent": "We focus on the underresearched problem of utilising external metadata in automatic translation of TV dialogue , proposing a case study where a wide range of approaches for controlling attributes in translation is employed in a multi - attribute scenario .", "forward": false, "src_ids": "2022.eamt-1.15_2234"}
+{"input": "external metadata is used for Task| context: unlike english , morphologically rich languages can reveal characteristics of speakers or their conversational partners , such as gender and number , via pronouns , morphological endings of words and syntax . when translating from english to such languages , a machine translation model needs to opt for a certain interpretation of textual context , which may lead to serious translation errors if extra - textual information is unavailable .", "entity": "external metadata", "output": "automatic translation of tv dialogue", "neg_sample": ["external metadata is used for Task", "unlike english , morphologically rich languages can reveal characteristics of speakers or their conversational partners , such as gender and number , via pronouns , morphological endings of words and syntax .", "when translating from english to such languages , a machine translation model needs to opt for a certain interpretation of textual context , which may lead to serious translation errors if extra - textual information is unavailable ."], "relation": "used for", "id": "2022.eamt-1.15", "year": 2022, "rel_sent": "We focus on the underresearched problem of utilising external metadata in automatic translation of TV dialogue , proposing a case study where a wide range of approaches for controlling attributes in translation is employed in a multi - attribute scenario .", "forward": true, "src_ids": "2022.eamt-1.15_2235"}
+{"input": "cancer immunology research is done by using Task| context: cancer immunology research involves several important cell and protein factors . extracting the information of such cells and proteins and the interactions between them from text are crucial in text mining for cancer immunology research . however , there are few available datasets for these entities , and the amount of annotated documents is not sufficient compared with other major named entity types .", "entity": "cancer immunology research", "output": "named entity recognition", "neg_sample": ["cancer immunology research is done by using Task", "cancer immunology research involves several important cell and protein factors .", "extracting the information of such cells and proteins and the interactions between them from text are crucial in text mining for cancer immunology research .", "however , there are few available datasets for these entities , and the amount of annotated documents is not sufficient compared with other major named entity types ."], "relation": "used for", "id": "2022.bionlp-1.17", "year": 2022, "rel_sent": "Named Entity Recognition for Cancer Immunology Research Using Distant Supervision.", "forward": false, "src_ids": "2022.bionlp-1.17_2236"}
+{"input": "named entity recognition is used for Task| context: however , there are few available datasets for these entities , and the amount of annotated documents is not sufficient compared with other major named entity types .", "entity": "named entity recognition", "output": "cancer immunology research", "neg_sample": ["named entity recognition is used for Task", "however , there are few available datasets for these entities , and the amount of annotated documents is not sufficient compared with other major named entity types ."], "relation": "used for", "id": "2022.bionlp-1.17", "year": 2022, "rel_sent": "Named Entity Recognition for Cancer Immunology Research Using Distant Supervision.", "forward": true, "src_ids": "2022.bionlp-1.17_2237"}
+{"input": "zero - shot multi - lingual extractive summarization is done by using Method| context: in zero - shot multilingual extractive text summarization , a model is typically trained on english summarization dataset and then applied on summarization datasets of other languages . given english gold summaries and documents , sentence - level labels for extractive summarization are usually generated using heuristics . however , these monolingual labels created on english datasets may not be optimal on datasets of other languages , for that there is the syntactic or semantic discrepancy between different languages .", "entity": "zero - shot multi - lingual extractive summarization", "output": "neural label search", "neg_sample": ["zero - shot multi - lingual extractive summarization is done by using Method", "in zero - shot multilingual extractive text summarization , a model is typically trained on english summarization dataset and then applied on summarization datasets of other languages .", "given english gold summaries and documents , sentence - level labels for extractive summarization are usually generated using heuristics .", "however , these monolingual labels created on english datasets may not be optimal on datasets of other languages , for that there is the syntactic or semantic discrepancy between different languages ."], "relation": "used for", "id": "2022.acl-long.42", "year": 2022, "rel_sent": "Neural Label Search for Zero - Shot Multi - Lingual Extractive Summarization.", "forward": false, "src_ids": "2022.acl-long.42_2238"}
+{"input": "neural label search is used for Task| context: in zero - shot multilingual extractive text summarization , a model is typically trained on english summarization dataset and then applied on summarization datasets of other languages . given english gold summaries and documents , sentence - level labels for extractive summarization are usually generated using heuristics . however , these monolingual labels created on english datasets may not be optimal on datasets of other languages , for that there is the syntactic or semantic discrepancy between different languages .", "entity": "neural label search", "output": "zero - shot multi - lingual extractive summarization", "neg_sample": ["neural label search is used for Task", "in zero - shot multilingual extractive text summarization , a model is typically trained on english summarization dataset and then applied on summarization datasets of other languages .", "given english gold summaries and documents , sentence - level labels for extractive summarization are usually generated using heuristics .", "however , these monolingual labels created on english datasets may not be optimal on datasets of other languages , for that there is the syntactic or semantic discrepancy between different languages ."], "relation": "used for", "id": "2022.acl-long.42", "year": 2022, "rel_sent": "Neural Label Search for Zero - Shot Multi - Lingual Extractive Summarization.", "forward": true, "src_ids": "2022.acl-long.42_2239"}
+{"input": "morphologically - rich and pro - drop languages is done by using Task| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "morphologically - rich and pro - drop languages", "output": "amr alignment", "neg_sample": ["morphologically - rich and pro - drop languages is done by using Task", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "AMR Alignment for Morphologically - rich and Pro - drop Languages.", "forward": false, "src_ids": "2022.acl-srw.13_2240"}
+{"input": "amr alignment is used for Task| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "amr alignment", "output": "morphologically - rich and pro - drop languages", "neg_sample": ["amr alignment is used for Task", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "AMR Alignment for Morphologically - rich and Pro - drop Languages.", "forward": true, "src_ids": "2022.acl-srw.13_2241"}
+{"input": "amr aligners is used for Task| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "amr aligners", "output": "morphologically - rich and pro - drop languages", "neg_sample": ["amr aligners is used for Task", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Although there exist high performing AMR aligners for English , unfortunately , these are not well suited for many languages where many concepts appear from morpho - semantic elements . For the first time in the literature , this paper presents an AMR aligner tailored for morphologically - rich and pro - drop languages by experimenting on the Turkish language being a prominent example of this language group . Our aligner focuses on the meaning considering the rich Turkish morphology and aligns AMR concepts that emerge from morphemes using a tree traversal approach without additional resources or rules .", "forward": true, "src_ids": "2022.acl-srw.13_2242"}
+{"input": "morphologically - rich and pro - drop languages is done by using Method| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "morphologically - rich and pro - drop languages", "output": "amr aligners", "neg_sample": ["morphologically - rich and pro - drop languages is done by using Method", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Although there exist high performing AMR aligners for English , unfortunately , these are not well suited for many languages where many concepts appear from morpho - semantic elements . For the first time in the literature , this paper presents an AMR aligner tailored for morphologically - rich and pro - drop languages by experimenting on the Turkish language being a prominent example of this language group . Our aligner focuses on the meaning considering the rich Turkish morphology and aligns AMR concepts that emerge from morphemes using a tree traversal approach without additional resources or rules .", "forward": false, "src_ids": "2022.acl-srw.13_2243"}
+{"input": "aligners is used for Material| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "aligners", "output": "english", "neg_sample": ["aligners is used for Material", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Our aligner outperforms the Turkish adaptations of the previously proposed aligners for English and Portuguese by an F1 score of 0.87 and provides a relative error reduction of up to 76 % .", "forward": true, "src_ids": "2022.acl-srw.13_2244"}
+{"input": "amr concepts is done by using Method| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "amr concepts", "output": "aligner", "neg_sample": ["amr concepts is done by using Method", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Although there exist high performing AMR aligners for English , unfortunately , these are not well suited for many languages where many concepts appear from morpho - semantic elements . For the first time in the literature , this paper presents an AMR aligner tailored for morphologically - rich and pro - drop languages by experimenting on the Turkish language being a prominent example of this language group . Our aligner focuses on the meaning considering the rich Turkish morphology and aligns AMR concepts that emerge from morphemes using a tree traversal approach without additional resources or rules .", "forward": false, "src_ids": "2022.acl-srw.13_2245"}
+{"input": "aligner is used for OtherScientificTerm| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "aligner", "output": "amr concepts", "neg_sample": ["aligner is used for OtherScientificTerm", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Although there exist high performing AMR aligners for English , unfortunately , these are not well suited for many languages where many concepts appear from morpho - semantic elements . For the first time in the literature , this paper presents an AMR aligner tailored for morphologically - rich and pro - drop languages by experimenting on the Turkish language being a prominent example of this language group . Our aligner focuses on the meaning considering the rich Turkish morphology and aligns AMR concepts that emerge from morphemes using a tree traversal approach without additional resources or rules .", "forward": true, "src_ids": "2022.acl-srw.13_2246"}
+{"input": "english is done by using Method| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "english", "output": "aligners", "neg_sample": ["english is done by using Method", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Our aligner outperforms the Turkish adaptations of the previously proposed aligners for English and Portuguese by an F1 score of 0.87 and provides a relative error reduction of up to 76 % .", "forward": false, "src_ids": "2022.acl-srw.13_2247"}
+{"input": "portuguese is done by using Method| context: alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing .", "entity": "portuguese", "output": "aligners", "neg_sample": ["portuguese is done by using Method", "alignment between concepts in an abstract meaning representation ( amr ) graph and the words within a sentence is one of the important stages of amr parsing ."], "relation": "used for", "id": "2022.acl-srw.13", "year": 2022, "rel_sent": "Our aligner outperforms the Turkish adaptations of the previously proposed aligners for English and Portuguese by an F1 score of 0.87 and provides a relative error reduction of up to 76 % .", "forward": false, "src_ids": "2022.acl-srw.13_2248"}
+{"input": "semantic parsing annotation is done by using Method| context: collecting data for conversational semantic parsing is a time - consuming and demanding process .", "entity": "semantic parsing annotation", "output": "guided k - best selection process", "neg_sample": ["semantic parsing annotation is done by using Method", "collecting data for conversational semantic parsing is a time - consuming and demanding process ."], "relation": "used for", "id": "2022.acl-demo.11", "year": 2022, "rel_sent": "Guided K - best Selection for Semantic Parsing Annotation.", "forward": false, "src_ids": "2022.acl-demo.11_2249"}
+{"input": "guided k - best selection process is used for Task| context: collecting data for conversational semantic parsing is a time - consuming and demanding process .", "entity": "guided k - best selection process", "output": "semantic parsing annotation", "neg_sample": ["guided k - best selection process is used for Task", "collecting data for conversational semantic parsing is a time - consuming and demanding process ."], "relation": "used for", "id": "2022.acl-demo.11", "year": 2022, "rel_sent": "Guided K - best Selection for Semantic Parsing Annotation.", "forward": true, "src_ids": "2022.acl-demo.11_2250"}
+{"input": "software tools is used for Task| context: in this paper we present the speech corpus for the siberian ingrian finnish language . the speech corpus includes audio data , annotations , software tools for data - processing , two databases and a web application .", "entity": "software tools", "output": "parsing annotation files", "neg_sample": ["software tools is used for Task", "in this paper we present the speech corpus for the siberian ingrian finnish language .", "the speech corpus includes audio data , annotations , software tools for data - processing , two databases and a web application ."], "relation": "used for", "id": "2022.computel-1.1", "year": 2022, "rel_sent": "The software tool for parsing annotation files and feeding a relational database is developed and published under a free license .", "forward": true, "src_ids": "2022.computel-1.1_2251"}
+{"input": "software tools is used for Material| context: in this paper we present the speech corpus for the siberian ingrian finnish language . the speech corpus includes audio data , annotations , software tools for data - processing , two databases and a web application .", "entity": "software tools", "output": "relational database", "neg_sample": ["software tools is used for Material", "in this paper we present the speech corpus for the siberian ingrian finnish language .", "the speech corpus includes audio data , annotations , software tools for data - processing , two databases and a web application ."], "relation": "used for", "id": "2022.computel-1.1", "year": 2022, "rel_sent": "The software tool for parsing annotation files and feeding a relational database is developed and published under a free license .", "forward": true, "src_ids": "2022.computel-1.1_2252"}
+{"input": "parsing annotation files is done by using Method| context: in this paper we present the speech corpus for the siberian ingrian finnish language .", "entity": "parsing annotation files", "output": "software tools", "neg_sample": ["parsing annotation files is done by using Method", "in this paper we present the speech corpus for the siberian ingrian finnish language ."], "relation": "used for", "id": "2022.computel-1.1", "year": 2022, "rel_sent": "The software tool for parsing annotation files and feeding a relational database is developed and published under a free license .", "forward": false, "src_ids": "2022.computel-1.1_2253"}
+{"input": "relational database is done by using Method| context: in this paper we present the speech corpus for the siberian ingrian finnish language .", "entity": "relational database", "output": "software tools", "neg_sample": ["relational database is done by using Method", "in this paper we present the speech corpus for the siberian ingrian finnish language ."], "relation": "used for", "id": "2022.computel-1.1", "year": 2022, "rel_sent": "The software tool for parsing annotation files and feeding a relational database is developed and published under a free license .", "forward": false, "src_ids": "2022.computel-1.1_2254"}
+{"input": "direct speech translation is done by using OtherScientificTerm| context: transformers have achieved state - of - the - art results across multiple nlp tasks .", "entity": "direct speech translation", "output": "self - attention", "neg_sample": ["direct speech translation is done by using OtherScientificTerm", "transformers have achieved state - of - the - art results across multiple nlp tasks ."], "relation": "used for", "id": "2022.acl-srw.32", "year": 2022, "rel_sent": "In this paper , we discuss the usefulness of self - attention for Direct Speech Translation .", "forward": false, "src_ids": "2022.acl-srw.32_2255"}
+{"input": "self - attention is used for Task| context: transformers have achieved state - of - the - art results across multiple nlp tasks . however , the self - attention mechanism complexity scales quadratically with the sequence length , creating an obstacle for tasks involving long sequences , like in the speech domain .", "entity": "self - attention", "output": "direct speech translation", "neg_sample": ["self - attention is used for Task", "transformers have achieved state - of - the - art results across multiple nlp tasks .", "however , the self - attention mechanism complexity scales quadratically with the sequence length , creating an obstacle for tasks involving long sequences , like in the speech domain ."], "relation": "used for", "id": "2022.acl-srw.32", "year": 2022, "rel_sent": "In this paper , we discuss the usefulness of self - attention for Direct Speech Translation .", "forward": true, "src_ids": "2022.acl-srw.32_2256"}
+{"input": "lid systems is used for Material| context: when using multilingual applications , users have their own language preferences , which can be regarded as external knowledge for lid . nevertheless , current studies do not consider the inter - personal variations due to the lack of user annotated training data .", "entity": "lid systems", "output": "ambiguous texts", "neg_sample": ["lid systems is used for Material", "when using multilingual applications , users have their own language preferences , which can be regarded as external knowledge for lid .", "nevertheless , current studies do not consider the inter - personal variations due to the lack of user annotated training data ."], "relation": "used for", "id": "2022.findings-acl.303", "year": 2022, "rel_sent": "Experimental results reveal that our model can incarnate user traits and significantly outperforms existing LID systems on handling ambiguous texts .", "forward": true, "src_ids": "2022.findings-acl.303_2257"}
+{"input": "ambiguous texts is done by using Method| context: recognizing the language of ambiguous texts has become a main challenge in language identification ( lid ) . when using multilingual applications , users have their own language preferences , which can be regarded as external knowledge for lid . nevertheless , current studies do not consider the inter - personal variations due to the lack of user annotated training data .", "entity": "ambiguous texts", "output": "lid systems", "neg_sample": ["ambiguous texts is done by using Method", "recognizing the language of ambiguous texts has become a main challenge in language identification ( lid ) .", "when using multilingual applications , users have their own language preferences , which can be regarded as external knowledge for lid .", "nevertheless , current studies do not consider the inter - personal variations due to the lack of user annotated training data ."], "relation": "used for", "id": "2022.findings-acl.303", "year": 2022, "rel_sent": "Experimental results reveal that our model can incarnate user traits and significantly outperforms existing LID systems on handling ambiguous texts .", "forward": false, "src_ids": "2022.findings-acl.303_2258"}
+{"input": "aspect - based sentiment analysis is done by using Method| context: aspect - based sentiment analysis ( absa ) predicts sentiment polarity towards a specific aspect in the given sentence .", "entity": "aspect - based sentiment analysis", "output": "pre - trained language models", "neg_sample": ["aspect - based sentiment analysis is done by using Method", "aspect - based sentiment analysis ( absa ) predicts sentiment polarity towards a specific aspect in the given sentence ."], "relation": "used for", "id": "2022.findings-acl.285", "year": 2022, "rel_sent": "Incorporating Dynamic Semantics into Pre - Trained Language Model for Aspect - based Sentiment Analysis.", "forward": false, "src_ids": "2022.findings-acl.285_2259"}
+{"input": "aspect - based sentiment analysis is done by using OtherScientificTerm| context: aspect - based sentiment analysis ( absa ) predicts sentiment polarity towards a specific aspect in the given sentence . while pre - trained language models such as bert have achieved great success , incorporating dynamic semantic changes into absa remains challenging .", "entity": "aspect - based sentiment analysis", "output": "dynamic aspect - oriented semantics", "neg_sample": ["aspect - based sentiment analysis is done by using OtherScientificTerm", "aspect - based sentiment analysis ( absa ) predicts sentiment polarity towards a specific aspect in the given sentence .", "while pre - trained language models such as bert have achieved great success , incorporating dynamic semantic changes into absa remains challenging ."], "relation": "used for", "id": "2022.findings-acl.285", "year": 2022, "rel_sent": "To this end , in this paper , we propose to address this problem by Dynamic Re - weighting BERT ( DR - BERT ) , a novel method designed to learn dynamic aspect - oriented semantics for ABSA .", "forward": false, "src_ids": "2022.findings-acl.285_2260"}
+{"input": "homophobia is done by using Method| context: hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion .", "entity": "homophobia", "output": "ensembled transformers", "neg_sample": ["homophobia is done by using Method", "hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion ."], "relation": "used for", "id": "2022.ltedi-1.39", "year": 2022, "rel_sent": "Sammaan@LT - EDI - ACL2022 : Ensembled Transformers Against Homophobia and Transphobia.", "forward": false, "src_ids": "2022.ltedi-1.39_2261"}
+{"input": "ensembled transformers is used for OtherScientificTerm| context: hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion .", "entity": "ensembled transformers", "output": "homophobia", "neg_sample": ["ensembled transformers is used for OtherScientificTerm", "hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion ."], "relation": "used for", "id": "2022.ltedi-1.39", "year": 2022, "rel_sent": "Sammaan@LT - EDI - ACL2022 : Ensembled Transformers Against Homophobia and Transphobia.", "forward": true, "src_ids": "2022.ltedi-1.39_2262"}
+{"input": "classifier is done by using Method| context: hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion .", "entity": "classifier", "output": "ensemble of transformer - based models", "neg_sample": ["classifier is done by using Method", "hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion ."], "relation": "used for", "id": "2022.ltedi-1.39", "year": 2022, "rel_sent": "We used an ensemble of transformer - based models to build our classifier .", "forward": false, "src_ids": "2022.ltedi-1.39_2263"}
+{"input": "ensemble of transformer - based models is used for Method| context: hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion .", "entity": "ensemble of transformer - based models", "output": "classifier", "neg_sample": ["ensemble of transformer - based models is used for Method", "hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality , diversity and inclusion ."], "relation": "used for", "id": "2022.ltedi-1.39", "year": 2022, "rel_sent": "We used an ensemble of transformer - based models to build our classifier .", "forward": true, "src_ids": "2022.ltedi-1.39_2264"}
+{"input": "multi - classification challenge is used for Material| context: social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups . in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour . however , these methods are still at an early stage in low - resource languages .", "entity": "multi - classification challenge", "output": "tamil", "neg_sample": ["multi - classification challenge is used for Material", "social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups .", "in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour .", "however , these methods are still at an early stage in low - resource languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.7", "year": 2022, "rel_sent": "With the aim of reducing this barrier , the TamilNLP shared task has proposed a multi - classification challenge for Tamil written in Tamil script and code - mixed to detect abusive comments and hope - speech .", "forward": true, "src_ids": "2022.dravidianlangtech-1.7_2265"}
+{"input": "abusive comments is done by using Task| context: social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups . in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour . however , these methods are still at an early stage in low - resource languages .", "entity": "abusive comments", "output": "multi - classification challenge", "neg_sample": ["abusive comments is done by using Task", "social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups .", "in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour .", "however , these methods are still at an early stage in low - resource languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.7", "year": 2022, "rel_sent": "With the aim of reducing this barrier , the TamilNLP shared task has proposed a multi - classification challenge for Tamil written in Tamil script and code - mixed to detect abusive comments and hope - speech .", "forward": false, "src_ids": "2022.dravidianlangtech-1.7_2266"}
+{"input": "hope - speech is done by using Task| context: social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups . in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour . however , these methods are still at an early stage in low - resource languages .", "entity": "hope - speech", "output": "multi - classification challenge", "neg_sample": ["hope - speech is done by using Task", "social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups .", "in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour .", "however , these methods are still at an early stage in low - resource languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.7", "year": 2022, "rel_sent": "With the aim of reducing this barrier , the TamilNLP shared task has proposed a multi - classification challenge for Tamil written in Tamil script and code - mixed to detect abusive comments and hope - speech .", "forward": false, "src_ids": "2022.dravidianlangtech-1.7_2267"}
+{"input": "tamil is done by using Task| context: social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups . in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour . however , these methods are still at an early stage in low - resource languages .", "entity": "tamil", "output": "multi - classification challenge", "neg_sample": ["tamil is done by using Task", "social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups .", "in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour .", "however , these methods are still at an early stage in low - resource languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.7", "year": 2022, "rel_sent": "With the aim of reducing this barrier , the TamilNLP shared task has proposed a multi - classification challenge for Tamil written in Tamil script and code - mixed to detect abusive comments and hope - speech .", "forward": false, "src_ids": "2022.dravidianlangtech-1.7_2268"}
+{"input": "multi - classification challenge is used for OtherScientificTerm| context: social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups . in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour . however , these methods are still at an early stage in low - resource languages .", "entity": "multi - classification challenge", "output": "abusive comments", "neg_sample": ["multi - classification challenge is used for OtherScientificTerm", "social media has become a dangerous place as bullies take advantage of the anonymity the internet provides to target and intimidate vulnerable individuals and groups .", "in the past few years , the research community has focused on developing automatic classification tools for detecting hate - speech , its variants , and other types of abusive behaviour .", "however , these methods are still at an early stage in low - resource languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.7", "year": 2022, "rel_sent": "With the aim of reducing this barrier , the TamilNLP shared task has proposed a multi - classification challenge for Tamil written in Tamil script and code - mixed to detect abusive comments and hope - speech .", "forward": true, "src_ids": "2022.dravidianlangtech-1.7_2269"}
+{"input": "chinese spell checking is done by using Method| context: recently , pre - trained language models ( plms ) promote the progress of csc task . however , there exists a gap between the learned knowledge of plms and the goal of csc task . plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections .", "entity": "chinese spell checking", "output": "error - driven contrastive probability optimization", "neg_sample": ["chinese spell checking is done by using Method", "recently , pre - trained language models ( plms ) promote the progress of csc task .", "however , there exists a gap between the learned knowledge of plms and the goal of csc task .", "plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections ."], "relation": "used for", "id": "2022.findings-acl.252", "year": 2022, "rel_sent": "The Past Mistake is the Future Wisdom : Error - driven Contrastive Probability Optimization for Chinese Spell Checking.", "forward": false, "src_ids": "2022.findings-acl.252_2270"}
+{"input": "knowledge representations of plms is done by using Method| context: recently , pre - trained language models ( plms ) promote the progress of csc task . however , there exists a gap between the learned knowledge of plms and the goal of csc task . plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections .", "entity": "knowledge representations of plms", "output": "error - driven contrastive probability optimization", "neg_sample": ["knowledge representations of plms is done by using Method", "recently , pre - trained language models ( plms ) promote the progress of csc task .", "however , there exists a gap between the learned knowledge of plms and the goal of csc task .", "plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections ."], "relation": "used for", "id": "2022.findings-acl.252", "year": 2022, "rel_sent": "ECOPO refines the knowledge representations of PLMs , and guides the model to avoid predicting these common characters through an error - driven way .", "forward": false, "src_ids": "2022.findings-acl.252_2271"}
+{"input": "error - driven contrastive probability optimization is used for Task| context: recently , pre - trained language models ( plms ) promote the progress of csc task . however , there exists a gap between the learned knowledge of plms and the goal of csc task . plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections .", "entity": "error - driven contrastive probability optimization", "output": "chinese spell checking", "neg_sample": ["error - driven contrastive probability optimization is used for Task", "recently , pre - trained language models ( plms ) promote the progress of csc task .", "however , there exists a gap between the learned knowledge of plms and the goal of csc task .", "plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections ."], "relation": "used for", "id": "2022.findings-acl.252", "year": 2022, "rel_sent": "The Past Mistake is the Future Wisdom : Error - driven Contrastive Probability Optimization for Chinese Spell Checking.", "forward": true, "src_ids": "2022.findings-acl.252_2272"}
+{"input": "chinese spelling errors is done by using Task| context: recently , pre - trained language models ( plms ) promote the progress of csc task . however , there exists a gap between the learned knowledge of plms and the goal of csc task . plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections .", "entity": "chinese spelling errors", "output": "chinese spell checking ( csc )", "neg_sample": ["chinese spelling errors is done by using Task", "recently , pre - trained language models ( plms ) promote the progress of csc task .", "however , there exists a gap between the learned knowledge of plms and the goal of csc task .", "plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections ."], "relation": "used for", "id": "2022.findings-acl.252", "year": 2022, "rel_sent": "Chinese Spell Checking ( CSC ) aims to detect and correct Chinese spelling errors , which are mainly caused by the phonological or visual similarity .", "forward": false, "src_ids": "2022.findings-acl.252_2273"}
+{"input": "chinese spell checking ( csc ) is used for OtherScientificTerm| context: recently , pre - trained language models ( plms ) promote the progress of csc task . however , there exists a gap between the learned knowledge of plms and the goal of csc task . plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections .", "entity": "chinese spell checking ( csc )", "output": "chinese spelling errors", "neg_sample": ["chinese spell checking ( csc ) is used for OtherScientificTerm", "recently , pre - trained language models ( plms ) promote the progress of csc task .", "however , there exists a gap between the learned knowledge of plms and the goal of csc task .", "plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections ."], "relation": "used for", "id": "2022.findings-acl.252", "year": 2022, "rel_sent": "Chinese Spell Checking ( CSC ) aims to detect and correct Chinese spelling errors , which are mainly caused by the phonological or visual similarity .", "forward": true, "src_ids": "2022.findings-acl.252_2274"}
+{"input": "error - driven contrastive probability optimization is used for Method| context: recently , pre - trained language models ( plms ) promote the progress of csc task . however , there exists a gap between the learned knowledge of plms and the goal of csc task . plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections .", "entity": "error - driven contrastive probability optimization", "output": "knowledge representations of plms", "neg_sample": ["error - driven contrastive probability optimization is used for Method", "recently , pre - trained language models ( plms ) promote the progress of csc task .", "however , there exists a gap between the learned knowledge of plms and the goal of csc task .", "plms focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones , but these are n't the ground - truth corrections ."], "relation": "used for", "id": "2022.findings-acl.252", "year": 2022, "rel_sent": "ECOPO refines the knowledge representations of PLMs , and guides the model to avoid predicting these common characters through an error - driven way .", "forward": true, "src_ids": "2022.findings-acl.252_2275"}
+{"input": "amr parsing is done by using Method| context: abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models . however , there exists a gap between their flat training objective ( i.e. , equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization .", "entity": "amr parsing", "output": "hierarchical curriculum learning", "neg_sample": ["amr parsing is done by using Method", "abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models .", "however , there exists a gap between their flat training objective ( i.e.", ", equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization ."], "relation": "used for", "id": "2022.acl-short.37", "year": 2022, "rel_sent": "Hierarchical Curriculum Learning for AMR Parsing.", "forward": false, "src_ids": "2022.acl-short.37_2276"}
+{"input": "complex structures is done by using Method| context: abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models . however , there exists a gap between their flat training objective ( i.e. , equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization .", "entity": "complex structures", "output": "hierarchical curriculum learning", "neg_sample": ["complex structures is done by using Method", "abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models .", "however , there exists a gap between their flat training objective ( i.e.", ", equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization ."], "relation": "used for", "id": "2022.acl-short.37", "year": 2022, "rel_sent": "Through these two warming - up processes , HCL reduces the difficulty of learning complex structures , thus the flat model can better adapt to the AMR hierarchy .", "forward": false, "src_ids": "2022.acl-short.37_2277"}
+{"input": "hierarchical curriculum learning is used for Task| context: abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models . however , there exists a gap between their flat training objective ( i.e. , equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization .", "entity": "hierarchical curriculum learning", "output": "amr parsing", "neg_sample": ["hierarchical curriculum learning is used for Task", "abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models .", "however , there exists a gap between their flat training objective ( i.e.", ", equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization ."], "relation": "used for", "id": "2022.acl-short.37", "year": 2022, "rel_sent": "Hierarchical Curriculum Learning for AMR Parsing.", "forward": true, "src_ids": "2022.acl-short.37_2278"}
+{"input": "hierarchical curriculum learning is used for OtherScientificTerm| context: abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models . however , there exists a gap between their flat training objective ( i.e. , equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization .", "entity": "hierarchical curriculum learning", "output": "complex structures", "neg_sample": ["hierarchical curriculum learning is used for OtherScientificTerm", "abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models .", "however , there exists a gap between their flat training objective ( i.e.", ", equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization ."], "relation": "used for", "id": "2022.acl-short.37", "year": 2022, "rel_sent": "Through these two warming - up processes , HCL reduces the difficulty of learning complex structures , thus the flat model can better adapt to the AMR hierarchy .", "forward": true, "src_ids": "2022.acl-short.37_2279"}
+{"input": "amr hierarchy is done by using Method| context: abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models . however , there exists a gap between their flat training objective ( i.e. , equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization .", "entity": "amr hierarchy", "output": "flat model", "neg_sample": ["amr hierarchy is done by using Method", "abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models .", "however , there exists a gap between their flat training objective ( i.e.", ", equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization ."], "relation": "used for", "id": "2022.acl-short.37", "year": 2022, "rel_sent": "Through these two warming - up processes , HCL reduces the difficulty of learning complex structures , thus the flat model can better adapt to the AMR hierarchy .", "forward": false, "src_ids": "2022.acl-short.37_2280"}
+{"input": "flat model is used for OtherScientificTerm| context: abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models . however , there exists a gap between their flat training objective ( i.e. , equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization .", "entity": "flat model", "output": "amr hierarchy", "neg_sample": ["flat model is used for OtherScientificTerm", "abstract meaning representation ( amr ) parsing aims to translate sentences to semantic representation with a hierarchical structure , and is recently empowered by pretrained sequence - to - sequence models .", "however , there exists a gap between their flat training objective ( i.e.", ", equally treats all output tokens ) and the hierarchical amr structure , which limits the model generalization ."], "relation": "used for", "id": "2022.acl-short.37", "year": 2022, "rel_sent": "Through these two warming - up processes , HCL reduces the difficulty of learning complex structures , thus the flat model can better adapt to the AMR hierarchy .", "forward": true, "src_ids": "2022.acl-short.37_2281"}
+{"input": "distantly supervised relation extraction is done by using Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "distantly supervised relation extraction", "output": "hierarchical contrastive learning framework", "neg_sample": ["distantly supervised relation extraction is done by using Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "HiCLRE : A Hierarchical Contrastive Learning Framework for Distantly Supervised Relation Extraction.", "forward": false, "src_ids": "2022.findings-acl.202_2282"}
+{"input": "distantly supervised relation extraction ( hiclre ) is done by using Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "distantly supervised relation extraction ( hiclre )", "output": "hierarchical contrastive learning framework", "neg_sample": ["distantly supervised relation extraction ( hiclre ) is done by using Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical contrastive learning Framework for Distantly Supervised relation extraction ( HiCLRE ) to reduce noisy sentences , which integrate the global structural information and local fine - grained interaction .", "forward": false, "src_ids": "2022.findings-acl.202_2283"}
+{"input": "noisy sentences is done by using Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "noisy sentences", "output": "hierarchical contrastive learning framework", "neg_sample": ["noisy sentences is done by using Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical contrastive learning Framework for Distantly Supervised relation extraction ( HiCLRE ) to reduce noisy sentences , which integrate the global structural information and local fine - grained interaction .", "forward": false, "src_ids": "2022.findings-acl.202_2284"}
+{"input": "hierarchical contrastive learning framework is used for Task| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "entity": "hierarchical contrastive learning framework", "output": "distantly supervised relation extraction", "neg_sample": ["hierarchical contrastive learning framework is used for Task", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "HiCLRE : A Hierarchical Contrastive Learning Framework for Distantly Supervised Relation Extraction.", "forward": true, "src_ids": "2022.findings-acl.202_2285"}
+{"input": "hierarchical contrastive learning framework is used for Task| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "hierarchical contrastive learning framework", "output": "distantly supervised relation extraction ( hiclre )", "neg_sample": ["hierarchical contrastive learning framework is used for Task", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical contrastive learning Framework for Distantly Supervised relation extraction ( HiCLRE ) to reduce noisy sentences , which integrate the global structural information and local fine - grained interaction .", "forward": true, "src_ids": "2022.findings-acl.202_2286"}
+{"input": "hierarchical contrastive learning framework is used for OtherScientificTerm| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "hierarchical contrastive learning framework", "output": "noisy sentences", "neg_sample": ["hierarchical contrastive learning framework is used for OtherScientificTerm", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical contrastive learning Framework for Distantly Supervised relation extraction ( HiCLRE ) to reduce noisy sentences , which integrate the global structural information and local fine - grained interaction .", "forward": true, "src_ids": "2022.findings-acl.202_2287"}
+{"input": "de - noising context - aware representations is done by using Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "de - noising context - aware representations", "output": "three - level hierarchical learning framework", "neg_sample": ["de - noising context - aware representations is done by using Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "Specifically , we propose a three - level hierarchical learning framework to interact with cross levels , generating the de - noising context - aware representations via adapting the existing multi - head self - attention , named Multi - Granularity Recontextualization .", "forward": false, "src_ids": "2022.findings-acl.202_2288"}
+{"input": "cross levels is done by using Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "cross levels", "output": "three - level hierarchical learning framework", "neg_sample": ["cross levels is done by using Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "Specifically , we propose a three - level hierarchical learning framework to interact with cross levels , generating the de - noising context - aware representations via adapting the existing multi - head self - attention , named Multi - Granularity Recontextualization .", "forward": false, "src_ids": "2022.findings-acl.202_2289"}
+{"input": "three - level hierarchical learning framework is used for OtherScientificTerm| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "entity": "three - level hierarchical learning framework", "output": "cross levels", "neg_sample": ["three - level hierarchical learning framework is used for OtherScientificTerm", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "Specifically , we propose a three - level hierarchical learning framework to interact with cross levels , generating the de - noising context - aware representations via adapting the existing multi - head self - attention , named Multi - Granularity Recontextualization .", "forward": true, "src_ids": "2022.findings-acl.202_2290"}
+{"input": "three - level hierarchical learning framework is used for Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "three - level hierarchical learning framework", "output": "de - noising context - aware representations", "neg_sample": ["three - level hierarchical learning framework is used for Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "Specifically , we propose a three - level hierarchical learning framework to interact with cross levels , generating the de - noising context - aware representations via adapting the existing multi - head self - attention , named Multi - Granularity Recontextualization .", "forward": true, "src_ids": "2022.findings-acl.202_2291"}
+{"input": "contrastive learning is done by using Material| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "contrastive learning", "output": "pseudo positive samples", "neg_sample": ["contrastive learning is done by using Material", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "Meanwhile , pseudo positive samples are also provided in the specific level for contrastive learning via a dynamic gradient - based data augmentation strategy , named Dynamic Gradient Adversarial Perturbation .", "forward": false, "src_ids": "2022.findings-acl.202_2292"}
+{"input": "pseudo positive samples is used for Method| context: distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships . previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels .", "entity": "pseudo positive samples", "output": "contrastive learning", "neg_sample": ["pseudo positive samples is used for Method", "distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships .", "previous works of distantly supervised relation extraction ( dsre ) task generally focus on sentence - level or bag - level de - noising techniques independently , neglecting the explicit interaction with cross levels ."], "relation": "used for", "id": "2022.findings-acl.202", "year": 2022, "rel_sent": "Meanwhile , pseudo positive samples are also provided in the specific level for contrastive learning via a dynamic gradient - based data augmentation strategy , named Dynamic Gradient Adversarial Perturbation .", "forward": true, "src_ids": "2022.findings-acl.202_2293"}
+{"input": "simultaneous machine translation is done by using OtherScientificTerm| context: simultaneous machine translation ( simt ) outputs translation while reading source sentence and hence requires a policy to decide whether to wait for the next source word ( read ) or generate a target word ( write ) , the actions of which form a read / write path . although the read / write path is essential to simt performance , no direct supervision is given to the path in the existing methods .", "entity": "simultaneous machine translation", "output": "dual read / write paths", "neg_sample": ["simultaneous machine translation is done by using OtherScientificTerm", "simultaneous machine translation ( simt ) outputs translation while reading source sentence and hence requires a policy to decide whether to wait for the next source word ( read ) or generate a target word ( write ) , the actions of which form a read / write path .", "although the read / write path is essential to simt performance , no direct supervision is given to the path in the existing methods ."], "relation": "used for", "id": "2022.acl-long.176", "year": 2022, "rel_sent": "Modeling Dual Read / Write Paths for Simultaneous Machine Translation.", "forward": false, "src_ids": "2022.acl-long.176_2294"}
+{"input": "read / write paths is done by using OtherScientificTerm| context: simultaneous machine translation ( simt ) outputs translation while reading source sentence and hence requires a policy to decide whether to wait for the next source word ( read ) or generate a target word ( write ) , the actions of which form a read / write path . although the read / write path is essential to simt performance , no direct supervision is given to the path in the existing methods .", "entity": "read / write paths", "output": "duality constraints", "neg_sample": ["read / write paths is done by using OtherScientificTerm", "simultaneous machine translation ( simt ) outputs translation while reading source sentence and hence requires a policy to decide whether to wait for the next source word ( read ) or generate a target word ( write ) , the actions of which form a read / write path .", "although the read / write path is essential to simt performance , no direct supervision is given to the path in the existing methods ."], "relation": "used for", "id": "2022.acl-long.176", "year": 2022, "rel_sent": "In this paper , we propose a method of dual - path SiMT which introduces duality constraints to direct the read / write path .", "forward": false, "src_ids": "2022.acl-long.176_2295"}
+{"input": "vqa task is done by using Method| context: clip has shown a remarkable zero - shot capability on a wide range of vision tasks . previously , clip is only regarded as a powerful visual encoder . however , after being pre - trained by language supervision from a large amount of image - caption pairs , clip itself should also have acquired some few - shot abilities for vision - language tasks .", "entity": "vqa task", "output": "parameter - efficient fine - tuning strategy", "neg_sample": ["vqa task is done by using Method", "clip has shown a remarkable zero - shot capability on a wide range of vision tasks .", "previously , clip is only regarded as a powerful visual encoder .", "however , after being pre - trained by language supervision from a large amount of image - caption pairs , clip itself should also have acquired some few - shot abilities for vision - language tasks ."], "relation": "used for", "id": "2022.acl-long.421", "year": 2022, "rel_sent": "Then we propose a parameter - efficient fine - tuning strategy to boost the few - shot performance on the vqa task .", "forward": false, "src_ids": "2022.acl-long.421_2296"}
+{"input": "parameter - efficient fine - tuning strategy is used for Task| context: clip has shown a remarkable zero - shot capability on a wide range of vision tasks . previously , clip is only regarded as a powerful visual encoder . however , after being pre - trained by language supervision from a large amount of image - caption pairs , clip itself should also have acquired some few - shot abilities for vision - language tasks .", "entity": "parameter - efficient fine - tuning strategy", "output": "vqa task", "neg_sample": ["parameter - efficient fine - tuning strategy is used for Task", "clip has shown a remarkable zero - shot capability on a wide range of vision tasks .", "previously , clip is only regarded as a powerful visual encoder .", "however , after being pre - trained by language supervision from a large amount of image - caption pairs , clip itself should also have acquired some few - shot abilities for vision - language tasks ."], "relation": "used for", "id": "2022.acl-long.421", "year": 2022, "rel_sent": "Then we propose a parameter - efficient fine - tuning strategy to boost the few - shot performance on the vqa task .", "forward": true, "src_ids": "2022.acl-long.421_2297"}
+{"input": "zero - shot entity linking is done by using Method| context: recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain . current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions . however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain .", "entity": "zero - shot entity linking", "output": "transformational biencoder", "neg_sample": ["zero - shot entity linking is done by using Method", "recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain .", "current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions .", "however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain ."], "relation": "used for", "id": "2022.findings-acl.114", "year": 2022, "rel_sent": "A Transformational Biencoder with In - Domain Negative Sampling for Zero - Shot Entity Linking.", "forward": false, "src_ids": "2022.findings-acl.114_2298"}
+{"input": "zero - shot transfer is done by using Method| context: recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain . current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions . however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain .", "entity": "zero - shot transfer", "output": "transformational biencoder", "neg_sample": ["zero - shot transfer is done by using Method", "recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain .", "current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions .", "however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain ."], "relation": "used for", "id": "2022.findings-acl.114", "year": 2022, "rel_sent": "We solve this problem by proposing a Transformational Biencoder that incorporates a transformation into BERT to perform a zero - shot transfer from the source domain during training .", "forward": false, "src_ids": "2022.findings-acl.114_2299"}
+{"input": "transformational biencoder is used for Task| context: recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain . current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions . however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain .", "entity": "transformational biencoder", "output": "zero - shot entity linking", "neg_sample": ["transformational biencoder is used for Task", "recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain .", "current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions .", "however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain ."], "relation": "used for", "id": "2022.findings-acl.114", "year": 2022, "rel_sent": "A Transformational Biencoder with In - Domain Negative Sampling for Zero - Shot Entity Linking.", "forward": true, "src_ids": "2022.findings-acl.114_2300"}
+{"input": "transformational biencoder is used for Task| context: recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain . current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions . however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain .", "entity": "transformational biencoder", "output": "zero - shot transfer", "neg_sample": ["transformational biencoder is used for Task", "recent interest in entity linking has focused in the zero - shot scenario , where at test time the entity mention to be labelled is never seen during training , or may belong to a different domain from the source domain .", "current work leverage pre - trained bert with the implicit assumption that it bridges the gap between the source and target domain distributions .", "however , fine - tuned bert has a considerable underperformance at zero - shot when applied in a different domain ."], "relation": "used for", "id": "2022.findings-acl.114", "year": 2022, "rel_sent": "We solve this problem by proposing a Transformational Biencoder that incorporates a transformation into BERT to perform a zero - shot transfer from the source domain during training .", "forward": true, "src_ids": "2022.findings-acl.114_2301"}
+{"input": "ue methods is used for Task| context: most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks . little attention has been paid to ue in natural language processing .", "entity": "ue methods", "output": "misclassification detection", "neg_sample": ["ue methods is used for Task", "most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks .", "little attention has been paid to ue in natural language processing ."], "relation": "used for", "id": "2022.acl-long.566", "year": 2022, "rel_sent": "Tofill this gap , we perform a vast empirical investigation of state - of - the - art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications , one of which approaches or even outperforms computationally intensive methods .", "forward": true, "src_ids": "2022.acl-long.566_2302"}
+{"input": "misclassification detection is done by using Method| context: uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc . most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks . little attention has been paid to ue in natural language processing .", "entity": "misclassification detection", "output": "ue methods", "neg_sample": ["misclassification detection is done by using Method", "uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc .", "most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks .", "little attention has been paid to ue in natural language processing ."], "relation": "used for", "id": "2022.acl-long.566", "year": 2022, "rel_sent": "Tofill this gap , we perform a vast empirical investigation of state - of - the - art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications , one of which approaches or even outperforms computationally intensive methods .", "forward": false, "src_ids": "2022.acl-long.566_2303"}
+{"input": "arpoca is used for OtherScientificTerm| context: automatic speech recognition ( asr ) has evolved from a pipeline architecture with pronunciation dictionaries , phonetic features and language models to the end - to - end systems performing a direct translation from a raw waveform into a word sequence . with the increase in accuracy and the availability of pre - trained models , the asr systems are now omnipresent in our daily applications . on the other hand , the models ' interpretability and their computational cost have become more challenging , particularly when dealing with less - common languages or identifying regional variations of speakers .", "entity": "arpoca", "output": "regional language variation", "neg_sample": ["arpoca is used for OtherScientificTerm", "automatic speech recognition ( asr ) has evolved from a pipeline architecture with pronunciation dictionaries , phonetic features and language models to the end - to - end systems performing a direct translation from a raw waveform into a word sequence .", "with the increase in accuracy and the availability of pre - trained models , the asr systems are now omnipresent in our daily applications .", "on the other hand , the models ' interpretability and their computational cost have become more challenging , particularly when dealing with less - common languages or identifying regional variations of speakers ."], "relation": "used for", "id": "2022.acl-srw.28", "year": 2022, "rel_sent": "This research proposal will follow a four - stage process : 1 ) Proving an overview of acoustic features and feature extraction algorithms ; 2 ) Exploring current ASR models , tools , and performance assessment techniques ; 3 ) Aligning features with interpretable phonetic transcripts ; and 4 ) Designing a prototype ARPOCA to increase awareness of regional language variation and improve models feedback by developing a semi - automatic acoustic features extraction using PRAAT in conjunction with phonetic transcription .", "forward": true, "src_ids": "2022.acl-srw.28_2304"}
+{"input": "regional language variation is done by using Method| context: automatic speech recognition ( asr ) has evolved from a pipeline architecture with pronunciation dictionaries , phonetic features and language models to the end - to - end systems performing a direct translation from a raw waveform into a word sequence . with the increase in accuracy and the availability of pre - trained models , the asr systems are now omnipresent in our daily applications . on the other hand , the models ' interpretability and their computational cost have become more challenging , particularly when dealing with less - common languages or identifying regional variations of speakers .", "entity": "regional language variation", "output": "arpoca", "neg_sample": ["regional language variation is done by using Method", "automatic speech recognition ( asr ) has evolved from a pipeline architecture with pronunciation dictionaries , phonetic features and language models to the end - to - end systems performing a direct translation from a raw waveform into a word sequence .", "with the increase in accuracy and the availability of pre - trained models , the asr systems are now omnipresent in our daily applications .", "on the other hand , the models ' interpretability and their computational cost have become more challenging , particularly when dealing with less - common languages or identifying regional variations of speakers ."], "relation": "used for", "id": "2022.acl-srw.28", "year": 2022, "rel_sent": "This research proposal will follow a four - stage process : 1 ) Proving an overview of acoustic features and feature extraction algorithms ; 2 ) Exploring current ASR models , tools , and performance assessment techniques ; 3 ) Aligning features with interpretable phonetic transcripts ; and 4 ) Designing a prototype ARPOCA to increase awareness of regional language variation and improve models feedback by developing a semi - automatic acoustic features extraction using PRAAT in conjunction with phonetic transcription .", "forward": false, "src_ids": "2022.acl-srw.28_2305"}
+{"input": "retrieval is done by using Method| context: machine translation models struggle when translating out - of - domain text , which makes domain adaptation a topic of critical importance . however , most domain adaptation methods focus on fine - tuning or training the entire or part of the model on every new domain , which can be costly . on the other hand , semi - parametric models have been shown to successfully perform domain adaptation by retrieving examples from an in - domain datastore ( khandelwal et al . , 2021 ) . a drawback of these retrieval - augmented models , however , is that they tend to be substantially slower .", "entity": "retrieval", "output": "caching strategy", "neg_sample": ["retrieval is done by using Method", "machine translation models struggle when translating out - of - domain text , which makes domain adaptation a topic of critical importance .", "however , most domain adaptation methods focus on fine - tuning or training the entire or part of the model on every new domain , which can be costly .", "on the other hand , semi - parametric models have been shown to successfully perform domain adaptation by retrieving examples from an in - domain datastore ( khandelwal et al .", ", 2021 ) .", "a drawback of these retrieval - augmented models , however , is that they tend to be substantially slower ."], "relation": "used for", "id": "2022.spanlp-1.3", "year": 2022, "rel_sent": "( 2021 ) for language modeling , and introduce a simple but effective caching strategy that avoids performing retrieval when similar contexts have been seen before .", "forward": false, "src_ids": "2022.spanlp-1.3_2306"}
+{"input": "caching strategy is used for Task| context: machine translation models struggle when translating out - of - domain text , which makes domain adaptation a topic of critical importance . however , most domain adaptation methods focus on fine - tuning or training the entire or part of the model on every new domain , which can be costly . on the other hand , semi - parametric models have been shown to successfully perform domain adaptation by retrieving examples from an in - domain datastore ( khandelwal et al . , 2021 ) .", "entity": "caching strategy", "output": "retrieval", "neg_sample": ["caching strategy is used for Task", "machine translation models struggle when translating out - of - domain text , which makes domain adaptation a topic of critical importance .", "however , most domain adaptation methods focus on fine - tuning or training the entire or part of the model on every new domain , which can be costly .", "on the other hand , semi - parametric models have been shown to successfully perform domain adaptation by retrieving examples from an in - domain datastore ( khandelwal et al .", ", 2021 ) ."], "relation": "used for", "id": "2022.spanlp-1.3", "year": 2022, "rel_sent": "( 2021 ) for language modeling , and introduce a simple but effective caching strategy that avoids performing retrieval when similar contexts have been seen before .", "forward": true, "src_ids": "2022.spanlp-1.3_2307"}
+{"input": "low - resource text classification and generation is done by using Method| context: building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available .", "entity": "low - resource text classification and generation", "output": "meta - learning", "neg_sample": ["low - resource text classification and generation is done by using Method", "building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available ."], "relation": "used for", "id": "2022.acl-long.44", "year": 2022, "rel_sent": "Improving Meta - learning for Low - resource Text Classification and Generation via Memory Imitation.", "forward": false, "src_ids": "2022.acl-long.44_2308"}
+{"input": "meta - learning is used for Task| context: building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available . optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks . nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks .", "entity": "meta - learning", "output": "low - resource text classification and generation", "neg_sample": ["meta - learning is used for Task", "building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available .", "optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks .", "nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks ."], "relation": "used for", "id": "2022.acl-long.44", "year": 2022, "rel_sent": "Improving Meta - learning for Low - resource Text Classification and Generation via Memory Imitation.", "forward": true, "src_ids": "2022.acl-long.44_2309"}
+{"input": "task adaptation is done by using OtherScientificTerm| context: building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available . optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks .", "entity": "task adaptation", "output": "support sets", "neg_sample": ["task adaptation is done by using OtherScientificTerm", "building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available .", "optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks ."], "relation": "used for", "id": "2022.acl-long.44", "year": 2022, "rel_sent": "To address this issue , we propose a memory imitation meta - learning ( MemIML ) method that enhances the model 's reliance on support sets for task adaptation .", "forward": false, "src_ids": "2022.acl-long.44_2310"}
+{"input": "support sets is used for Task| context: building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available . optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks . nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks .", "entity": "support sets", "output": "task adaptation", "neg_sample": ["support sets is used for Task", "building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available .", "optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks .", "nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks ."], "relation": "used for", "id": "2022.acl-long.44", "year": 2022, "rel_sent": "To address this issue , we propose a memory imitation meta - learning ( MemIML ) method that enhances the model 's reliance on support sets for task adaptation .", "forward": true, "src_ids": "2022.acl-long.44_2311"}
+{"input": "support set information is done by using Method| context: building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available . optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks . nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks .", "entity": "support set information", "output": "task - specific memory module", "neg_sample": ["support set information is done by using Method", "building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available .", "optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks .", "nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks ."], "relation": "used for", "id": "2022.acl-long.44", "year": 2022, "rel_sent": "Specifically , we introduce a task - specific memory module to store support set information and construct an imitation module toforce query sets to imitate the behaviors of support sets stored in the memory .", "forward": false, "src_ids": "2022.acl-long.44_2312"}
+{"input": "task - specific memory module is used for OtherScientificTerm| context: building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available . optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks . nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks .", "entity": "task - specific memory module", "output": "support set information", "neg_sample": ["task - specific memory module is used for OtherScientificTerm", "building models of natural language processing ( nlp ) is challenging in low - resource scenarios where limited data are available .", "optimization - based meta - learning algorithms achieve promising results in low - resource scenarios by adapting a well - generalized model initialization to handle new tasks .", "nonetheless , these approaches suffer from the memorization overfitting issue , where the model tends to memorize the meta - training tasks while ignoring support sets when adapting to new tasks ."], "relation": "used for", "id": "2022.acl-long.44", "year": 2022, "rel_sent": "Specifically , we introduce a task - specific memory module to store support set information and construct an imitation module toforce query sets to imitate the behaviors of support sets stored in the memory .", "forward": true, "src_ids": "2022.acl-long.44_2313"}
+{"input": "inductive bias is done by using Task| context: pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks . such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations .", "entity": "inductive bias", "output": "probing", "neg_sample": ["inductive bias is done by using Task", "pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks .", "such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations ."], "relation": "used for", "id": "2022.acl-long.129", "year": 2022, "rel_sent": "In the theoretical portion of this paper , we take the position that the goal of probing ought to be measuring the amount of inductive bias that the representations encode on a specific task .", "forward": false, "src_ids": "2022.acl-long.129_2314"}
+{"input": "probing is used for OtherScientificTerm| context: pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks . such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations . in general , researchers quantify the amount of linguistic information through probing , an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations . unfortunately , this definition of probing has been subject to extensive criticism in the literature , and has been observed to lead to paradoxical and counter - intuitive results .", "entity": "probing", "output": "inductive bias", "neg_sample": ["probing is used for OtherScientificTerm", "pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks .", "such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations .", "in general , researchers quantify the amount of linguistic information through probing , an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations .", "unfortunately , this definition of probing has been subject to extensive criticism in the literature , and has been observed to lead to paradoxical and counter - intuitive results ."], "relation": "used for", "id": "2022.acl-long.129", "year": 2022, "rel_sent": "In the theoretical portion of this paper , we take the position that the goal of probing ought to be measuring the amount of inductive bias that the representations encode on a specific task .", "forward": true, "src_ids": "2022.acl-long.129_2315"}
+{"input": "bayesian framework is used for OtherScientificTerm| context: pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks . such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations . in general , researchers quantify the amount of linguistic information through probing , an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations . unfortunately , this definition of probing has been subject to extensive criticism in the literature , and has been observed to lead to paradoxical and counter - intuitive results .", "entity": "bayesian framework", "output": "inductive bias", "neg_sample": ["bayesian framework is used for OtherScientificTerm", "pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks .", "such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations .", "in general , researchers quantify the amount of linguistic information through probing , an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations .", "unfortunately , this definition of probing has been subject to extensive criticism in the literature , and has been observed to lead to paradoxical and counter - intuitive results ."], "relation": "used for", "id": "2022.acl-long.129", "year": 2022, "rel_sent": "We further describe a Bayesian framework that operationalizes this goal and allows us to quantify the representations ' inductive bias .", "forward": true, "src_ids": "2022.acl-long.129_2316"}
+{"input": "inductive bias is done by using Method| context: pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks . such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations . in general , researchers quantify the amount of linguistic information through probing , an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations . unfortunately , this definition of probing has been subject to extensive criticism in the literature , and has been observed to lead to paradoxical and counter - intuitive results .", "entity": "inductive bias", "output": "bayesian framework", "neg_sample": ["inductive bias is done by using Method", "pre - trained contextual representations have led to dramatic performance improvements on a range of downstream tasks .", "such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations .", "in general , researchers quantify the amount of linguistic information through probing , an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations .", "unfortunately , this definition of probing has been subject to extensive criticism in the literature , and has been observed to lead to paradoxical and counter - intuitive results ."], "relation": "used for", "id": "2022.acl-long.129", "year": 2022, "rel_sent": "We further describe a Bayesian framework that operationalizes this goal and allows us to quantify the representations ' inductive bias .", "forward": false, "src_ids": "2022.acl-long.129_2317"}
+{"input": "few - shot fine - tuning is done by using Method| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "few - shot fine - tuning", "output": "perfect", "neg_sample": ["few - shot fine - tuning is done by using Method", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "In this work , we propose Perfect , a simple and efficient method for few - shot fine - tuning of PLMs without relying on any such handcrafting , which is highly effective given as few as 32 data points .", "forward": false, "src_ids": "2022.acl-long.254_2318"}
+{"input": "sample - efficient fine - tuning is done by using OtherScientificTerm| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "sample - efficient fine - tuning", "output": "task - specific adapters", "neg_sample": ["sample - efficient fine - tuning is done by using OtherScientificTerm", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "Perfect makes two key design choices : First , we show that manually engineered task prompts can be replaced with task - specific adapters that enable sample - efficient fine - tuning and reduce memory and storage costs by roughly factors of 5 and 100 , respectively .", "forward": false, "src_ids": "2022.acl-long.254_2319"}
+{"input": "task - specific adapters is used for Task| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "task - specific adapters", "output": "sample - efficient fine - tuning", "neg_sample": ["task - specific adapters is used for Task", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "Perfect makes two key design choices : First , we show that manually engineered task prompts can be replaced with task - specific adapters that enable sample - efficient fine - tuning and reduce memory and storage costs by roughly factors of 5 and 100 , respectively .", "forward": true, "src_ids": "2022.acl-long.254_2320"}
+{"input": "fine - tuning is done by using OtherScientificTerm| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "fine - tuning", "output": "multi - token label embeddings", "neg_sample": ["fine - tuning is done by using OtherScientificTerm", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "Second , instead of using handcrafted verbalizers , we learn new multi - token label embeddings during fine - tuning , which are not tied to the model vocabulary and which allow us to avoid complex auto - regressive decoding .", "forward": false, "src_ids": "2022.acl-long.254_2321"}
+{"input": "training is done by using OtherScientificTerm| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "training", "output": "embeddings", "neg_sample": ["training is done by using OtherScientificTerm", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "These embeddings are not only learnable from limited data but also enable nearly 100x faster training and inference .", "forward": false, "src_ids": "2022.acl-long.254_2322"}
+{"input": "inference is done by using OtherScientificTerm| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "inference", "output": "embeddings", "neg_sample": ["inference is done by using OtherScientificTerm", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "These embeddings are not only learnable from limited data but also enable nearly 100x faster training and inference .", "forward": false, "src_ids": "2022.acl-long.254_2323"}
+{"input": "embeddings is used for Task| context: current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score .", "entity": "embeddings", "output": "training", "neg_sample": ["embeddings is used for Task", "current methods for few - shot fine - tuning of pretrained masked language models ( plms ) require carefully engineered prompts and verbalizers for each new task to convert examples into a cloze - format that the plm can score ."], "relation": "used for", "id": "2022.acl-long.254", "year": 2022, "rel_sent": "These embeddings are not only learnable from limited data but also enable nearly 100x faster training and inference .", "forward": true, "src_ids": "2022.acl-long.254_2324"}
+{"input": "real - world mt applications is done by using Method| context: though nearest neighbor machine translation ( knn - mt ) ( citation ) has proved to introduce significant performance boosts over standard neural mt systems , it is prohibitively slow since it uses the entire reference corpus as the datastore for the nearest neighbor search . this means each step for each beam in the beam search has to search over the entire reference corpus . knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services .", "entity": "real - world mt applications", "output": "knn - mt systems", "neg_sample": ["real - world mt applications is done by using Method", "though nearest neighbor machine translation ( knn - mt ) ( citation ) has proved to introduce significant performance boosts over standard neural mt systems , it is prohibitively slow since it uses the entire reference corpus as the datastore for the nearest neighbor search .", "this means each step for each beam in the beam search has to search over the entire reference corpus .", "knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services ."], "relation": "used for", "id": "2022.findings-acl.47", "year": 2022, "rel_sent": "Fast kNN - MT enables the practical use of kNN - MT systems in real - world MT applications .", "forward": false, "src_ids": "2022.findings-acl.47_2325"}
+{"input": "knn - mt systems is used for Task| context: though nearest neighbor machine translation ( knn - mt ) ( citation ) has proved to introduce significant performance boosts over standard neural mt systems , it is prohibitively slow since it uses the entire reference corpus as the datastore for the nearest neighbor search . this means each step for each beam in the beam search has to search over the entire reference corpus . knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services .", "entity": "knn - mt systems", "output": "real - world mt applications", "neg_sample": ["knn - mt systems is used for Task", "though nearest neighbor machine translation ( knn - mt ) ( citation ) has proved to introduce significant performance boosts over standard neural mt systems , it is prohibitively slow since it uses the entire reference corpus as the datastore for the nearest neighbor search .", "this means each step for each beam in the beam search has to search over the entire reference corpus .", "knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services ."], "relation": "used for", "id": "2022.findings-acl.47", "year": 2022, "rel_sent": "Fast kNN - MT enables the practical use of kNN - MT systems in real - world MT applications .", "forward": true, "src_ids": "2022.findings-acl.47_2326"}
+{"input": "knn - mt is used for Task| context: though nearest neighbor machine translation ( knn - mt ) ( citation ) has proved to introduce significant performance boosts over standard neural mt systems , it is prohibitively slow since it uses the entire reference corpus as the datastore for the nearest neighbor search . this means each step for each beam in the beam search has to search over the entire reference corpus . knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services .", "entity": "knn - mt", "output": "real - world mt applications", "neg_sample": ["knn - mt is used for Task", "though nearest neighbor machine translation ( knn - mt ) ( citation ) has proved to introduce significant performance boosts over standard neural mt systems , it is prohibitively slow since it uses the entire reference corpus as the datastore for the nearest neighbor search .", "this means each step for each beam in the beam search has to search over the entire reference corpus .", "knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services ."], "relation": "used for", "id": "2022.findings-acl.47", "year": 2022, "rel_sent": "Fast kNN - MT enables the practical use of kNN - MT systems in real - world MT applications .", "forward": true, "src_ids": "2022.findings-acl.47_2327"}
+{"input": "knn - mt is used for Method| context: this means each step for each beam in the beam search has to search over the entire reference corpus . knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services .", "entity": "knn - mt", "output": "nearest neighbor search", "neg_sample": ["knn - mt is used for Method", "this means each step for each beam in the beam search has to search over the entire reference corpus .", "knn - mt is thus two - orders slower than vanilla mt models , making it hard to be applied to real - world applications , especially online services ."], "relation": "used for", "id": "2022.findings-acl.47", "year": 2022, "rel_sent": "Fast kNN - MT constructs a significantly smaller datastore for the nearest neighbor search : for each word in a source sentence , Fast kNN - MT first selects its nearest token - level neighbors , which is limited to tokens that are the same as the query token .", "forward": true, "src_ids": "2022.findings-acl.47_2328"}
+{"input": "nearest neighbor search is done by using Method| context: this means each step for each beam in the beam search has to search over the entire reference corpus .", "entity": "nearest neighbor search", "output": "knn - mt", "neg_sample": ["nearest neighbor search is done by using Method", "this means each step for each beam in the beam search has to search over the entire reference corpus ."], "relation": "used for", "id": "2022.findings-acl.47", "year": 2022, "rel_sent": "Fast kNN - MT constructs a significantly smaller datastore for the nearest neighbor search : for each word in a source sentence , Fast kNN - MT first selects its nearest token - level neighbors , which is limited to tokens that are the same as the query token .", "forward": false, "src_ids": "2022.findings-acl.47_2329"}
+{"input": "real - world mt applications is done by using Method| context: this means each step for each beam in the beam search has to search over the entire reference corpus .", "entity": "real - world mt applications", "output": "knn - mt", "neg_sample": ["real - world mt applications is done by using Method", "this means each step for each beam in the beam search has to search over the entire reference corpus ."], "relation": "used for", "id": "2022.findings-acl.47", "year": 2022, "rel_sent": "Fast kNN - MT enables the practical use of kNN - MT systems in real - world MT applications .", "forward": false, "src_ids": "2022.findings-acl.47_2330"}
+{"input": "social media data is done by using Method| context: detecting emotion in text allows social and computational scientists to study how people behave and react to online events . however , developing these tools for different languages requires data that is not always available .", "entity": "social media data", "output": "multilingual emotion prediction model", "neg_sample": ["social media data is done by using Method", "detecting emotion in text allows social and computational scientists to study how people behave and react to online events .", "however , developing these tools for different languages requires data that is not always available ."], "relation": "used for", "id": "2022.wassa-1.18", "year": 2022, "rel_sent": "We train a multilingual emotion prediction model for social media data , XLM - EMO .", "forward": false, "src_ids": "2022.wassa-1.18_2331"}
+{"input": "multilingual emotion prediction model is used for Material| context: detecting emotion in text allows social and computational scientists to study how people behave and react to online events . however , developing these tools for different languages requires data that is not always available .", "entity": "multilingual emotion prediction model", "output": "social media data", "neg_sample": ["multilingual emotion prediction model is used for Material", "detecting emotion in text allows social and computational scientists to study how people behave and react to online events .", "however , developing these tools for different languages requires data that is not always available ."], "relation": "used for", "id": "2022.wassa-1.18", "year": 2022, "rel_sent": "We train a multilingual emotion prediction model for social media data , XLM - EMO .", "forward": true, "src_ids": "2022.wassa-1.18_2332"}
+{"input": "top - notch multilingual parsing and generation is done by using OtherScientificTerm| context: a language - independent representation of meaning is one of the most coveted dreams in natural language understanding . with this goal in mind , several formalisms have been proposed as frameworks for meaning representation in semantic parsing . and yet , the dependencies these formalisms share with respect to language - specific repositories of knowledge make the objective of closing the gap between high- and low - resourced languages hard to accomplish .", "entity": "top - notch multilingual parsing and generation", "output": "fully - semantic framing", "neg_sample": ["top - notch multilingual parsing and generation is done by using OtherScientificTerm", "a language - independent representation of meaning is one of the most coveted dreams in natural language understanding .", "with this goal in mind , several formalisms have been proposed as frameworks for meaning representation in semantic parsing .", "and yet , the dependencies these formalisms share with respect to language - specific repositories of knowledge make the objective of closing the gap between high- and low - resourced languages hard to accomplish ."], "relation": "used for", "id": "2022.acl-long.121", "year": 2022, "rel_sent": "Moreover , we show how BMR is able to outperform previous formalisms thanks to its fully - semantic framing , which enables top - notch multilingual parsing and generation .", "forward": false, "src_ids": "2022.acl-long.121_2333"}
+{"input": "fully - semantic framing is used for Task| context: a language - independent representation of meaning is one of the most coveted dreams in natural language understanding . with this goal in mind , several formalisms have been proposed as frameworks for meaning representation in semantic parsing . and yet , the dependencies these formalisms share with respect to language - specific repositories of knowledge make the objective of closing the gap between high- and low - resourced languages hard to accomplish .", "entity": "fully - semantic framing", "output": "top - notch multilingual parsing and generation", "neg_sample": ["fully - semantic framing is used for Task", "a language - independent representation of meaning is one of the most coveted dreams in natural language understanding .", "with this goal in mind , several formalisms have been proposed as frameworks for meaning representation in semantic parsing .", "and yet , the dependencies these formalisms share with respect to language - specific repositories of knowledge make the objective of closing the gap between high- and low - resourced languages hard to accomplish ."], "relation": "used for", "id": "2022.acl-long.121", "year": 2022, "rel_sent": "Moreover , we show how BMR is able to outperform previous formalisms thanks to its fully - semantic framing , which enables top - notch multilingual parsing and generation .", "forward": true, "src_ids": "2022.acl-long.121_2334"}
+{"input": "multi - modality problem is done by using Method| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "multi - modality problem", "output": "glat", "neg_sample": ["multi - modality problem is done by using Method", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "In this paper , we propose GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": false, "src_ids": "2022.acl-long.575_2335"}
+{"input": "parallel text generation is done by using Task| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "parallel text generation", "output": "glancing at latent variables", "neg_sample": ["parallel text generation is done by using Task", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "GLAT : Glancing at Latent Variables for Parallel Text Generation.", "forward": false, "src_ids": "2022.acl-long.575_2336"}
+{"input": "glancing at latent variables is used for Task| context: although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "glancing at latent variables", "output": "parallel text generation", "neg_sample": ["glancing at latent variables is used for Task", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "GLAT : Glancing at Latent Variables for Parallel Text Generation.", "forward": true, "src_ids": "2022.acl-long.575_2337"}
+{"input": "word categorical information is done by using OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "word categorical information", "output": "discrete latent variables", "neg_sample": ["word categorical information is done by using OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "In this paper , we propose GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": false, "src_ids": "2022.acl-long.575_2338"}
+{"input": "discrete latent variables is used for OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "discrete latent variables", "output": "word categorical information", "neg_sample": ["discrete latent variables is used for OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "In this paper , we propose GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": true, "src_ids": "2022.acl-long.575_2339"}
+{"input": "curriculum learning technique is used for OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "curriculum learning technique", "output": "multi - modality problem", "neg_sample": ["curriculum learning technique is used for OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "In this paper , we propose GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": true, "src_ids": "2022.acl-long.575_2340"}
+{"input": "glat is used for OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "glat", "output": "multi - modality problem", "neg_sample": ["glat is used for OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.575", "year": 2022, "rel_sent": "In this paper , we propose GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": true, "src_ids": "2022.acl-long.575_2341"}
+{"input": "rules - based quantifiers is used for OtherScientificTerm| context: a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents . transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies .", "entity": "rules - based quantifiers", "output": "seizure frequencies", "neg_sample": ["rules - based quantifiers is used for OtherScientificTerm", "a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents .", "transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies ."], "relation": "used for", "id": "2022.bionlp-1.36", "year": 2022, "rel_sent": "We used text summarization models to convert unstructured sentences into specific formats , and then employed rules - based quantifiers to calculate seizure frequencies and dates of last seizure .", "forward": true, "src_ids": "2022.bionlp-1.36_2342"}
+{"input": "unstructured sentences is done by using Method| context: a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents . for patients with epilepsy , such information includes outcome measures like seizure frequency and dates of last seizure , key parameters that guide all therapy for these patients . transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies .", "entity": "unstructured sentences", "output": "text summarization models", "neg_sample": ["unstructured sentences is done by using Method", "a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents .", "for patients with epilepsy , such information includes outcome measures like seizure frequency and dates of last seizure , key parameters that guide all therapy for these patients .", "transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies ."], "relation": "used for", "id": "2022.bionlp-1.36", "year": 2022, "rel_sent": "We used text summarization models to convert unstructured sentences into specific formats , and then employed rules - based quantifiers to calculate seizure frequencies and dates of last seizure .", "forward": false, "src_ids": "2022.bionlp-1.36_2343"}
+{"input": "text summarization models is used for Material| context: a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents . for patients with epilepsy , such information includes outcome measures like seizure frequency and dates of last seizure , key parameters that guide all therapy for these patients . transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies .", "entity": "text summarization models", "output": "unstructured sentences", "neg_sample": ["text summarization models is used for Material", "a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents .", "for patients with epilepsy , such information includes outcome measures like seizure frequency and dates of last seizure , key parameters that guide all therapy for these patients .", "transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies ."], "relation": "used for", "id": "2022.bionlp-1.36", "year": 2022, "rel_sent": "We used text summarization models to convert unstructured sentences into specific formats , and then employed rules - based quantifiers to calculate seizure frequencies and dates of last seizure .", "forward": true, "src_ids": "2022.bionlp-1.36_2344"}
+{"input": "seizure frequencies is done by using Method| context: a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents . for patients with epilepsy , such information includes outcome measures like seizure frequency and dates of last seizure , key parameters that guide all therapy for these patients . transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies .", "entity": "seizure frequencies", "output": "rules - based quantifiers", "neg_sample": ["seizure frequencies is done by using Method", "a wealth of important clinical information lies untouched in the electronic health record , often in the form of unstructured textual documents .", "for patients with epilepsy , such information includes outcome measures like seizure frequency and dates of last seizure , key parameters that guide all therapy for these patients .", "transformer models have been able to extract such outcome measures from unstructured clinical note text as sentences with human - like accuracy ; however , these sentences are not yet usable in a quantitative analysis for large - scale studies ."], "relation": "used for", "id": "2022.bionlp-1.36", "year": 2022, "rel_sent": "We used text summarization models to convert unstructured sentences into specific formats , and then employed rules - based quantifiers to calculate seizure frequencies and dates of last seizure .", "forward": false, "src_ids": "2022.bionlp-1.36_2345"}
+{"input": "feature extractor is done by using Method| context: fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise . existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias .", "entity": "feature extractor", "output": "coarse backbone model", "neg_sample": ["feature extractor is done by using Method", "fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise .", "existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias ."], "relation": "used for", "id": "2022.acl-long.141", "year": 2022, "rel_sent": "FCLC first train a coarse backbone model as a feature extractor and noise estimator .", "forward": false, "src_ids": "2022.acl-long.141_2346"}
+{"input": "noise estimator is done by using Method| context: fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise . existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias .", "entity": "noise estimator", "output": "coarse backbone model", "neg_sample": ["noise estimator is done by using Method", "fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise .", "existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias ."], "relation": "used for", "id": "2022.acl-long.141", "year": 2022, "rel_sent": "FCLC first train a coarse backbone model as a feature extractor and noise estimator .", "forward": false, "src_ids": "2022.acl-long.141_2347"}
+{"input": "coarse backbone model is used for Method| context: fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise . existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias .", "entity": "coarse backbone model", "output": "feature extractor", "neg_sample": ["coarse backbone model is used for Method", "fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise .", "existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias ."], "relation": "used for", "id": "2022.acl-long.141", "year": 2022, "rel_sent": "FCLC first train a coarse backbone model as a feature extractor and noise estimator .", "forward": true, "src_ids": "2022.acl-long.141_2348"}
+{"input": "feature cluster is done by using Method| context: fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise . existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias .", "entity": "feature cluster", "output": "loss correction", "neg_sample": ["feature cluster is done by using Method", "fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise .", "existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias ."], "relation": "used for", "id": "2022.acl-long.141", "year": 2022, "rel_sent": "Loss correction is then applied to each feature cluster , learning directly from the noisy labels .", "forward": false, "src_ids": "2022.acl-long.141_2349"}
+{"input": "loss correction is used for OtherScientificTerm| context: fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise . existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias .", "entity": "loss correction", "output": "feature cluster", "neg_sample": ["loss correction is used for OtherScientificTerm", "fine - grained entity typing ( fet ) has made great progress based on distant supervision but still suffers from label noise .", "existing fet noise learning methods rely on prediction distributions in an instance - independent manner , which causes the problem of confirmation bias ."], "relation": "used for", "id": "2022.acl-long.141", "year": 2022, "rel_sent": "Loss correction is then applied to each feature cluster , learning directly from the noisy labels .", "forward": true, "src_ids": "2022.acl-long.141_2350"}
+{"input": "commonsense pragmatics is done by using Method| context: more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "commonsense pragmatics", "output": "neural language models", "neg_sample": ["commonsense pragmatics is done by using Method", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "This paper analyzed neural language models ' understanding of commonsense pragmatics ( i.e. , implied meanings ) through human behavioral and neurophysiological data .", "forward": false, "src_ids": "2022.csrr-1.3_2351"}
+{"input": "neural language models is used for OtherScientificTerm| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "neural language models", "output": "commonsense pragmatics", "neg_sample": ["neural language models is used for OtherScientificTerm", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "This paper analyzed neural language models ' understanding of commonsense pragmatics ( i.e. , implied meanings ) through human behavioral and neurophysiological data .", "forward": true, "src_ids": "2022.csrr-1.3_2352"}
+{"input": "word - prediction models is done by using OtherScientificTerm| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "word - prediction models", "output": "psycholinguistic tests", "neg_sample": ["word - prediction models is done by using OtherScientificTerm", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "These psycholinguistic tests are designed to draw conclusions based on predictive responses in context , making them very well suited to test word - prediction models such as BERT in natural settings .", "forward": false, "src_ids": "2022.csrr-1.3_2353"}
+{"input": "psycholinguistic tests is used for Method| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "psycholinguistic tests", "output": "word - prediction models", "neg_sample": ["psycholinguistic tests is used for Method", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "These psycholinguistic tests are designed to draw conclusions based on predictive responses in context , making them very well suited to test word - prediction models such as BERT in natural settings .", "forward": true, "src_ids": "2022.csrr-1.3_2354"}
+{"input": "predictive responses is done by using OtherScientificTerm| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "predictive responses", "output": "linguistic mechanisms", "neg_sample": ["predictive responses is done by using OtherScientificTerm", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "They can provide the appropriate prompts and tasks to answer questions about linguistic mechanisms underlying predictive responses .", "forward": false, "src_ids": "2022.csrr-1.3_2355"}
+{"input": "linguistic mechanisms is used for OtherScientificTerm| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "linguistic mechanisms", "output": "predictive responses", "neg_sample": ["linguistic mechanisms is used for OtherScientificTerm", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "They can provide the appropriate prompts and tasks to answer questions about linguistic mechanisms underlying predictive responses .", "forward": true, "src_ids": "2022.csrr-1.3_2356"}
+{"input": "language models is done by using Material| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "language models", "output": "psycholinguistic datasets", "neg_sample": ["language models is done by using Material", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "This paper adopted psycholinguistic datasets to probe language models ' commonsense reasoning .", "forward": false, "src_ids": "2022.csrr-1.3_2357"}
+{"input": "commonsense reasoning is done by using Material| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "commonsense reasoning", "output": "psycholinguistic datasets", "neg_sample": ["commonsense reasoning is done by using Material", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "This paper adopted psycholinguistic datasets to probe language models ' commonsense reasoning .", "forward": false, "src_ids": "2022.csrr-1.3_2358"}
+{"input": "language models is used for Task| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "language models", "output": "commonsense reasoning", "neg_sample": ["language models is used for Task", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "This paper adopted psycholinguistic datasets to probe language models ' commonsense reasoning .", "forward": true, "src_ids": "2022.csrr-1.3_2359"}
+{"input": "psycholinguistic datasets is used for Task| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "psycholinguistic datasets", "output": "commonsense reasoning", "neg_sample": ["psycholinguistic datasets is used for Task", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "This paper adopted psycholinguistic datasets to probe language models ' commonsense reasoning .", "forward": true, "src_ids": "2022.csrr-1.3_2360"}
+{"input": "human - level is done by using Method| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "human - level", "output": "fine - tuning", "neg_sample": ["human - level is done by using Method", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "Whether or not fine - tuning improved its performance to human - level depends on the type of commonsense reasoning .", "forward": false, "src_ids": "2022.csrr-1.3_2361"}
+{"input": "fine - tuning is used for Metric| context: neural language models have attracted a lot of attention in the past few years . more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do .", "entity": "fine - tuning", "output": "human - level", "neg_sample": ["fine - tuning is used for Metric", "neural language models have attracted a lot of attention in the past few years .", "more and more researchers are getting intrigued by how language models encode commonsense , specifically what kind of commonsense they understand , and why they do ."], "relation": "used for", "id": "2022.csrr-1.3", "year": 2022, "rel_sent": "Whether or not fine - tuning improved its performance to human - level depends on the type of commonsense reasoning .", "forward": true, "src_ids": "2022.csrr-1.3_2362"}
+{"input": "full - length sequences is done by using Generic| context: transformer - based models generally allocate the same amount of computation for each token in a given sequence .", "entity": "full - length sequences", "output": "token dropping", "neg_sample": ["full - length sequences is done by using Generic", "transformer - based models generally allocate the same amount of computation for each token in a given sequence ."], "relation": "used for", "id": "2022.acl-long.262", "year": 2022, "rel_sent": "The dropped tokens are later picked up by the last layer of the model so that the model still produces full - length sequences .", "forward": false, "src_ids": "2022.acl-long.262_2363"}
+{"input": "token dropping is used for OtherScientificTerm| context: transformer - based models generally allocate the same amount of computation for each token in a given sequence .", "entity": "token dropping", "output": "full - length sequences", "neg_sample": ["token dropping is used for OtherScientificTerm", "transformer - based models generally allocate the same amount of computation for each token in a given sequence ."], "relation": "used for", "id": "2022.acl-long.262", "year": 2022, "rel_sent": "The dropped tokens are later picked up by the last layer of the model so that the model still produces full - length sequences .", "forward": true, "src_ids": "2022.acl-long.262_2364"}
+{"input": "pretraining of transformer models is done by using Method| context: transformer - based models generally allocate the same amount of computation for each token in a given sequence .", "entity": "pretraining of transformer models", "output": "token dropping ' method", "neg_sample": ["pretraining of transformer models is done by using Method", "transformer - based models generally allocate the same amount of computation for each token in a given sequence ."], "relation": "used for", "id": "2022.acl-long.262", "year": 2022, "rel_sent": "We develop a simple but effective ' token dropping ' method to accelerate the pretraining of transformer models , such as BERT , without degrading its performance on downstream tasks .", "forward": false, "src_ids": "2022.acl-long.262_2365"}
+{"input": "token dropping ' method is used for Task| context: transformer - based models generally allocate the same amount of computation for each token in a given sequence .", "entity": "token dropping ' method", "output": "pretraining of transformer models", "neg_sample": ["token dropping ' method is used for Task", "transformer - based models generally allocate the same amount of computation for each token in a given sequence ."], "relation": "used for", "id": "2022.acl-long.262", "year": 2022, "rel_sent": "We develop a simple but effective ' token dropping ' method to accelerate the pretraining of transformer models , such as BERT , without degrading its performance on downstream tasks .", "forward": true, "src_ids": "2022.acl-long.262_2366"}
+{"input": "human semantic knowledge is done by using OtherScientificTerm| context: distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning . the majority of these models are purely text based , even though the human sensory experience is much richer .", "entity": "human semantic knowledge", "output": "word embeddings", "neg_sample": ["human semantic knowledge is done by using OtherScientificTerm", "distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning .", "the majority of these models are purely text based , even though the human sensory experience is much richer ."], "relation": "used for", "id": "2022.cmcl-1.1", "year": 2022, "rel_sent": "Seeing the advantage : visually grounding word embeddings to better capture human semantic knowledge.", "forward": false, "src_ids": "2022.cmcl-1.1_2367"}
+{"input": "word embeddings is used for OtherScientificTerm| context: distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning . the majority of these models are purely text based , even though the human sensory experience is much richer .", "entity": "word embeddings", "output": "human semantic knowledge", "neg_sample": ["word embeddings is used for OtherScientificTerm", "distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning .", "the majority of these models are purely text based , even though the human sensory experience is much richer ."], "relation": "used for", "id": "2022.cmcl-1.1", "year": 2022, "rel_sent": "Seeing the advantage : visually grounding word embeddings to better capture human semantic knowledge.", "forward": true, "src_ids": "2022.cmcl-1.1_2368"}
+{"input": "visual information is used for OtherScientificTerm| context: the majority of these models are purely text based , even though the human sensory experience is much richer .", "entity": "visual information", "output": "cognitive aspects of word meaning", "neg_sample": ["visual information is used for OtherScientificTerm", "the majority of these models are purely text based , even though the human sensory experience is much richer ."], "relation": "used for", "id": "2022.cmcl-1.1", "year": 2022, "rel_sent": "In this paper we create visually grounded word embeddings by combining English text and images and compare them to popular text - based methods , to see if visual information allows our model to better capture cognitive aspects of word meaning .", "forward": true, "src_ids": "2022.cmcl-1.1_2369"}
+{"input": "cognitive aspects of word meaning is done by using OtherScientificTerm| context: distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning . the majority of these models are purely text based , even though the human sensory experience is much richer .", "entity": "cognitive aspects of word meaning", "output": "visual information", "neg_sample": ["cognitive aspects of word meaning is done by using OtherScientificTerm", "distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning .", "the majority of these models are purely text based , even though the human sensory experience is much richer ."], "relation": "used for", "id": "2022.cmcl-1.1", "year": 2022, "rel_sent": "In this paper we create visually grounded word embeddings by combining English text and images and compare them to popular text - based methods , to see if visual information allows our model to better capture cognitive aspects of word meaning .", "forward": false, "src_ids": "2022.cmcl-1.1_2370"}
+{"input": "human reaction times is done by using OtherScientificTerm| context: distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning . the majority of these models are purely text based , even though the human sensory experience is much richer .", "entity": "human reaction times", "output": "visually grounded embedding similarities", "neg_sample": ["human reaction times is done by using OtherScientificTerm", "distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning .", "the majority of these models are purely text based , even though the human sensory experience is much richer ."], "relation": "used for", "id": "2022.cmcl-1.1", "year": 2022, "rel_sent": "Our analysis shows that visually grounded embedding similarities are more predictive of the human reaction times in a large priming experiment than the purely text - based embeddings .", "forward": false, "src_ids": "2022.cmcl-1.1_2371"}
+{"input": "visually grounded embedding similarities is used for Metric| context: distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning . the majority of these models are purely text based , even though the human sensory experience is much richer .", "entity": "visually grounded embedding similarities", "output": "human reaction times", "neg_sample": ["visually grounded embedding similarities is used for Metric", "distributional semantic models capture word - level meaning that is useful in many natural language processing tasks and have even been shown to capture cognitive aspects of word meaning .", "the majority of these models are purely text based , even though the human sensory experience is much richer ."], "relation": "used for", "id": "2022.cmcl-1.1", "year": 2022, "rel_sent": "Our analysis shows that visually grounded embedding similarities are more predictive of the human reaction times in a large priming experiment than the purely text - based embeddings .", "forward": true, "src_ids": "2022.cmcl-1.1_2372"}
+{"input": "qr datasets is used for Material| context: resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) . the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form .", "entity": "qr datasets", "output": "cqa dataset", "neg_sample": ["qr datasets is used for Material", "resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) .", "the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form ."], "relation": "used for", "id": "2022.acl-srw.6", "year": 2022, "rel_sent": "However , QR datasets are limited and existing methods tend to depend on the assumption of the existence of corresponding QR datasets for every CQA dataset . This paper proposes a reinforcement learning approach that integrates QR and CQA tasks without corresponding labeled QR datasets .", "forward": true, "src_ids": "2022.acl-srw.6_2373"}
+{"input": "qr and cqa tasks is done by using Method| context: resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) . the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form .", "entity": "qr and cqa tasks", "output": "reinforcement learning approach", "neg_sample": ["qr and cqa tasks is done by using Method", "resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) .", "the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form ."], "relation": "used for", "id": "2022.acl-srw.6", "year": 2022, "rel_sent": "However , QR datasets are limited and existing methods tend to depend on the assumption of the existence of corresponding QR datasets for every CQA dataset . This paper proposes a reinforcement learning approach that integrates QR and CQA tasks without corresponding labeled QR datasets .", "forward": false, "src_ids": "2022.acl-srw.6_2374"}
+{"input": "reinforcement learning approach is used for Task| context: resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) . the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form .", "entity": "reinforcement learning approach", "output": "qr and cqa tasks", "neg_sample": ["reinforcement learning approach is used for Task", "resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) .", "the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form ."], "relation": "used for", "id": "2022.acl-srw.6", "year": 2022, "rel_sent": "However , QR datasets are limited and existing methods tend to depend on the assumption of the existence of corresponding QR datasets for every CQA dataset . This paper proposes a reinforcement learning approach that integrates QR and CQA tasks without corresponding labeled QR datasets .", "forward": true, "src_ids": "2022.acl-srw.6_2375"}
+{"input": "cqa dataset is done by using Material| context: resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) . the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form .", "entity": "cqa dataset", "output": "qr datasets", "neg_sample": ["cqa dataset is done by using Material", "resolving dependencies among dialogue history is one of the main obstacles in the research on conversational question answering ( qa ) .", "the conversational question rewrites ( qr ) task has been shown to be effective to solve this problem by reformulating questions in a self - contained form ."], "relation": "used for", "id": "2022.acl-srw.6", "year": 2022, "rel_sent": "However , QR datasets are limited and existing methods tend to depend on the assumption of the existence of corresponding QR datasets for every CQA dataset . This paper proposes a reinforcement learning approach that integrates QR and CQA tasks without corresponding labeled QR datasets .", "forward": false, "src_ids": "2022.acl-srw.6_2376"}
+{"input": "large language models is done by using OtherScientificTerm| context: bert ) with the aim of learning better language representations .", "entity": "large language models", "output": "pre - training objectives", "neg_sample": ["large language models is done by using OtherScientificTerm", "bert ) with the aim of learning better language representations ."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "How does the pre - training objective affect what large language models learn about linguistic properties ?.", "forward": false, "src_ids": "2022.acl-short.16_2377"}
+{"input": "pre - training objectives is used for Method| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g. bert ) with the aim of learning better language representations . however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties .", "entity": "pre - training objectives", "output": "large language models", "neg_sample": ["pre - training objectives is used for Method", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "bert ) with the aim of learning better language representations .", "however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties ."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "How does the pre - training objective affect what large language models learn about linguistic properties ?.", "forward": true, "src_ids": "2022.acl-short.16_2378"}
+{"input": "linguistic knowledge is done by using Method| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "entity": "linguistic knowledge", "output": "bert", "neg_sample": ["linguistic knowledge is done by using Method", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non - linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted .", "forward": false, "src_ids": "2022.acl-short.16_2379"}
+{"input": "linguistically motivated objectives is used for Method| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "entity": "linguistically motivated objectives", "output": "bert", "neg_sample": ["linguistically motivated objectives is used for Method", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non - linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted .", "forward": true, "src_ids": "2022.acl-short.16_2380"}
+{"input": "bert is done by using OtherScientificTerm| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g. bert ) with the aim of learning better language representations . however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties .", "entity": "bert", "output": "linguistically motivated objectives", "neg_sample": ["bert is done by using OtherScientificTerm", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "bert ) with the aim of learning better language representations .", "however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties ."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non - linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted .", "forward": false, "src_ids": "2022.acl-short.16_2381"}
+{"input": "linguistic knowledge is done by using OtherScientificTerm| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g. bert ) with the aim of learning better language representations . however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties .", "entity": "linguistic knowledge", "output": "linguistically motivated objectives", "neg_sample": ["linguistic knowledge is done by using OtherScientificTerm", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "bert ) with the aim of learning better language representations .", "however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties ."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non - linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted .", "forward": false, "src_ids": "2022.acl-short.16_2382"}
+{"input": "linguistically motivated objectives is used for OtherScientificTerm| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g. bert ) with the aim of learning better language representations . however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties .", "entity": "linguistically motivated objectives", "output": "linguistic knowledge", "neg_sample": ["linguistically motivated objectives is used for OtherScientificTerm", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "bert ) with the aim of learning better language representations .", "however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties ."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non - linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted .", "forward": true, "src_ids": "2022.acl-short.16_2383"}
+{"input": "bert is used for OtherScientificTerm| context: several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g. bert ) with the aim of learning better language representations . however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties .", "entity": "bert", "output": "linguistic knowledge", "neg_sample": ["bert is used for OtherScientificTerm", "several pre - training objectives , such as masked language modeling ( mlm ) , have been proposed to pre - train language models ( e.g.", "bert ) with the aim of learning better language representations .", "however , to the best of our knowledge , no previous work sofar has investigated how different pre - training objectives affect what bert learns about linguistics properties ."], "relation": "used for", "id": "2022.acl-short.16", "year": 2022, "rel_sent": "We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non - linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted .", "forward": true, "src_ids": "2022.acl-short.16_2384"}
+{"input": "multi - domain imbalanced data is done by using Task| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "multi - domain imbalanced data", "output": "domain - aware contrastive knowledge transfer", "neg_sample": ["multi - domain imbalanced data is done by using Task", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "Domain - Aware Contrastive Knowledge Transfer for Multi - domain Imbalanced Data.", "forward": false, "src_ids": "2022.wassa-1.3_2385"}
+{"input": "domain - aware contrastive knowledge transfer is used for Material| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "domain - aware contrastive knowledge transfer", "output": "multi - domain imbalanced data", "neg_sample": ["domain - aware contrastive knowledge transfer is used for Material", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "Domain - Aware Contrastive Knowledge Transfer for Multi - domain Imbalanced Data.", "forward": true, "src_ids": "2022.wassa-1.3_2386"}
+{"input": "shared domain knowledge is done by using Method| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "shared domain knowledge", "output": "domain - aware contrastive knowledge transfer method", "neg_sample": ["shared domain knowledge is done by using Method", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "In the MIL setting , different domains exhibit different patterns and there is a varying degree of similarity and divergence among domains posing opportunities and challenges for transfer learning especially when faced with limited or insufficient training data . We propose a novel domain - aware contrastive knowledge transfer method called DCMI to ( 1 ) identify the shared domain knowledge to encourage positive transfer among similar domains ( in particular from head domains to tail domains ) ; ( 2 ) isolate the domain - specific knowledge to minimize the negative transfer from dissimilar domains .", "forward": false, "src_ids": "2022.wassa-1.3_2387"}
+{"input": "positive transfer is done by using OtherScientificTerm| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "positive transfer", "output": "shared domain knowledge", "neg_sample": ["positive transfer is done by using OtherScientificTerm", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "In the MIL setting , different domains exhibit different patterns and there is a varying degree of similarity and divergence among domains posing opportunities and challenges for transfer learning especially when faced with limited or insufficient training data . We propose a novel domain - aware contrastive knowledge transfer method called DCMI to ( 1 ) identify the shared domain knowledge to encourage positive transfer among similar domains ( in particular from head domains to tail domains ) ; ( 2 ) isolate the domain - specific knowledge to minimize the negative transfer from dissimilar domains .", "forward": false, "src_ids": "2022.wassa-1.3_2388"}
+{"input": "domain - aware contrastive knowledge transfer method is used for OtherScientificTerm| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "domain - aware contrastive knowledge transfer method", "output": "shared domain knowledge", "neg_sample": ["domain - aware contrastive knowledge transfer method is used for OtherScientificTerm", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "In the MIL setting , different domains exhibit different patterns and there is a varying degree of similarity and divergence among domains posing opportunities and challenges for transfer learning especially when faced with limited or insufficient training data . We propose a novel domain - aware contrastive knowledge transfer method called DCMI to ( 1 ) identify the shared domain knowledge to encourage positive transfer among similar domains ( in particular from head domains to tail domains ) ; ( 2 ) isolate the domain - specific knowledge to minimize the negative transfer from dissimilar domains .", "forward": true, "src_ids": "2022.wassa-1.3_2389"}
+{"input": "shared domain knowledge is used for OtherScientificTerm| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "shared domain knowledge", "output": "positive transfer", "neg_sample": ["shared domain knowledge is used for OtherScientificTerm", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "In the MIL setting , different domains exhibit different patterns and there is a varying degree of similarity and divergence among domains posing opportunities and challenges for transfer learning especially when faced with limited or insufficient training data . We propose a novel domain - aware contrastive knowledge transfer method called DCMI to ( 1 ) identify the shared domain knowledge to encourage positive transfer among similar domains ( in particular from head domains to tail domains ) ; ( 2 ) isolate the domain - specific knowledge to minimize the negative transfer from dissimilar domains .", "forward": true, "src_ids": "2022.wassa-1.3_2390"}
+{"input": "negative transfer is done by using OtherScientificTerm| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "negative transfer", "output": "domain - specific knowledge", "neg_sample": ["negative transfer is done by using OtherScientificTerm", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "In the MIL setting , different domains exhibit different patterns and there is a varying degree of similarity and divergence among domains posing opportunities and challenges for transfer learning especially when faced with limited or insufficient training data . We propose a novel domain - aware contrastive knowledge transfer method called DCMI to ( 1 ) identify the shared domain knowledge to encourage positive transfer among similar domains ( in particular from head domains to tail domains ) ; ( 2 ) isolate the domain - specific knowledge to minimize the negative transfer from dissimilar domains .", "forward": false, "src_ids": "2022.wassa-1.3_2391"}
+{"input": "domain - specific knowledge is used for OtherScientificTerm| context: in many real - world machine learning applications , samples belong to a set of domains e.g. , for product reviews each review belongs to a product category . in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains .", "entity": "domain - specific knowledge", "output": "negative transfer", "neg_sample": ["domain - specific knowledge is used for OtherScientificTerm", "in many real - world machine learning applications , samples belong to a set of domains e.g.", ", for product reviews each review belongs to a product category .", "in this paper , we study multi - domain imbalanced learning ( mil ) , the scenario that there is imbalance not only in classes but also in domains ."], "relation": "used for", "id": "2022.wassa-1.3", "year": 2022, "rel_sent": "In the MIL setting , different domains exhibit different patterns and there is a varying degree of similarity and divergence among domains posing opportunities and challenges for transfer learning especially when faced with limited or insufficient training data . We propose a novel domain - aware contrastive knowledge transfer method called DCMI to ( 1 ) identify the shared domain knowledge to encourage positive transfer among similar domains ( in particular from head domains to tail domains ) ; ( 2 ) isolate the domain - specific knowledge to minimize the negative transfer from dissimilar domains .", "forward": true, "src_ids": "2022.wassa-1.3_2392"}
+{"input": "equality is done by using Task| context: in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms . the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms .", "the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire ."], "relation": "used for", "id": "2022.ltedi-1.30", "year": 2022, "rel_sent": "SSNCSE_NLP@LT - EDI - ACL2022 : Hope Speech Detection for Equality , Diversity and Inclusion using sentence transformers.", "forward": false, "src_ids": "2022.ltedi-1.30_2393"}
+{"input": "equality is done by using Task| context: in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms . the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms .", "the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire ."], "relation": "used for", "id": "2022.ltedi-1.30", "year": 2022, "rel_sent": "This paper depicts our work for the Shared Task on Hope Speech Detection for Equality , Diversity , and Inclusion at LTEDI 2021 .", "forward": false, "src_ids": "2022.ltedi-1.30_2394"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms . the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms .", "the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire ."], "relation": "used for", "id": "2022.ltedi-1.30", "year": 2022, "rel_sent": "SSNCSE_NLP@LT - EDI - ACL2022 : Hope Speech Detection for Equality , Diversity and Inclusion using sentence transformers.", "forward": true, "src_ids": "2022.ltedi-1.30_2395"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms . the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "in recent times , applications have been developed to regulate and control the spread of negativity and toxicity on online platforms .", "the world is filled with serious problems like political & religious conflicts , wars , pandemics , and offensive hate speech is the last thing we desire ."], "relation": "used for", "id": "2022.ltedi-1.30", "year": 2022, "rel_sent": "This paper depicts our work for the Shared Task on Hope Speech Detection for Equality , Diversity , and Inclusion at LTEDI 2021 .", "forward": true, "src_ids": "2022.ltedi-1.30_2396"}
+{"input": "bert serving is done by using Method| context: we present an efficient bert - based multi - task ( mt ) framework that is particularly suitable for iterative and incremental development of the tasks . the proposed framework is based on the idea of partial fine - tuning , i.e.", "entity": "bert serving", "output": "flexible multi - task model", "neg_sample": ["bert serving is done by using Method", "we present an efficient bert - based multi - task ( mt ) framework that is particularly suitable for iterative and incremental development of the tasks .", "the proposed framework is based on the idea of partial fine - tuning , i.e."], "relation": "used for", "id": "2022.acl-short.89", "year": 2022, "rel_sent": "A Flexible Multi - Task Model for BERT Serving.", "forward": false, "src_ids": "2022.acl-short.89_2397"}
+{"input": "flexible multi - task model is used for Task| context: we present an efficient bert - based multi - task ( mt ) framework that is particularly suitable for iterative and incremental development of the tasks . the proposed framework is based on the idea of partial fine - tuning , i.e.", "entity": "flexible multi - task model", "output": "bert serving", "neg_sample": ["flexible multi - task model is used for Task", "we present an efficient bert - based multi - task ( mt ) framework that is particularly suitable for iterative and incremental development of the tasks .", "the proposed framework is based on the idea of partial fine - tuning , i.e."], "relation": "used for", "id": "2022.acl-short.89", "year": 2022, "rel_sent": "A Flexible Multi - Task Model for BERT Serving.", "forward": true, "src_ids": "2022.acl-short.89_2398"}
+{"input": "argument patterninformation is done by using Material| context: cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language .", "entity": "argument patterninformation", "output": "lexi - cal database", "neg_sample": ["argument patterninformation is done by using Material", "cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language ."], "relation": "used for", "id": "2022.deelio-1.1", "year": 2022, "rel_sent": "In this work we introduce a lexi - cal database calledValency Patterns Leipzig(ValPal)which provides the argument patterninformation about various verb - forms in mul - tiple languages including low - resource langua - ges .", "forward": false, "src_ids": "2022.deelio-1.1_2399"}
+{"input": "lexi - cal database is used for OtherScientificTerm| context: cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language .", "entity": "lexi - cal database", "output": "argument patterninformation", "neg_sample": ["lexi - cal database is used for OtherScientificTerm", "cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language ."], "relation": "used for", "id": "2022.deelio-1.1", "year": 2022, "rel_sent": "In this work we introduce a lexi - cal database calledValency Patterns Leipzig(ValPal)which provides the argument patterninformation about various verb - forms in mul - tiple languages including low - resource langua - ges .", "forward": true, "src_ids": "2022.deelio-1.1_2400"}
+{"input": "lstm based model is done by using OtherScientificTerm| context: cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language .", "entity": "lstm based model", "output": "valpal database knowledge", "neg_sample": ["lstm based model is done by using OtherScientificTerm", "cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language ."], "relation": "used for", "id": "2022.deelio-1.1", "year": 2022, "rel_sent": "We also provide a framework to integratethe ValPal database knowledge into the state - of - the - art LSTM based model for cross - lingualsemantic role labelling .", "forward": false, "src_ids": "2022.deelio-1.1_2401"}
+{"input": "cross - lingualsemantic role labelling is done by using Method| context: cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language .", "entity": "cross - lingualsemantic role labelling", "output": "lstm based model", "neg_sample": ["cross - lingualsemantic role labelling is done by using Method", "cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language ."], "relation": "used for", "id": "2022.deelio-1.1", "year": 2022, "rel_sent": "We also provide a framework to integratethe ValPal database knowledge into the state - of - the - art LSTM based model for cross - lingualsemantic role labelling .", "forward": false, "src_ids": "2022.deelio-1.1_2402"}
+{"input": "valpal database knowledge is used for Method| context: cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language .", "entity": "valpal database knowledge", "output": "lstm based model", "neg_sample": ["valpal database knowledge is used for Method", "cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language ."], "relation": "used for", "id": "2022.deelio-1.1", "year": 2022, "rel_sent": "We also provide a framework to integratethe ValPal database knowledge into the state - of - the - art LSTM based model for cross - lingualsemantic role labelling .", "forward": true, "src_ids": "2022.deelio-1.1_2403"}
+{"input": "lstm based model is used for Task| context: cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language .", "entity": "lstm based model", "output": "cross - lingualsemantic role labelling", "neg_sample": ["lstm based model is used for Task", "cross - lingual transfer learning typically involves training a model on a high - resource sourcelanguage and applying it to a low - resource tar - get language ."], "relation": "used for", "id": "2022.deelio-1.1", "year": 2022, "rel_sent": "We also provide a framework to integratethe ValPal database knowledge into the state - of - the - art LSTM based model for cross - lingualsemantic role labelling .", "forward": true, "src_ids": "2022.deelio-1.1_2404"}
+{"input": "uncertainty is done by using Method| context: we explore the notion of uncertainty in the context of modern abstractive summarization models , using the tools of bayesian deep learning .", "entity": "uncertainty", "output": "bayesian inference", "neg_sample": ["uncertainty is done by using Method", "we explore the notion of uncertainty in the context of modern abstractive summarization models , using the tools of bayesian deep learning ."], "relation": "used for", "id": "2022.findings-acl.325", "year": 2022, "rel_sent": "Based on Bayesian inference we are able to effectively quantify uncertainty at prediction time .", "forward": false, "src_ids": "2022.findings-acl.325_2405"}
+{"input": "bayesian summary is done by using Method| context: we explore the notion of uncertainty in the context of modern abstractive summarization models , using the tools of bayesian deep learning .", "entity": "bayesian summary", "output": "bayesian inference", "neg_sample": ["bayesian summary is done by using Method", "we explore the notion of uncertainty in the context of modern abstractive summarization models , using the tools of bayesian deep learning ."], "relation": "used for", "id": "2022.findings-acl.325", "year": 2022, "rel_sent": "Finally , Bayesian inference enables us tofind a Bayesian summary which performs better than a deterministic one and is more robust to uncertainty .", "forward": false, "src_ids": "2022.findings-acl.325_2406"}
+{"input": "bayesian inference is used for OtherScientificTerm| context: we explore the notion of uncertainty in the context of modern abstractive summarization models , using the tools of bayesian deep learning .", "entity": "bayesian inference", "output": "bayesian summary", "neg_sample": ["bayesian inference is used for OtherScientificTerm", "we explore the notion of uncertainty in the context of modern abstractive summarization models , using the tools of bayesian deep learning ."], "relation": "used for", "id": "2022.findings-acl.325", "year": 2022, "rel_sent": "Finally , Bayesian inference enables us tofind a Bayesian summary which performs better than a deterministic one and is more robust to uncertainty .", "forward": true, "src_ids": "2022.findings-acl.325_2407"}
+{"input": "natural language inference is done by using Method| context: natural language inference ( nli ) datasets contain examples with highly ambiguous labels due to its subjectivity . several recent efforts have been made to acknowledge and embrace the existence of ambiguity , and explore how to capture the human disagreement distribution .", "entity": "natural language inference", "output": "calibrated networks", "neg_sample": ["natural language inference is done by using Method", "natural language inference ( nli ) datasets contain examples with highly ambiguous labels due to its subjectivity .", "several recent efforts have been made to acknowledge and embrace the existence of ambiguity , and explore how to capture the human disagreement distribution ."], "relation": "used for", "id": "2022.findings-acl.120", "year": 2022, "rel_sent": "Capture Human Disagreement Distributions by Calibrated Networks for Natural Language Inference.", "forward": false, "src_ids": "2022.findings-acl.120_2408"}
+{"input": "calibrated networks is used for Task| context: several recent efforts have been made to acknowledge and embrace the existence of ambiguity , and explore how to capture the human disagreement distribution .", "entity": "calibrated networks", "output": "natural language inference", "neg_sample": ["calibrated networks is used for Task", "several recent efforts have been made to acknowledge and embrace the existence of ambiguity , and explore how to capture the human disagreement distribution ."], "relation": "used for", "id": "2022.findings-acl.120", "year": 2022, "rel_sent": "Capture Human Disagreement Distributions by Calibrated Networks for Natural Language Inference.", "forward": true, "src_ids": "2022.findings-acl.120_2409"}
+{"input": "compositional word embeddings is done by using Method| context: it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "compositional word embeddings", "output": "neural model", "neg_sample": ["compositional word embeddings is done by using Method", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "A Neural Model for Compositional Word Embeddings and Sentence Processing.", "forward": false, "src_ids": "2022.cmcl-1.2_2410"}
+{"input": "sentence processing is done by using Method| context: it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "sentence processing", "output": "neural model", "neg_sample": ["sentence processing is done by using Method", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "A Neural Model for Compositional Word Embeddings and Sentence Processing.", "forward": false, "src_ids": "2022.cmcl-1.2_2411"}
+{"input": "deep learning systems is done by using Method| context: it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "deep learning systems", "output": "neural model", "neg_sample": ["deep learning systems is done by using Method", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "Our proposed model goes some way towards offering a class of computationally powerful deep learning systems that can be fully understood and compared to human cognitive processes for natural language learning and representation .", "forward": false, "src_ids": "2022.cmcl-1.2_2412"}
+{"input": "neural model is used for Task| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information . it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "neural model", "output": "compositional word embeddings", "neg_sample": ["neural model is used for Task", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "A Neural Model for Compositional Word Embeddings and Sentence Processing.", "forward": true, "src_ids": "2022.cmcl-1.2_2413"}
+{"input": "neural model is used for Task| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "entity": "neural model", "output": "sentence processing", "neg_sample": ["neural model is used for Task", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "A Neural Model for Compositional Word Embeddings and Sentence Processing.", "forward": true, "src_ids": "2022.cmcl-1.2_2414"}
+{"input": "natural language learning and representation is done by using Method| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information . it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "natural language learning and representation", "output": "deep learning systems", "neg_sample": ["natural language learning and representation is done by using Method", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "Our proposed model goes some way towards offering a class of computationally powerful deep learning systems that can be fully understood and compared to human cognitive processes for natural language learning and representation .", "forward": false, "src_ids": "2022.cmcl-1.2_2415"}
+{"input": "neural model is used for Method| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information . it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "neural model", "output": "deep learning systems", "neg_sample": ["neural model is used for Method", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "Our proposed model goes some way towards offering a class of computationally powerful deep learning systems that can be fully understood and compared to human cognitive processes for natural language learning and representation .", "forward": true, "src_ids": "2022.cmcl-1.2_2416"}
+{"input": "natural language learning and representation is done by using OtherScientificTerm| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information . it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "natural language learning and representation", "output": "human cognitive processes", "neg_sample": ["natural language learning and representation is done by using OtherScientificTerm", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "Our proposed model goes some way towards offering a class of computationally powerful deep learning systems that can be fully understood and compared to human cognitive processes for natural language learning and representation .", "forward": false, "src_ids": "2022.cmcl-1.2_2417"}
+{"input": "human cognitive processes is used for Task| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information . it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "human cognitive processes", "output": "natural language learning and representation", "neg_sample": ["human cognitive processes is used for Task", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "Our proposed model goes some way towards offering a class of computationally powerful deep learning systems that can be fully understood and compared to human cognitive processes for natural language learning and representation .", "forward": true, "src_ids": "2022.cmcl-1.2_2418"}
+{"input": "deep learning systems is used for Task| context: we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information . it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context .", "entity": "deep learning systems", "output": "natural language learning and representation", "neg_sample": ["deep learning systems is used for Task", "we propose a new neural model for word embeddings , which uses unitary matrices as the primary device for encoding lexical information .", "it uses simple matrix multiplication to derive matrices for large units , yielding a sentence processing model that is strictly compositional , does not lose information over time steps , and is transparent , in the sense that word embeddings can be analysed regardless of context ."], "relation": "used for", "id": "2022.cmcl-1.2", "year": 2022, "rel_sent": "Our proposed model goes some way towards offering a class of computationally powerful deep learning systems that can be fully understood and compared to human cognitive processes for natural language learning and representation .", "forward": true, "src_ids": "2022.cmcl-1.2_2419"}
+{"input": "dense retrieval is done by using Method| context: dense retrieval models , which aim at retrieving the most relevant document for an input query on a dense representation space , have gained considerable attention for their remarkable success . yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans .", "entity": "dense retrieval", "output": "representations of documents", "neg_sample": ["dense retrieval is done by using Method", "dense retrieval models , which aim at retrieving the most relevant document for an input query on a dense representation space , have gained considerable attention for their remarkable success .", "yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans ."], "relation": "used for", "id": "2022.acl-short.48", "year": 2022, "rel_sent": "Augmenting Document Representations for Dense Retrieval with Interpolation and Perturbation.", "forward": false, "src_ids": "2022.acl-short.48_2420"}
+{"input": "representations of documents is used for Task| context: yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans .", "entity": "representations of documents", "output": "dense retrieval", "neg_sample": ["representations of documents is used for Task", "yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans ."], "relation": "used for", "id": "2022.acl-short.48", "year": 2022, "rel_sent": "Augmenting Document Representations for Dense Retrieval with Interpolation and Perturbation.", "forward": true, "src_ids": "2022.acl-short.48_2421"}
+{"input": "document augmentation for dense retrieval is used for Task| context: yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans .", "entity": "document augmentation for dense retrieval", "output": "dense retrieval", "neg_sample": ["document augmentation for dense retrieval is used for Task", "yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans ."], "relation": "used for", "id": "2022.acl-short.48", "year": 2022, "rel_sent": "We validate the performance of DAR on retrieval tasks with two benchmark datasets , showing that the proposed DAR significantly outperforms relevant baselines on the dense retrieval of both the labeled and unlabeled documents .", "forward": true, "src_ids": "2022.acl-short.48_2422"}
+{"input": "relevant baselines is used for Task| context: yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans .", "entity": "relevant baselines", "output": "dense retrieval", "neg_sample": ["relevant baselines is used for Task", "yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans ."], "relation": "used for", "id": "2022.acl-short.48", "year": 2022, "rel_sent": "We validate the performance of DAR on retrieval tasks with two benchmark datasets , showing that the proposed DAR significantly outperforms relevant baselines on the dense retrieval of both the labeled and unlabeled documents .", "forward": true, "src_ids": "2022.acl-short.48_2423"}
+{"input": "dense retrieval is done by using Generic| context: dense retrieval models , which aim at retrieving the most relevant document for an input query on a dense representation space , have gained considerable attention for their remarkable success . yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans .", "entity": "dense retrieval", "output": "relevant baselines", "neg_sample": ["dense retrieval is done by using Generic", "dense retrieval models , which aim at retrieving the most relevant document for an input query on a dense representation space , have gained considerable attention for their remarkable success .", "yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans ."], "relation": "used for", "id": "2022.acl-short.48", "year": 2022, "rel_sent": "We validate the performance of DAR on retrieval tasks with two benchmark datasets , showing that the proposed DAR significantly outperforms relevant baselines on the dense retrieval of both the labeled and unlabeled documents .", "forward": false, "src_ids": "2022.acl-short.48_2424"}
+{"input": "dense retrieval is done by using Method| context: dense retrieval models , which aim at retrieving the most relevant document for an input query on a dense representation space , have gained considerable attention for their remarkable success . yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans .", "entity": "dense retrieval", "output": "document augmentation for dense retrieval", "neg_sample": ["dense retrieval is done by using Method", "dense retrieval models , which aim at retrieving the most relevant document for an input query on a dense representation space , have gained considerable attention for their remarkable success .", "yet , dense models require a vast amount of labeled training data for notable performance , whereas it is often challenging to acquire query - document pairs annotated by humans ."], "relation": "used for", "id": "2022.acl-short.48", "year": 2022, "rel_sent": "We validate the performance of DAR on retrieval tasks with two benchmark datasets , showing that the proposed DAR significantly outperforms relevant baselines on the dense retrieval of both the labeled and unlabeled documents .", "forward": false, "src_ids": "2022.acl-short.48_2425"}
+{"input": "low - resource sentence retrieval is done by using Method| context: aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation .", "entity": "low - resource sentence retrieval", "output": "onealigner", "neg_sample": ["low - resource sentence retrieval is done by using Method", "aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation ."], "relation": "used for", "id": "2022.findings-acl.226", "year": 2022, "rel_sent": "OneAligner : Zero - shot Cross - lingual Transfer with One Rich - Resource Language Pair for Low - Resource Sentence Retrieval.", "forward": false, "src_ids": "2022.findings-acl.226_2426"}
+{"input": "low - resource sentence retrieval is done by using Task| context: aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation .", "entity": "low - resource sentence retrieval", "output": "zero - shot cross - lingual transfer", "neg_sample": ["low - resource sentence retrieval is done by using Task", "aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation ."], "relation": "used for", "id": "2022.findings-acl.226", "year": 2022, "rel_sent": "OneAligner : Zero - shot Cross - lingual Transfer with One Rich - Resource Language Pair for Low - Resource Sentence Retrieval.", "forward": false, "src_ids": "2022.findings-acl.226_2427"}
+{"input": "zero - shot cross - lingual transfer is used for Task| context: aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation .", "entity": "zero - shot cross - lingual transfer", "output": "low - resource sentence retrieval", "neg_sample": ["zero - shot cross - lingual transfer is used for Task", "aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation ."], "relation": "used for", "id": "2022.findings-acl.226", "year": 2022, "rel_sent": "OneAligner : Zero - shot Cross - lingual Transfer with One Rich - Resource Language Pair for Low - Resource Sentence Retrieval.", "forward": true, "src_ids": "2022.findings-acl.226_2428"}
+{"input": "onealigner is used for Task| context: aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation .", "entity": "onealigner", "output": "low - resource sentence retrieval", "neg_sample": ["onealigner is used for Task", "aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation ."], "relation": "used for", "id": "2022.findings-acl.226", "year": 2022, "rel_sent": "OneAligner : Zero - shot Cross - lingual Transfer with One Rich - Resource Language Pair for Low - Resource Sentence Retrieval.", "forward": true, "src_ids": "2022.findings-acl.226_2429"}
+{"input": "sentence retrieval tasks is done by using Method| context: aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation .", "entity": "sentence retrieval tasks", "output": "alignment model", "neg_sample": ["sentence retrieval tasks is done by using Method", "aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation ."], "relation": "used for", "id": "2022.findings-acl.226", "year": 2022, "rel_sent": "In this work , we present OneAligner , an alignment model specially designed for sentence retrieval tasks .", "forward": false, "src_ids": "2022.findings-acl.226_2430"}
+{"input": "alignment model is used for Task| context: aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation .", "entity": "alignment model", "output": "sentence retrieval tasks", "neg_sample": ["alignment model is used for Task", "aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as machine translation ."], "relation": "used for", "id": "2022.findings-acl.226", "year": 2022, "rel_sent": "In this work , we present OneAligner , an alignment model specially designed for sentence retrieval tasks .", "forward": true, "src_ids": "2022.findings-acl.226_2431"}
+{"input": "hope speech content is done by using Method| context: with the increase of users on social media platforms , manipulating or provoking masses of people has become a piece of cake . this spread of hatred among people , which has become a loophole for freedom of speech , must be minimized . hence , it is essential to have a system that automatically classifies the hatred content , especially on social media , to take it down .", "entity": "hope speech content", "output": "modular pipeline classifier", "neg_sample": ["hope speech content is done by using Method", "with the increase of users on social media platforms , manipulating or provoking masses of people has become a piece of cake .", "this spread of hatred among people , which has become a loophole for freedom of speech , must be minimized .", "hence , it is essential to have a system that automatically classifies the hatred content , especially on social media , to take it down ."], "relation": "used for", "id": "2022.ltedi-1.49", "year": 2022, "rel_sent": "This paper presents a simple modular pipeline classifier with BERT embeddings and attention mechanism to classify hope speech content in the Hope Speech Detection shared task for Equality , Diversity , and Inclusion - ACL 2022 .", "forward": false, "src_ids": "2022.ltedi-1.49_2432"}
+{"input": "modular pipeline classifier is used for Material| context: with the increase of users on social media platforms , manipulating or provoking masses of people has become a piece of cake . this spread of hatred among people , which has become a loophole for freedom of speech , must be minimized . hence , it is essential to have a system that automatically classifies the hatred content , especially on social media , to take it down .", "entity": "modular pipeline classifier", "output": "hope speech content", "neg_sample": ["modular pipeline classifier is used for Material", "with the increase of users on social media platforms , manipulating or provoking masses of people has become a piece of cake .", "this spread of hatred among people , which has become a loophole for freedom of speech , must be minimized .", "hence , it is essential to have a system that automatically classifies the hatred content , especially on social media , to take it down ."], "relation": "used for", "id": "2022.ltedi-1.49", "year": 2022, "rel_sent": "This paper presents a simple modular pipeline classifier with BERT embeddings and attention mechanism to classify hope speech content in the Hope Speech Detection shared task for Equality , Diversity , and Inclusion - ACL 2022 .", "forward": true, "src_ids": "2022.ltedi-1.49_2433"}
+{"input": "attention mechanism is used for Material| context: with the increase of users on social media platforms , manipulating or provoking masses of people has become a piece of cake . this spread of hatred among people , which has become a loophole for freedom of speech , must be minimized . hence , it is essential to have a system that automatically classifies the hatred content , especially on social media , to take it down .", "entity": "attention mechanism", "output": "hope speech content", "neg_sample": ["attention mechanism is used for Material", "with the increase of users on social media platforms , manipulating or provoking masses of people has become a piece of cake .", "this spread of hatred among people , which has become a loophole for freedom of speech , must be minimized .", "hence , it is essential to have a system that automatically classifies the hatred content , especially on social media , to take it down ."], "relation": "used for", "id": "2022.ltedi-1.49", "year": 2022, "rel_sent": "This paper presents a simple modular pipeline classifier with BERT embeddings and attention mechanism to classify hope speech content in the Hope Speech Detection shared task for Equality , Diversity , and Inclusion - ACL 2022 .", "forward": true, "src_ids": "2022.ltedi-1.49_2434"}
+{"input": "emotions is done by using Method| context: emotion analysis ( ea ) is the process of automatically analyzing and categorizing the input text into one of the predefined sets of emotions . in recent years , people have turned to social media to express their emotions , opinions or feelings about news , movies , products , services , and so on . these users ' emotions may help the public , governments , business organizations , film producers , and others in devising strategies , making decisions , and so on . the increasing number of social media users and the increasing amount of user generated text containing emotions on social media demands automated tools for the analysis of such data as handling this data manually is labor intensive and error prone . further , the characteristics of social media data makes the ea challenging . most of the ea research works have focused on english language leaving several indian languages including tamil unexplored for this task .", "entity": "emotions", "output": "ensemble of logistic regression penalties", "neg_sample": ["emotions is done by using Method", "emotion analysis ( ea ) is the process of automatically analyzing and categorizing the input text into one of the predefined sets of emotions .", "in recent years , people have turned to social media to express their emotions , opinions or feelings about news , movies , products , services , and so on .", "these users ' emotions may help the public , governments , business organizations , film producers , and others in devising strategies , making decisions , and so on .", "the increasing number of social media users and the increasing amount of user generated text containing emotions on social media demands automated tools for the analysis of such data as handling this data manually is labor intensive and error prone .", "further , the characteristics of social media data makes the ea challenging .", "most of the ea research works have focused on english language leaving several indian languages including tamil unexplored for this task ."], "relation": "used for", "id": "2022.dravidianlangtech-1.23", "year": 2022, "rel_sent": "MUCS@DravidianLangTech@ACL2022 : Ensemble of Logistic Regression Penalties to Identify Emotions in Tamil Text.", "forward": false, "src_ids": "2022.dravidianlangtech-1.23_2435"}
+{"input": "ensemble of logistic regression penalties is used for OtherScientificTerm| context: further , the characteristics of social media data makes the ea challenging . most of the ea research works have focused on english language leaving several indian languages including tamil unexplored for this task .", "entity": "ensemble of logistic regression penalties", "output": "emotions", "neg_sample": ["ensemble of logistic regression penalties is used for OtherScientificTerm", "further , the characteristics of social media data makes the ea challenging .", "most of the ea research works have focused on english language leaving several indian languages including tamil unexplored for this task ."], "relation": "used for", "id": "2022.dravidianlangtech-1.23", "year": 2022, "rel_sent": "MUCS@DravidianLangTech@ACL2022 : Ensemble of Logistic Regression Penalties to Identify Emotions in Tamil Text.", "forward": true, "src_ids": "2022.dravidianlangtech-1.23_2436"}
+{"input": "intelligent agents is done by using Task| context: vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities .", "entity": "intelligent agents", "output": "ai research", "neg_sample": ["intelligent agents is done by using Task", "vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities ."], "relation": "used for", "id": "2022.acl-long.524", "year": 2022, "rel_sent": "A long - term goal of AI research is to build intelligent agents that can communicate with humans in natural language , perceive the environment , and perform real - world tasks .", "forward": false, "src_ids": "2022.acl-long.524_2437"}
+{"input": "real - world tasks is done by using Method| context: vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities .", "entity": "real - world tasks", "output": "intelligent agents", "neg_sample": ["real - world tasks is done by using Method", "vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities ."], "relation": "used for", "id": "2022.acl-long.524", "year": 2022, "rel_sent": "A long - term goal of AI research is to build intelligent agents that can communicate with humans in natural language , perceive the environment , and perform real - world tasks .", "forward": false, "src_ids": "2022.acl-long.524_2438"}
+{"input": "ai research is used for Method| context: vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities .", "entity": "ai research", "output": "intelligent agents", "neg_sample": ["ai research is used for Method", "vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities ."], "relation": "used for", "id": "2022.acl-long.524", "year": 2022, "rel_sent": "A long - term goal of AI research is to build intelligent agents that can communicate with humans in natural language , perceive the environment , and perform real - world tasks .", "forward": true, "src_ids": "2022.acl-long.524_2439"}
+{"input": "intelligent agents is used for Task| context: vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities .", "entity": "intelligent agents", "output": "real - world tasks", "neg_sample": ["intelligent agents is used for Task", "vision - and - language navigation ( vln ) is a fundamental and interdisciplinary research topic towards this goal , and receives increasing attention from natural language processing , computer vision , robotics , and machine learning communities ."], "relation": "used for", "id": "2022.acl-long.524", "year": 2022, "rel_sent": "A long - term goal of AI research is to build intelligent agents that can communicate with humans in natural language , perceive the environment , and perform real - world tasks .", "forward": true, "src_ids": "2022.acl-long.524_2440"}
+{"input": "specialized module is used for Method| context: to successfully account for language , computational models need to take into account both the linguistic context ( the content of the utterances ) and the extra - linguistic context ( for instance , the participants in a dialogue ) .", "entity": "specialized module", "output": "character representation", "neg_sample": ["specialized module is used for Method", "to successfully account for language , computational models need to take into account both the linguistic context ( the content of the utterances ) and the extra - linguistic context ( for instance , the participants in a dialogue ) ."], "relation": "used for", "id": "2022.insights-1.18", "year": 2022, "rel_sent": "In particular , our architecture combines a previously proposed specialized module ( an ' entity library ' ) for character representation with transfer learning from a pre - trained language model .", "forward": true, "src_ids": "2022.insights-1.18_2441"}
+{"input": "character representation is done by using Method| context: to successfully account for language , computational models need to take into account both the linguistic context ( the content of the utterances ) and the extra - linguistic context ( for instance , the participants in a dialogue ) .", "entity": "character representation", "output": "specialized module", "neg_sample": ["character representation is done by using Method", "to successfully account for language , computational models need to take into account both the linguistic context ( the content of the utterances ) and the extra - linguistic context ( for instance , the participants in a dialogue ) ."], "relation": "used for", "id": "2022.insights-1.18", "year": 2022, "rel_sent": "In particular , our architecture combines a previously proposed specialized module ( an ' entity library ' ) for character representation with transfer learning from a pre - trained language model .", "forward": false, "src_ids": "2022.insights-1.18_2442"}
+{"input": "automatically measuring factual consistency is done by using Method| context: despite recent success , large neural models often generate factually incorrect text . compounding this is the lack of a standard automatic evaluation for factuality - it can not be meaningfully improved if it can not be measured . grounded generation promises a path to solving both of these problems : models draw on a reliable external document ( grounding ) for factual information , simplifying the challenge of factuality . measuring factuality is also simplified - tofactual consistency , testing whether the generation agrees with the grounding , rather than all facts . yet , without a standard automatic metric for factual consistency , factually grounded generation remains an open problem .", "entity": "automatically measuring factual consistency", "output": "factual ablation", "neg_sample": ["automatically measuring factual consistency is done by using Method", "despite recent success , large neural models often generate factually incorrect text .", "compounding this is the lack of a standard automatic evaluation for factuality - it can not be meaningfully improved if it can not be measured .", "grounded generation promises a path to solving both of these problems : models draw on a reliable external document ( grounding ) for factual information , simplifying the challenge of factuality .", "measuring factuality is also simplified - tofactual consistency , testing whether the generation agrees with the grounding , rather than all facts .", "yet , without a standard automatic metric for factual consistency , factually grounded generation remains an open problem ."], "relation": "used for", "id": "2022.findings-acl.294", "year": 2022, "rel_sent": "Particularly , this domain allows us to introduce the notion of factual ablation for automatically measuring factual consistency : this captures the intuition that the model should be less likely to produce an output given a less relevant grounding document .", "forward": false, "src_ids": "2022.findings-acl.294_2443"}
+{"input": "factual ablation is used for Task| context: despite recent success , large neural models often generate factually incorrect text . compounding this is the lack of a standard automatic evaluation for factuality - it can not be meaningfully improved if it can not be measured . grounded generation promises a path to solving both of these problems : models draw on a reliable external document ( grounding ) for factual information , simplifying the challenge of factuality . measuring factuality is also simplified - tofactual consistency , testing whether the generation agrees with the grounding , rather than all facts . yet , without a standard automatic metric for factual consistency , factually grounded generation remains an open problem .", "entity": "factual ablation", "output": "automatically measuring factual consistency", "neg_sample": ["factual ablation is used for Task", "despite recent success , large neural models often generate factually incorrect text .", "compounding this is the lack of a standard automatic evaluation for factuality - it can not be meaningfully improved if it can not be measured .", "grounded generation promises a path to solving both of these problems : models draw on a reliable external document ( grounding ) for factual information , simplifying the challenge of factuality .", "measuring factuality is also simplified - tofactual consistency , testing whether the generation agrees with the grounding , rather than all facts .", "yet , without a standard automatic metric for factual consistency , factually grounded generation remains an open problem ."], "relation": "used for", "id": "2022.findings-acl.294", "year": 2022, "rel_sent": "Particularly , this domain allows us to introduce the notion of factual ablation for automatically measuring factual consistency : this captures the intuition that the model should be less likely to produce an output given a less relevant grounding document .", "forward": true, "src_ids": "2022.findings-acl.294_2444"}
+{"input": "integrated argument mining tasks is done by using Material| context: traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc . as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system .", "entity": "integrated argument mining tasks", "output": "iam", "neg_sample": ["integrated argument mining tasks is done by using Material", "traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc .", "as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system ."], "relation": "used for", "id": "2022.acl-long.162", "year": 2022, "rel_sent": "IAM : A Comprehensive and Large - Scale Dataset for Integrated Argument Mining Tasks.", "forward": false, "src_ids": "2022.acl-long.162_2445"}
+{"input": "iam is used for Task| context: traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc . as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system .", "entity": "iam", "output": "integrated argument mining tasks", "neg_sample": ["iam is used for Task", "traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc .", "as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system ."], "relation": "used for", "id": "2022.acl-long.162", "year": 2022, "rel_sent": "IAM : A Comprehensive and Large - Scale Dataset for Integrated Argument Mining Tasks.", "forward": true, "src_ids": "2022.acl-long.162_2446"}
+{"input": "integrated task is done by using Method| context: traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc . as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system .", "entity": "integrated task", "output": "pipeline approach", "neg_sample": ["integrated task is done by using Method", "traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc .", "as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system ."], "relation": "used for", "id": "2022.acl-long.162", "year": 2022, "rel_sent": "We adopt a pipeline approach and an end - to - end method for each integrated task separately .", "forward": false, "src_ids": "2022.acl-long.162_2447"}
+{"input": "end - to - end method is used for Generic| context: traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc . as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system .", "entity": "end - to - end method", "output": "integrated task", "neg_sample": ["end - to - end method is used for Generic", "traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc .", "as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system ."], "relation": "used for", "id": "2022.acl-long.162", "year": 2022, "rel_sent": "We adopt a pipeline approach and an end - to - end method for each integrated task separately .", "forward": true, "src_ids": "2022.acl-long.162_2448"}
+{"input": "pipeline approach is used for Generic| context: traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc . as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system .", "entity": "pipeline approach", "output": "integrated task", "neg_sample": ["pipeline approach is used for Generic", "traditionally , a debate usually requires a manual preparation process , including reading plenty of articles , selecting the claims , identifying the stances of the claims , seeking the evidence for the claims , etc .", "as the ai debate attracts more attention these years , it is worth exploring the methods to automate the tedious process involved in the debating system ."], "relation": "used for", "id": "2022.acl-long.162", "year": 2022, "rel_sent": "We adopt a pipeline approach and an end - to - end method for each integrated task separately .", "forward": true, "src_ids": "2022.acl-long.162_2449"}
+{"input": "linguistic search is done by using Task| context: we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax .", "entity": "linguistic search", "output": "parsing early modern english", "neg_sample": ["linguistic search is done by using Task", "we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax ."], "relation": "used for", "id": "2022.scil-1.12", "year": 2022, "rel_sent": "Parsing Early Modern English for Linguistic Search.", "forward": false, "src_ids": "2022.scil-1.12_2450"}
+{"input": "parsing early modern english is used for Task| context: we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax .", "entity": "parsing early modern english", "output": "linguistic search", "neg_sample": ["parsing early modern english is used for Task", "we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax ."], "relation": "used for", "id": "2022.scil-1.12", "year": 2022, "rel_sent": "Parsing Early Modern English for Linguistic Search.", "forward": true, "src_ids": "2022.scil-1.12_2451"}
+{"input": "linguistic queries is done by using Task| context: we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax .", "entity": "linguistic queries", "output": "parsing", "neg_sample": ["linguistic queries is done by using Task", "we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax ."], "relation": "used for", "id": "2022.scil-1.12", "year": 2022, "rel_sent": "This brings together many of the usual tools in NLP - word embeddings , tagging , and parsing - in the service of linguistic queries over automatically annotated corpora .", "forward": false, "src_ids": "2022.scil-1.12_2452"}
+{"input": "parsing is used for Task| context: we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax .", "entity": "parsing", "output": "linguistic queries", "neg_sample": ["parsing is used for Task", "we investigate the question of whether advances in nlp over the last few years make it possible to vastly increase the size of data usable for research in historical syntax ."], "relation": "used for", "id": "2022.scil-1.12", "year": 2022, "rel_sent": "This brings together many of the usual tools in NLP - word embeddings , tagging , and parsing - in the service of linguistic queries over automatically annotated corpora .", "forward": true, "src_ids": "2022.scil-1.12_2453"}
+{"input": "transfer learning setup is used for Task| context: cross - lingual transfer learning with large multilingual pre - trained models can be an effective approach for low - resource languages with no labeled training data . existing evaluations of zero - shot cross - lingual generalisability of large pre - trained models use datasets with english training data , and test data in a selection of target languages .", "entity": "transfer learning setup", "output": "part - of - speech tagging", "neg_sample": ["transfer learning setup is used for Task", "cross - lingual transfer learning with large multilingual pre - trained models can be an effective approach for low - resource languages with no labeled training data .", "existing evaluations of zero - shot cross - lingual generalisability of large pre - trained models use datasets with english training data , and test data in a selection of target languages ."], "relation": "used for", "id": "2022.acl-long.529", "year": 2022, "rel_sent": "We explore a more extensive transfer learning setup with 65 different source languages and 105 target languages for part - of - speech tagging .", "forward": true, "src_ids": "2022.acl-long.529_2454"}
+{"input": "part - of - speech tagging is done by using Method| context: cross - lingual transfer learning with large multilingual pre - trained models can be an effective approach for low - resource languages with no labeled training data . existing evaluations of zero - shot cross - lingual generalisability of large pre - trained models use datasets with english training data , and test data in a selection of target languages .", "entity": "part - of - speech tagging", "output": "transfer learning setup", "neg_sample": ["part - of - speech tagging is done by using Method", "cross - lingual transfer learning with large multilingual pre - trained models can be an effective approach for low - resource languages with no labeled training data .", "existing evaluations of zero - shot cross - lingual generalisability of large pre - trained models use datasets with english training data , and test data in a selection of target languages ."], "relation": "used for", "id": "2022.acl-long.529", "year": 2022, "rel_sent": "We explore a more extensive transfer learning setup with 65 different source languages and 105 target languages for part - of - speech tagging .", "forward": false, "src_ids": "2022.acl-long.529_2455"}
+{"input": "radiology image report generation is done by using Method| context: automatic generating the clinically accurate radiology report from x - ray images is important but challenging . the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models .", "entity": "radiology image report generation", "output": "memory - aligned knowledge graph", "neg_sample": ["radiology image report generation is done by using Method", "automatic generating the clinically accurate radiology report from x - ray images is important but challenging .", "the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models ."], "relation": "used for", "id": "2022.bionlp-1.11", "year": 2022, "rel_sent": "Memory - aligned Knowledge Graph for Clinically Accurate Radiology Image Report Generation.", "forward": false, "src_ids": "2022.bionlp-1.11_2456"}
+{"input": "memory - aligned knowledge graph is used for Task| context: automatic generating the clinically accurate radiology report from x - ray images is important but challenging . the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models .", "entity": "memory - aligned knowledge graph", "output": "radiology image report generation", "neg_sample": ["memory - aligned knowledge graph is used for Task", "automatic generating the clinically accurate radiology report from x - ray images is important but challenging .", "the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models ."], "relation": "used for", "id": "2022.bionlp-1.11", "year": 2022, "rel_sent": "Memory - aligned Knowledge Graph for Clinically Accurate Radiology Image Report Generation.", "forward": true, "src_ids": "2022.bionlp-1.11_2457"}
+{"input": "visual patterns of abnormalities is done by using Method| context: automatic generating the clinically accurate radiology report from x - ray images is important but challenging . the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models .", "entity": "visual patterns of abnormalities", "output": "memory - aligned knowledge graph ( makg )", "neg_sample": ["visual patterns of abnormalities is done by using Method", "automatic generating the clinically accurate radiology report from x - ray images is important but challenging .", "the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models ."], "relation": "used for", "id": "2022.bionlp-1.11", "year": 2022, "rel_sent": "In this work , we introduce a Memory - aligned Knowledge Graph ( MaKG ) of clinical abnormalities to better learn the visual patterns of abnormalities and their relationships by integrating it into a deep model architecture for the report generation .", "forward": false, "src_ids": "2022.bionlp-1.11_2458"}
+{"input": "memory - aligned knowledge graph ( makg ) is used for OtherScientificTerm| context: automatic generating the clinically accurate radiology report from x - ray images is important but challenging . the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models .", "entity": "memory - aligned knowledge graph ( makg )", "output": "visual patterns of abnormalities", "neg_sample": ["memory - aligned knowledge graph ( makg ) is used for OtherScientificTerm", "automatic generating the clinically accurate radiology report from x - ray images is important but challenging .", "the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models ."], "relation": "used for", "id": "2022.bionlp-1.11", "year": 2022, "rel_sent": "In this work , we introduce a Memory - aligned Knowledge Graph ( MaKG ) of clinical abnormalities to better learn the visual patterns of abnormalities and their relationships by integrating it into a deep model architecture for the report generation .", "forward": true, "src_ids": "2022.bionlp-1.11_2459"}
+{"input": "report generation is done by using Method| context: automatic generating the clinically accurate radiology report from x - ray images is important but challenging . the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models .", "entity": "report generation", "output": "deep model architecture", "neg_sample": ["report generation is done by using Method", "automatic generating the clinically accurate radiology report from x - ray images is important but challenging .", "the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models ."], "relation": "used for", "id": "2022.bionlp-1.11", "year": 2022, "rel_sent": "In this work , we introduce a Memory - aligned Knowledge Graph ( MaKG ) of clinical abnormalities to better learn the visual patterns of abnormalities and their relationships by integrating it into a deep model architecture for the report generation .", "forward": false, "src_ids": "2022.bionlp-1.11_2460"}
+{"input": "deep model architecture is used for Task| context: automatic generating the clinically accurate radiology report from x - ray images is important but challenging . the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models .", "entity": "deep model architecture", "output": "report generation", "neg_sample": ["deep model architecture is used for Task", "automatic generating the clinically accurate radiology report from x - ray images is important but challenging .", "the identification of multi - grained abnormal regions in image and corresponding abnormalities is difficult for data - driven neural models ."], "relation": "used for", "id": "2022.bionlp-1.11", "year": 2022, "rel_sent": "In this work , we introduce a Memory - aligned Knowledge Graph ( MaKG ) of clinical abnormalities to better learn the visual patterns of abnormalities and their relationships by integrating it into a deep model architecture for the report generation .", "forward": true, "src_ids": "2022.bionlp-1.11_2461"}
+{"input": "grammar checker is done by using OtherScientificTerm| context: grammar checkers ( gec ) are needed for digital language survival . very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools .", "entity": "grammar checker", "output": "multi - lingual setup", "neg_sample": ["grammar checker is done by using OtherScientificTerm", "grammar checkers ( gec ) are needed for digital language survival .", "very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "Reusing a Multi - lingual Setup to Bootstrap a Grammar Checker for a Very Low Resource Language without Data.", "forward": false, "src_ids": "2022.computel-1.19_2462"}
+{"input": "low resource languages is done by using Method| context: very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools .", "entity": "low resource languages", "output": "grammar checker", "neg_sample": ["low resource languages is done by using Method", "very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "Reusing a Multi - lingual Setup to Bootstrap a Grammar Checker for a Very Low Resource Language without Data.", "forward": false, "src_ids": "2022.computel-1.19_2463"}
+{"input": "multi - lingual setup is used for Method| context: very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools .", "entity": "multi - lingual setup", "output": "grammar checker", "neg_sample": ["multi - lingual setup is used for Method", "very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "Reusing a Multi - lingual Setup to Bootstrap a Grammar Checker for a Very Low Resource Language without Data.", "forward": true, "src_ids": "2022.computel-1.19_2464"}
+{"input": "grammar checker is used for Material| context: grammar checkers ( gec ) are needed for digital language survival .", "entity": "grammar checker", "output": "low resource languages", "neg_sample": ["grammar checker is used for Material", "grammar checkers ( gec ) are needed for digital language survival ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "Reusing a Multi - lingual Setup to Bootstrap a Grammar Checker for a Very Low Resource Language without Data.", "forward": true, "src_ids": "2022.computel-1.19_2465"}
+{"input": "test case is used for Material| context: grammar checkers ( gec ) are needed for digital language survival .", "entity": "test case", "output": "lule sami", "neg_sample": ["test case is used for Material", "grammar checkers ( gec ) are needed for digital language survival ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "We present a test case for Lule Sami reusing resources from North Sami , show how we achieve a categorisation of the most frequent errors , and present a preliminary evaluation of the system .", "forward": true, "src_ids": "2022.computel-1.19_2466"}
+{"input": "rules is done by using Method| context: very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools .", "entity": "rules", "output": "grammar", "neg_sample": ["rules is done by using Method", "very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "We use an existing grammar to infer rules for the new language , and we do not need a large gold corpus of annotated grammar errors , but a smaller corpus of regression tests is built while developing the tool .", "forward": false, "src_ids": "2022.computel-1.19_2467"}
+{"input": "grammar is used for OtherScientificTerm| context: grammar checkers ( gec ) are needed for digital language survival . very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools .", "entity": "grammar", "output": "rules", "neg_sample": ["grammar is used for OtherScientificTerm", "grammar checkers ( gec ) are needed for digital language survival .", "very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "We use an existing grammar to infer rules for the new language , and we do not need a large gold corpus of annotated grammar errors , but a smaller corpus of regression tests is built while developing the tool .", "forward": true, "src_ids": "2022.computel-1.19_2468"}
+{"input": "lule sami is done by using Generic| context: grammar checkers ( gec ) are needed for digital language survival . very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools .", "entity": "lule sami", "output": "test case", "neg_sample": ["lule sami is done by using Generic", "grammar checkers ( gec ) are needed for digital language survival .", "very low resource languages like lule sami with less than 3,000 speakers need to hurry to build these tools , but do not have the big corpus data that are required for the construction of machine learning tools ."], "relation": "used for", "id": "2022.computel-1.19", "year": 2022, "rel_sent": "We present a test case for Lule Sami reusing resources from North Sami , show how we achieve a categorisation of the most frequent errors , and present a preliminary evaluation of the system .", "forward": false, "src_ids": "2022.computel-1.19_2469"}
+{"input": "opinion expression identification is done by using Method| context: recent works of opinion expression identification ( oei ) rely heavily on the quality and scale of the manually - constructed training corpus , which could be extremely difficult to satisfy .", "entity": "opinion expression identification", "output": "crowdsourcing", "neg_sample": ["opinion expression identification is done by using Method", "recent works of opinion expression identification ( oei ) rely heavily on the quality and scale of the manually - constructed training corpus , which could be extremely difficult to satisfy ."], "relation": "used for", "id": "2022.acl-long.200", "year": 2022, "rel_sent": "The simulation experiments on our constructed dataset show that crowdsourcing is highly promising for OEI , and our proposed annotator - mixup can further enhance the crowdsourcing modeling .", "forward": false, "src_ids": "2022.acl-long.200_2470"}
+{"input": "crowdsourcing is used for Task| context: crowdsourcing is one practical solution for this problem , aiming to create a large - scale but quality - unguaranteed corpus .", "entity": "crowdsourcing", "output": "opinion expression identification", "neg_sample": ["crowdsourcing is used for Task", "crowdsourcing is one practical solution for this problem , aiming to create a large - scale but quality - unguaranteed corpus ."], "relation": "used for", "id": "2022.acl-long.200", "year": 2022, "rel_sent": "The simulation experiments on our constructed dataset show that crowdsourcing is highly promising for OEI , and our proposed annotator - mixup can further enhance the crowdsourcing modeling .", "forward": true, "src_ids": "2022.acl-long.200_2471"}
+{"input": "crowdsourcing modeling is done by using Method| context: recent works of opinion expression identification ( oei ) rely heavily on the quality and scale of the manually - constructed training corpus , which could be extremely difficult to satisfy . crowdsourcing is one practical solution for this problem , aiming to create a large - scale but quality - unguaranteed corpus .", "entity": "crowdsourcing modeling", "output": "annotator - mixup", "neg_sample": ["crowdsourcing modeling is done by using Method", "recent works of opinion expression identification ( oei ) rely heavily on the quality and scale of the manually - constructed training corpus , which could be extremely difficult to satisfy .", "crowdsourcing is one practical solution for this problem , aiming to create a large - scale but quality - unguaranteed corpus ."], "relation": "used for", "id": "2022.acl-long.200", "year": 2022, "rel_sent": "The simulation experiments on our constructed dataset show that crowdsourcing is highly promising for OEI , and our proposed annotator - mixup can further enhance the crowdsourcing modeling .", "forward": false, "src_ids": "2022.acl-long.200_2472"}
+{"input": "annotator - mixup is used for Method| context: recent works of opinion expression identification ( oei ) rely heavily on the quality and scale of the manually - constructed training corpus , which could be extremely difficult to satisfy . crowdsourcing is one practical solution for this problem , aiming to create a large - scale but quality - unguaranteed corpus .", "entity": "annotator - mixup", "output": "crowdsourcing modeling", "neg_sample": ["annotator - mixup is used for Method", "recent works of opinion expression identification ( oei ) rely heavily on the quality and scale of the manually - constructed training corpus , which could be extremely difficult to satisfy .", "crowdsourcing is one practical solution for this problem , aiming to create a large - scale but quality - unguaranteed corpus ."], "relation": "used for", "id": "2022.acl-long.200", "year": 2022, "rel_sent": "The simulation experiments on our constructed dataset show that crowdsourcing is highly promising for OEI , and our proposed annotator - mixup can further enhance the crowdsourcing modeling .", "forward": true, "src_ids": "2022.acl-long.200_2473"}
+{"input": "supervision is used for OtherScientificTerm| context: one important question in today 's research is how to extend neural entity linking systems to new domains .", "entity": "supervision", "output": "entities", "neg_sample": ["supervision is used for OtherScientificTerm", "one important question in today 's research is how to extend neural entity linking systems to new domains ."], "relation": "used for", "id": "2022.repl4nlp-1.19", "year": 2022, "rel_sent": "While additional supervision on entities that appear in both KGs performs best in an intrinsic evaluation of the vector space , it has less impact on the downstream task of entity linking .", "forward": true, "src_ids": "2022.repl4nlp-1.19_2474"}
+{"input": "entities is done by using OtherScientificTerm| context: entity linking disambiguates mentions by mapping them to entities in a knowledge graph ( kg ) . one important question in today 's research is how to extend neural entity linking systems to new domains .", "entity": "entities", "output": "supervision", "neg_sample": ["entities is done by using OtherScientificTerm", "entity linking disambiguates mentions by mapping them to entities in a knowledge graph ( kg ) .", "one important question in today 's research is how to extend neural entity linking systems to new domains ."], "relation": "used for", "id": "2022.repl4nlp-1.19", "year": 2022, "rel_sent": "While additional supervision on entities that appear in both KGs performs best in an intrinsic evaluation of the vector space , it has less impact on the downstream task of entity linking .", "forward": false, "src_ids": "2022.repl4nlp-1.19_2475"}
+{"input": "intrinsic evaluation of the vector space is done by using OtherScientificTerm| context: entity linking disambiguates mentions by mapping them to entities in a knowledge graph ( kg ) . one important question in today 's research is how to extend neural entity linking systems to new domains .", "entity": "intrinsic evaluation of the vector space", "output": "supervision", "neg_sample": ["intrinsic evaluation of the vector space is done by using OtherScientificTerm", "entity linking disambiguates mentions by mapping them to entities in a knowledge graph ( kg ) .", "one important question in today 's research is how to extend neural entity linking systems to new domains ."], "relation": "used for", "id": "2022.repl4nlp-1.19", "year": 2022, "rel_sent": "While additional supervision on entities that appear in both KGs performs best in an intrinsic evaluation of the vector space , it has less impact on the downstream task of entity linking .", "forward": false, "src_ids": "2022.repl4nlp-1.19_2476"}
+{"input": "supervision is used for Task| context: entity linking disambiguates mentions by mapping them to entities in a knowledge graph ( kg ) . one important question in today 's research is how to extend neural entity linking systems to new domains .", "entity": "supervision", "output": "intrinsic evaluation of the vector space", "neg_sample": ["supervision is used for Task", "entity linking disambiguates mentions by mapping them to entities in a knowledge graph ( kg ) .", "one important question in today 's research is how to extend neural entity linking systems to new domains ."], "relation": "used for", "id": "2022.repl4nlp-1.19", "year": 2022, "rel_sent": "While additional supervision on entities that appear in both KGs performs best in an intrinsic evaluation of the vector space , it has less impact on the downstream task of entity linking .", "forward": true, "src_ids": "2022.repl4nlp-1.19_2477"}
+{"input": "biomedical nlp research is done by using Material| context: recognizing biomedical entities in the text has significance in biomedical and health science research , as it benefits myriad downstream tasks , including entity linking , relation extraction , or entity resolution . while english and a few other widely used languages enjoy ample resources for automatic biomedical entity recognition , it is not the case for bangla , a low - resource language .", "entity": "biomedical nlp research", "output": "bangla nlp resource", "neg_sample": ["biomedical nlp research is done by using Material", "recognizing biomedical entities in the text has significance in biomedical and health science research , as it benefits myriad downstream tasks , including entity linking , relation extraction , or entity resolution .", "while english and a few other widely used languages enjoy ample resources for automatic biomedical entity recognition , it is not the case for bangla , a low - resource language ."], "relation": "used for", "id": "2022.bionlp-1.31", "year": 2022, "rel_sent": "Our developed corpus is a much - needed addition to the Bangla NLP resource that will facilitate biomedical NLP research in Bangla .", "forward": false, "src_ids": "2022.bionlp-1.31_2478"}
+{"input": "bangla nlp resource is used for Task| context: recognizing biomedical entities in the text has significance in biomedical and health science research , as it benefits myriad downstream tasks , including entity linking , relation extraction , or entity resolution . while english and a few other widely used languages enjoy ample resources for automatic biomedical entity recognition , it is not the case for bangla , a low - resource language .", "entity": "bangla nlp resource", "output": "biomedical nlp research", "neg_sample": ["bangla nlp resource is used for Task", "recognizing biomedical entities in the text has significance in biomedical and health science research , as it benefits myriad downstream tasks , including entity linking , relation extraction , or entity resolution .", "while english and a few other widely used languages enjoy ample resources for automatic biomedical entity recognition , it is not the case for bangla , a low - resource language ."], "relation": "used for", "id": "2022.bionlp-1.31", "year": 2022, "rel_sent": "Our developed corpus is a much - needed addition to the Bangla NLP resource that will facilitate biomedical NLP research in Bangla .", "forward": true, "src_ids": "2022.bionlp-1.31_2479"}
+{"input": "open - domain dense retrieval is done by using Method| context: dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document . however , a document can usually answer multiple potential queries from different views . so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem .", "entity": "open - domain dense retrieval", "output": "multi - view document representation learning framework", "neg_sample": ["open - domain dense retrieval is done by using Method", "dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document .", "however , a document can usually answer multiple potential queries from different views .", "so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem ."], "relation": "used for", "id": "2022.acl-long.414", "year": 2022, "rel_sent": "Multi - View Document Representation Learning for Open - Domain Dense Retrieval.", "forward": false, "src_ids": "2022.acl-long.414_2480"}
+{"input": "multi - view embeddings is done by using Method| context: dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document . however , a document can usually answer multiple potential queries from different views . so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem .", "entity": "multi - view embeddings", "output": "multi - view document representation learning framework", "neg_sample": ["multi - view embeddings is done by using Method", "dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document .", "however , a document can usually answer multiple potential queries from different views .", "so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem ."], "relation": "used for", "id": "2022.acl-long.414", "year": 2022, "rel_sent": "This paper proposes a multi - view document representation learning framework , aiming to produce multi - view embeddings to represent documents and enforce them to align with different queries .", "forward": false, "src_ids": "2022.acl-long.414_2481"}
+{"input": "multi - view document representation learning framework is used for Task| context: dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document . however , a document can usually answer multiple potential queries from different views . so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem .", "entity": "multi - view document representation learning framework", "output": "open - domain dense retrieval", "neg_sample": ["multi - view document representation learning framework is used for Task", "dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document .", "however , a document can usually answer multiple potential queries from different views .", "so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem ."], "relation": "used for", "id": "2022.acl-long.414", "year": 2022, "rel_sent": "Multi - View Document Representation Learning for Open - Domain Dense Retrieval.", "forward": true, "src_ids": "2022.acl-long.414_2482"}
+{"input": "multi - view document representation learning framework is used for OtherScientificTerm| context: dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document . however , a document can usually answer multiple potential queries from different views . so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem .", "entity": "multi - view document representation learning framework", "output": "multi - view embeddings", "neg_sample": ["multi - view document representation learning framework is used for OtherScientificTerm", "dense retrieval has achieved impressive advances in first - stage retrieval from a large - scale document collection , which is built on bi - encoder architecture to produce single vector representation of query and document .", "however , a document can usually answer multiple potential queries from different views .", "so the single vector representation of a document is hard to match with multi - view queries , and faces a semantic mismatch problem ."], "relation": "used for", "id": "2022.acl-long.414", "year": 2022, "rel_sent": "This paper proposes a multi - view document representation learning framework , aiming to produce multi - view embeddings to represent documents and enforce them to align with different queries .", "forward": true, "src_ids": "2022.acl-long.414_2483"}
+{"input": "nlp tasks is done by using Task| context: although recent massively multilingual language models ( mmlms ) like mbert and xlmr support around 100 languages , most existing multilingual nlp benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity . we argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of mmlms across the linguistic landscape .", "entity": "nlp tasks", "output": "performance prediction", "neg_sample": ["nlp tasks is done by using Task", "although recent massively multilingual language models ( mmlms ) like mbert and xlmr support around 100 languages , most existing multilingual nlp benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity .", "we argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of mmlms across the linguistic landscape ."], "relation": "used for", "id": "2022.nlppower-1.7", "year": 2022, "rel_sent": "We propose that the recent work done in Performance Prediction for NLP tasks can serve as a potential solution in fixing benchmarking in Multilingual NLP by utilizing features related to data and language typology to estimate the performance of an MMLM on different languages .", "forward": false, "src_ids": "2022.nlppower-1.7_2484"}
+{"input": "performance prediction is used for Task| context: although recent massively multilingual language models ( mmlms ) like mbert and xlmr support around 100 languages , most existing multilingual nlp benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity . we argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of mmlms across the linguistic landscape .", "entity": "performance prediction", "output": "nlp tasks", "neg_sample": ["performance prediction is used for Task", "although recent massively multilingual language models ( mmlms ) like mbert and xlmr support around 100 languages , most existing multilingual nlp benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity .", "we argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of mmlms across the linguistic landscape ."], "relation": "used for", "id": "2022.nlppower-1.7", "year": 2022, "rel_sent": "We propose that the recent work done in Performance Prediction for NLP tasks can serve as a potential solution in fixing benchmarking in Multilingual NLP by utilizing features related to data and language typology to estimate the performance of an MMLM on different languages .", "forward": true, "src_ids": "2022.nlppower-1.7_2485"}
+{"input": "mmlm is done by using OtherScientificTerm| context: although recent massively multilingual language models ( mmlms ) like mbert and xlmr support around 100 languages , most existing multilingual nlp benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity . we argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of mmlms across the linguistic landscape .", "entity": "mmlm", "output": "features", "neg_sample": ["mmlm is done by using OtherScientificTerm", "although recent massively multilingual language models ( mmlms ) like mbert and xlmr support around 100 languages , most existing multilingual nlp benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity .", "we argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of mmlms across the linguistic landscape ."], "relation": "used for", "id": "2022.nlppower-1.7", "year": 2022, "rel_sent": "We propose that the recent work done in Performance Prediction for NLP tasks can serve as a potential solution in fixing benchmarking in Multilingual NLP by utilizing features related to data and language typology to estimate the performance of an MMLM on different languages .", "forward": false, "src_ids": "2022.nlppower-1.7_2486"}
+{"input": "word discovery is done by using Method| context: processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario . no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain .", "entity": "word discovery", "output": "topwords - seg", "neg_sample": ["word discovery is done by using Method", "processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario .", "no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain ."], "relation": "used for", "id": "2022.acl-long.13", "year": 2022, "rel_sent": "TopWORDS - Seg : Simultaneous Text Segmentation and Word Discovery for Open - Domain Chinese Texts via Bayesian Inference.", "forward": false, "src_ids": "2022.acl-long.13_2487"}
+{"input": "simultaneous text segmentation is done by using Method| context: processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario . no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain .", "entity": "simultaneous text segmentation", "output": "topwords - seg", "neg_sample": ["simultaneous text segmentation is done by using Method", "processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario .", "no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain ."], "relation": "used for", "id": "2022.acl-long.13", "year": 2022, "rel_sent": "TopWORDS - Seg : Simultaneous Text Segmentation and Word Discovery for Open - Domain Chinese Texts via Bayesian Inference.", "forward": false, "src_ids": "2022.acl-long.13_2488"}
+{"input": "open - domain chinese texts is done by using Task| context: processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario . no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain .", "entity": "open - domain chinese texts", "output": "simultaneous text segmentation", "neg_sample": ["open - domain chinese texts is done by using Task", "processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario .", "no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain ."], "relation": "used for", "id": "2022.acl-long.13", "year": 2022, "rel_sent": "TopWORDS - Seg : Simultaneous Text Segmentation and Word Discovery for Open - Domain Chinese Texts via Bayesian Inference.", "forward": false, "src_ids": "2022.acl-long.13_2489"}
+{"input": "topwords - seg is used for Task| context: processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario . no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain .", "entity": "topwords - seg", "output": "simultaneous text segmentation", "neg_sample": ["topwords - seg is used for Task", "processing open - domain chinese texts has been a critical bottleneck in computational linguistics for decades , partially because text segmentation and word discovery often entangle with each other in this challenging scenario .", "no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain ."], "relation": "used for", "id": "2022.acl-long.13", "year": 2022, "rel_sent": "TopWORDS - Seg : Simultaneous Text Segmentation and Word Discovery for Open - Domain Chinese Texts via Bayesian Inference.", "forward": true, "src_ids": "2022.acl-long.13_2490"}
+{"input": "simultaneous text segmentation is used for Material| context: no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain .", "entity": "simultaneous text segmentation", "output": "open - domain chinese texts", "neg_sample": ["simultaneous text segmentation is used for Material", "no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain ."], "relation": "used for", "id": "2022.acl-long.13", "year": 2022, "rel_sent": "TopWORDS - Seg : Simultaneous Text Segmentation and Word Discovery for Open - Domain Chinese Texts via Bayesian Inference.", "forward": true, "src_ids": "2022.acl-long.13_2491"}
+{"input": "word discovery is used for Material| context: no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain .", "entity": "word discovery", "output": "open - domain chinese texts", "neg_sample": ["word discovery is used for Material", "no existing methods yet can achieve effective text segmentation and word discovery simultaneously in open domain ."], "relation": "used for", "id": "2022.acl-long.13", "year": 2022, "rel_sent": "TopWORDS - Seg : Simultaneous Text Segmentation and Word Discovery for Open - Domain Chinese Texts via Bayesian Inference.", "forward": true, "src_ids": "2022.acl-long.13_2492"}
+{"input": "phonological information is done by using Method| context: character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue .", "entity": "phonological information", "output": "probing classifiers", "neg_sample": ["phonological information is done by using Method", "character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue ."], "relation": "used for", "id": "2022.acl-long.470", "year": 2022, "rel_sent": "This cross - lingual analysis shows that textual character representations correlate strongly with sound representations for languages using an alphabetic script , while shape correlates with featural scripts . We further develop a set of probing classifiers to intrinsically evaluate what phonological information is encoded in character embeddings .", "forward": false, "src_ids": "2022.acl-long.470_2493"}
+{"input": "probing classifiers is used for OtherScientificTerm| context: character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue .", "entity": "probing classifiers", "output": "phonological information", "neg_sample": ["probing classifiers is used for OtherScientificTerm", "character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue ."], "relation": "used for", "id": "2022.acl-long.470", "year": 2022, "rel_sent": "This cross - lingual analysis shows that textual character representations correlate strongly with sound representations for languages using an alphabetic script , while shape correlates with featural scripts . We further develop a set of probing classifiers to intrinsically evaluate what phonological information is encoded in character embeddings .", "forward": true, "src_ids": "2022.acl-long.470_2494"}
+{"input": "sociofillmore is used for Task| context: sociofillmore is a multilingual tool which helps to bring to the fore the focus or the perspective that a text expresses in depicting an event .", "entity": "sociofillmore", "output": "discovering perspectives", "neg_sample": ["sociofillmore is used for Task", "sociofillmore is a multilingual tool which helps to bring to the fore the focus or the perspective that a text expresses in depicting an event ."], "relation": "used for", "id": "2022.acl-demo.24", "year": 2022, "rel_sent": "SocioFillmore : A Tool for Discovering Perspectives.", "forward": true, "src_ids": "2022.acl-demo.24_2495"}
+{"input": "paraphrase evaluation is done by using Method| context: we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation .", "entity": "paraphrase evaluation", "output": "generative pretraining", "neg_sample": ["paraphrase evaluation is done by using Method", "we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation ."], "relation": "used for", "id": "2022.acl-long.280", "year": 2022, "rel_sent": "Generative Pretraining for Paraphrase Evaluation.", "forward": false, "src_ids": "2022.acl-long.280_2496"}
+{"input": "generative pretraining is used for Task| context: we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation .", "entity": "generative pretraining", "output": "paraphrase evaluation", "neg_sample": ["generative pretraining is used for Task", "we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation ."], "relation": "used for", "id": "2022.acl-long.280", "year": 2022, "rel_sent": "Generative Pretraining for Paraphrase Evaluation.", "forward": true, "src_ids": "2022.acl-long.280_2497"}
+{"input": "parableu is used for OtherScientificTerm| context: we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation .", "entity": "parableu", "output": "paraphrasis", "neg_sample": ["parableu is used for OtherScientificTerm", "we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation ."], "relation": "used for", "id": "2022.acl-long.280", "year": 2022, "rel_sent": "Unlike previous approaches , ParaBLEU learns to understand paraphrasis using generative conditioning as a pretraining objective .", "forward": true, "src_ids": "2022.acl-long.280_2498"}
+{"input": "parableu is used for OtherScientificTerm| context: we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation .", "entity": "parableu", "output": "paraphrases", "neg_sample": ["parableu is used for OtherScientificTerm", "we introduce parableu , a paraphrase representation learning model and evaluation metric for text generation ."], "relation": "used for", "id": "2022.acl-long.280", "year": 2022, "rel_sent": "Finally , we demonstrate that ParaBLEU can be used to conditionally generate novel paraphrases from a single demonstration , which we use to confirm our hypothesis that it learns abstract , generalized paraphrase representations .", "forward": true, "src_ids": "2022.acl-long.280_2499"}
+{"input": "inductive biases is done by using OtherScientificTerm| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "inductive biases", "output": "speaker information", "neg_sample": ["inductive biases is done by using OtherScientificTerm", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "Speaker Information Can Guide Models to Better Inductive Biases : A Case Study On Predicting Code - Switching.", "forward": false, "src_ids": "2022.acl-long.267_2500"}
+{"input": "speaker information is used for OtherScientificTerm| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "speaker information", "output": "inductive biases", "neg_sample": ["speaker information is used for OtherScientificTerm", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "Speaker Information Can Guide Models to Better Inductive Biases : A Case Study On Predicting Code - Switching.", "forward": true, "src_ids": "2022.acl-long.267_2501"}
+{"input": "predicting code - switching points is done by using Task| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "predicting code - switching points", "output": "speaker - driven task", "neg_sample": ["predicting code - switching points is done by using Task", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "For the speaker - driven task of predicting code - switching points in English - Spanish bilingual dialogues , we show that adding sociolinguistically - grounded speaker features as prepended prompts significantly improves accuracy .", "forward": false, "src_ids": "2022.acl-long.267_2502"}
+{"input": "speaker - driven task is used for Task| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "speaker - driven task", "output": "predicting code - switching points", "neg_sample": ["speaker - driven task is used for Task", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "For the speaker - driven task of predicting code - switching points in English - Spanish bilingual dialogues , we show that adding sociolinguistically - grounded speaker features as prepended prompts significantly improves accuracy .", "forward": true, "src_ids": "2022.acl-long.267_2503"}
+{"input": "speaker - informed models is done by using OtherScientificTerm| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "speaker - informed models", "output": "influential phrases", "neg_sample": ["speaker - informed models is done by using OtherScientificTerm", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "We find that by adding influential phrases to the input , speaker - informed models learn useful and explainable linguistic information .", "forward": false, "src_ids": "2022.acl-long.267_2504"}
+{"input": "linguistic information is done by using Method| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "linguistic information", "output": "speaker - informed models", "neg_sample": ["linguistic information is done by using Method", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "We find that by adding influential phrases to the input , speaker - informed models learn useful and explainable linguistic information .", "forward": false, "src_ids": "2022.acl-long.267_2505"}
+{"input": "influential phrases is used for Method| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "influential phrases", "output": "speaker - informed models", "neg_sample": ["influential phrases is used for Method", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "We find that by adding influential phrases to the input , speaker - informed models learn useful and explainable linguistic information .", "forward": true, "src_ids": "2022.acl-long.267_2506"}
+{"input": "speaker - informed models is used for OtherScientificTerm| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "speaker - informed models", "output": "linguistic information", "neg_sample": ["speaker - informed models is used for OtherScientificTerm", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "We find that by adding influential phrases to the input , speaker - informed models learn useful and explainable linguistic information .", "forward": true, "src_ids": "2022.acl-long.267_2507"}
+{"input": "code - switching is done by using Method| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "code - switching", "output": "neural model", "neg_sample": ["code - switching is done by using Method", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "To our knowledge , we are the first to incorporate speaker characteristics in a neural model for code - switching , and more generally , take a step towards developing transparent , personalized models that use speaker information in a controlled way .", "forward": false, "src_ids": "2022.acl-long.267_2508"}
+{"input": "neural model is used for Task| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "neural model", "output": "code - switching", "neg_sample": ["neural model is used for Task", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.267", "year": 2022, "rel_sent": "To our knowledge , we are the first to incorporate speaker characteristics in a neural model for code - switching , and more generally , take a step towards developing transparent , personalized models that use speaker information in a controlled way .", "forward": true, "src_ids": "2022.acl-long.267_2509"}
+{"input": "out - of - vocabulary named entity recognition is done by using Method| context: ner model has achieved promising performance on standard ner benchmarks . however , recent studies show that previous approaches may over - rely on entity mention information , resulting in poor performance on out - of - vocabulary(oov ) entity recognition .", "entity": "out - of - vocabulary named entity recognition", "output": "miner", "neg_sample": ["out - of - vocabulary named entity recognition is done by using Method", "ner model has achieved promising performance on standard ner benchmarks .", "however , recent studies show that previous approaches may over - rely on entity mention information , resulting in poor performance on out - of - vocabulary(oov ) entity recognition ."], "relation": "used for", "id": "2022.acl-long.383", "year": 2022, "rel_sent": "MINER : Improving Out - of - Vocabulary Named Entity Recognition from an Information Theoretic Perspective.", "forward": false, "src_ids": "2022.acl-long.383_2510"}
+{"input": "miner is used for Task| context: ner model has achieved promising performance on standard ner benchmarks . however , recent studies show that previous approaches may over - rely on entity mention information , resulting in poor performance on out - of - vocabulary(oov ) entity recognition .", "entity": "miner", "output": "out - of - vocabulary named entity recognition", "neg_sample": ["miner is used for Task", "ner model has achieved promising performance on standard ner benchmarks .", "however , recent studies show that previous approaches may over - rely on entity mention information , resulting in poor performance on out - of - vocabulary(oov ) entity recognition ."], "relation": "used for", "id": "2022.acl-long.383", "year": 2022, "rel_sent": "MINER : Improving Out - of - Vocabulary Named Entity Recognition from an Information Theoretic Perspective.", "forward": true, "src_ids": "2022.acl-long.383_2511"}
+{"input": "identifying semantic - level non - novelty is done by using Task| context: the quest for new information is an inborn human trait and has always been quintessential for human survival and progress . novelty drives curiosity , which in turn drives innovation . in natural language processing ( nlp ) , novelty detection refers tofinding text that has some new information to offer with respect to whatever is earlier seen or known . with the exponential growth of information all across the web , there is an accompanying menace of redundancy . a considerable portion of the web contents are duplicates , and we need efficient mechanisms to retain new information and filter out redundant information . however , detecting redundancy at the semantic level and identifying novel text is not straightforward because the text may have less lexical overlap yet convey the same information . on top of that , non - novel / redundant information in a document may have assimilated from multiple source documents , not just one . the problem surmounts when the subject of the discourse is documents , and numerous prior documents need to be processed to ascertain the novelty / non - novelty of the current one in concern .", "entity": "identifying semantic - level non - novelty", "output": "multipremise entailment task", "neg_sample": ["identifying semantic - level non - novelty is done by using Task", "the quest for new information is an inborn human trait and has always been quintessential for human survival and progress .", "novelty drives curiosity , which in turn drives innovation .", "in natural language processing ( nlp ) , novelty detection refers tofinding text that has some new information to offer with respect to whatever is earlier seen or known .", "with the exponential growth of information all across the web , there is an accompanying menace of redundancy .", "a considerable portion of the web contents are duplicates , and we need efficient mechanisms to retain new information and filter out redundant information .", "however , detecting redundancy at the semantic level and identifying novel text is not straightforward because the text may have less lexical overlap yet convey the same information .", "on top of that , non - novel / redundant information in a document may have assimilated from multiple source documents , not just one .", "the problem surmounts when the subject of the discourse is documents , and numerous prior documents need to be processed to ascertain the novelty / non - novelty of the current one in concern ."], "relation": "used for", "id": "2022.cl-1.3", "year": 2022, "rel_sent": "We argue that a multipremise entailment task is one close approximation toward identifying semantic - level non - novelty .", "forward": false, "src_ids": "2022.cl-1.3_2512"}
+{"input": "multipremise entailment task is used for Task| context: the quest for new information is an inborn human trait and has always been quintessential for human survival and progress . novelty drives curiosity , which in turn drives innovation . in natural language processing ( nlp ) , novelty detection refers tofinding text that has some new information to offer with respect to whatever is earlier seen or known . with the exponential growth of information all across the web , there is an accompanying menace of redundancy . a considerable portion of the web contents are duplicates , and we need efficient mechanisms to retain new information and filter out redundant information . however , detecting redundancy at the semantic level and identifying novel text is not straightforward because the text may have less lexical overlap yet convey the same information . on top of that , non - novel / redundant information in a document may have assimilated from multiple source documents , not just one . the problem surmounts when the subject of the discourse is documents , and numerous prior documents need to be processed to ascertain the novelty / non - novelty of the current one in concern .", "entity": "multipremise entailment task", "output": "identifying semantic - level non - novelty", "neg_sample": ["multipremise entailment task is used for Task", "the quest for new information is an inborn human trait and has always been quintessential for human survival and progress .", "novelty drives curiosity , which in turn drives innovation .", "in natural language processing ( nlp ) , novelty detection refers tofinding text that has some new information to offer with respect to whatever is earlier seen or known .", "with the exponential growth of information all across the web , there is an accompanying menace of redundancy .", "a considerable portion of the web contents are duplicates , and we need efficient mechanisms to retain new information and filter out redundant information .", "however , detecting redundancy at the semantic level and identifying novel text is not straightforward because the text may have less lexical overlap yet convey the same information .", "on top of that , non - novel / redundant information in a document may have assimilated from multiple source documents , not just one .", "the problem surmounts when the subject of the discourse is documents , and numerous prior documents need to be processed to ascertain the novelty / non - novelty of the current one in concern ."], "relation": "used for", "id": "2022.cl-1.3", "year": 2022, "rel_sent": "We argue that a multipremise entailment task is one close approximation toward identifying semantic - level non - novelty .", "forward": true, "src_ids": "2022.cl-1.3_2513"}
+{"input": "mathematical text is done by using OtherScientificTerm| context: the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge . however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines . one of the fundamental requirements towards mathematical language understanding , is the creation of models able to meaningfully represent variables . this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e. , the representation of a variable should come from its context .", "entity": "mathematical text", "output": "encoding variables", "neg_sample": ["mathematical text is done by using OtherScientificTerm", "the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge .", "however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines .", "one of the fundamental requirements towards mathematical language understanding , is the creation of models able to meaningfully represent variables .", "this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e.", ", the representation of a variable should come from its context ."], "relation": "used for", "id": "2022.findings-acl.76", "year": 2022, "rel_sent": "To be or not to be an Integer ? Encoding Variables for Mathematical Text.", "forward": false, "src_ids": "2022.findings-acl.76_2514"}
+{"input": "encoding variables is used for Material| context: the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge . however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines . one of the fundamental requirements towards mathematical language understanding , is the creation of models able to meaningfully represent variables . this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e. , the representation of a variable should come from its context .", "entity": "encoding variables", "output": "mathematical text", "neg_sample": ["encoding variables is used for Material", "the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge .", "however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines .", "one of the fundamental requirements towards mathematical language understanding , is the creation of models able to meaningfully represent variables .", "this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e.", ", the representation of a variable should come from its context ."], "relation": "used for", "id": "2022.findings-acl.76", "year": 2022, "rel_sent": "To be or not to be an Integer ? Encoding Variables for Mathematical Text.", "forward": true, "src_ids": "2022.findings-acl.76_2515"}
+{"input": "context - based representations is used for OtherScientificTerm| context: the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge . however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines . this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e. , the representation of a variable should come from its context .", "entity": "context - based representations", "output": "variables", "neg_sample": ["context - based representations is used for OtherScientificTerm", "the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge .", "however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines .", "this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e.", ", the representation of a variable should come from its context ."], "relation": "used for", "id": "2022.findings-acl.76", "year": 2022, "rel_sent": "In this work , we propose VarSlot , a Variable Slot - based approach , which not only delivers state - of - the - art results in the task of variable typing , but is also able to create context - based representations for variables .", "forward": true, "src_ids": "2022.findings-acl.76_2516"}
+{"input": "variables is done by using Method| context: the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge . however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines . one of the fundamental requirements towards mathematical language understanding , is the creation of models able to meaningfully represent variables . this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e. , the representation of a variable should come from its context .", "entity": "variables", "output": "context - based representations", "neg_sample": ["variables is done by using Method", "the application of natural language inference ( nli ) methods over large textual corpora can facilitate scientific discovery , reducing the gap between current research and the available large - scale scientific knowledge .", "however , contemporary nli models are still limited in interpreting mathematical knowledge written in natural language , even though mathematics is an integral part of scientific argumentation for many disciplines .", "one of the fundamental requirements towards mathematical language understanding , is the creation of models able to meaningfully represent variables .", "this problem is particularly challenging since the meaning of a variable should be assigned exclusively from its defining type , i.e.", ", the representation of a variable should come from its context ."], "relation": "used for", "id": "2022.findings-acl.76", "year": 2022, "rel_sent": "In this work , we propose VarSlot , a Variable Slot - based approach , which not only delivers state - of - the - art results in the task of variable typing , but is also able to create context - based representations for variables .", "forward": false, "src_ids": "2022.findings-acl.76_2517"}
+{"input": "student model is used for OtherScientificTerm| context: machine learning models that offer excellent predictive performance often lack the interpretability necessary to support integrated human machine decision - making . in clinical medicine and other high - risk settings , domain experts may be unwilling to trust model predictions without explanations . work in explainable ai must balance competing objectives along two different axes : 1 ) models should ideally be both accurate and simple . 2 ) explanations must balance faithfulness to the model 's decision - making with their plausibility to a domain expert .", "entity": "student model", "output": "natural language explanations", "neg_sample": ["student model is used for OtherScientificTerm", "machine learning models that offer excellent predictive performance often lack the interpretability necessary to support integrated human machine decision - making .", "in clinical medicine and other high - risk settings , domain experts may be unwilling to trust model predictions without explanations .", "work in explainable ai must balance competing objectives along two different axes : 1 ) models should ideally be both accurate and simple .", "2 ) explanations must balance faithfulness to the model 's decision - making with their plausibility to a domain expert ."], "relation": "used for", "id": "2022.bionlp-1.41", "year": 2022, "rel_sent": "We evaluate our approach on the task of assigning ICD codes to clinical notes to demonstrate that the student model is faithful to the teacher model 's behavior and produces quality natural language explanations .", "forward": true, "src_ids": "2022.bionlp-1.41_2518"}
+{"input": "natural language explanations is done by using Method| context: machine learning models that offer excellent predictive performance often lack the interpretability necessary to support integrated human machine decision - making . in clinical medicine and other high - risk settings , domain experts may be unwilling to trust model predictions without explanations . work in explainable ai must balance competing objectives along two different axes : 1 ) models should ideally be both accurate and simple . 2 ) explanations must balance faithfulness to the model 's decision - making with their plausibility to a domain expert .", "entity": "natural language explanations", "output": "student model", "neg_sample": ["natural language explanations is done by using Method", "machine learning models that offer excellent predictive performance often lack the interpretability necessary to support integrated human machine decision - making .", "in clinical medicine and other high - risk settings , domain experts may be unwilling to trust model predictions without explanations .", "work in explainable ai must balance competing objectives along two different axes : 1 ) models should ideally be both accurate and simple .", "2 ) explanations must balance faithfulness to the model 's decision - making with their plausibility to a domain expert ."], "relation": "used for", "id": "2022.bionlp-1.41", "year": 2022, "rel_sent": "We evaluate our approach on the task of assigning ICD codes to clinical notes to demonstrate that the student model is faithful to the teacher model 's behavior and produces quality natural language explanations .", "forward": false, "src_ids": "2022.bionlp-1.41_2519"}
+{"input": "prompt tuning approach is done by using Method| context: there has been growing interest in parameter - efficient methods to apply pre - trained language models to downstream tasks . building on the prompt tuning approach of lester et al .", "entity": "prompt tuning approach", "output": "spot", "neg_sample": ["prompt tuning approach is done by using Method", "there has been growing interest in parameter - efficient methods to apply pre - trained language models to downstream tasks .", "building on the prompt tuning approach of lester et al ."], "relation": "used for", "id": "2022.acl-long.346", "year": 2022, "rel_sent": "We show that SPoT significantly boosts the performance of Prompt Tuning across many tasks .", "forward": false, "src_ids": "2022.acl-long.346_2520"}
+{"input": "spot is used for Method| context: there has been growing interest in parameter - efficient methods to apply pre - trained language models to downstream tasks .", "entity": "spot", "output": "prompt tuning approach", "neg_sample": ["spot is used for Method", "there has been growing interest in parameter - efficient methods to apply pre - trained language models to downstream tasks ."], "relation": "used for", "id": "2022.acl-long.346", "year": 2022, "rel_sent": "We show that SPoT significantly boosts the performance of Prompt Tuning across many tasks .", "forward": true, "src_ids": "2022.acl-long.346_2521"}
+{"input": "sentence embeddings is done by using Method| context: contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g. , simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures .", "entity": "sentence embeddings", "output": "semantic - aware contrastive learning framework", "neg_sample": ["sentence embeddings is done by using Method", "contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g.", ", simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures ."], "relation": "used for", "id": "2022.findings-acl.22", "year": 2022, "rel_sent": "In this paper , we propose a semantic - aware contrastive learning framework for sentence embeddings , termed Pseudo - Token BERT ( PT - BERT ) , which is able to explore the pseudo - token space ( i.e. , latent semantic space ) representation of a sentence while eliminating the impact of superficial features such as sentence length and syntax .", "forward": false, "src_ids": "2022.findings-acl.22_2522"}
+{"input": "semantic - aware contrastive learning framework is used for OtherScientificTerm| context: contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g. , simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures .", "entity": "semantic - aware contrastive learning framework", "output": "sentence embeddings", "neg_sample": ["semantic - aware contrastive learning framework is used for OtherScientificTerm", "contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g.", ", simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures ."], "relation": "used for", "id": "2022.findings-acl.22", "year": 2022, "rel_sent": "In this paper , we propose a semantic - aware contrastive learning framework for sentence embeddings , termed Pseudo - Token BERT ( PT - BERT ) , which is able to explore the pseudo - token space ( i.e. , latent semantic space ) representation of a sentence while eliminating the impact of superficial features such as sentence length and syntax .", "forward": true, "src_ids": "2022.findings-acl.22_2523"}
+{"input": "contrastive learning is done by using Method| context: contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g. , simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures .", "entity": "contrastive learning", "output": "attention mechanism", "neg_sample": ["contrastive learning is done by using Method", "contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g.", ", simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures ."], "relation": "used for", "id": "2022.findings-acl.22", "year": 2022, "rel_sent": "Leveraging these pseudo sequences , we are able to construct same - length positive and negative pairs based on the attention mechanism to perform contrastive learning .", "forward": false, "src_ids": "2022.findings-acl.22_2524"}
+{"input": "representation of sentence embeddings is done by using Generic| context: contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g. , simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures .", "entity": "representation of sentence embeddings", "output": "queue", "neg_sample": ["representation of sentence embeddings is done by using Generic", "contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g.", ", simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures ."], "relation": "used for", "id": "2022.findings-acl.22", "year": 2022, "rel_sent": "In addition , we utilize both the gradient - updating and momentum - updating encoders to encode instances while dynamically maintaining an additional queue to store the representation of sentence embeddings , enhancing the encoder 's learning performance for negative examples .", "forward": false, "src_ids": "2022.findings-acl.22_2525"}
+{"input": "queue is used for OtherScientificTerm| context: contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g. , simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures .", "entity": "queue", "output": "representation of sentence embeddings", "neg_sample": ["queue is used for OtherScientificTerm", "contrastive learning has shown great potential in unsupervised sentence embedding tasks , e.g.", ", simcse ( citation).however , these existing solutions are heavily affected by superficial features like the length of sentences or syntactic structures ."], "relation": "used for", "id": "2022.findings-acl.22", "year": 2022, "rel_sent": "In addition , we utilize both the gradient - updating and momentum - updating encoders to encode instances while dynamically maintaining an additional queue to store the representation of sentence embeddings , enhancing the encoder 's learning performance for negative examples .", "forward": true, "src_ids": "2022.findings-acl.22_2526"}
+{"input": "brain - to - word decoding is done by using Task| context: decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing . due to the noisy nature of brain recordings , existing work has simplified brain - to - word decoding as a binary classification task which is to discriminate a brain signal between its corresponding word and a wrong one . this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons .", "entity": "brain - to - word decoding", "output": "cross - modal cloze ( cmc ) task", "neg_sample": ["brain - to - word decoding is done by using Task", "decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing .", "due to the noisy nature of brain recordings , existing work has simplified brain - to - word decoding as a binary classification task which is to discriminate a brain signal between its corresponding word and a wrong one .", "this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons ."], "relation": "used for", "id": "2022.findings-acl.54", "year": 2022, "rel_sent": "Cross - Modal Cloze Task : A New Task to Brain - to - Word Decoding.", "forward": false, "src_ids": "2022.findings-acl.54_2527"}
+{"input": "brain - to - word decoding is done by using Generic| context: decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing . this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons .", "entity": "brain - to - word decoding", "output": "brain - to - word decoding", "neg_sample": ["brain - to - word decoding is done by using Generic", "decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing .", "this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons ."], "relation": "used for", "id": "2022.findings-acl.54", "year": 2022, "rel_sent": "Cross - Modal Cloze Task : A New Task to Brain - to - Word Decoding.", "forward": false, "src_ids": "2022.findings-acl.54_2528"}
+{"input": "brain - to - word decoding is used for Generic| context: decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing . this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons .", "entity": "brain - to - word decoding", "output": "brain - to - word decoding", "neg_sample": ["brain - to - word decoding is used for Generic", "decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing .", "this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons ."], "relation": "used for", "id": "2022.findings-acl.54", "year": 2022, "rel_sent": "Cross - Modal Cloze Task : A New Task to Brain - to - Word Decoding.", "forward": true, "src_ids": "2022.findings-acl.54_2529"}
+{"input": "cross - modal cloze ( cmc ) task is used for Generic| context: decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing . this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons .", "entity": "cross - modal cloze ( cmc ) task", "output": "brain - to - word decoding", "neg_sample": ["cross - modal cloze ( cmc ) task is used for Generic", "decoding language from non - invasive brain activity has attracted increasing attention from both researchers in neuroscience and natural language processing .", "this pairwise classification task , however , can not promote the development of practical neural decoders for two reasons ."], "relation": "used for", "id": "2022.findings-acl.54", "year": 2022, "rel_sent": "Cross - Modal Cloze Task : A New Task to Brain - to - Word Decoding.", "forward": true, "src_ids": "2022.findings-acl.54_2530"}
+{"input": "text and speech translation models is done by using Method| context: pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches .", "entity": "text and speech translation models", "output": "wav2vec", "neg_sample": ["text and speech translation models is done by using Method", "pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches ."], "relation": "used for", "id": "2022.iwslt-1.14", "year": 2022, "rel_sent": "In this evaluation , we aim at empirically looking for the answer by using the wav2vec , mBART50 and DeltaLM models to improve text and speech translation models .", "forward": false, "src_ids": "2022.iwslt-1.14_2531"}
+{"input": "deltalm models is used for Method| context: pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches .", "entity": "deltalm models", "output": "text and speech translation models", "neg_sample": ["deltalm models is used for Method", "pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches ."], "relation": "used for", "id": "2022.iwslt-1.14", "year": 2022, "rel_sent": "In this evaluation , we aim at empirically looking for the answer by using the wav2vec , mBART50 and DeltaLM models to improve text and speech translation models .", "forward": true, "src_ids": "2022.iwslt-1.14_2532"}
+{"input": "mbart50 is used for Method| context: pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches .", "entity": "mbart50", "output": "text and speech translation models", "neg_sample": ["mbart50 is used for Method", "pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches ."], "relation": "used for", "id": "2022.iwslt-1.14", "year": 2022, "rel_sent": "In this evaluation , we aim at empirically looking for the answer by using the wav2vec , mBART50 and DeltaLM models to improve text and speech translation models .", "forward": true, "src_ids": "2022.iwslt-1.14_2533"}
+{"input": "wav2vec is used for Method| context: pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches .", "entity": "wav2vec", "output": "text and speech translation models", "neg_sample": ["wav2vec is used for Method", "pretrained models in acoustic and textual modalities can potentially improve speech translation for both cascade and end - to - end approaches ."], "relation": "used for", "id": "2022.iwslt-1.14", "year": 2022, "rel_sent": "In this evaluation , we aim at empirically looking for the answer by using the wav2vec , mBART50 and DeltaLM models to improve text and speech translation models .", "forward": true, "src_ids": "2022.iwslt-1.14_2534"}
+{"input": "latent clusters is done by using OtherScientificTerm| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "latent clusters", "output": "prototype tensors", "neg_sample": ["latent clusters is done by using OtherScientificTerm", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "ProtoTEx faithfully explains model decisions based on prototype tensors that encode latent clusters of training examples .", "forward": false, "src_ids": "2022.acl-long.213_2535"}
+{"input": "prototex is used for Task| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "prototex", "output": "model decisions", "neg_sample": ["prototex is used for Task", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "ProtoTEx faithfully explains model decisions based on prototype tensors that encode latent clusters of training examples .", "forward": true, "src_ids": "2022.acl-long.213_2536"}
+{"input": "prototype tensors is used for OtherScientificTerm| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "prototype tensors", "output": "latent clusters", "neg_sample": ["prototype tensors is used for OtherScientificTerm", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "ProtoTEx faithfully explains model decisions based on prototype tensors that encode latent clusters of training examples .", "forward": true, "src_ids": "2022.acl-long.213_2537"}
+{"input": "propaganda is done by using OtherScientificTerm| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "propaganda", "output": "prototype - based explanations", "neg_sample": ["propaganda is done by using OtherScientificTerm", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "A user study also shows that prototype - based explanations help non - experts to better recognize propaganda in online news .", "forward": false, "src_ids": "2022.acl-long.213_2538"}
+{"input": "non - experts is done by using OtherScientificTerm| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "non - experts", "output": "prototype - based explanations", "neg_sample": ["non - experts is done by using OtherScientificTerm", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "A user study also shows that prototype - based explanations help non - experts to better recognize propaganda in online news .", "forward": false, "src_ids": "2022.acl-long.213_2539"}
+{"input": "prototype - based explanations is used for OtherScientificTerm| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "prototype - based explanations", "output": "non - experts", "neg_sample": ["prototype - based explanations is used for OtherScientificTerm", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "A user study also shows that prototype - based explanations help non - experts to better recognize propaganda in online news .", "forward": true, "src_ids": "2022.acl-long.213_2540"}
+{"input": "non - experts is used for OtherScientificTerm| context: we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al . , 2018 ) .", "entity": "non - experts", "output": "propaganda", "neg_sample": ["non - experts is used for OtherScientificTerm", "we present prototex , a novel white - box nlp classification architecture based on prototype networks ( li et al .", ", 2018 ) ."], "relation": "used for", "id": "2022.acl-long.213", "year": 2022, "rel_sent": "A user study also shows that prototype - based explanations help non - experts to better recognize propaganda in online news .", "forward": true, "src_ids": "2022.acl-long.213_2541"}
+{"input": "controlled text generation is done by using Metric| context: existing reference - free metrics have obvious limitations for evaluating controlled text generation models . unsupervised metrics can only provide a task - agnostic evaluation result which correlates weakly with human judgments , whereas supervised ones may overfit task - specific data with poor generalization ability to other datasets .", "entity": "controlled text generation", "output": "unsupervised reference - free metric", "neg_sample": ["controlled text generation is done by using Metric", "existing reference - free metrics have obvious limitations for evaluating controlled text generation models .", "unsupervised metrics can only provide a task - agnostic evaluation result which correlates weakly with human judgments , whereas supervised ones may overfit task - specific data with poor generalization ability to other datasets ."], "relation": "used for", "id": "2022.acl-long.164", "year": 2022, "rel_sent": "In this paper , we propose an unsupervised reference - free metric called CTRLEval , which evaluates controlled text generation from different aspects by formulating each aspect into multiple text infilling tasks .", "forward": false, "src_ids": "2022.acl-long.164_2542"}
+{"input": "unsupervised reference - free metric is used for Task| context: unsupervised metrics can only provide a task - agnostic evaluation result which correlates weakly with human judgments , whereas supervised ones may overfit task - specific data with poor generalization ability to other datasets .", "entity": "unsupervised reference - free metric", "output": "controlled text generation", "neg_sample": ["unsupervised reference - free metric is used for Task", "unsupervised metrics can only provide a task - agnostic evaluation result which correlates weakly with human judgments , whereas supervised ones may overfit task - specific data with poor generalization ability to other datasets ."], "relation": "used for", "id": "2022.acl-long.164", "year": 2022, "rel_sent": "In this paper , we propose an unsupervised reference - free metric called CTRLEval , which evaluates controlled text generation from different aspects by formulating each aspect into multiple text infilling tasks .", "forward": true, "src_ids": "2022.acl-long.164_2543"}
+{"input": "static and contextualised sense embeddings is used for OtherScientificTerm| context: sense embedding learning methods learn different embeddings for the different senses of an ambiguous word . one sense of an ambiguous word might be socially biased while its other senses remain unbiased .", "entity": "static and contextualised sense embeddings", "output": "social biases", "neg_sample": ["static and contextualised sense embeddings is used for OtherScientificTerm", "sense embedding learning methods learn different embeddings for the different senses of an ambiguous word .", "one sense of an ambiguous word might be socially biased while its other senses remain unbiased ."], "relation": "used for", "id": "2022.acl-long.135", "year": 2022, "rel_sent": "We conduct an extensive evaluation of multiple static and contextualised sense embeddings for various types of social biases using the proposed measures .", "forward": true, "src_ids": "2022.acl-long.135_2544"}
+{"input": "social biases is done by using Method| context: sense embedding learning methods learn different embeddings for the different senses of an ambiguous word . one sense of an ambiguous word might be socially biased while its other senses remain unbiased . in comparison to the numerous prior work evaluating the social biases in pretrained word embeddings , the biases in sense embeddings have been relatively understudied .", "entity": "social biases", "output": "static and contextualised sense embeddings", "neg_sample": ["social biases is done by using Method", "sense embedding learning methods learn different embeddings for the different senses of an ambiguous word .", "one sense of an ambiguous word might be socially biased while its other senses remain unbiased .", "in comparison to the numerous prior work evaluating the social biases in pretrained word embeddings , the biases in sense embeddings have been relatively understudied ."], "relation": "used for", "id": "2022.acl-long.135", "year": 2022, "rel_sent": "We conduct an extensive evaluation of multiple static and contextualised sense embeddings for various types of social biases using the proposed measures .", "forward": false, "src_ids": "2022.acl-long.135_2545"}
+{"input": "argument mining is done by using Task| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "argument mining", "output": "unsupervised knowledge transfer", "neg_sample": ["argument mining is done by using Task", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "Can Unsupervised Knowledge Transfer from Social Discussions Help Argument Mining ?.", "forward": false, "src_ids": "2022.acl-long.536_2546"}
+{"input": "unsupervised knowledge transfer is used for Task| context: the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "unsupervised knowledge transfer", "output": "argument mining", "neg_sample": ["unsupervised knowledge transfer is used for Task", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "Can Unsupervised Knowledge Transfer from Social Discussions Help Argument Mining ?.", "forward": true, "src_ids": "2022.acl-long.536_2547"}
+{"input": "unsupervised , argumentative discourse - aware knowledge is done by using Material| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "unsupervised , argumentative discourse - aware knowledge", "output": "argumentation - rich social discussions", "neg_sample": ["unsupervised , argumentative discourse - aware knowledge is done by using Material", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "We utilize argumentation - rich social discussions from the ChangeMyView subreddit as a source of unsupervised , argumentative discourse - aware knowledge by finetuning pretrained LMs on a selectively masked language modeling task .", "forward": false, "src_ids": "2022.acl-long.536_2548"}
+{"input": "argumentation - rich social discussions is used for OtherScientificTerm| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "argumentation - rich social discussions", "output": "unsupervised , argumentative discourse - aware knowledge", "neg_sample": ["argumentation - rich social discussions is used for OtherScientificTerm", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "We utilize argumentation - rich social discussions from the ChangeMyView subreddit as a source of unsupervised , argumentative discourse - aware knowledge by finetuning pretrained LMs on a selectively masked language modeling task .", "forward": true, "src_ids": "2022.acl-long.536_2549"}
+{"input": "selectively masked language modeling task is done by using Method| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "selectively masked language modeling task", "output": "finetuning pretrained lms", "neg_sample": ["selectively masked language modeling task is done by using Method", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "We utilize argumentation - rich social discussions from the ChangeMyView subreddit as a source of unsupervised , argumentative discourse - aware knowledge by finetuning pretrained LMs on a selectively masked language modeling task .", "forward": false, "src_ids": "2022.acl-long.536_2550"}
+{"input": "finetuning pretrained lms is used for Task| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "finetuning pretrained lms", "output": "selectively masked language modeling task", "neg_sample": ["finetuning pretrained lms is used for Task", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "We utilize argumentation - rich social discussions from the ChangeMyView subreddit as a source of unsupervised , argumentative discourse - aware knowledge by finetuning pretrained LMs on a selectively masked language modeling task .", "forward": true, "src_ids": "2022.acl-long.536_2551"}
+{"input": "inter - component relation prediction is done by using Method| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "inter - component relation prediction", "output": "prompt - based strategy", "neg_sample": ["inter - component relation prediction is done by using Method", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "Furthermore , we introduce a novel prompt - based strategy for inter - component relation prediction that compliments our proposed finetuning method while leveraging on the discourse context .", "forward": false, "src_ids": "2022.acl-long.536_2552"}
+{"input": "prompt - based strategy is used for Task| context: identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining . the intrinsic complexity of these tasks demands powerful learning models . while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models .", "entity": "prompt - based strategy", "output": "inter - component relation prediction", "neg_sample": ["prompt - based strategy is used for Task", "identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining .", "the intrinsic complexity of these tasks demands powerful learning models .", "while pretrained transformer - based language models ( lm ) have been shown to provide state - of - the - art results over different nlp tasks , the scarcity of manually annotated data and the highly domain - dependent nature of argumentation restrict the capabilities of such models ."], "relation": "used for", "id": "2022.acl-long.536", "year": 2022, "rel_sent": "Furthermore , we introduce a novel prompt - based strategy for inter - component relation prediction that compliments our proposed finetuning method while leveraging on the discourse context .", "forward": true, "src_ids": "2022.acl-long.536_2553"}
+{"input": "sequence - to - sequence models is used for Method| context: pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) . different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt .", "entity": "sequence - to - sequence models", "output": "cemat", "neg_sample": ["sequence - to - sequence models is used for Method", "pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) .", "different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt ."], "relation": "used for", "id": "2022.acl-long.442", "year": 2022, "rel_sent": "We also introduce two simple but effective methods to enhance the CeMAT , aligned code - switching & masking and dynamic dual - masking .", "forward": true, "src_ids": "2022.acl-long.442_2554"}
+{"input": "fine - tuning is done by using Method| context: pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) . different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt .", "entity": "fine - tuning", "output": "unified model", "neg_sample": ["fine - tuning is done by using Method", "pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) .", "different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt ."], "relation": "used for", "id": "2022.acl-long.442", "year": 2022, "rel_sent": "To the best of our knowledge , this is the first work to pre - train a unified model for fine - tuning on both NMT tasks .", "forward": false, "src_ids": "2022.acl-long.442_2555"}
+{"input": "nmt tasks is done by using Method| context: pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) . different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt .", "entity": "nmt tasks", "output": "unified model", "neg_sample": ["nmt tasks is done by using Method", "pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) .", "different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt ."], "relation": "used for", "id": "2022.acl-long.442", "year": 2022, "rel_sent": "To the best of our knowledge , this is the first work to pre - train a unified model for fine - tuning on both NMT tasks .", "forward": false, "src_ids": "2022.acl-long.442_2556"}
+{"input": "unified model is used for Method| context: pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) . different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt .", "entity": "unified model", "output": "fine - tuning", "neg_sample": ["unified model is used for Method", "pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) .", "different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt ."], "relation": "used for", "id": "2022.acl-long.442", "year": 2022, "rel_sent": "To the best of our knowledge , this is the first work to pre - train a unified model for fine - tuning on both NMT tasks .", "forward": true, "src_ids": "2022.acl-long.442_2557"}
+{"input": "unified model is used for Task| context: pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) . different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt .", "entity": "unified model", "output": "nmt tasks", "neg_sample": ["unified model is used for Task", "pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) .", "different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt ."], "relation": "used for", "id": "2022.acl-long.442", "year": 2022, "rel_sent": "To the best of our knowledge , this is the first work to pre - train a unified model for fine - tuning on both NMT tasks .", "forward": true, "src_ids": "2022.acl-long.442_2558"}
+{"input": "fine - tuning is used for Task| context: pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) . different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt .", "entity": "fine - tuning", "output": "nmt tasks", "neg_sample": ["fine - tuning is used for Task", "pre - trained sequence - to - sequence models have significantly improved neural machine translation ( nmt ) .", "different from prior works where pre - trained models usually adopt an unidirectional decoder , this paper demonstrates that pre - training a sequence - to - sequence model but with a bidirectional decoder can produce notable performance gains for both autoregressive and non - autoregressive nmt ."], "relation": "used for", "id": "2022.acl-long.442", "year": 2022, "rel_sent": "To the best of our knowledge , this is the first work to pre - train a unified model for fine - tuning on both NMT tasks .", "forward": true, "src_ids": "2022.acl-long.442_2559"}
+{"input": "sentence representations is done by using Method| context: learning high - quality sentence representations is a fundamental problem of natural language processing which could benefit a wide range of downstream tasks . though the bert - like pre - trained language models have achieved great success , using their sentence representations directly often results in poor performance on the semantic textual similarity task . recently , several contrastive learning methods have been proposed for learning sentence representations and have shown promising results . however , most of them focus on the constitution of positive and negative representation pairs and pay little attention to the training objective like nt - xent , which is not sufficient enough to acquire the discriminating power and is unable to model the partial order of semantics between sentences .", "entity": "sentence representations", "output": "contrastive framework", "neg_sample": ["sentence representations is done by using Method", "learning high - quality sentence representations is a fundamental problem of natural language processing which could benefit a wide range of downstream tasks .", "though the bert - like pre - trained language models have achieved great success , using their sentence representations directly often results in poor performance on the semantic textual similarity task .", "recently , several contrastive learning methods have been proposed for learning sentence representations and have shown promising results .", "however , most of them focus on the constitution of positive and negative representation pairs and pay little attention to the training objective like nt - xent , which is not sufficient enough to acquire the discriminating power and is unable to model the partial order of semantics between sentences ."], "relation": "used for", "id": "2022.acl-long.336", "year": 2022, "rel_sent": "A Contrastive Framework for Learning Sentence Representations from Pairwise and Triple - wise Perspective in Angular Space.", "forward": false, "src_ids": "2022.acl-long.336_2560"}
+{"input": "contrastive framework is used for Method| context: however , most of them focus on the constitution of positive and negative representation pairs and pay little attention to the training objective like nt - xent , which is not sufficient enough to acquire the discriminating power and is unable to model the partial order of semantics between sentences .", "entity": "contrastive framework", "output": "sentence representations", "neg_sample": ["contrastive framework is used for Method", "however , most of them focus on the constitution of positive and negative representation pairs and pay little attention to the training objective like nt - xent , which is not sufficient enough to acquire the discriminating power and is unable to model the partial order of semantics between sentences ."], "relation": "used for", "id": "2022.acl-long.336", "year": 2022, "rel_sent": "A Contrastive Framework for Learning Sentence Representations from Pairwise and Triple - wise Perspective in Angular Space.", "forward": true, "src_ids": "2022.acl-long.336_2561"}
+{"input": "pairwise discriminative power is done by using OtherScientificTerm| context: learning high - quality sentence representations is a fundamental problem of natural language processing which could benefit a wide range of downstream tasks . though the bert - like pre - trained language models have achieved great success , using their sentence representations directly often results in poor performance on the semantic textual similarity task . recently , several contrastive learning methods have been proposed for learning sentence representations and have shown promising results .", "entity": "pairwise discriminative power", "output": "training objectives", "neg_sample": ["pairwise discriminative power is done by using OtherScientificTerm", "learning high - quality sentence representations is a fundamental problem of natural language processing which could benefit a wide range of downstream tasks .", "though the bert - like pre - trained language models have achieved great success , using their sentence representations directly often results in poor performance on the semantic textual similarity task .", "recently , several contrastive learning methods have been proposed for learning sentence representations and have shown promising results ."], "relation": "used for", "id": "2022.acl-long.336", "year": 2022, "rel_sent": "So in this paper , we propose a new method ArcCSE , with training objectives designed to enhance the pairwise discriminative power and model the entailment relation of triplet sentences .", "forward": false, "src_ids": "2022.acl-long.336_2562"}
+{"input": "training objectives is used for OtherScientificTerm| context: learning high - quality sentence representations is a fundamental problem of natural language processing which could benefit a wide range of downstream tasks . though the bert - like pre - trained language models have achieved great success , using their sentence representations directly often results in poor performance on the semantic textual similarity task . recently , several contrastive learning methods have been proposed for learning sentence representations and have shown promising results . however , most of them focus on the constitution of positive and negative representation pairs and pay little attention to the training objective like nt - xent , which is not sufficient enough to acquire the discriminating power and is unable to model the partial order of semantics between sentences .", "entity": "training objectives", "output": "pairwise discriminative power", "neg_sample": ["training objectives is used for OtherScientificTerm", "learning high - quality sentence representations is a fundamental problem of natural language processing which could benefit a wide range of downstream tasks .", "though the bert - like pre - trained language models have achieved great success , using their sentence representations directly often results in poor performance on the semantic textual similarity task .", "recently , several contrastive learning methods have been proposed for learning sentence representations and have shown promising results .", "however , most of them focus on the constitution of positive and negative representation pairs and pay little attention to the training objective like nt - xent , which is not sufficient enough to acquire the discriminating power and is unable to model the partial order of semantics between sentences ."], "relation": "used for", "id": "2022.acl-long.336", "year": 2022, "rel_sent": "So in this paper , we propose a new method ArcCSE , with training objectives designed to enhance the pairwise discriminative power and model the entailment relation of triplet sentences .", "forward": true, "src_ids": "2022.acl-long.336_2563"}
+{"input": "story completion is done by using Method| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "story completion", "output": "plug - and - play controller", "neg_sample": ["story completion is done by using Method", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "Plug - and - Play Controller for Story Completion : A Pilot Study toward Emotion - aware Story Writing Assistance.", "forward": false, "src_ids": "2022.in2writing-1.6_2564"}
+{"input": "plug - and - play controller is used for Task| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "plug - and - play controller", "output": "story completion", "neg_sample": ["plug - and - play controller is used for Task", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "Plug - and - Play Controller for Story Completion : A Pilot Study toward Emotion - aware Story Writing Assistance.", "forward": true, "src_ids": "2022.in2writing-1.6_2565"}
+{"input": "novelists is done by using Task| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "novelists", "output": "natural language processing", "neg_sample": ["novelists is done by using Task", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "The authors of this paper , including a professional novelist , have examined the use of natural language processing to address the problems of novelists from the perspective of practical creative writing .", "forward": false, "src_ids": "2022.in2writing-1.6_2566"}
+{"input": "practical creative writing is done by using Task| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "practical creative writing", "output": "natural language processing", "neg_sample": ["practical creative writing is done by using Task", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "The authors of this paper , including a professional novelist , have examined the use of natural language processing to address the problems of novelists from the perspective of practical creative writing .", "forward": false, "src_ids": "2022.in2writing-1.6_2567"}
+{"input": "natural language processing is used for OtherScientificTerm| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "natural language processing", "output": "novelists", "neg_sample": ["natural language processing is used for OtherScientificTerm", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "The authors of this paper , including a professional novelist , have examined the use of natural language processing to address the problems of novelists from the perspective of practical creative writing .", "forward": true, "src_ids": "2022.in2writing-1.6_2568"}
+{"input": "natural language processing is used for Task| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "natural language processing", "output": "practical creative writing", "neg_sample": ["natural language processing is used for Task", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "The authors of this paper , including a professional novelist , have examined the use of natural language processing to address the problems of novelists from the perspective of practical creative writing .", "forward": true, "src_ids": "2022.in2writing-1.6_2569"}
+{"input": "controllable text generation is done by using Method| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "controllable text generation", "output": "plug - and - play module", "neg_sample": ["controllable text generation is done by using Method", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "Furthermore , based on the plug - and - play module for controllable text generation using GPT-2 , an additional module was implemented to consider emotions .", "forward": false, "src_ids": "2022.in2writing-1.6_2570"}
+{"input": "plug - and - play module is used for Task| context: emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied .", "entity": "plug - and - play module", "output": "controllable text generation", "neg_sample": ["plug - and - play module is used for Task", "emotions are essential for storytelling and narrative generation , and as such , the relationship between stories and emotions has been extensively studied ."], "relation": "used for", "id": "2022.in2writing-1.6", "year": 2022, "rel_sent": "Furthermore , based on the plug - and - play module for controllable text generation using GPT-2 , an additional module was implemented to consider emotions .", "forward": true, "src_ids": "2022.in2writing-1.6_2571"}
+{"input": "augmented zero - shot learning is used for Task| context: in this paper , we leverage large language models ( llms ) to perform zero - shot text style transfer .", "entity": "augmented zero - shot learning", "output": "style transfer tasks", "neg_sample": ["augmented zero - shot learning is used for Task", "in this paper , we leverage large language models ( llms ) to perform zero - shot text style transfer ."], "relation": "used for", "id": "2022.acl-short.94", "year": 2022, "rel_sent": "Augmented zero - shot learning is simple and demonstrates promising results not just on standard style transfer tasks such as sentiment , but also on arbitrary transformations such as ' make this melodramatic ' or ' insert a metaphor . '", "forward": true, "src_ids": "2022.acl-short.94_2572"}
+{"input": "style transfer tasks is done by using Method| context: in this paper , we leverage large language models ( llms ) to perform zero - shot text style transfer .", "entity": "style transfer tasks", "output": "augmented zero - shot learning", "neg_sample": ["style transfer tasks is done by using Method", "in this paper , we leverage large language models ( llms ) to perform zero - shot text style transfer ."], "relation": "used for", "id": "2022.acl-short.94", "year": 2022, "rel_sent": "Augmented zero - shot learning is simple and demonstrates promising results not just on standard style transfer tasks such as sentiment , but also on arbitrary transformations such as ' make this melodramatic ' or ' insert a metaphor . '", "forward": false, "src_ids": "2022.acl-short.94_2573"}
+{"input": "st graphs is done by using Method| context: predicting the effects of unexpected situations is an important reasoning task , e.g. , would cloudy skies help or hinder plant growth ?", "entity": "st graphs", "output": "curie", "neg_sample": ["st graphs is done by using Method", "predicting the effects of unexpected situations is an important reasoning task , e.g.", ", would cloudy skies help or hinder plant growth ?"], "relation": "used for", "id": "2022.csrr-1.7", "year": 2022, "rel_sent": "Across multiple domains , CURIE generates st graphs that humans find relevant and meaningful in eliciting the consequences of a new situation ( 75 % of the graphs were judged correct by humans ) .", "forward": false, "src_ids": "2022.csrr-1.7_2574"}
+{"input": "curie is used for OtherScientificTerm| context: predicting the effects of unexpected situations is an important reasoning task , e.g. , would cloudy skies help or hinder plant growth ?", "entity": "curie", "output": "st graphs", "neg_sample": ["curie is used for OtherScientificTerm", "predicting the effects of unexpected situations is an important reasoning task , e.g.", ", would cloudy skies help or hinder plant growth ?"], "relation": "used for", "id": "2022.csrr-1.7", "year": 2022, "rel_sent": "Across multiple domains , CURIE generates st graphs that humans find relevant and meaningful in eliciting the consequences of a new situation ( 75 % of the graphs were judged correct by humans ) .", "forward": true, "src_ids": "2022.csrr-1.7_2575"}
+{"input": "inner structure of metamorphic relations is done by using Method| context: metamorphic testing has recently been used to check the safety of neural nlp models . its main advantage is that it does not rely on a ground truth to generate test cases . however , existing studies are mostly concerned with robustness - like metamorphic relations , limiting the scope of linguistic properties they can test .", "entity": "inner structure of metamorphic relations", "output": "graphical notation", "neg_sample": ["inner structure of metamorphic relations is done by using Method", "metamorphic testing has recently been used to check the safety of neural nlp models .", "its main advantage is that it does not rely on a ground truth to generate test cases .", "however , existing studies are mostly concerned with robustness - like metamorphic relations , limiting the scope of linguistic properties they can test ."], "relation": "used for", "id": "2022.findings-acl.185", "year": 2022, "rel_sent": "Lastly , we introduce a novel graphical notation that efficiently summarises the inner structure of metamorphic relations .", "forward": false, "src_ids": "2022.findings-acl.185_2576"}
+{"input": "graphical notation is used for OtherScientificTerm| context: metamorphic testing has recently been used to check the safety of neural nlp models . its main advantage is that it does not rely on a ground truth to generate test cases . however , existing studies are mostly concerned with robustness - like metamorphic relations , limiting the scope of linguistic properties they can test .", "entity": "graphical notation", "output": "inner structure of metamorphic relations", "neg_sample": ["graphical notation is used for OtherScientificTerm", "metamorphic testing has recently been used to check the safety of neural nlp models .", "its main advantage is that it does not rely on a ground truth to generate test cases .", "however , existing studies are mostly concerned with robustness - like metamorphic relations , limiting the scope of linguistic properties they can test ."], "relation": "used for", "id": "2022.findings-acl.185", "year": 2022, "rel_sent": "Lastly , we introduce a novel graphical notation that efficiently summarises the inner structure of metamorphic relations .", "forward": true, "src_ids": "2022.findings-acl.185_2577"}
+{"input": "self - supervised tasks is used for Method| context: in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort .", "entity": "self - supervised tasks", "output": "representation learning", "neg_sample": ["self - supervised tasks is used for Method", "in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort ."], "relation": "used for", "id": "2022.acl-long.596", "year": 2022, "rel_sent": "In addition , several self - supervised tasks are proposed based on the information tree to improve the representation learning under insufficient labeling .", "forward": true, "src_ids": "2022.acl-long.596_2578"}
+{"input": "one - shot natural language spatial video grounding is done by using Method| context: in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort . another challenge relates to the limited supervision , which might result in ineffective representation learning .", "entity": "one - shot natural language spatial video grounding", "output": "information tree", "neg_sample": ["one - shot natural language spatial video grounding is done by using Method", "in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort .", "another challenge relates to the limited supervision , which might result in ineffective representation learning ."], "relation": "used for", "id": "2022.acl-long.596", "year": 2022, "rel_sent": "End - to - End Modeling via Information Tree for One - Shot Natural Language Spatial Video Grounding.", "forward": false, "src_ids": "2022.acl-long.596_2579"}
+{"input": "information tree is used for Task| context: in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort . another challenge relates to the limited supervision , which might result in ineffective representation learning .", "entity": "information tree", "output": "one - shot natural language spatial video grounding", "neg_sample": ["information tree is used for Task", "in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort .", "another challenge relates to the limited supervision , which might result in ineffective representation learning ."], "relation": "used for", "id": "2022.acl-long.596", "year": 2022, "rel_sent": "End - to - End Modeling via Information Tree for One - Shot Natural Language Spatial Video Grounding.", "forward": true, "src_ids": "2022.acl-long.596_2580"}
+{"input": "representation learning is done by using Task| context: in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort . another challenge relates to the limited supervision , which might result in ineffective representation learning .", "entity": "representation learning", "output": "self - supervised tasks", "neg_sample": ["representation learning is done by using Task", "in spite of the great advances , most existing methods rely on dense videoframe annotations , which require a tremendous amount of human effort .", "another challenge relates to the limited supervision , which might result in ineffective representation learning ."], "relation": "used for", "id": "2022.acl-long.596", "year": 2022, "rel_sent": "In addition , several self - supervised tasks are proposed based on the information tree to improve the representation learning under insufficient labeling .", "forward": false, "src_ids": "2022.acl-long.596_2581"}
+{"input": "paraphrase generation is done by using Method| context: paraphrase generation using deep learning has been a research hotspot of natural language processing in the past few years . while previous studies tackle the problem from different aspects , the essence of paraphrase generation is to retain the key semantics of the source sentence and rewrite the rest of the content .", "entity": "paraphrase generation", "output": "multi - task learning", "neg_sample": ["paraphrase generation is done by using Method", "paraphrase generation using deep learning has been a research hotspot of natural language processing in the past few years .", "while previous studies tackle the problem from different aspects , the essence of paraphrase generation is to retain the key semantics of the source sentence and rewrite the rest of the content ."], "relation": "used for", "id": "2022.findings-acl.97", "year": 2022, "rel_sent": "Multi - task Learning for Paraphrase Generation With Keyword and Part - of - Speech Reconstruction.", "forward": false, "src_ids": "2022.findings-acl.97_2582"}
+{"input": "paraphrase generation is done by using Method| context: paraphrase generation using deep learning has been a research hotspot of natural language processing in the past few years . while previous studies tackle the problem from different aspects , the essence of paraphrase generation is to retain the key semantics of the source sentence and rewrite the rest of the content .", "entity": "paraphrase generation", "output": "multi - task learning", "neg_sample": ["paraphrase generation is done by using Method", "paraphrase generation using deep learning has been a research hotspot of natural language processing in the past few years .", "while previous studies tackle the problem from different aspects , the essence of paraphrase generation is to retain the key semantics of the source sentence and rewrite the rest of the content ."], "relation": "used for", "id": "2022.findings-acl.97", "year": 2022, "rel_sent": "In the second stage , we train a transformer - based model via multi - task learning for paraphrase generation .", "forward": false, "src_ids": "2022.findings-acl.97_2583"}
+{"input": "paraphrase generation is done by using Method| context: paraphrase generation using deep learning has been a research hotspot of natural language processing in the past few years . while previous studies tackle the problem from different aspects , the essence of paraphrase generation is to retain the key semantics of the source sentence and rewrite the rest of the content .", "entity": "paraphrase generation", "output": "pgkpr", "neg_sample": ["paraphrase generation is done by using Method", "paraphrase generation using deep learning has been a research hotspot of natural language processing in the past few years .", "while previous studies tackle the problem from different aspects , the essence of paraphrase generation is to retain the key semantics of the source sentence and rewrite the rest of the content ."], "relation": "used for", "id": "2022.findings-acl.97", "year": 2022, "rel_sent": "Inspired by this observation , we propose a novel two - stage model , PGKPR , for paraphrase generation with keyword and part - of - speech reconstruction .", "forward": false, "src_ids": "2022.findings-acl.97_2584"}
+{"input": "ml models is used for Task| context: despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e. , the way people split datasets into training , validation , and test sets , were not well studied . this may lead to evaluations that are inconsistent with the intended use cases .", "entity": "ml models", "output": "code summarization tasks", "neg_sample": ["ml models is used for Task", "despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e.", ", the way people split datasets into training , validation , and test sets , were not well studied .", "this may lead to evaluations that are inconsistent with the intended use cases ."], "relation": "used for", "id": "2022.acl-long.339", "year": 2022, "rel_sent": "Each methodology can be mapped to some use cases , and the time - segmented methodology should be adopted in the evaluation of ML models for code summarization .", "forward": true, "src_ids": "2022.acl-long.339_2585"}
+{"input": "ml models is used for Task| context: despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e. , the way people split datasets into training , validation , and test sets , were not well studied . this may lead to evaluations that are inconsistent with the intended use cases .", "entity": "ml models", "output": "code summarization tasks", "neg_sample": ["ml models is used for Task", "despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e.", ", the way people split datasets into training , validation , and test sets , were not well studied .", "this may lead to evaluations that are inconsistent with the intended use cases ."], "relation": "used for", "id": "2022.acl-long.339", "year": 2022, "rel_sent": "To assess the impact of methodologies , we collect a dataset of ( code , comment ) pairs with timestamps to train and evaluate several recent ML models for code summarization .", "forward": true, "src_ids": "2022.acl-long.339_2586"}
+{"input": "code summarization tasks is done by using Method| context: there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g. , comment generation and method naming . specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation . this may lead to evaluations that are inconsistent with the intended use cases .", "entity": "code summarization tasks", "output": "ml models", "neg_sample": ["code summarization tasks is done by using Method", "there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g.", ", comment generation and method naming .", "specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation .", "this may lead to evaluations that are inconsistent with the intended use cases ."], "relation": "used for", "id": "2022.acl-long.339", "year": 2022, "rel_sent": "Each methodology can be mapped to some use cases , and the time - segmented methodology should be adopted in the evaluation of ML models for code summarization .", "forward": false, "src_ids": "2022.acl-long.339_2587"}
+{"input": "code summarization tasks is done by using Method| context: there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g. , comment generation and method naming . specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation . this may lead to evaluations that are inconsistent with the intended use cases .", "entity": "code summarization tasks", "output": "ml models", "neg_sample": ["code summarization tasks is done by using Method", "there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g.", ", comment generation and method naming .", "specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation .", "this may lead to evaluations that are inconsistent with the intended use cases ."], "relation": "used for", "id": "2022.acl-long.339", "year": 2022, "rel_sent": "To assess the impact of methodologies , we collect a dataset of ( code , comment ) pairs with timestamps to train and evaluate several recent ML models for code summarization .", "forward": false, "src_ids": "2022.acl-long.339_2588"}
+{"input": "code summarization research community is done by using Method| context: there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g. , comment generation and method naming . despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e. , the way people split datasets into training , validation , and test sets , were not well studied . specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation . this may lead to evaluations that are inconsistent with the intended use cases .", "entity": "code summarization research community", "output": "time - segmented evaluation methodology", "neg_sample": ["code summarization research community is done by using Method", "there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g.", ", comment generation and method naming .", "despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e.", ", the way people split datasets into training , validation , and test sets , were not well studied .", "specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation .", "this may lead to evaluations that are inconsistent with the intended use cases ."], "relation": "used for", "id": "2022.acl-long.339", "year": 2022, "rel_sent": "In this paper , we introduce the time - segmented evaluation methodology , which is novel to the code summarization research community , and compare it with the mixed - project and cross - project methodologies that have been commonly used .", "forward": false, "src_ids": "2022.acl-long.339_2589"}
+{"input": "time - segmented evaluation methodology is used for Task| context: there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g. , comment generation and method naming . despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e. , the way people split datasets into training , validation , and test sets , were not well studied . specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation . this may lead to evaluations that are inconsistent with the intended use cases .", "entity": "time - segmented evaluation methodology", "output": "code summarization research community", "neg_sample": ["time - segmented evaluation methodology is used for Task", "there has been a growing interest in developing machine learning ( ml ) models for code summarization tasks , e.g.", ", comment generation and method naming .", "despite substantial increase in the effectiveness of ml models , the evaluation methodologies , i.e.", ", the way people split datasets into training , validation , and test sets , were not well studied .", "specifically , no prior work on code summarization considered the timestamps of code and comments during evaluation .", "this may lead to evaluations that are inconsistent with the intended use cases ."], "relation": "used for", "id": "2022.acl-long.339", "year": 2022, "rel_sent": "In this paper , we introduce the time - segmented evaluation methodology , which is novel to the code summarization research community , and compare it with the mixed - project and cross - project methodologies that have been commonly used .", "forward": true, "src_ids": "2022.acl-long.339_2590"}
+{"input": "label noise is done by using Method| context: incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision . it has been shown that complex noise - handling techniques - by modeling , cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise .", "entity": "label noise", "output": "bert", "neg_sample": ["label noise is done by using Method", "incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision .", "it has been shown that complex noise - handling techniques - by modeling , cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise ."], "relation": "used for", "id": "2022.insights-1.8", "year": 2022, "rel_sent": "Is BERT Robust to Label Noise ? A Study on Learning with Noisy Labels in Text Classification.", "forward": false, "src_ids": "2022.insights-1.8_2591"}
+{"input": "bert is used for OtherScientificTerm| context: incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision .", "entity": "bert", "output": "label noise", "neg_sample": ["bert is used for OtherScientificTerm", "incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision ."], "relation": "used for", "id": "2022.insights-1.8", "year": 2022, "rel_sent": "Is BERT Robust to Label Noise ? A Study on Learning with Noisy Labels in Text Classification.", "forward": true, "src_ids": "2022.insights-1.8_2592"}
+{"input": "text classification is done by using Task| context: incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision . it has been shown that complex noise - handling techniques - by modeling , cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise .", "entity": "text classification", "output": "learning with noisy labels", "neg_sample": ["text classification is done by using Task", "incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision .", "it has been shown that complex noise - handling techniques - by modeling , cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise ."], "relation": "used for", "id": "2022.insights-1.8", "year": 2022, "rel_sent": "Is BERT Robust to Label Noise ? A Study on Learning with Noisy Labels in Text Classification.", "forward": false, "src_ids": "2022.insights-1.8_2593"}
+{"input": "learning with noisy labels is used for Task| context: incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision . it has been shown that complex noise - handling techniques - by modeling , cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise .", "entity": "learning with noisy labels", "output": "text classification", "neg_sample": ["learning with noisy labels is used for Task", "incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision .", "it has been shown that complex noise - handling techniques - by modeling , cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise ."], "relation": "used for", "id": "2022.insights-1.8", "year": 2022, "rel_sent": "Is BERT Robust to Label Noise ? A Study on Learning with Noisy Labels in Text Classification.", "forward": true, "src_ids": "2022.insights-1.8_2594"}
+{"input": "induction of dementia - related linguistic anomalies is done by using Method| context: deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) . however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research .", "entity": "induction of dementia - related linguistic anomalies", "output": "gpt - d )", "neg_sample": ["induction of dementia - related linguistic anomalies is done by using Method", "deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) .", "however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research ."], "relation": "used for", "id": "2022.acl-long.131", "year": 2022, "rel_sent": "GPT - D : Inducing Dementia - related Linguistic Anomalies by Deliberate Degradation of Artificial Neural Language Models.", "forward": false, "src_ids": "2022.acl-long.131_2595"}
+{"input": "induction of dementia - related linguistic anomalies is done by using Method| context: deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) . however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research .", "entity": "induction of dementia - related linguistic anomalies", "output": "gpt - d )", "neg_sample": ["induction of dementia - related linguistic anomalies is done by using Method", "deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) .", "however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research ."], "relation": "used for", "id": "2022.acl-long.131", "year": 2022, "rel_sent": "Furthermore , GPT - D generates text with characteristics known to be associated with AD , demonstrating the induction of dementia - related linguistic anomalies .", "forward": false, "src_ids": "2022.acl-long.131_2596"}
+{"input": "gpt - d ) is used for Task| context: deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) . however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research .", "entity": "gpt - d )", "output": "induction of dementia - related linguistic anomalies", "neg_sample": ["gpt - d ) is used for Task", "deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) .", "however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research ."], "relation": "used for", "id": "2022.acl-long.131", "year": 2022, "rel_sent": "GPT - D : Inducing Dementia - related Linguistic Anomalies by Deliberate Degradation of Artificial Neural Language Models.", "forward": true, "src_ids": "2022.acl-long.131_2597"}
+{"input": "gpt - d ) is used for Task| context: deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) . however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research .", "entity": "gpt - d )", "output": "induction of dementia - related linguistic anomalies", "neg_sample": ["gpt - d ) is used for Task", "deep learning ( dl ) techniques involving fine - tuning large numbers of model parameters have delivered impressive performance on the task of discriminating between language produced by cognitively healthy individuals , and those with alzheimer 's disease ( ad ) .", "however , questions remain about their ability to generalize beyond the small reference sets that are publicly available for research ."], "relation": "used for", "id": "2022.acl-long.131", "year": 2022, "rel_sent": "Furthermore , GPT - D generates text with characteristics known to be associated with AD , demonstrating the induction of dementia - related linguistic anomalies .", "forward": true, "src_ids": "2022.acl-long.131_2598"}
+{"input": "multimodal additive late - fusion is done by using OtherScientificTerm| context: in multimodal machine learning , additive late - fusion is a straightforward approach to combine the feature representations from different modalities , in which the final prediction can be formulated as the sum of unimodal predictions . while it has been found that certain late - fusion models can achieve competitive performance with lower computational costs compared to complex multimodal interactive models , how to effectively search for a good late - fusion model is still an open question . moreover , for different modalities , the best unimodal models may work under significantly different learning rates due to the nature of the modality and the computational flow of the model ; thus , selecting a global learning rate for late - fusion models can result in a vanishing gradient for some modalities .", "entity": "multimodal additive late - fusion", "output": "modality - specific learning rates", "neg_sample": ["multimodal additive late - fusion is done by using OtherScientificTerm", "in multimodal machine learning , additive late - fusion is a straightforward approach to combine the feature representations from different modalities , in which the final prediction can be formulated as the sum of unimodal predictions .", "while it has been found that certain late - fusion models can achieve competitive performance with lower computational costs compared to complex multimodal interactive models , how to effectively search for a good late - fusion model is still an open question .", "moreover , for different modalities , the best unimodal models may work under significantly different learning rates due to the nature of the modality and the computational flow of the model ; thus , selecting a global learning rate for late - fusion models can result in a vanishing gradient for some modalities ."], "relation": "used for", "id": "2022.findings-acl.143", "year": 2022, "rel_sent": "Modality - specific Learning Rates for Effective Multimodal Additive Late - fusion.", "forward": false, "src_ids": "2022.findings-acl.143_2599"}
+{"input": "modality - specific learning rates is used for Method| context: in multimodal machine learning , additive late - fusion is a straightforward approach to combine the feature representations from different modalities , in which the final prediction can be formulated as the sum of unimodal predictions . while it has been found that certain late - fusion models can achieve competitive performance with lower computational costs compared to complex multimodal interactive models , how to effectively search for a good late - fusion model is still an open question . moreover , for different modalities , the best unimodal models may work under significantly different learning rates due to the nature of the modality and the computational flow of the model ; thus , selecting a global learning rate for late - fusion models can result in a vanishing gradient for some modalities .", "entity": "modality - specific learning rates", "output": "multimodal additive late - fusion", "neg_sample": ["modality - specific learning rates is used for Method", "in multimodal machine learning , additive late - fusion is a straightforward approach to combine the feature representations from different modalities , in which the final prediction can be formulated as the sum of unimodal predictions .", "while it has been found that certain late - fusion models can achieve competitive performance with lower computational costs compared to complex multimodal interactive models , how to effectively search for a good late - fusion model is still an open question .", "moreover , for different modalities , the best unimodal models may work under significantly different learning rates due to the nature of the modality and the computational flow of the model ; thus , selecting a global learning rate for late - fusion models can result in a vanishing gradient for some modalities ."], "relation": "used for", "id": "2022.findings-acl.143", "year": 2022, "rel_sent": "Modality - specific Learning Rates for Effective Multimodal Additive Late - fusion.", "forward": true, "src_ids": "2022.findings-acl.143_2600"}
+{"input": "emotional load is done by using Method| context: dieting is a behaviour change task that is difficult for many people to conduct successfully . this is due to many factors , including stress and cost . mobile applications offer an alternative to traditional coaching . however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits .", "entity": "emotional load", "output": "tailored communication", "neg_sample": ["emotional load is done by using Method", "dieting is a behaviour change task that is difficult for many people to conduct successfully .", "this is due to many factors , including stress and cost .", "mobile applications offer an alternative to traditional coaching .", "however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits ."], "relation": "used for", "id": "2022.humeval-1.5", "year": 2022, "rel_sent": "Beyond calories : evaluating how tailored communication reduces emotional load in diet - coaching.", "forward": false, "src_ids": "2022.humeval-1.5_2601"}
+{"input": "dieting is done by using Method| context: dieting is a behaviour change task that is difficult for many people to conduct successfully . this is due to many factors , including stress and cost . mobile applications offer an alternative to traditional coaching . however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits .", "entity": "dieting", "output": "tailored communication", "neg_sample": ["dieting is done by using Method", "dieting is a behaviour change task that is difficult for many people to conduct successfully .", "this is due to many factors , including stress and cost .", "mobile applications offer an alternative to traditional coaching .", "however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits ."], "relation": "used for", "id": "2022.humeval-1.5", "year": 2022, "rel_sent": "In this work , we introduce a novel evaluation of the effects that tailored communication can have on the emotional load of dieting .", "forward": false, "src_ids": "2022.humeval-1.5_2602"}
+{"input": "tailored communication is used for OtherScientificTerm| context: dieting is a behaviour change task that is difficult for many people to conduct successfully . this is due to many factors , including stress and cost . mobile applications offer an alternative to traditional coaching . however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits .", "entity": "tailored communication", "output": "emotional load", "neg_sample": ["tailored communication is used for OtherScientificTerm", "dieting is a behaviour change task that is difficult for many people to conduct successfully .", "this is due to many factors , including stress and cost .", "mobile applications offer an alternative to traditional coaching .", "however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits ."], "relation": "used for", "id": "2022.humeval-1.5", "year": 2022, "rel_sent": "Beyond calories : evaluating how tailored communication reduces emotional load in diet - coaching.", "forward": true, "src_ids": "2022.humeval-1.5_2603"}
+{"input": "tailored communication is used for Task| context: this is due to many factors , including stress and cost . mobile applications offer an alternative to traditional coaching . however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits .", "entity": "tailored communication", "output": "dieting", "neg_sample": ["tailored communication is used for Task", "this is due to many factors , including stress and cost .", "mobile applications offer an alternative to traditional coaching .", "however , previous work on apps evaluation only focused on dietary outcomes , ignoring users ' emotional state despite its influence on eating habits ."], "relation": "used for", "id": "2022.humeval-1.5", "year": 2022, "rel_sent": "In this work , we introduce a novel evaluation of the effects that tailored communication can have on the emotional load of dieting .", "forward": true, "src_ids": "2022.humeval-1.5_2604"}
+{"input": "autoregres- sive models is done by using OtherScientificTerm| context: existing pre - trained transformer analysis works usually focus only on one or two model families at a time , overlooking the variability of the architecture and pre - training objectives .", "entity": "autoregres- sive models", "output": "olmpics zero - shot setup", "neg_sample": ["autoregres- sive models is done by using OtherScientificTerm", "existing pre - trained transformer analysis works usually focus only on one or two model families at a time , overlooking the variability of the architecture and pre - training objectives ."], "relation": "used for", "id": "2022.acl-long.227", "year": 2022, "rel_sent": "Additionally , we adapt the oLMpics zero - shot setup for autoregres- sive models and evaluate GPT networks of different sizes .", "forward": false, "src_ids": "2022.acl-long.227_2605"}
+{"input": "olmpics zero - shot setup is used for Method| context: existing pre - trained transformer analysis works usually focus only on one or two model families at a time , overlooking the variability of the architecture and pre - training objectives .", "entity": "olmpics zero - shot setup", "output": "autoregres- sive models", "neg_sample": ["olmpics zero - shot setup is used for Method", "existing pre - trained transformer analysis works usually focus only on one or two model families at a time , overlooking the variability of the architecture and pre - training objectives ."], "relation": "used for", "id": "2022.acl-long.227", "year": 2022, "rel_sent": "Additionally , we adapt the oLMpics zero - shot setup for autoregres- sive models and evaluate GPT networks of different sizes .", "forward": true, "src_ids": "2022.acl-long.227_2606"}
+{"input": "open - domain document visual question answering is done by using Material| context: open - domain question answering has been used in a wide range of applications , such as web search and enterprise search , which usually takes clean texts extracted from various formats of documents ( e.g. , web pages , pdfs , or word documents ) as the information source . however , designing different text extraction approaches is time - consuming and not scalable .", "entity": "open - domain document visual question answering", "output": "chinese dataset", "neg_sample": ["open - domain document visual question answering is done by using Material", "open - domain question answering has been used in a wide range of applications , such as web search and enterprise search , which usually takes clean texts extracted from various formats of documents ( e.g.", ", web pages , pdfs , or word documents ) as the information source .", "however , designing different text extraction approaches is time - consuming and not scalable ."], "relation": "used for", "id": "2022.findings-acl.105", "year": 2022, "rel_sent": "DuReader_vis : A Chinese Dataset for Open - domain Document Visual Question Answering.", "forward": false, "src_ids": "2022.findings-acl.105_2607"}
+{"input": "chinese dataset is used for Task| context: open - domain question answering has been used in a wide range of applications , such as web search and enterprise search , which usually takes clean texts extracted from various formats of documents ( e.g. , web pages , pdfs , or word documents ) as the information source . however , designing different text extraction approaches is time - consuming and not scalable .", "entity": "chinese dataset", "output": "open - domain document visual question answering", "neg_sample": ["chinese dataset is used for Task", "open - domain question answering has been used in a wide range of applications , such as web search and enterprise search , which usually takes clean texts extracted from various formats of documents ( e.g.", ", web pages , pdfs , or word documents ) as the information source .", "however , designing different text extraction approaches is time - consuming and not scalable ."], "relation": "used for", "id": "2022.findings-acl.105", "year": 2022, "rel_sent": "DuReader_vis : A Chinese Dataset for Open - domain Document Visual Question Answering.", "forward": true, "src_ids": "2022.findings-acl.105_2608"}
+{"input": "machine - generated text is done by using OtherScientificTerm| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "machine - generated text", "output": "perceptual quality dimensions", "neg_sample": ["machine - generated text is done by using OtherScientificTerm", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "We present a study that aims to uncover relevant perceptual quality dimensions for one type of machine - generated text , that is , Machine Translation .", "forward": false, "src_ids": "2022.humeval-1.3_2609"}
+{"input": "machine - generated text is used for Task| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "machine - generated text", "output": "machine translation", "neg_sample": ["machine - generated text is used for Task", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "Perceptual Quality Dimensions of Machine - Generated Text with a Focus on Machine Translation.", "forward": true, "src_ids": "2022.humeval-1.3_2610"}
+{"input": "attribute ratings is done by using Material| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "attribute ratings", "output": "crowdsourcing survey", "neg_sample": ["attribute ratings is done by using Material", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "We conducted a crowdsourcing survey in the style of a Semantic Differential to collect attribute ratings for German MT outputs .", "forward": false, "src_ids": "2022.humeval-1.3_2611"}
+{"input": "german mt outputs is done by using OtherScientificTerm| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "german mt outputs", "output": "attribute ratings", "neg_sample": ["german mt outputs is done by using OtherScientificTerm", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "We conducted a crowdsourcing survey in the style of a Semantic Differential to collect attribute ratings for German MT outputs .", "forward": false, "src_ids": "2022.humeval-1.3_2612"}
+{"input": "crowdsourcing survey is used for OtherScientificTerm| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "crowdsourcing survey", "output": "attribute ratings", "neg_sample": ["crowdsourcing survey is used for OtherScientificTerm", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "We conducted a crowdsourcing survey in the style of a Semantic Differential to collect attribute ratings for German MT outputs .", "forward": true, "src_ids": "2022.humeval-1.3_2613"}
+{"input": "attribute ratings is used for OtherScientificTerm| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "attribute ratings", "output": "german mt outputs", "neg_sample": ["attribute ratings is used for OtherScientificTerm", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "We conducted a crowdsourcing survey in the style of a Semantic Differential to collect attribute ratings for German MT outputs .", "forward": true, "src_ids": "2022.humeval-1.3_2614"}
+{"input": "perceptual dimensions is done by using Method| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "perceptual dimensions", "output": "exploratory factor analysis", "neg_sample": ["perceptual dimensions is done by using Method", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "An Exploratory Factor Analysis revealed the underlying perceptual dimensions .", "forward": false, "src_ids": "2022.humeval-1.3_2615"}
+{"input": "exploratory factor analysis is used for OtherScientificTerm| context: the quality of machine - generated text is a complex construct consisting of various aspects and dimensions .", "entity": "exploratory factor analysis", "output": "perceptual dimensions", "neg_sample": ["exploratory factor analysis is used for OtherScientificTerm", "the quality of machine - generated text is a complex construct consisting of various aspects and dimensions ."], "relation": "used for", "id": "2022.humeval-1.3", "year": 2022, "rel_sent": "An Exploratory Factor Analysis revealed the underlying perceptual dimensions .", "forward": true, "src_ids": "2022.humeval-1.3_2616"}
+{"input": "word embeddings is done by using Method| context: pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases .", "entity": "word embeddings", "output": "train - time debiasing algorithm", "neg_sample": ["word embeddings is done by using Method", "pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases ."], "relation": "used for", "id": "2022.findings-acl.90", "year": 2022, "rel_sent": "To address this problem , we propose DD - GloVe , a train - time debiasing algorithm to learn word embeddings by leveraging dictionary definitions .", "forward": false, "src_ids": "2022.findings-acl.90_2617"}
+{"input": "word embeddings is done by using Method| context: pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases .", "entity": "word embeddings", "output": "dictionary - guided loss functions", "neg_sample": ["word embeddings is done by using Method", "pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases ."], "relation": "used for", "id": "2022.findings-acl.90", "year": 2022, "rel_sent": "We introduce dictionary - guided loss functions that encourage word embeddings to be similar to their relatively neutral dictionary definition representations .", "forward": false, "src_ids": "2022.findings-acl.90_2618"}
+{"input": "bias direction is done by using OtherScientificTerm| context: pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases .", "entity": "bias direction", "output": "seed words", "neg_sample": ["bias direction is done by using OtherScientificTerm", "pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases ."], "relation": "used for", "id": "2022.findings-acl.90", "year": 2022, "rel_sent": "Existing debiasing algorithms typically need a pre - compiled list of seed words to represent the bias direction , along which biased information gets removed .", "forward": false, "src_ids": "2022.findings-acl.90_2619"}
+{"input": "seed words is used for OtherScientificTerm| context: pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases .", "entity": "seed words", "output": "bias direction", "neg_sample": ["seed words is used for OtherScientificTerm", "pre - trained word embeddings , such as glove , have shown undesirable gender , racial , and religious biases ."], "relation": "used for", "id": "2022.findings-acl.90", "year": 2022, "rel_sent": "Existing debiasing algorithms typically need a pre - compiled list of seed words to represent the bias direction , along which biased information gets removed .", "forward": true, "src_ids": "2022.findings-acl.90_2620"}
+{"input": "lstms is done by using OtherScientificTerm| context: we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al . , 2007 ) .", "entity": "lstms", "output": "predictors", "neg_sample": ["lstms is done by using OtherScientificTerm", "we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al .", ", 2007 ) ."], "relation": "used for", "id": "2022.tacl-1.1", "year": 2022, "rel_sent": "Drawing on studies of word acquisition in children , we evaluate multiple predictors for words ' ages of acquisition in LSTMs , BERT , and GPT-2 .", "forward": false, "src_ids": "2022.tacl-1.1_2621"}
+{"input": "gpt-2 is done by using OtherScientificTerm| context: we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al . , 2007 ) .", "entity": "gpt-2", "output": "predictors", "neg_sample": ["gpt-2 is done by using OtherScientificTerm", "we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al .", ", 2007 ) ."], "relation": "used for", "id": "2022.tacl-1.1", "year": 2022, "rel_sent": "Drawing on studies of word acquisition in children , we evaluate multiple predictors for words ' ages of acquisition in LSTMs , BERT , and GPT-2 .", "forward": false, "src_ids": "2022.tacl-1.1_2622"}
+{"input": "predictors is used for Method| context: we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al . , 2007 ) .", "entity": "predictors", "output": "lstms", "neg_sample": ["predictors is used for Method", "we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al .", ", 2007 ) ."], "relation": "used for", "id": "2022.tacl-1.1", "year": 2022, "rel_sent": "Drawing on studies of word acquisition in children , we evaluate multiple predictors for words ' ages of acquisition in LSTMs , BERT , and GPT-2 .", "forward": true, "src_ids": "2022.tacl-1.1_2623"}
+{"input": "child language acquisition is done by using OtherScientificTerm| context: we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al . , 2007 ) .", "entity": "child language acquisition", "output": "interaction", "neg_sample": ["child language acquisition is done by using OtherScientificTerm", "we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al .", ", 2007 ) ."], "relation": "used for", "id": "2022.tacl-1.1", "year": 2022, "rel_sent": "We find that the effects of concreteness , word length , and lexical class are pointedly different in children and language models , reinforcing the importance of interaction and sensorimotor experience in child language acquisition .", "forward": false, "src_ids": "2022.tacl-1.1_2624"}
+{"input": "interaction is used for Task| context: we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al . , 2007 ) .", "entity": "interaction", "output": "child language acquisition", "neg_sample": ["interaction is used for Task", "we investigate how neural language models acquire individual words during training , extracting learning curves and ages of acquisition for over 600 words on the macarthur - bates communicative development inventory ( fenson et al .", ", 2007 ) ."], "relation": "used for", "id": "2022.tacl-1.1", "year": 2022, "rel_sent": "We find that the effects of concreteness , word length , and lexical class are pointedly different in children and language models , reinforcing the importance of interaction and sensorimotor experience in child language acquisition .", "forward": true, "src_ids": "2022.tacl-1.1_2625"}
+{"input": "explainable medical code prediction is done by using Generic| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "explainable medical code prediction", "output": "medical concept driven attention", "neg_sample": ["explainable medical code prediction is done by using Generic", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "A Novel Framework Based on Medical Concept Driven Attention for Explainable Medical Code Prediction via External Knowledge.", "forward": false, "src_ids": "2022.findings-acl.110_2626"}
+{"input": "explainable medical code prediction is done by using Generic| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "explainable medical code prediction", "output": "medical concept driven attention", "neg_sample": ["explainable medical code prediction is done by using Generic", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "Therefore , in this paper , we propose a novel framework based on medical concept driven attention to incorporate external knowledge for explainable medical code prediction .", "forward": false, "src_ids": "2022.findings-acl.110_2627"}
+{"input": "medical concept driven attention is used for Task| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "medical concept driven attention", "output": "explainable medical code prediction", "neg_sample": ["medical concept driven attention is used for Task", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "A Novel Framework Based on Medical Concept Driven Attention for Explainable Medical Code Prediction via External Knowledge.", "forward": true, "src_ids": "2022.findings-acl.110_2628"}
+{"input": "external knowledge is used for Task| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "external knowledge", "output": "explainable medical code prediction", "neg_sample": ["external knowledge is used for Task", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "Therefore , in this paper , we propose a novel framework based on medical concept driven attention to incorporate external knowledge for explainable medical code prediction .", "forward": true, "src_ids": "2022.findings-acl.110_2629"}
+{"input": "medical concept driven attention is used for Task| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "medical concept driven attention", "output": "explainable medical code prediction", "neg_sample": ["medical concept driven attention is used for Task", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "Therefore , in this paper , we propose a novel framework based on medical concept driven attention to incorporate external knowledge for explainable medical code prediction .", "forward": true, "src_ids": "2022.findings-acl.110_2630"}
+{"input": "explainable medical code prediction is done by using OtherScientificTerm| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "explainable medical code prediction", "output": "external knowledge", "neg_sample": ["explainable medical code prediction is done by using OtherScientificTerm", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "Therefore , in this paper , we propose a novel framework based on medical concept driven attention to incorporate external knowledge for explainable medical code prediction .", "forward": false, "src_ids": "2022.findings-acl.110_2631"}
+{"input": "medical concepts is done by using OtherScientificTerm| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "medical concepts", "output": "topic space", "neg_sample": ["medical concepts is done by using OtherScientificTerm", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "In specific , both the clinical notes and Wikipedia documents are aligned into topic space to extract medical concepts using topic modeling .", "forward": false, "src_ids": "2022.findings-acl.110_2632"}
+{"input": "topic space is used for OtherScientificTerm| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "topic space", "output": "medical concepts", "neg_sample": ["topic space is used for OtherScientificTerm", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "In specific , both the clinical notes and Wikipedia documents are aligned into topic space to extract medical concepts using topic modeling .", "forward": true, "src_ids": "2022.findings-acl.110_2633"}
+{"input": "medical code related concepts is done by using Method| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "medical code related concepts", "output": "medical concept - driven attention mechanism", "neg_sample": ["medical code related concepts is done by using Method", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "Then , the medical concept - driven attention mechanism is applied to uncover the medical code related concepts which provide explanations for medical code prediction .", "forward": false, "src_ids": "2022.findings-acl.110_2634"}
+{"input": "medical concept - driven attention mechanism is used for OtherScientificTerm| context: rare code problem , the medical codes with low occurrences , is prominent in medical code prediction . recent studies employ deep neural networks and the external knowledge to tackle it . however , such approaches lack interpretability which is a vital issue in medical application . moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results .", "entity": "medical concept - driven attention mechanism", "output": "medical code related concepts", "neg_sample": ["medical concept - driven attention mechanism is used for OtherScientificTerm", "rare code problem , the medical codes with low occurrences , is prominent in medical code prediction .", "recent studies employ deep neural networks and the external knowledge to tackle it .", "however , such approaches lack interpretability which is a vital issue in medical application .", "moreover , due to the lengthy and noisy clinical notes , such approaches fail to achieve satisfactory results ."], "relation": "used for", "id": "2022.findings-acl.110", "year": 2022, "rel_sent": "Then , the medical concept - driven attention mechanism is applied to uncover the medical code related concepts which provide explanations for medical code prediction .", "forward": true, "src_ids": "2022.findings-acl.110_2635"}
+{"input": "resource - lean content flagging is done by using Method| context: we propose a novel framework for cross- lingual content flagging with limited target- language data , which significantly outperforms prior work in terms of predictive performance .", "entity": "resource - lean content flagging", "output": "neighborhood framework", "neg_sample": ["resource - lean content flagging is done by using Method", "we propose a novel framework for cross- lingual content flagging with limited target- language data , which significantly outperforms prior work in terms of predictive performance ."], "relation": "used for", "id": "2022.tacl-1.28", "year": 2022, "rel_sent": "A Neighborhood Framework for Resource - Lean Content Flagging.", "forward": false, "src_ids": "2022.tacl-1.28_2636"}
+{"input": "neighborhood framework is used for Task| context: we propose a novel framework for cross- lingual content flagging with limited target- language data , which significantly outperforms prior work in terms of predictive performance .", "entity": "neighborhood framework", "output": "resource - lean content flagging", "neg_sample": ["neighborhood framework is used for Task", "we propose a novel framework for cross- lingual content flagging with limited target- language data , which significantly outperforms prior work in terms of predictive performance ."], "relation": "used for", "id": "2022.tacl-1.28", "year": 2022, "rel_sent": "A Neighborhood Framework for Resource - Lean Content Flagging.", "forward": true, "src_ids": "2022.tacl-1.28_2637"}
+{"input": "state bills is done by using Method| context: decisions on state - level policies have a deep effect on many aspects of our everyday life , such as health - care and education access . however , there is little understanding of how these policies and decisions are being formed in the legislative process .", "entity": "state bills", "output": "textual graph - based model", "neg_sample": ["state bills is done by using Method", "decisions on state - level policies have a deep effect on many aspects of our everyday life , such as health - care and education access .", "however , there is little understanding of how these policies and decisions are being formed in the legislative process ."], "relation": "used for", "id": "2022.acl-long.22", "year": 2022, "rel_sent": "Next , we develop a textual graph - based model to embed and analyze state bills .", "forward": false, "src_ids": "2022.acl-long.22_2638"}
+{"input": "textual graph - based model is used for Material| context: decisions on state - level policies have a deep effect on many aspects of our everyday life , such as health - care and education access . however , there is little understanding of how these policies and decisions are being formed in the legislative process .", "entity": "textual graph - based model", "output": "state bills", "neg_sample": ["textual graph - based model is used for Material", "decisions on state - level policies have a deep effect on many aspects of our everyday life , such as health - care and education access .", "however , there is little understanding of how these policies and decisions are being formed in the legislative process ."], "relation": "used for", "id": "2022.acl-long.22", "year": 2022, "rel_sent": "Next , we develop a textual graph - based model to embed and analyze state bills .", "forward": true, "src_ids": "2022.acl-long.22_2639"}
+{"input": "transfer learning is used for Task| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "transfer learning", "output": "resolution of anaphora", "neg_sample": ["transfer learning is used for Task", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.275", "year": 2022, "rel_sent": "We demonstrate empirically that transfer learning from the chemical domain improves resolution of anaphora in recipes , suggesting transferability of general procedural knowledge .", "forward": true, "src_ids": "2022.findings-acl.275_2640"}
+{"input": "resolution of anaphora is done by using Method| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "resolution of anaphora", "output": "transfer learning", "neg_sample": ["resolution of anaphora is done by using Method", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.275", "year": 2022, "rel_sent": "We demonstrate empirically that transfer learning from the chemical domain improves resolution of anaphora in recipes , suggesting transferability of general procedural knowledge .", "forward": false, "src_ids": "2022.findings-acl.275_2641"}
+{"input": "transferability of general procedural knowledge is done by using Method| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "transferability of general procedural knowledge", "output": "transfer learning", "neg_sample": ["transferability of general procedural knowledge is done by using Method", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.275", "year": 2022, "rel_sent": "We demonstrate empirically that transfer learning from the chemical domain improves resolution of anaphora in recipes , suggesting transferability of general procedural knowledge .", "forward": false, "src_ids": "2022.findings-acl.275_2642"}
+{"input": "transfer learning is used for OtherScientificTerm| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "transfer learning", "output": "transferability of general procedural knowledge", "neg_sample": ["transfer learning is used for OtherScientificTerm", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.275", "year": 2022, "rel_sent": "We demonstrate empirically that transfer learning from the chemical domain improves resolution of anaphora in recipes , suggesting transferability of general procedural knowledge .", "forward": true, "src_ids": "2022.findings-acl.275_2643"}
+{"input": "detecting signs of depression is done by using Method| context: depression is a common and serious mental illness that early detection can improve the patient 's symptoms and make depression easier to treat .", "entity": "detecting signs of depression", "output": "bert - based ensembles", "neg_sample": ["detecting signs of depression is done by using Method", "depression is a common and serious mental illness that early detection can improve the patient 's symptoms and make depression easier to treat ."], "relation": "used for", "id": "2022.ltedi-1.38", "year": 2022, "rel_sent": "KADO@LT - EDI - ACL2022 : BERT - based Ensembles for Detecting Signs of Depression from Social Media Text.", "forward": false, "src_ids": "2022.ltedi-1.38_2644"}
+{"input": "bert - based ensembles is used for Task| context: depression is a common and serious mental illness that early detection can improve the patient 's symptoms and make depression easier to treat .", "entity": "bert - based ensembles", "output": "detecting signs of depression", "neg_sample": ["bert - based ensembles is used for Task", "depression is a common and serious mental illness that early detection can improve the patient 's symptoms and make depression easier to treat ."], "relation": "used for", "id": "2022.ltedi-1.38", "year": 2022, "rel_sent": "KADO@LT - EDI - ACL2022 : BERT - based Ensembles for Detecting Signs of Depression from Social Media Text.", "forward": true, "src_ids": "2022.ltedi-1.38_2645"}
+{"input": "nlp is done by using OtherScientificTerm| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "nlp", "output": "linguistic expertise", "neg_sample": ["nlp is done by using OtherScientificTerm", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in NLP for Under - Documented Languages.", "forward": false, "src_ids": "2022.findings-acl.167_2646"}
+{"input": "under - documented languages is done by using OtherScientificTerm| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "under - documented languages", "output": "linguistic expertise", "neg_sample": ["under - documented languages is done by using OtherScientificTerm", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in NLP for Under - Documented Languages.", "forward": false, "src_ids": "2022.findings-acl.167_2647"}
+{"input": "under - documented languages is done by using Task| context: technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "entity": "under - documented languages", "output": "nlp", "neg_sample": ["under - documented languages is done by using Task", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in NLP for Under - Documented Languages.", "forward": false, "src_ids": "2022.findings-acl.167_2648"}
+{"input": "linguistic expertise is used for Task| context: technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "entity": "linguistic expertise", "output": "nlp", "neg_sample": ["linguistic expertise is used for Task", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in NLP for Under - Documented Languages.", "forward": true, "src_ids": "2022.findings-acl.167_2649"}
+{"input": "structured formats is used for Task| context: technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "entity": "structured formats", "output": "nlp", "neg_sample": ["structured formats is used for Task", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "( 2 ) Great care and target language expertise is required when converting the data into structured formats commonly employed in NLP .", "forward": true, "src_ids": "2022.findings-acl.167_2650"}
+{"input": "linguistic expertise is used for Material| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "linguistic expertise", "output": "under - documented languages", "neg_sample": ["linguistic expertise is used for Material", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in NLP for Under - Documented Languages.", "forward": true, "src_ids": "2022.findings-acl.167_2651"}
+{"input": "nlp is used for Material| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "nlp", "output": "under - documented languages", "neg_sample": ["nlp is used for Material", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in NLP for Under - Documented Languages.", "forward": true, "src_ids": "2022.findings-acl.167_2652"}
+{"input": "nlp is done by using OtherScientificTerm| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "nlp", "output": "structured formats", "neg_sample": ["nlp is done by using OtherScientificTerm", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "( 2 ) Great care and target language expertise is required when converting the data into structured formats commonly employed in NLP .", "forward": false, "src_ids": "2022.findings-acl.167_2653"}
+{"input": "tsimchianic language gitksan is done by using Method| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "tsimchianic language gitksan", "output": "morphological reinflection system", "neg_sample": ["tsimchianic language gitksan is done by using Method", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "We illustrate each step through a case study on developing a morphological reinflection system for the Tsimchianic language Gitksan .", "forward": false, "src_ids": "2022.findings-acl.167_2654"}
+{"input": "morphological reinflection system is used for Material| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "entity": "morphological reinflection system", "output": "tsimchianic language gitksan", "neg_sample": ["morphological reinflection system is used for Material", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific ."], "relation": "used for", "id": "2022.findings-acl.167", "year": 2022, "rel_sent": "We illustrate each step through a case study on developing a morphological reinflection system for the Tsimchianic language Gitksan .", "forward": true, "src_ids": "2022.findings-acl.167_2655"}
+{"input": "neural - symbolic semantic parsing is done by using Method| context: recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations . however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference .", "entity": "neural - symbolic semantic parsing", "output": "collaborative framework", "neg_sample": ["neural - symbolic semantic parsing is done by using Method", "recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations .", "however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference ."], "relation": "used for", "id": "2022.findings-acl.328", "year": 2022, "rel_sent": "To address this , we further propose a simple yet principled collaborative framework for neural - symbolic semantic parsing , by designing a decision criterion for beam search that incorporates the prior knowledge from a symbolic parser and accounts for model uncertainty .", "forward": false, "src_ids": "2022.findings-acl.328_2656"}
+{"input": "collaborative framework is used for Task| context: recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations . however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference .", "entity": "collaborative framework", "output": "neural - symbolic semantic parsing", "neg_sample": ["collaborative framework is used for Task", "recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations .", "however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference ."], "relation": "used for", "id": "2022.findings-acl.328", "year": 2022, "rel_sent": "To address this , we further propose a simple yet principled collaborative framework for neural - symbolic semantic parsing , by designing a decision criterion for beam search that incorporates the prior knowledge from a symbolic parser and accounts for model uncertainty .", "forward": true, "src_ids": "2022.findings-acl.328_2657"}
+{"input": "beam search is done by using OtherScientificTerm| context: recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations . however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference .", "entity": "beam search", "output": "decision criterion", "neg_sample": ["beam search is done by using OtherScientificTerm", "recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations .", "however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference ."], "relation": "used for", "id": "2022.findings-acl.328", "year": 2022, "rel_sent": "To address this , we further propose a simple yet principled collaborative framework for neural - symbolic semantic parsing , by designing a decision criterion for beam search that incorporates the prior knowledge from a symbolic parser and accounts for model uncertainty .", "forward": false, "src_ids": "2022.findings-acl.328_2658"}
+{"input": "model uncertainty is done by using OtherScientificTerm| context: recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations . however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference .", "entity": "model uncertainty", "output": "decision criterion", "neg_sample": ["model uncertainty is done by using OtherScientificTerm", "recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations .", "however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference ."], "relation": "used for", "id": "2022.findings-acl.328", "year": 2022, "rel_sent": "To address this , we further propose a simple yet principled collaborative framework for neural - symbolic semantic parsing , by designing a decision criterion for beam search that incorporates the prior knowledge from a symbolic parser and accounts for model uncertainty .", "forward": false, "src_ids": "2022.findings-acl.328_2659"}
+{"input": "decision criterion is used for Method| context: recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations . however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference .", "entity": "decision criterion", "output": "beam search", "neg_sample": ["decision criterion is used for Method", "recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations .", "however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference ."], "relation": "used for", "id": "2022.findings-acl.328", "year": 2022, "rel_sent": "To address this , we further propose a simple yet principled collaborative framework for neural - symbolic semantic parsing , by designing a decision criterion for beam search that incorporates the prior knowledge from a symbolic parser and accounts for model uncertainty .", "forward": true, "src_ids": "2022.findings-acl.328_2660"}
+{"input": "decision criterion is used for OtherScientificTerm| context: recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations . however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference .", "entity": "decision criterion", "output": "model uncertainty", "neg_sample": ["decision criterion is used for OtherScientificTerm", "recent work in task - independent graph semantic parsing has shifted from grammar - based symbolic approaches to neural models , showing strong performance on different types of meaning representations .", "however , it is still unclear that what are the limitations of these neural parsers , and whether these limitations can be compensated by incorporating symbolic knowledge into model inference ."], "relation": "used for", "id": "2022.findings-acl.328", "year": 2022, "rel_sent": "To address this , we further propose a simple yet principled collaborative framework for neural - symbolic semantic parsing , by designing a decision criterion for beam search that incorporates the prior knowledge from a symbolic parser and accounts for model uncertainty .", "forward": true, "src_ids": "2022.findings-acl.328_2661"}
+{"input": "language generation process is done by using Material| context: existing studies have investigated the tendency of autoregressive language models to generate contexts that exhibit undesired biases and toxicity . various debiasing approaches have been proposed , which are primarily categorized into data - based and decoding - based .", "entity": "language generation process", "output": "toxic corpus", "neg_sample": ["language generation process is done by using Material", "existing studies have investigated the tendency of autoregressive language models to generate contexts that exhibit undesired biases and toxicity .", "various debiasing approaches have been proposed , which are primarily categorized into data - based and decoding - based ."], "relation": "used for", "id": "2022.ltedi-1.6", "year": 2022, "rel_sent": "Our result shows that toxic corpus can indeed help to reduce the toxicity of the language generation process substantially , complementing the existing debiasing methods .", "forward": false, "src_ids": "2022.ltedi-1.6_2662"}
+{"input": "toxic corpus is used for Task| context: existing studies have investigated the tendency of autoregressive language models to generate contexts that exhibit undesired biases and toxicity . various debiasing approaches have been proposed , which are primarily categorized into data - based and decoding - based .", "entity": "toxic corpus", "output": "language generation process", "neg_sample": ["toxic corpus is used for Task", "existing studies have investigated the tendency of autoregressive language models to generate contexts that exhibit undesired biases and toxicity .", "various debiasing approaches have been proposed , which are primarily categorized into data - based and decoding - based ."], "relation": "used for", "id": "2022.ltedi-1.6", "year": 2022, "rel_sent": "Our result shows that toxic corpus can indeed help to reduce the toxicity of the language generation process substantially , complementing the existing debiasing methods .", "forward": true, "src_ids": "2022.ltedi-1.6_2663"}
+{"input": "restricted - access corpora is done by using Method| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "restricted - access corpora", "output": "automated speech tools", "neg_sample": ["restricted - access corpora is done by using Method", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Automated speech tools for helping communities process restricted - access corpora for language revival efforts.", "forward": false, "src_ids": "2022.computel-1.6_2664"}
+{"input": "language revival efforts is done by using Method| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "language revival efforts", "output": "automated speech tools", "neg_sample": ["language revival efforts is done by using Method", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Automated speech tools for helping communities process restricted - access corpora for language revival efforts.", "forward": false, "src_ids": "2022.computel-1.6_2665"}
+{"input": "language revival efforts is done by using Material| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "language revival efforts", "output": "restricted - access corpora", "neg_sample": ["language revival efforts is done by using Material", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Automated speech tools for helping communities process restricted - access corpora for language revival efforts.", "forward": false, "src_ids": "2022.computel-1.6_2666"}
+{"input": "automated speech tools is used for Material| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "automated speech tools", "output": "restricted - access corpora", "neg_sample": ["automated speech tools is used for Material", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Automated speech tools for helping communities process restricted - access corpora for language revival efforts.", "forward": true, "src_ids": "2022.computel-1.6_2667"}
+{"input": "automated speech tools is used for Task| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "automated speech tools", "output": "language revival efforts", "neg_sample": ["automated speech tools is used for Task", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Automated speech tools for helping communities process restricted - access corpora for language revival efforts.", "forward": true, "src_ids": "2022.computel-1.6_2668"}
+{"input": "restricted - access corpora is used for Task| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "restricted - access corpora", "output": "language revival efforts", "neg_sample": ["restricted - access corpora is used for Task", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Automated speech tools for helping communities process restricted - access corpora for language revival efforts.", "forward": true, "src_ids": "2022.computel-1.6_2669"}
+{"input": "meta - linguistic commentary is done by using Material| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "meta - linguistic commentary", "output": "english", "neg_sample": ["meta - linguistic commentary is done by using Material", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We propose a privacy - preserving workflow to widen both bottlenecks for recordings where speech in the endangered language is intermixed with a more widely - used language such as English for meta - linguistic commentary and questions ( e.g. What is the word for ' tree ' ? ) .", "forward": false, "src_ids": "2022.computel-1.6_2670"}
+{"input": "english is used for OtherScientificTerm| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "english", "output": "meta - linguistic commentary", "neg_sample": ["english is used for OtherScientificTerm", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We propose a privacy - preserving workflow to widen both bottlenecks for recordings where speech in the endangered language is intermixed with a more widely - used language such as English for meta - linguistic commentary and questions ( e.g. What is the word for ' tree ' ? ) .", "forward": true, "src_ids": "2022.computel-1.6_2671"}
+{"input": "metalinguistic content is done by using Method| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "metalinguistic content", "output": "voice activity detection ( vad )", "neg_sample": ["metalinguistic content is done by using Method", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We integrate voice activity detection ( VAD ) , spoken language identification ( SLI ) , and automatic speech recognition ( ASR ) to transcribe the metalinguistic content , which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access .", "forward": false, "src_ids": "2022.computel-1.6_2672"}
+{"input": "metalinguistic content is done by using Task| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "metalinguistic content", "output": "automatic speech recognition ( asr )", "neg_sample": ["metalinguistic content is done by using Task", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We integrate voice activity detection ( VAD ) , spoken language identification ( SLI ) , and automatic speech recognition ( ASR ) to transcribe the metalinguistic content , which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access .", "forward": false, "src_ids": "2022.computel-1.6_2673"}
+{"input": "automatic speech recognition ( asr ) is used for OtherScientificTerm| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "automatic speech recognition ( asr )", "output": "metalinguistic content", "neg_sample": ["automatic speech recognition ( asr ) is used for OtherScientificTerm", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We integrate voice activity detection ( VAD ) , spoken language identification ( SLI ) , and automatic speech recognition ( ASR ) to transcribe the metalinguistic content , which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access .", "forward": true, "src_ids": "2022.computel-1.6_2674"}
+{"input": "voice activity detection ( vad ) is used for OtherScientificTerm| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "voice activity detection ( vad )", "output": "metalinguistic content", "neg_sample": ["voice activity detection ( vad ) is used for OtherScientificTerm", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We integrate voice activity detection ( VAD ) , spoken language identification ( SLI ) , and automatic speech recognition ( ASR ) to transcribe the metalinguistic content , which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access .", "forward": true, "src_ids": "2022.computel-1.6_2675"}
+{"input": "spoken language identification ( sli ) is used for OtherScientificTerm| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "spoken language identification ( sli )", "output": "metalinguistic content", "neg_sample": ["spoken language identification ( sli ) is used for OtherScientificTerm", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "We integrate voice activity detection ( VAD ) , spoken language identification ( SLI ) , and automatic speech recognition ( ASR ) to transcribe the metalinguistic content , which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access .", "forward": true, "src_ids": "2022.computel-1.6_2676"}
+{"input": "spoken language identification is done by using Material| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "spoken language identification", "output": "annotated training data", "neg_sample": ["spoken language identification is done by using Material", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Our collaborative work with the Muruwari custodian of the archival materials show that this workflow reduces metalanguage transcription time by 20 % even given only minimal amounts of annotated training data , 10 utterances per language for SLI and for ASR at most 39 minutes , and possibly as little as 39 seconds .", "forward": false, "src_ids": "2022.computel-1.6_2677"}
+{"input": "annotated training data is used for Task| context: many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs . one bottleneck is the time - intensive nature of annotation . an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin .", "entity": "annotated training data", "output": "spoken language identification", "neg_sample": ["annotated training data is used for Task", "many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs .", "one bottleneck is the time - intensive nature of annotation .", "an even narrower bottleneck occurs for recordings with access constraints , such as language that must be vetted or filtered by authorised community members before annotation can begin ."], "relation": "used for", "id": "2022.computel-1.6", "year": 2022, "rel_sent": "Our collaborative work with the Muruwari custodian of the archival materials show that this workflow reduces metalanguage transcription time by 20 % even given only minimal amounts of annotated training data , 10 utterances per language for SLI and for ASR at most 39 minutes , and possibly as little as 39 seconds .", "forward": true, "src_ids": "2022.computel-1.6_2678"}
+{"input": "multimodal co - learning is used for Task| context: as humans , we experience the world with all our senses or modalities ( sound , sight , touch , smell , and taste ) . we use these modalities , particularly sight and touch , to convey and interpret specific meanings . multimodal expressions are central to conversations ; a rich set of modalities amplify and often compensate for each other . a multimodal conversational ai system answers questions , fulfills tasks , and emulates human conversations by understanding and expressing itself via multiple modalities .", "entity": "multimodal co - learning", "output": "multimodal conversational ai research", "neg_sample": ["multimodal co - learning is used for Task", "as humans , we experience the world with all our senses or modalities ( sound , sight , touch , smell , and taste ) .", "we use these modalities , particularly sight and touch , to convey and interpret specific meanings .", "multimodal expressions are central to conversations ; a rich set of modalities amplify and often compensate for each other .", "a multimodal conversational ai system answers questions , fulfills tasks , and emulates human conversations by understanding and expressing itself via multiple modalities ."], "relation": "used for", "id": "2022.nlp4convai-1.12", "year": 2022, "rel_sent": "Finally , we identify multimodal co - learning as a promising direction for multimodal conversational AI research .", "forward": true, "src_ids": "2022.nlp4convai-1.12_2679"}
+{"input": "multimodal conversational ai research is done by using Method| context: as humans , we experience the world with all our senses or modalities ( sound , sight , touch , smell , and taste ) . we use these modalities , particularly sight and touch , to convey and interpret specific meanings . multimodal expressions are central to conversations ; a rich set of modalities amplify and often compensate for each other . a multimodal conversational ai system answers questions , fulfills tasks , and emulates human conversations by understanding and expressing itself via multiple modalities .", "entity": "multimodal conversational ai research", "output": "multimodal co - learning", "neg_sample": ["multimodal conversational ai research is done by using Method", "as humans , we experience the world with all our senses or modalities ( sound , sight , touch , smell , and taste ) .", "we use these modalities , particularly sight and touch , to convey and interpret specific meanings .", "multimodal expressions are central to conversations ; a rich set of modalities amplify and often compensate for each other .", "a multimodal conversational ai system answers questions , fulfills tasks , and emulates human conversations by understanding and expressing itself via multiple modalities ."], "relation": "used for", "id": "2022.nlp4convai-1.12", "year": 2022, "rel_sent": "Finally , we identify multimodal co - learning as a promising direction for multimodal conversational AI research .", "forward": false, "src_ids": "2022.nlp4convai-1.12_2680"}
+{"input": "out - of - distribution data streams is done by using Task| context: real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting . however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario .", "entity": "out - of - distribution data streams", "output": "continual model refinement", "neg_sample": ["out - of - distribution data streams is done by using Task", "real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting .", "however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario ."], "relation": "used for", "id": "2022.acl-long.223", "year": 2022, "rel_sent": "On Continual Model Refinement in Out - of - Distribution Data Streams.", "forward": false, "src_ids": "2022.acl-long.223_2681"}
+{"input": "continual model refinement is used for Material| context: real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting . however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario .", "entity": "continual model refinement", "output": "out - of - distribution data streams", "neg_sample": ["continual model refinement is used for Material", "real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting .", "however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario ."], "relation": "used for", "id": "2022.acl-long.223", "year": 2022, "rel_sent": "On Continual Model Refinement in Out - of - Distribution Data Streams.", "forward": true, "src_ids": "2022.acl-long.223_2682"}
+{"input": "cmr setting is done by using Method| context: real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting . however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario .", "entity": "cmr setting", "output": "cl approaches", "neg_sample": ["cmr setting is done by using Method", "real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting .", "however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario ."], "relation": "used for", "id": "2022.acl-long.223", "year": 2022, "rel_sent": "We extend several existing CL approaches to the CMR setting and evaluate them extensively .", "forward": false, "src_ids": "2022.acl-long.223_2683"}
+{"input": "cl approaches is used for Material| context: real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting . however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario .", "entity": "cl approaches", "output": "cmr setting", "neg_sample": ["cl approaches is used for Material", "real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting .", "however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario ."], "relation": "used for", "id": "2022.acl-long.223", "year": 2022, "rel_sent": "We extend several existing CL approaches to the CMR setting and evaluate them extensively .", "forward": true, "src_ids": "2022.acl-long.223_2684"}
+{"input": "dynamic ood data streams is done by using Method| context: real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting . however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario .", "entity": "dynamic ood data streams", "output": "sampling algorithm", "neg_sample": ["dynamic ood data streams is done by using Method", "real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting .", "however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario ."], "relation": "used for", "id": "2022.acl-long.223", "year": 2022, "rel_sent": "For benchmarking and analysis , we propose a general sampling algorithm to obtain dynamic OOD data streams with controllable non - stationarity , as well as a suite of metrics measuring various aspects of online performance .", "forward": false, "src_ids": "2022.acl-long.223_2685"}
+{"input": "sampling algorithm is used for OtherScientificTerm| context: real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting . however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario .", "entity": "sampling algorithm", "output": "dynamic ood data streams", "neg_sample": ["sampling algorithm is used for OtherScientificTerm", "real - world natural language processing ( nlp ) models need to be continually updated tofix the prediction errors in out - of - distribution ( ood ) data streams while overcoming catastrophic forgetting .", "however , existing continual learning ( cl ) problem setups can not cover such a realistic and complex scenario ."], "relation": "used for", "id": "2022.acl-long.223", "year": 2022, "rel_sent": "For benchmarking and analysis , we propose a general sampling algorithm to obtain dynamic OOD data streams with controllable non - stationarity , as well as a suite of metrics measuring various aspects of online performance .", "forward": true, "src_ids": "2022.acl-long.223_2686"}
+{"input": "implicit discourse relation recognition ( idrr ) is done by using OtherScientificTerm| context: prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) . however , diverse relation senses may benefit from different attention mechanisms . we also argue that some linguistic relation in between two words can be further exploited for idrr .", "entity": "implicit discourse relation recognition ( idrr )", "output": "semantic connection", "neg_sample": ["implicit discourse relation recognition ( idrr ) is done by using OtherScientificTerm", "prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) .", "however , diverse relation senses may benefit from different attention mechanisms .", "we also argue that some linguistic relation in between two words can be further exploited for idrr ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "Encoding and Fusing Semantic Connection and Linguistic Evidence for Implicit Discourse Relation Recognition.", "forward": false, "src_ids": "2022.findings-acl.256_2687"}
+{"input": "implicit discourse relation recognition ( idrr ) is done by using OtherScientificTerm| context: prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) . however , diverse relation senses may benefit from different attention mechanisms . we also argue that some linguistic relation in between two words can be further exploited for idrr .", "entity": "implicit discourse relation recognition ( idrr )", "output": "semantic connection", "neg_sample": ["implicit discourse relation recognition ( idrr ) is done by using OtherScientificTerm", "prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) .", "however , diverse relation senses may benefit from different attention mechanisms .", "we also argue that some linguistic relation in between two words can be further exploited for idrr ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "This paper proposes a Multi - Attentive Neural Fusion ( MANF ) model to encode and fuse both semantic connection and linguistic evidence for IDRR .", "forward": false, "src_ids": "2022.findings-acl.256_2688"}
+{"input": "linguistic evidence is used for Task| context: however , diverse relation senses may benefit from different attention mechanisms .", "entity": "linguistic evidence", "output": "implicit discourse relation recognition ( idrr )", "neg_sample": ["linguistic evidence is used for Task", "however , diverse relation senses may benefit from different attention mechanisms ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "Encoding and Fusing Semantic Connection and Linguistic Evidence for Implicit Discourse Relation Recognition.", "forward": true, "src_ids": "2022.findings-acl.256_2689"}
+{"input": "semantic connection is used for Task| context: however , diverse relation senses may benefit from different attention mechanisms .", "entity": "semantic connection", "output": "implicit discourse relation recognition ( idrr )", "neg_sample": ["semantic connection is used for Task", "however , diverse relation senses may benefit from different attention mechanisms ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "Encoding and Fusing Semantic Connection and Linguistic Evidence for Implicit Discourse Relation Recognition.", "forward": true, "src_ids": "2022.findings-acl.256_2690"}
+{"input": "linguistic evidence is used for Task| context: however , diverse relation senses may benefit from different attention mechanisms .", "entity": "linguistic evidence", "output": "implicit discourse relation recognition ( idrr )", "neg_sample": ["linguistic evidence is used for Task", "however , diverse relation senses may benefit from different attention mechanisms ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "This paper proposes a Multi - Attentive Neural Fusion ( MANF ) model to encode and fuse both semantic connection and linguistic evidence for IDRR .", "forward": true, "src_ids": "2022.findings-acl.256_2691"}
+{"input": "semantic connection is used for Task| context: however , diverse relation senses may benefit from different attention mechanisms .", "entity": "semantic connection", "output": "implicit discourse relation recognition ( idrr )", "neg_sample": ["semantic connection is used for Task", "however , diverse relation senses may benefit from different attention mechanisms ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "This paper proposes a Multi - Attentive Neural Fusion ( MANF ) model to encode and fuse both semantic connection and linguistic evidence for IDRR .", "forward": true, "src_ids": "2022.findings-acl.256_2692"}
+{"input": "offset matrix network ( omn ) is used for OtherScientificTerm| context: prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) . however , diverse relation senses may benefit from different attention mechanisms .", "entity": "offset matrix network ( omn )", "output": "linguistic relations", "neg_sample": ["offset matrix network ( omn ) is used for OtherScientificTerm", "prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) .", "however , diverse relation senses may benefit from different attention mechanisms ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "We also propose an Offset Matrix Network ( OMN ) to encode the linguistic relations of word - pairs as linguistic evidence .", "forward": true, "src_ids": "2022.findings-acl.256_2693"}
+{"input": "attentive representation is done by using Method| context: prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) . however , diverse relation senses may benefit from different attention mechanisms . we also argue that some linguistic relation in between two words can be further exploited for idrr .", "entity": "attentive representation", "output": "dual attention network ( dan )", "neg_sample": ["attentive representation is done by using Method", "prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) .", "however , diverse relation senses may benefit from different attention mechanisms .", "we also argue that some linguistic relation in between two words can be further exploited for idrr ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "In MANF , we design a Dual Attention Network ( DAN ) to learn and fuse two kinds of attentive representation for arguments as its semantic connection .", "forward": false, "src_ids": "2022.findings-acl.256_2694"}
+{"input": "dual attention network ( dan ) is used for Method| context: prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) . however , diverse relation senses may benefit from different attention mechanisms . we also argue that some linguistic relation in between two words can be further exploited for idrr .", "entity": "dual attention network ( dan )", "output": "attentive representation", "neg_sample": ["dual attention network ( dan ) is used for Method", "prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) .", "however , diverse relation senses may benefit from different attention mechanisms .", "we also argue that some linguistic relation in between two words can be further exploited for idrr ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "In MANF , we design a Dual Attention Network ( DAN ) to learn and fuse two kinds of attentive representation for arguments as its semantic connection .", "forward": true, "src_ids": "2022.findings-acl.256_2695"}
+{"input": "linguistic relations is done by using Method| context: prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) . however , diverse relation senses may benefit from different attention mechanisms . we also argue that some linguistic relation in between two words can be further exploited for idrr .", "entity": "linguistic relations", "output": "offset matrix network ( omn )", "neg_sample": ["linguistic relations is done by using Method", "prior studies use one attention mechanism to improve contextual semantic representation learning for implicit discourse relation recognition ( idrr ) .", "however , diverse relation senses may benefit from different attention mechanisms .", "we also argue that some linguistic relation in between two words can be further exploited for idrr ."], "relation": "used for", "id": "2022.findings-acl.256", "year": 2022, "rel_sent": "We also propose an Offset Matrix Network ( OMN ) to encode the linguistic relations of word - pairs as linguistic evidence .", "forward": false, "src_ids": "2022.findings-acl.256_2696"}
+{"input": "intent categories is done by using Task| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature . existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate .", "entity": "intent categories", "output": "intent discovery", "neg_sample": ["intent categories is done by using Task", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature .", "existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "New intent discovery aims to uncover novel intent categories from user utterances to expand the set of supported intent classes .", "forward": false, "src_ids": "2022.acl-long.21_2697"}
+{"input": "intent discovery is used for OtherScientificTerm| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature . existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate .", "entity": "intent discovery", "output": "intent categories", "neg_sample": ["intent discovery is used for OtherScientificTerm", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature .", "existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "New intent discovery aims to uncover novel intent categories from user utterances to expand the set of supported intent classes .", "forward": true, "src_ids": "2022.acl-long.21_2698"}
+{"input": "representation learning is done by using Material| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature . existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate .", "entity": "representation learning", "output": "external labeled data", "neg_sample": ["representation learning is done by using Material", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature .", "existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "Particularly , we first propose a multi - task pre - training strategy to leverage rich unlabeled data along with external labeled data for representation learning .", "forward": false, "src_ids": "2022.acl-long.21_2699"}
+{"input": "external labeled data is used for Method| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature .", "entity": "external labeled data", "output": "representation learning", "neg_sample": ["external labeled data is used for Method", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "Particularly , we first propose a multi - task pre - training strategy to leverage rich unlabeled data along with external labeled data for representation learning .", "forward": true, "src_ids": "2022.acl-long.21_2700"}
+{"input": "self - supervisory signals is done by using OtherScientificTerm| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature . existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate .", "entity": "self - supervisory signals", "output": "contrastive loss", "neg_sample": ["self - supervisory signals is done by using OtherScientificTerm", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature .", "existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "Then , we design a new contrastive loss to exploit self - supervisory signals in unlabeled data for clustering .", "forward": false, "src_ids": "2022.acl-long.21_2701"}
+{"input": "clustering is done by using OtherScientificTerm| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature . existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate .", "entity": "clustering", "output": "self - supervisory signals", "neg_sample": ["clustering is done by using OtherScientificTerm", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature .", "existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "Then , we design a new contrastive loss to exploit self - supervisory signals in unlabeled data for clustering .", "forward": false, "src_ids": "2022.acl-long.21_2702"}
+{"input": "contrastive loss is used for OtherScientificTerm| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature . existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate .", "entity": "contrastive loss", "output": "self - supervisory signals", "neg_sample": ["contrastive loss is used for OtherScientificTerm", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature .", "existing approaches typically rely on a large amount of labeled utterances and employ pseudo - labeling methods for representation learning and clustering , which are label - intensive , inefficient , and inaccurate ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "Then , we design a new contrastive loss to exploit self - supervisory signals in unlabeled data for clustering .", "forward": true, "src_ids": "2022.acl-long.21_2703"}
+{"input": "self - supervisory signals is used for Task| context: it is a critical task for the development and service expansion of a practical dialogue system . despite its importance , this problem remains under - explored in the literature .", "entity": "self - supervisory signals", "output": "clustering", "neg_sample": ["self - supervisory signals is used for Task", "it is a critical task for the development and service expansion of a practical dialogue system .", "despite its importance , this problem remains under - explored in the literature ."], "relation": "used for", "id": "2022.acl-long.21", "year": 2022, "rel_sent": "Then , we design a new contrastive loss to exploit self - supervisory signals in unlabeled data for clustering .", "forward": true, "src_ids": "2022.acl-long.21_2704"}
+{"input": "compositional tasks is done by using Method| context: several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing .", "entity": "compositional tasks", "output": "transformers", "neg_sample": ["compositional tasks is done by using Method", "several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing ."], "relation": "used for", "id": "2022.acl-long.251", "year": 2022, "rel_sent": "Making Transformers Solve Compositional Tasks.", "forward": false, "src_ids": "2022.acl-long.251_2705"}
+{"input": "transformers is used for Task| context: several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing .", "entity": "transformers", "output": "compositional tasks", "neg_sample": ["transformers is used for Task", "several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing ."], "relation": "used for", "id": "2022.acl-long.251", "year": 2022, "rel_sent": "Making Transformers Solve Compositional Tasks.", "forward": true, "src_ids": "2022.acl-long.251_2706"}
+{"input": "transformer configurations is used for Task| context: several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing .", "entity": "transformer configurations", "output": "compositional tasks", "neg_sample": ["transformer configurations is used for Task", "several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing ."], "relation": "used for", "id": "2022.acl-long.251", "year": 2022, "rel_sent": "We identified Transformer configurations that generalize compositionally significantly better than previously reported in the literature in many compositional tasks .", "forward": true, "src_ids": "2022.acl-long.251_2707"}
+{"input": "compositional tasks is done by using OtherScientificTerm| context: several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing .", "entity": "compositional tasks", "output": "transformer configurations", "neg_sample": ["compositional tasks is done by using OtherScientificTerm", "several studies have reported the inability of transformer models to generalize compositionally , a key type of generalization in many nlp tasks such as semantic parsing ."], "relation": "used for", "id": "2022.acl-long.251", "year": 2022, "rel_sent": "We identified Transformer configurations that generalize compositionally significantly better than previously reported in the literature in many compositional tasks .", "forward": false, "src_ids": "2022.acl-long.251_2708"}
+{"input": "speech translation is done by using Method| context: how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ? existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities .", "entity": "speech translation", "output": "speech - text manifold mixup", "neg_sample": ["speech translation is done by using Method", "how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ?", "existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities ."], "relation": "used for", "id": "2022.acl-long.486", "year": 2022, "rel_sent": "STEMM : Self - learning with Speech - text Manifold Mixup for Speech Translation.", "forward": false, "src_ids": "2022.acl-long.486_2709"}
+{"input": "speech - text manifold mixup is used for Task| context: how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ? existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities .", "entity": "speech - text manifold mixup", "output": "speech translation", "neg_sample": ["speech - text manifold mixup is used for Task", "how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ?", "existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities ."], "relation": "used for", "id": "2022.acl-long.486", "year": 2022, "rel_sent": "STEMM : Self - learning with Speech - text Manifold Mixup for Speech Translation.", "forward": true, "src_ids": "2022.acl-long.486_2710"}
+{"input": "self - learning is used for Task| context: how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ? existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities .", "entity": "self - learning", "output": "speech translation", "neg_sample": ["self - learning is used for Task", "how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ?", "existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities ."], "relation": "used for", "id": "2022.acl-long.486", "year": 2022, "rel_sent": "STEMM : Self - learning with Speech - text Manifold Mixup for Speech Translation.", "forward": true, "src_ids": "2022.acl-long.486_2711"}
+{"input": "translation model is done by using Material| context: how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ? existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities .", "entity": "translation model", "output": "unimodal speech sequences", "neg_sample": ["translation model is done by using Material", "how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ?", "existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities ."], "relation": "used for", "id": "2022.acl-long.486", "year": 2022, "rel_sent": "Specifically , we mix up the representation sequences of different modalities , and take both unimodal speech sequences and multimodal mixed sequences as input to the translation model in parallel , and regularize their output predictions with a self - learning framework .", "forward": false, "src_ids": "2022.acl-long.486_2712"}
+{"input": "multimodal mixed sequences is used for Method| context: how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ? existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities .", "entity": "multimodal mixed sequences", "output": "translation model", "neg_sample": ["multimodal mixed sequences is used for Method", "how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ?", "existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities ."], "relation": "used for", "id": "2022.acl-long.486", "year": 2022, "rel_sent": "Specifically , we mix up the representation sequences of different modalities , and take both unimodal speech sequences and multimodal mixed sequences as input to the translation model in parallel , and regularize their output predictions with a self - learning framework .", "forward": true, "src_ids": "2022.acl-long.486_2713"}
+{"input": "unimodal speech sequences is used for Method| context: how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ? existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities .", "entity": "unimodal speech sequences", "output": "translation model", "neg_sample": ["unimodal speech sequences is used for Method", "how to learn a better speech representation for end - to - end speech - to - text translation ( st ) with limited labeled data ?", "existing techniques often attempt to transfer powerful machine translation ( mt ) capabilities to st , but neglect the representation discrepancy across modalities ."], "relation": "used for", "id": "2022.acl-long.486", "year": 2022, "rel_sent": "Specifically , we mix up the representation sequences of different modalities , and take both unimodal speech sequences and multimodal mixed sequences as input to the translation model in parallel , and regularize their output predictions with a self - learning framework .", "forward": true, "src_ids": "2022.acl-long.486_2714"}
+{"input": "dynamic learning process is done by using Method| context: detecting disclosures of individuals ' employment status on social media can provide valuable information to match job seekers with suitable vacancies , offer social protection , or measure labor market flows . however , identifying such personal disclosures is a challenging task due to their rarity in a sea of social media content and the variety of linguistic forms used to describe them .", "entity": "dynamic learning process", "output": "interpretable models", "neg_sample": ["dynamic learning process is done by using Method", "detecting disclosures of individuals ' employment status on social media can provide valuable information to match job seekers with suitable vacancies , offer social protection , or measure labor market flows .", "however , identifying such personal disclosures is a challenging task due to their rarity in a sea of social media content and the variety of linguistic forms used to describe them ."], "relation": "used for", "id": "2022.acl-long.453", "year": 2022, "rel_sent": "Qualitative analysis suggests that AL helps focus the attention mechanism of BERT on core terms and adjust the boundaries of semantic expansion , highlighting the importance of interpretable models to provide greater control and visibility into this dynamic learning process .", "forward": false, "src_ids": "2022.acl-long.453_2715"}
+{"input": "interpretable models is used for Task| context: detecting disclosures of individuals ' employment status on social media can provide valuable information to match job seekers with suitable vacancies , offer social protection , or measure labor market flows . however , identifying such personal disclosures is a challenging task due to their rarity in a sea of social media content and the variety of linguistic forms used to describe them .", "entity": "interpretable models", "output": "dynamic learning process", "neg_sample": ["interpretable models is used for Task", "detecting disclosures of individuals ' employment status on social media can provide valuable information to match job seekers with suitable vacancies , offer social protection , or measure labor market flows .", "however , identifying such personal disclosures is a challenging task due to their rarity in a sea of social media content and the variety of linguistic forms used to describe them ."], "relation": "used for", "id": "2022.acl-long.453", "year": 2022, "rel_sent": "Qualitative analysis suggests that AL helps focus the attention mechanism of BERT on core terms and adjust the boundaries of semantic expansion , highlighting the importance of interpretable models to provide greater control and visibility into this dynamic learning process .", "forward": true, "src_ids": "2022.acl-long.453_2716"}
+{"input": "neural network is done by using OtherScientificTerm| context: the evolution of language follows the rule of gradual change . grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap . as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation . although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era .", "entity": "neural network", "output": "switch - memory", "neg_sample": ["neural network is done by using OtherScientificTerm", "the evolution of language follows the rule of gradual change .", "grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap .", "as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation .", "although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era ."], "relation": "used for", "id": "2022.acl-long.540", "year": 2022, "rel_sent": "Further analyses also demonstrate that the SM can effectively integrate the knowledge of the eras into the neural network .", "forward": false, "src_ids": "2022.acl-long.540_2717"}
+{"input": "chinese word segmentation ( cws ) is done by using Method| context: the evolution of language follows the rule of gradual change . grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap . as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation . although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era .", "entity": "chinese word segmentation ( cws )", "output": "cross - era learning framework", "neg_sample": ["chinese word segmentation ( cws ) is done by using Method", "the evolution of language follows the rule of gradual change .", "grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap .", "as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation .", "although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era ."], "relation": "used for", "id": "2022.acl-long.540", "year": 2022, "rel_sent": "Therefore , we propose a cross - era learning framework for Chinese word segmentation ( CWS ) , CROSSWISE , which uses the Switch - memory ( SM ) module to incorporate era - specific linguistic knowledge .", "forward": false, "src_ids": "2022.acl-long.540_2718"}
+{"input": "cross - era learning framework is used for Task| context: the evolution of language follows the rule of gradual change . grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap . as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation . although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era .", "entity": "cross - era learning framework", "output": "chinese word segmentation ( cws )", "neg_sample": ["cross - era learning framework is used for Task", "the evolution of language follows the rule of gradual change .", "grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap .", "as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation .", "although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era ."], "relation": "used for", "id": "2022.acl-long.540", "year": 2022, "rel_sent": "Therefore , we propose a cross - era learning framework for Chinese word segmentation ( CWS ) , CROSSWISE , which uses the Switch - memory ( SM ) module to incorporate era - specific linguistic knowledge .", "forward": true, "src_ids": "2022.acl-long.540_2719"}
+{"input": "era - specific linguistic knowledge is done by using Method| context: the evolution of language follows the rule of gradual change . grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap . as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation . although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era .", "entity": "era - specific linguistic knowledge", "output": "switch - memory ( sm ) module", "neg_sample": ["era - specific linguistic knowledge is done by using Method", "the evolution of language follows the rule of gradual change .", "grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap .", "as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation .", "although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era ."], "relation": "used for", "id": "2022.acl-long.540", "year": 2022, "rel_sent": "Therefore , we propose a cross - era learning framework for Chinese word segmentation ( CWS ) , CROSSWISE , which uses the Switch - memory ( SM ) module to incorporate era - specific linguistic knowledge .", "forward": false, "src_ids": "2022.acl-long.540_2720"}
+{"input": "switch - memory ( sm ) module is used for OtherScientificTerm| context: the evolution of language follows the rule of gradual change . grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap . as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation . although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era .", "entity": "switch - memory ( sm ) module", "output": "era - specific linguistic knowledge", "neg_sample": ["switch - memory ( sm ) module is used for OtherScientificTerm", "the evolution of language follows the rule of gradual change .", "grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap .", "as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation .", "although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era ."], "relation": "used for", "id": "2022.acl-long.540", "year": 2022, "rel_sent": "Therefore , we propose a cross - era learning framework for Chinese word segmentation ( CWS ) , CROSSWISE , which uses the Switch - memory ( SM ) module to incorporate era - specific linguistic knowledge .", "forward": true, "src_ids": "2022.acl-long.540_2721"}
+{"input": "switch - memory is used for Method| context: the evolution of language follows the rule of gradual change . grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap . as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation . although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era .", "entity": "switch - memory", "output": "neural network", "neg_sample": ["switch - memory is used for Method", "the evolution of language follows the rule of gradual change .", "grammar , vocabulary , and lexical semantic shifts take place over time , resulting in a diachronic linguistic gap .", "as such , a considerable amount of texts are written in languages of different eras , which creates obstacles for natural language processing tasks , such as word segmentation and machine translation .", "although the chinese language has a long history , previous chinese natural language processing research has primarily focused on tasks within a specific era ."], "relation": "used for", "id": "2022.acl-long.540", "year": 2022, "rel_sent": "Further analyses also demonstrate that the SM can effectively integrate the knowledge of the eras into the neural network .", "forward": true, "src_ids": "2022.acl-long.540_2722"}
+{"input": "prompt - learning is done by using Method| context: prompt - learning has become a new paradigm in modern natural language processing , which directly adapts pre - trained language models ( plms ) to cloze - style prediction , autoregressive modeling , or sequence to sequence generation , resulting in promising performances on various tasks . however , no standard implementation framework of prompt - learning is proposed yet , and most existing prompt- learning codebases , often unregulated , only provide limited implementations for specific scenarios . since there are many details such as templating strategy , initializing strategy , verbalizing strategy , etc . , that need to be considered in prompt - learning , practitioners face impediments to quickly adapting the de - sired prompt learning methods to their applications .", "entity": "prompt - learning", "output": "open- prompt", "neg_sample": ["prompt - learning is done by using Method", "prompt - learning has become a new paradigm in modern natural language processing , which directly adapts pre - trained language models ( plms ) to cloze - style prediction , autoregressive modeling , or sequence to sequence generation , resulting in promising performances on various tasks .", "however , no standard implementation framework of prompt - learning is proposed yet , and most existing prompt- learning codebases , often unregulated , only provide limited implementations for specific scenarios .", "since there are many details such as templating strategy , initializing strategy , verbalizing strategy , etc .", ", that need to be considered in prompt - learning , practitioners face impediments to quickly adapting the de - sired prompt learning methods to their applications ."], "relation": "used for", "id": "2022.acl-demo.10", "year": 2022, "rel_sent": "In this paper , we present Open- Prompt , a unified easy - to - use toolkit to conduct prompt - learning over PLMs .", "forward": false, "src_ids": "2022.acl-demo.10_2723"}
+{"input": "named entity recognition is done by using Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "named entity recognition", "output": "boundary smoothing", "neg_sample": ["named entity recognition is done by using Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Boundary Smoothing for Named Entity Recognition.", "forward": false, "src_ids": "2022.acl-long.490_2724"}
+{"input": "span - based neural ner models is done by using Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "span - based neural ner models", "output": "boundary smoothing", "neg_sample": ["span - based neural ner models is done by using Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Inspired by label smoothing and driven by the ambiguity of boundary annotation in NER engineering , we propose boundary smoothing as a regularization technique for span - based neural NER models .", "forward": false, "src_ids": "2022.acl-long.490_2725"}
+{"input": "regularization technique is done by using Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "regularization technique", "output": "boundary smoothing", "neg_sample": ["regularization technique is done by using Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Inspired by label smoothing and driven by the ambiguity of boundary annotation in NER engineering , we propose boundary smoothing as a regularization technique for span - based neural NER models .", "forward": false, "src_ids": "2022.acl-long.490_2726"}
+{"input": "over - confidence is done by using Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "over - confidence", "output": "boundary smoothing", "neg_sample": ["over - confidence is done by using Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Further empirical analysis suggests that boundary smoothing effectively mitigates over - confidence , improves model calibration , and brings flatter neural minima and more smoothed loss landscapes .", "forward": false, "src_ids": "2022.acl-long.490_2727"}
+{"input": "model calibration is done by using Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "model calibration", "output": "boundary smoothing", "neg_sample": ["model calibration is done by using Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Further empirical analysis suggests that boundary smoothing effectively mitigates over - confidence , improves model calibration , and brings flatter neural minima and more smoothed loss landscapes .", "forward": false, "src_ids": "2022.acl-long.490_2728"}
+{"input": "boundary smoothing is used for Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "boundary smoothing", "output": "regularization technique", "neg_sample": ["boundary smoothing is used for Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Inspired by label smoothing and driven by the ambiguity of boundary annotation in NER engineering , we propose boundary smoothing as a regularization technique for span - based neural NER models .", "forward": true, "src_ids": "2022.acl-long.490_2729"}
+{"input": "regularization technique is used for Method| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "regularization technique", "output": "span - based neural ner models", "neg_sample": ["regularization technique is used for Method", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Inspired by label smoothing and driven by the ambiguity of boundary annotation in NER engineering , we propose boundary smoothing as a regularization technique for span - based neural NER models .", "forward": true, "src_ids": "2022.acl-long.490_2730"}
+{"input": "boundary smoothing is used for Task| context: neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration .", "entity": "boundary smoothing", "output": "model calibration", "neg_sample": ["boundary smoothing is used for Task", "neural named entity recognition ( ner ) models may easily encounter the over - confidence issue , which degrades the performance and calibration ."], "relation": "used for", "id": "2022.acl-long.490", "year": 2022, "rel_sent": "Further empirical analysis suggests that boundary smoothing effectively mitigates over - confidence , improves model calibration , and brings flatter neural minima and more smoothed loss landscapes .", "forward": true, "src_ids": "2022.acl-long.490_2731"}
+{"input": "targeted sentiment analysis is done by using Method| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "targeted sentiment analysis", "output": "bert - based models", "neg_sample": ["targeted sentiment analysis is done by using Method", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "A Dataset and BERT - based Models for Targeted Sentiment Analysis on Turkish Texts.", "forward": false, "src_ids": "2022.acl-srw.39_2732"}
+{"input": "targeted sentiment analysis is done by using Method| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "targeted sentiment analysis", "output": "bert - based models", "neg_sample": ["targeted sentiment analysis is done by using Method", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "We also propose BERT - based models with different architectures to accomplish the task of targeted sentiment analysis .", "forward": false, "src_ids": "2022.acl-srw.39_2733"}
+{"input": "sentiment is done by using Task| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "sentiment", "output": "targeted sentiment analysis", "neg_sample": ["sentiment is done by using Task", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "Targeted Sentiment Analysis aims to extract sentiment towards a particular target from a given text .", "forward": false, "src_ids": "2022.acl-srw.39_2734"}
+{"input": "bert - based models is used for Task| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "bert - based models", "output": "targeted sentiment analysis", "neg_sample": ["bert - based models is used for Task", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "A Dataset and BERT - based Models for Targeted Sentiment Analysis on Turkish Texts.", "forward": true, "src_ids": "2022.acl-srw.39_2735"}
+{"input": "annotated turkish dataset is used for Task| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "annotated turkish dataset", "output": "targeted sentiment analysis", "neg_sample": ["annotated turkish dataset is used for Task", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "We present an annotated Turkish dataset suitable for targeted sentiment analysis .", "forward": true, "src_ids": "2022.acl-srw.39_2736"}
+{"input": "bert - based models is used for Task| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "bert - based models", "output": "targeted sentiment analysis", "neg_sample": ["bert - based models is used for Task", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "We also propose BERT - based models with different architectures to accomplish the task of targeted sentiment analysis .", "forward": true, "src_ids": "2022.acl-srw.39_2737"}
+{"input": "targeted sentiment analysis is used for OtherScientificTerm| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "targeted sentiment analysis", "output": "sentiment", "neg_sample": ["targeted sentiment analysis is used for OtherScientificTerm", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "Targeted Sentiment Analysis aims to extract sentiment towards a particular target from a given text .", "forward": true, "src_ids": "2022.acl-srw.39_2738"}
+{"input": "targeted sentiment analysis is done by using Material| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "targeted sentiment analysis", "output": "annotated turkish dataset", "neg_sample": ["targeted sentiment analysis is done by using Material", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "We present an annotated Turkish dataset suitable for targeted sentiment analysis .", "forward": false, "src_ids": "2022.acl-srw.39_2739"}
+{"input": "targeted sentiment analysis task is done by using Method| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "targeted sentiment analysis task", "output": "sentiment analysis models", "neg_sample": ["targeted sentiment analysis task is done by using Method", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "The results demonstrate that the proposed models outperform the traditional sentiment analysis models for the targeted sentiment analysis task .", "forward": false, "src_ids": "2022.acl-srw.39_2740"}
+{"input": "sentiment analysis models is used for Task| context: it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data . sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english . for low - resource languages such as turkish , there is a lack of such annotated data .", "entity": "sentiment analysis models", "output": "targeted sentiment analysis task", "neg_sample": ["sentiment analysis models is used for Task", "it is a field that is attracting attention due to the increasing accessibility of the internet , which leads people to generate an enormous amount of data .", "sentiment analysis , which in general requires annotated data for training , is a well - researched area for widely studied languages such as english .", "for low - resource languages such as turkish , there is a lack of such annotated data ."], "relation": "used for", "id": "2022.acl-srw.39", "year": 2022, "rel_sent": "The results demonstrate that the proposed models outperform the traditional sentiment analysis models for the targeted sentiment analysis task .", "forward": true, "src_ids": "2022.acl-srw.39_2741"}
+{"input": "linear mixed model is used for OtherScientificTerm| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "linear mixed model", "output": "word durations", "neg_sample": ["linear mixed model is used for OtherScientificTerm", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Using a Linear Mixed Model containing memory and surprisal costs as predictors of word duration in read aloud speech ( parts - of - speech and speakers being intercept terms ) , we investigate the following hypotheses : 1 .", "forward": true, "src_ids": "2022.scil-1.10_2742"}
+{"input": "linguistic complexity measures is used for OtherScientificTerm| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "linguistic complexity measures", "output": "word durations", "neg_sample": ["linguistic complexity measures is used for OtherScientificTerm", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "High values of linguistic complexity measures ( lex - ical+PCFG surprisal and DLT memory costs ) lead to high word durations .", "forward": true, "src_ids": "2022.scil-1.10_2743"}
+{"input": "articulatory planning is done by using OtherScientificTerm| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "articulatory planning", "output": "forward surprisal", "neg_sample": ["articulatory planning is done by using OtherScientificTerm", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Forward surprisal aims to capture articulatory planning when readers incorporate parafoveal viewing during reading aloud .", "forward": false, "src_ids": "2022.scil-1.10_2744"}
+{"input": "forward surprisal is used for Task| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "forward surprisal", "output": "articulatory planning", "neg_sample": ["forward surprisal is used for Task", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Forward surprisal aims to capture articulatory planning when readers incorporate parafoveal viewing during reading aloud .", "forward": true, "src_ids": "2022.scil-1.10_2745"}
+{"input": "reading aloud is done by using OtherScientificTerm| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "reading aloud", "output": "parafoveal viewing", "neg_sample": ["reading aloud is done by using OtherScientificTerm", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Forward surprisal aims to capture articulatory planning when readers incorporate parafoveal viewing during reading aloud .", "forward": false, "src_ids": "2022.scil-1.10_2746"}
+{"input": "word durations is done by using Method| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "word durations", "output": "linear mixed model", "neg_sample": ["word durations is done by using Method", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Using a Linear Mixed Model containing memory and surprisal costs as predictors of word duration in read aloud speech ( parts - of - speech and speakers being intercept terms ) , we investigate the following hypotheses : 1 .", "forward": false, "src_ids": "2022.scil-1.10_2747"}
+{"input": "word durations is done by using Metric| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "word durations", "output": "linguistic complexity measures", "neg_sample": ["word durations is done by using Metric", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "High values of linguistic complexity measures ( lex - ical+PCFG surprisal and DLT memory costs ) lead to high word durations .", "forward": false, "src_ids": "2022.scil-1.10_2748"}
+{"input": "content and function word labels is done by using Method| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "content and function word labels", "output": "generalized linear model", "neg_sample": ["content and function word labels is done by using Method", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Further , using a Generalized Linear Model to predict content and function word labels we show that lexical surprisal measures do not help distinguish between these 2 classes .", "forward": false, "src_ids": "2022.scil-1.10_2749"}
+{"input": "generalized linear model is used for OtherScientificTerm| context: reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words .", "entity": "generalized linear model", "output": "content and function word labels", "neg_sample": ["generalized linear model is used for OtherScientificTerm", "reading aloud involves both comprehension and production processes , and we use measures defined by two influential theories of sentence comprehension , surprisal theory and dependency locality theory , to model the time taken to enunciate individual words ."], "relation": "used for", "id": "2022.scil-1.10", "year": 2022, "rel_sent": "Further , using a Generalized Linear Model to predict content and function word labels we show that lexical surprisal measures do not help distinguish between these 2 classes .", "forward": true, "src_ids": "2022.scil-1.10_2750"}
+{"input": "multi - class classifier is done by using Method| context: in this paper , we study the named entity recognition ( ner ) problem under distant supervision . due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate .", "entity": "multi - class classifier", "output": "conf - mpu risk estimation", "neg_sample": ["multi - class classifier is done by using Method", "in this paper , we study the named entity recognition ( ner ) problem under distant supervision .", "due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate ."], "relation": "used for", "id": "2022.acl-long.498", "year": 2022, "rel_sent": "Then , the proposed Conf - MPU risk estimation is applied to train a multi - class classifier for the NER task .", "forward": false, "src_ids": "2022.acl-long.498_2751"}
+{"input": "ner task is done by using Method| context: in this paper , we study the named entity recognition ( ner ) problem under distant supervision . due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate .", "entity": "ner task", "output": "conf - mpu risk estimation", "neg_sample": ["ner task is done by using Method", "in this paper , we study the named entity recognition ( ner ) problem under distant supervision .", "due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate ."], "relation": "used for", "id": "2022.acl-long.498", "year": 2022, "rel_sent": "Then , the proposed Conf - MPU risk estimation is applied to train a multi - class classifier for the NER task .", "forward": false, "src_ids": "2022.acl-long.498_2752"}
+{"input": "conf - mpu risk estimation is used for Method| context: in this paper , we study the named entity recognition ( ner ) problem under distant supervision . due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate .", "entity": "conf - mpu risk estimation", "output": "multi - class classifier", "neg_sample": ["conf - mpu risk estimation is used for Method", "in this paper , we study the named entity recognition ( ner ) problem under distant supervision .", "due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate ."], "relation": "used for", "id": "2022.acl-long.498", "year": 2022, "rel_sent": "Then , the proposed Conf - MPU risk estimation is applied to train a multi - class classifier for the NER task .", "forward": true, "src_ids": "2022.acl-long.498_2753"}
+{"input": "multi - class classifier is used for Task| context: in this paper , we study the named entity recognition ( ner ) problem under distant supervision . due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate .", "entity": "multi - class classifier", "output": "ner task", "neg_sample": ["multi - class classifier is used for Task", "in this paper , we study the named entity recognition ( ner ) problem under distant supervision .", "due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate ."], "relation": "used for", "id": "2022.acl-long.498", "year": 2022, "rel_sent": "Then , the proposed Conf - MPU risk estimation is applied to train a multi - class classifier for the NER task .", "forward": true, "src_ids": "2022.acl-long.498_2754"}
+{"input": "conf - mpu risk estimation is used for Task| context: in this paper , we study the named entity recognition ( ner ) problem under distant supervision . due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate .", "entity": "conf - mpu risk estimation", "output": "ner task", "neg_sample": ["conf - mpu risk estimation is used for Task", "in this paper , we study the named entity recognition ( ner ) problem under distant supervision .", "due to the incompleteness of the external dictionaries and/or knowledge bases , such distantly annotated training data usually suffer from a high false negative rate ."], "relation": "used for", "id": "2022.acl-long.498", "year": 2022, "rel_sent": "Then , the proposed Conf - MPU risk estimation is applied to train a multi - class classifier for the NER task .", "forward": true, "src_ids": "2022.acl-long.498_2755"}
+{"input": "language models is done by using Method| context: multi - modal techniques offer significant untapped potential to unlock improved nlp technology for local languages . however , many advances in language model pre - training are focused on text , a fact that only increases systematic inequalities in the performance of nlp tasks across the world 's languages .", "entity": "language models", "output": "multi - modal approach", "neg_sample": ["language models is done by using Method", "multi - modal techniques offer significant untapped potential to unlock improved nlp technology for local languages .", "however , many advances in language model pre - training are focused on text , a fact that only increases systematic inequalities in the performance of nlp tasks across the world 's languages ."], "relation": "used for", "id": "2022.acl-long.364", "year": 2022, "rel_sent": "In this work , we propose a multi - modal approach to train language models using whatever text and/or audio data might be available in a language .", "forward": false, "src_ids": "2022.acl-long.364_2756"}
+{"input": "multi - modal approach is used for Method| context: multi - modal techniques offer significant untapped potential to unlock improved nlp technology for local languages . however , many advances in language model pre - training are focused on text , a fact that only increases systematic inequalities in the performance of nlp tasks across the world 's languages .", "entity": "multi - modal approach", "output": "language models", "neg_sample": ["multi - modal approach is used for Method", "multi - modal techniques offer significant untapped potential to unlock improved nlp technology for local languages .", "however , many advances in language model pre - training are focused on text , a fact that only increases systematic inequalities in the performance of nlp tasks across the world 's languages ."], "relation": "used for", "id": "2022.acl-long.364", "year": 2022, "rel_sent": "In this work , we propose a multi - modal approach to train language models using whatever text and/or audio data might be available in a language .", "forward": true, "src_ids": "2022.acl-long.364_2757"}
+{"input": "summary templates is used for Task| context: annotating task - oriented dialogues is notorious for the expensive and difficult data collection process . few - shot dialogue state tracking ( dst ) is a realistic solution to this problem .", "entity": "summary templates", "output": "training", "neg_sample": ["summary templates is used for Task", "annotating task - oriented dialogues is notorious for the expensive and difficult data collection process .", "few - shot dialogue state tracking ( dst ) is a realistic solution to this problem ."], "relation": "used for", "id": "2022.findings-acl.302", "year": 2022, "rel_sent": "Our method also exhibits vast speedup during both training and inference as it can generate all states at once . Finally , based on our analysis , we discover that the naturalness of the summary templates plays a key role for successful training .", "forward": true, "src_ids": "2022.findings-acl.302_2758"}
+{"input": "few - shot dialogue state tracking is done by using Task| context: annotating task - oriented dialogues is notorious for the expensive and difficult data collection process . few - shot dialogue state tracking ( dst ) is a realistic solution to this problem .", "entity": "few - shot dialogue state tracking", "output": "template - guided summarization", "neg_sample": ["few - shot dialogue state tracking is done by using Task", "annotating task - oriented dialogues is notorious for the expensive and difficult data collection process .", "few - shot dialogue state tracking ( dst ) is a realistic solution to this problem ."], "relation": "used for", "id": "2022.findings-acl.302", "year": 2022, "rel_sent": "Dialogue Summaries as Dialogue States ( DS2 ) , Template - Guided Summarization for Few - shot Dialogue State Tracking.", "forward": false, "src_ids": "2022.findings-acl.302_2759"}
+{"input": "template - guided summarization is used for Task| context: annotating task - oriented dialogues is notorious for the expensive and difficult data collection process .", "entity": "template - guided summarization", "output": "few - shot dialogue state tracking", "neg_sample": ["template - guided summarization is used for Task", "annotating task - oriented dialogues is notorious for the expensive and difficult data collection process ."], "relation": "used for", "id": "2022.findings-acl.302", "year": 2022, "rel_sent": "Dialogue Summaries as Dialogue States ( DS2 ) , Template - Guided Summarization for Few - shot Dialogue State Tracking.", "forward": true, "src_ids": "2022.findings-acl.302_2760"}
+{"input": "training is done by using OtherScientificTerm| context: annotating task - oriented dialogues is notorious for the expensive and difficult data collection process . few - shot dialogue state tracking ( dst ) is a realistic solution to this problem .", "entity": "training", "output": "summary templates", "neg_sample": ["training is done by using OtherScientificTerm", "annotating task - oriented dialogues is notorious for the expensive and difficult data collection process .", "few - shot dialogue state tracking ( dst ) is a realistic solution to this problem ."], "relation": "used for", "id": "2022.findings-acl.302", "year": 2022, "rel_sent": "Our method also exhibits vast speedup during both training and inference as it can generate all states at once . Finally , based on our analysis , we discover that the naturalness of the summary templates plays a key role for successful training .", "forward": false, "src_ids": "2022.findings-acl.302_2761"}
+{"input": "feature - based model selection is done by using Method| context: selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning . to accelerate this process , researchers propose feature - based model selection ( fms ) methods , which assess ptms ' transferability to a specific task in a fast way without fine - tuning .", "entity": "feature - based model selection", "output": "backdoor attack", "neg_sample": ["feature - based model selection is done by using Method", "selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning .", "to accelerate this process , researchers propose feature - based model selection ( fms ) methods , which assess ptms ' transferability to a specific task in a fast way without fine - tuning ."], "relation": "used for", "id": "2022.acl-long.347", "year": 2022, "rel_sent": "Moreover , we find that these two methods can further be combined with the backdoor attack to misguide the FMS to select poisoned models .", "forward": false, "src_ids": "2022.acl-long.347_2762"}
+{"input": "poisoned models is done by using Method| context: selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning .", "entity": "poisoned models", "output": "feature - based model selection", "neg_sample": ["poisoned models is done by using Method", "selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning ."], "relation": "used for", "id": "2022.acl-long.347", "year": 2022, "rel_sent": "Moreover , we find that these two methods can further be combined with the backdoor attack to misguide the FMS to select poisoned models .", "forward": false, "src_ids": "2022.acl-long.347_2763"}
+{"input": "backdoor attack is used for Method| context: selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning .", "entity": "backdoor attack", "output": "feature - based model selection", "neg_sample": ["backdoor attack is used for Method", "selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning ."], "relation": "used for", "id": "2022.acl-long.347", "year": 2022, "rel_sent": "Moreover , we find that these two methods can further be combined with the backdoor attack to misguide the FMS to select poisoned models .", "forward": true, "src_ids": "2022.acl-long.347_2764"}
+{"input": "feature - based model selection is used for OtherScientificTerm| context: selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning . to accelerate this process , researchers propose feature - based model selection ( fms ) methods , which assess ptms ' transferability to a specific task in a fast way without fine - tuning .", "entity": "feature - based model selection", "output": "poisoned models", "neg_sample": ["feature - based model selection is used for OtherScientificTerm", "selecting an appropriate pre - trained model ( ptm ) for a specific downstream task typically requires significant efforts of fine - tuning .", "to accelerate this process , researchers propose feature - based model selection ( fms ) methods , which assess ptms ' transferability to a specific task in a fast way without fine - tuning ."], "relation": "used for", "id": "2022.acl-long.347", "year": 2022, "rel_sent": "Moreover , we find that these two methods can further be combined with the backdoor attack to misguide the FMS to select poisoned models .", "forward": true, "src_ids": "2022.acl-long.347_2765"}
+{"input": "multilingual media monitoring is done by using Method| context: monitio is a real - time crosslingual global media monitoring platform which delivers actionable insights beyond human scale and capabilities .", "entity": "multilingual media monitoring", "output": "monitio - large scale mt", "neg_sample": ["multilingual media monitoring is done by using Method", "monitio is a real - time crosslingual global media monitoring platform which delivers actionable insights beyond human scale and capabilities ."], "relation": "used for", "id": "2022.eamt-1.71", "year": 2022, "rel_sent": "Monitio - Large Scale MT for Multilingual Media Monitoring.", "forward": false, "src_ids": "2022.eamt-1.71_2766"}
+{"input": "monitio - large scale mt is used for Task| context: monitio is a real - time crosslingual global media monitoring platform which delivers actionable insights beyond human scale and capabilities .", "entity": "monitio - large scale mt", "output": "multilingual media monitoring", "neg_sample": ["monitio - large scale mt is used for Task", "monitio is a real - time crosslingual global media monitoring platform which delivers actionable insights beyond human scale and capabilities ."], "relation": "used for", "id": "2022.eamt-1.71", "year": 2022, "rel_sent": "Monitio - Large Scale MT for Multilingual Media Monitoring.", "forward": true, "src_ids": "2022.eamt-1.71_2767"}
+{"input": "compression methods is used for Method| context: the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression . we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights .", "entity": "compression methods", "output": "generative plms", "neg_sample": ["compression methods is used for Method", "the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression .", "we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights ."], "relation": "used for", "id": "2022.acl-long.331", "year": 2022, "rel_sent": "Empirical results on various tasks show that our proposed method outperforms the state - of - the - art compression methods on generative PLMs by a clear margin .", "forward": true, "src_ids": "2022.acl-long.331_2768"}
+{"input": "distinguishable word embeddings is done by using Method| context: the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression . despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear . in this paper , we compress generative plms by quantization . we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights .", "entity": "distinguishable word embeddings", "output": "token - level contrastive distillation", "neg_sample": ["distinguishable word embeddings is done by using Method", "the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression .", "despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear .", "in this paper , we compress generative plms by quantization .", "we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights ."], "relation": "used for", "id": "2022.acl-long.331", "year": 2022, "rel_sent": "Correspondingly , we propose a token - level contrastive distillation to learn distinguishable word embeddings , and a module - wise dynamic scaling to make quantizers adaptive to different modules .", "forward": false, "src_ids": "2022.acl-long.331_2769"}
+{"input": "token - level contrastive distillation is used for OtherScientificTerm| context: the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression . despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear . in this paper , we compress generative plms by quantization . we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights .", "entity": "token - level contrastive distillation", "output": "distinguishable word embeddings", "neg_sample": ["token - level contrastive distillation is used for OtherScientificTerm", "the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression .", "despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear .", "in this paper , we compress generative plms by quantization .", "we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights ."], "relation": "used for", "id": "2022.acl-long.331", "year": 2022, "rel_sent": "Correspondingly , we propose a token - level contrastive distillation to learn distinguishable word embeddings , and a module - wise dynamic scaling to make quantizers adaptive to different modules .", "forward": true, "src_ids": "2022.acl-long.331_2770"}
+{"input": "quantizers is done by using Method| context: the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression . despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear . in this paper , we compress generative plms by quantization . we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights .", "entity": "quantizers", "output": "module - wise dynamic scaling", "neg_sample": ["quantizers is done by using Method", "the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression .", "despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear .", "in this paper , we compress generative plms by quantization .", "we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights ."], "relation": "used for", "id": "2022.acl-long.331", "year": 2022, "rel_sent": "Correspondingly , we propose a token - level contrastive distillation to learn distinguishable word embeddings , and a module - wise dynamic scaling to make quantizers adaptive to different modules .", "forward": false, "src_ids": "2022.acl-long.331_2771"}
+{"input": "module - wise dynamic scaling is used for OtherScientificTerm| context: the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression . despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear . in this paper , we compress generative plms by quantization . we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights .", "entity": "module - wise dynamic scaling", "output": "quantizers", "neg_sample": ["module - wise dynamic scaling is used for OtherScientificTerm", "the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression .", "despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear .", "in this paper , we compress generative plms by quantization .", "we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights ."], "relation": "used for", "id": "2022.acl-long.331", "year": 2022, "rel_sent": "Correspondingly , we propose a token - level contrastive distillation to learn distinguishable word embeddings , and a module - wise dynamic scaling to make quantizers adaptive to different modules .", "forward": true, "src_ids": "2022.acl-long.331_2772"}
+{"input": "generative plms is done by using Method| context: the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression . despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear . in this paper , we compress generative plms by quantization . we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights .", "entity": "generative plms", "output": "compression methods", "neg_sample": ["generative plms is done by using Method", "the increasing size of generative pre - trained language models ( plms ) have greatly increased the demand for model compression .", "despite various methods to compress bert or its variants , there are few attempts to compress generative plms , and the underlying difficulty remains unclear .", "in this paper , we compress generative plms by quantization .", "we find that previous quantization methods fail on generative tasks due to the homogeneous word embeddings caused by reduced capacity and the varied distribution of weights ."], "relation": "used for", "id": "2022.acl-long.331", "year": 2022, "rel_sent": "Empirical results on various tasks show that our proposed method outperforms the state - of - the - art compression methods on generative PLMs by a clear margin .", "forward": false, "src_ids": "2022.acl-long.331_2773"}
+{"input": "legal language understanding is done by using Material| context: laws and their interpretations , legal arguments and agreements are typically expressed in writing , leading to the production of vast corpora of legal text . their analysis , which is at the center of legal practice , becomes increasingly elaborate as these collections grow in size . natural language understanding ( nlu ) technologies can be a valuable tool to support legal practitioners in these endeavors . their usefulness , however , largely depends on whether current state - of - the - art models can generalize across various tasks in the legal domain .", "entity": "legal language understanding", "output": "lexglue", "neg_sample": ["legal language understanding is done by using Material", "laws and their interpretations , legal arguments and agreements are typically expressed in writing , leading to the production of vast corpora of legal text .", "their analysis , which is at the center of legal practice , becomes increasingly elaborate as these collections grow in size .", "natural language understanding ( nlu ) technologies can be a valuable tool to support legal practitioners in these endeavors .", "their usefulness , however , largely depends on whether current state - of - the - art models can generalize across various tasks in the legal domain ."], "relation": "used for", "id": "2022.acl-long.297", "year": 2022, "rel_sent": "LexGLUE : A Benchmark Dataset for Legal Language Understanding in English.", "forward": false, "src_ids": "2022.acl-long.297_2774"}
+{"input": "lexglue is used for Task| context: laws and their interpretations , legal arguments and agreements are typically expressed in writing , leading to the production of vast corpora of legal text . their analysis , which is at the center of legal practice , becomes increasingly elaborate as these collections grow in size . natural language understanding ( nlu ) technologies can be a valuable tool to support legal practitioners in these endeavors . their usefulness , however , largely depends on whether current state - of - the - art models can generalize across various tasks in the legal domain .", "entity": "lexglue", "output": "legal language understanding", "neg_sample": ["lexglue is used for Task", "laws and their interpretations , legal arguments and agreements are typically expressed in writing , leading to the production of vast corpora of legal text .", "their analysis , which is at the center of legal practice , becomes increasingly elaborate as these collections grow in size .", "natural language understanding ( nlu ) technologies can be a valuable tool to support legal practitioners in these endeavors .", "their usefulness , however , largely depends on whether current state - of - the - art models can generalize across various tasks in the legal domain ."], "relation": "used for", "id": "2022.acl-long.297", "year": 2022, "rel_sent": "LexGLUE : A Benchmark Dataset for Legal Language Understanding in English.", "forward": true, "src_ids": "2022.acl-long.297_2775"}
+{"input": "program induction is done by using OtherScientificTerm| context: learning to induce programs relies on a large number of parallel question - program pairs for the given kb .", "entity": "program induction", "output": "program annotations", "neg_sample": ["program induction is done by using OtherScientificTerm", "learning to induce programs relies on a large number of parallel question - program pairs for the given kb ."], "relation": "used for", "id": "2022.acl-long.559", "year": 2022, "rel_sent": "In this paper , we propose the approach of program transfer , which aims to leverage the valuable program annotations on the rich - resourced KBs as external supervision signals to aid program induction for the low - resourced KBs that lack program annotations .", "forward": false, "src_ids": "2022.acl-long.559_2776"}
+{"input": "program annotations is used for Task| context: learning to induce programs relies on a large number of parallel question - program pairs for the given kb . however , for most kbs , the gold program annotations are usually lacking , making learning difficult .", "entity": "program annotations", "output": "program induction", "neg_sample": ["program annotations is used for Task", "learning to induce programs relies on a large number of parallel question - program pairs for the given kb .", "however , for most kbs , the gold program annotations are usually lacking , making learning difficult ."], "relation": "used for", "id": "2022.acl-long.559", "year": 2022, "rel_sent": "In this paper , we propose the approach of program transfer , which aims to leverage the valuable program annotations on the rich - resourced KBs as external supervision signals to aid program induction for the low - resourced KBs that lack program annotations .", "forward": true, "src_ids": "2022.acl-long.559_2777"}
+{"input": "high - level program sketch is done by using Method| context: learning to induce programs relies on a large number of parallel question - program pairs for the given kb . however , for most kbs , the gold program annotations are usually lacking , making learning difficult .", "entity": "high - level program sketch", "output": "sketch parser", "neg_sample": ["high - level program sketch is done by using Method", "learning to induce programs relies on a large number of parallel question - program pairs for the given kb .", "however , for most kbs , the gold program annotations are usually lacking , making learning difficult ."], "relation": "used for", "id": "2022.acl-long.559", "year": 2022, "rel_sent": "First , a sketch parser translates the question into a high - level program sketch , which is the composition of functions .", "forward": false, "src_ids": "2022.acl-long.559_2778"}
+{"input": "sketch parser is used for OtherScientificTerm| context: learning to induce programs relies on a large number of parallel question - program pairs for the given kb . however , for most kbs , the gold program annotations are usually lacking , making learning difficult .", "entity": "sketch parser", "output": "high - level program sketch", "neg_sample": ["sketch parser is used for OtherScientificTerm", "learning to induce programs relies on a large number of parallel question - program pairs for the given kb .", "however , for most kbs , the gold program annotations are usually lacking , making learning difficult ."], "relation": "used for", "id": "2022.acl-long.559", "year": 2022, "rel_sent": "First , a sketch parser translates the question into a high - level program sketch , which is the composition of functions .", "forward": true, "src_ids": "2022.acl-long.559_2779"}
+{"input": "search space is done by using OtherScientificTerm| context: learning to induce programs relies on a large number of parallel question - program pairs for the given kb . however , for most kbs , the gold program annotations are usually lacking , making learning difficult .", "entity": "search space", "output": "kb ontology", "neg_sample": ["search space is done by using OtherScientificTerm", "learning to induce programs relies on a large number of parallel question - program pairs for the given kb .", "however , for most kbs , the gold program annotations are usually lacking , making learning difficult ."], "relation": "used for", "id": "2022.acl-long.559", "year": 2022, "rel_sent": "During the searching , we incorporate the KB ontology to prune the search space .", "forward": false, "src_ids": "2022.acl-long.559_2780"}
+{"input": "kb ontology is used for OtherScientificTerm| context: learning to induce programs relies on a large number of parallel question - program pairs for the given kb . however , for most kbs , the gold program annotations are usually lacking , making learning difficult .", "entity": "kb ontology", "output": "search space", "neg_sample": ["kb ontology is used for OtherScientificTerm", "learning to induce programs relies on a large number of parallel question - program pairs for the given kb .", "however , for most kbs , the gold program annotations are usually lacking , making learning difficult ."], "relation": "used for", "id": "2022.acl-long.559", "year": 2022, "rel_sent": "During the searching , we incorporate the KB ontology to prune the search space .", "forward": true, "src_ids": "2022.acl-long.559_2781"}
+{"input": "event annotations is used for Method| context: event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols . these approaches are usually limited to a set of pre - defined types .", "entity": "event annotations", "output": "unified model", "neg_sample": ["event annotations is used for Method", "event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols .", "these approaches are usually limited to a set of pre - defined types ."], "relation": "used for", "id": "2022.findings-acl.16", "year": 2022, "rel_sent": "Furthermore , the query - and - extract formulation allows our approach to leverage all available event annotations from various ontologies as a unified model .", "forward": true, "src_ids": "2022.findings-acl.16_2782"}
+{"input": "semantic correlation is done by using Method| context: event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols . these approaches are usually limited to a set of pre - defined types .", "entity": "semantic correlation", "output": "attention mechanisms", "neg_sample": ["semantic correlation is done by using Method", "event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols .", "these approaches are usually limited to a set of pre - defined types ."], "relation": "used for", "id": "2022.findings-acl.16", "year": 2022, "rel_sent": "With the rich semantics in the queries , our framework benefits from the attention mechanisms to better capture the semantic correlation between the event types or argument roles and the input text .", "forward": false, "src_ids": "2022.findings-acl.16_2783"}
+{"input": "attention mechanisms is used for OtherScientificTerm| context: event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols . these approaches are usually limited to a set of pre - defined types .", "entity": "attention mechanisms", "output": "semantic correlation", "neg_sample": ["attention mechanisms is used for OtherScientificTerm", "event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols .", "these approaches are usually limited to a set of pre - defined types ."], "relation": "used for", "id": "2022.findings-acl.16", "year": 2022, "rel_sent": "With the rich semantics in the queries , our framework benefits from the attention mechanisms to better capture the semantic correlation between the event types or argument roles and the input text .", "forward": true, "src_ids": "2022.findings-acl.16_2784"}
+{"input": "unified model is done by using OtherScientificTerm| context: event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols . these approaches are usually limited to a set of pre - defined types .", "entity": "unified model", "output": "event annotations", "neg_sample": ["unified model is done by using OtherScientificTerm", "event extraction is typically modeled as a multi - class classification problem where event types and argument roles are treated as atomic symbols .", "these approaches are usually limited to a set of pre - defined types ."], "relation": "used for", "id": "2022.findings-acl.16", "year": 2022, "rel_sent": "Furthermore , the query - and - extract formulation allows our approach to leverage all available event annotations from various ontologies as a unified model .", "forward": false, "src_ids": "2022.findings-acl.16_2785"}
+{"input": "general - domain fine - tuning is used for Task| context: recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products .", "entity": "general - domain fine - tuning", "output": "generalization", "neg_sample": ["general - domain fine - tuning is used for Task", "recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products ."], "relation": "used for", "id": "2022.ecnlp-1.26", "year": 2022, "rel_sent": "In this paper , we examine several general - domain and domain - specific pre - trained Roberta variants and discover that general - domain fine - tuning does not really help generalization which aligns with the discovery of prior art , yet proper domain - specific fine - tuning with clickstream data can lead to better model generalization , based on a bucketed analysis of a manually annotated query - product relevance data .", "forward": true, "src_ids": "2022.ecnlp-1.26_2786"}
+{"input": "generalization is done by using Method| context: recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products . yet , whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far .", "entity": "generalization", "output": "general - domain fine - tuning", "neg_sample": ["generalization is done by using Method", "recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products .", "yet , whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far ."], "relation": "used for", "id": "2022.ecnlp-1.26", "year": 2022, "rel_sent": "In this paper , we examine several general - domain and domain - specific pre - trained Roberta variants and discover that general - domain fine - tuning does not really help generalization which aligns with the discovery of prior art , yet proper domain - specific fine - tuning with clickstream data can lead to better model generalization , based on a bucketed analysis of a manually annotated query - product relevance data .", "forward": false, "src_ids": "2022.ecnlp-1.26_2787"}
+{"input": "model generalization is done by using OtherScientificTerm| context: recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products . yet , whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far .", "entity": "model generalization", "output": "domain - specific fine - tuning", "neg_sample": ["model generalization is done by using OtherScientificTerm", "recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products .", "yet , whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far ."], "relation": "used for", "id": "2022.ecnlp-1.26", "year": 2022, "rel_sent": "In this paper , we examine several general - domain and domain - specific pre - trained Roberta variants and discover that general - domain fine - tuning does not really help generalization which aligns with the discovery of prior art , yet proper domain - specific fine - tuning with clickstream data can lead to better model generalization , based on a bucketed analysis of a manually annotated query - product relevance data .", "forward": false, "src_ids": "2022.ecnlp-1.26_2788"}
+{"input": "domain - specific fine - tuning is used for Task| context: recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products . yet , whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far .", "entity": "domain - specific fine - tuning", "output": "model generalization", "neg_sample": ["domain - specific fine - tuning is used for Task", "recently , semantic search has been successfully applied to e - commerce product search and the learned semantic space for query and product encoding are expected to generalize well to unseen queries or products .", "yet , whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far ."], "relation": "used for", "id": "2022.ecnlp-1.26", "year": 2022, "rel_sent": "In this paper , we examine several general - domain and domain - specific pre - trained Roberta variants and discover that general - domain fine - tuning does not really help generalization which aligns with the discovery of prior art , yet proper domain - specific fine - tuning with clickstream data can lead to better model generalization , based on a bucketed analysis of a manually annotated query - product relevance data .", "forward": true, "src_ids": "2022.ecnlp-1.26_2789"}
+{"input": "spurious correlations is done by using Task| context: natural language processing models often exploit spurious correlations between task - independent features and labels in datasets to perform well only within the distributions they are trained on , while not generalising to different task distributions .", "entity": "spurious correlations", "output": "generating data", "neg_sample": ["spurious correlations is done by using Task", "natural language processing models often exploit spurious correlations between task - independent features and labels in datasets to perform well only within the distributions they are trained on , while not generalising to different task distributions ."], "relation": "used for", "id": "2022.acl-long.190", "year": 2022, "rel_sent": "Generating Data to Mitigate Spurious Correlations in Natural Language Inference Datasets.", "forward": false, "src_ids": "2022.acl-long.190_2790"}
+{"input": "ok - transformer is used for Task| context: we point out that commonsense has the nature of domain discrepancy . namely , commonsense has different data formats and is domain - independent from the downstream task . this nature brings challenges to introducing commonsense in general text understanding tasks . a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus . however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy .", "entity": "ok - transformer", "output": "text representation", "neg_sample": ["ok - transformer is used for Task", "we point out that commonsense has the nature of domain discrepancy .", "namely , commonsense has different data formats and is domain - independent from the downstream task .", "this nature brings challenges to introducing commonsense in general text understanding tasks .", "a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus .", "however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy ."], "relation": "used for", "id": "2022.findings-acl.138", "year": 2022, "rel_sent": "OK - Transformer effectively integrates commonsense descriptions and enhances them to the target text representation .", "forward": true, "src_ids": "2022.findings-acl.138_2791"}
+{"input": "text representation is done by using Method| context: we study how to enhance text representation via textual commonsense . we point out that commonsense has the nature of domain discrepancy . namely , commonsense has different data formats and is domain - independent from the downstream task . this nature brings challenges to introducing commonsense in general text understanding tasks . a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus . however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy .", "entity": "text representation", "output": "ok - transformer", "neg_sample": ["text representation is done by using Method", "we study how to enhance text representation via textual commonsense .", "we point out that commonsense has the nature of domain discrepancy .", "namely , commonsense has different data formats and is domain - independent from the downstream task .", "this nature brings challenges to introducing commonsense in general text understanding tasks .", "a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus .", "however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy ."], "relation": "used for", "id": "2022.findings-acl.138", "year": 2022, "rel_sent": "OK - Transformer effectively integrates commonsense descriptions and enhances them to the target text representation .", "forward": false, "src_ids": "2022.findings-acl.138_2792"}
+{"input": "transformer - based language models is done by using Method| context: we study how to enhance text representation via textual commonsense . we point out that commonsense has the nature of domain discrepancy . namely , commonsense has different data formats and is domain - independent from the downstream task . this nature brings challenges to introducing commonsense in general text understanding tasks . a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus . however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy .", "entity": "transformer - based language models", "output": "ok - transformer", "neg_sample": ["transformer - based language models is done by using Method", "we study how to enhance text representation via textual commonsense .", "we point out that commonsense has the nature of domain discrepancy .", "namely , commonsense has different data formats and is domain - independent from the downstream task .", "this nature brings challenges to introducing commonsense in general text understanding tasks .", "a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus .", "however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy ."], "relation": "used for", "id": "2022.findings-acl.138", "year": 2022, "rel_sent": "In addition , OK - Transformer can adapt to the Transformer - based language models ( e.g.", "forward": false, "src_ids": "2022.findings-acl.138_2793"}
+{"input": "commonsense reasoning is done by using Method| context: we study how to enhance text representation via textual commonsense . we point out that commonsense has the nature of domain discrepancy . namely , commonsense has different data formats and is domain - independent from the downstream task . this nature brings challenges to introducing commonsense in general text understanding tasks . a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus . however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy .", "entity": "commonsense reasoning", "output": "ok - transformer", "neg_sample": ["commonsense reasoning is done by using Method", "we study how to enhance text representation via textual commonsense .", "we point out that commonsense has the nature of domain discrepancy .", "namely , commonsense has different data formats and is domain - independent from the downstream task .", "this nature brings challenges to introducing commonsense in general text understanding tasks .", "a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus .", "however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy ."], "relation": "used for", "id": "2022.findings-acl.138", "year": 2022, "rel_sent": "We have verified the effectiveness of OK - Transformer in multiple applications such as commonsense reasoning , general text classification , and low - resource commonsense settings .", "forward": false, "src_ids": "2022.findings-acl.138_2794"}
+{"input": "ok - transformer is used for Method| context: we study how to enhance text representation via textual commonsense . we point out that commonsense has the nature of domain discrepancy . namely , commonsense has different data formats and is domain - independent from the downstream task . this nature brings challenges to introducing commonsense in general text understanding tasks . a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus . however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy .", "entity": "ok - transformer", "output": "transformer - based language models", "neg_sample": ["ok - transformer is used for Method", "we study how to enhance text representation via textual commonsense .", "we point out that commonsense has the nature of domain discrepancy .", "namely , commonsense has different data formats and is domain - independent from the downstream task .", "this nature brings challenges to introducing commonsense in general text understanding tasks .", "a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus .", "however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy ."], "relation": "used for", "id": "2022.findings-acl.138", "year": 2022, "rel_sent": "In addition , OK - Transformer can adapt to the Transformer - based language models ( e.g.", "forward": true, "src_ids": "2022.findings-acl.138_2795"}
+{"input": "ok - transformer is used for Task| context: we study how to enhance text representation via textual commonsense . we point out that commonsense has the nature of domain discrepancy . namely , commonsense has different data formats and is domain - independent from the downstream task . this nature brings challenges to introducing commonsense in general text understanding tasks . a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus . however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy .", "entity": "ok - transformer", "output": "commonsense reasoning", "neg_sample": ["ok - transformer is used for Task", "we study how to enhance text representation via textual commonsense .", "we point out that commonsense has the nature of domain discrepancy .", "namely , commonsense has different data formats and is domain - independent from the downstream task .", "this nature brings challenges to introducing commonsense in general text understanding tasks .", "a typical method of introducing textual knowledge is continuing pre - training over the commonsense corpus .", "however , it will cause catastrophic forgetting to the downstream task due to the domain discrepancy ."], "relation": "used for", "id": "2022.findings-acl.138", "year": 2022, "rel_sent": "We have verified the effectiveness of OK - Transformer in multiple applications such as commonsense reasoning , general text classification , and low - resource commonsense settings .", "forward": true, "src_ids": "2022.findings-acl.138_2796"}
+{"input": "annotator errors is done by using Method| context: annotation errors that stem from various sources are usually unavoidable when performing large - scale annotation of linguistic data .", "entity": "annotator errors", "output": "transformer model", "neg_sample": ["annotator errors is done by using Method", "annotation errors that stem from various sources are usually unavoidable when performing large - scale annotation of linguistic data ."], "relation": "used for", "id": "2022.acl-short.19", "year": 2022, "rel_sent": "In this paper , we evaluate the feasibility of using the Transformer model to detect various types of annotator errors in morphological data sets that contain inflected word forms .", "forward": false, "src_ids": "2022.acl-short.19_2797"}
+{"input": "transformer model is used for Task| context: annotation errors that stem from various sources are usually unavoidable when performing large - scale annotation of linguistic data .", "entity": "transformer model", "output": "annotator errors", "neg_sample": ["transformer model is used for Task", "annotation errors that stem from various sources are usually unavoidable when performing large - scale annotation of linguistic data ."], "relation": "used for", "id": "2022.acl-short.19", "year": 2022, "rel_sent": "In this paper , we evaluate the feasibility of using the Transformer model to detect various types of annotator errors in morphological data sets that contain inflected word forms .", "forward": true, "src_ids": "2022.acl-short.19_2798"}
+{"input": "taco is used for Method| context: in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms .", "entity": "taco", "output": "contextualized representations", "neg_sample": ["taco is used for Method", "in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms ."], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To be specific , TACO extracts and aligns contextual semantics hidden in contextualized representations to encourage models to attend global semantics when generating contextualized representations .", "forward": true, "src_ids": "2022.acl-long.193_2799"}
+{"input": "taco is used for OtherScientificTerm| context: currently , masked language modeling ( e.g. , bert ) is the prime choice to learn contextualized representations . due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?", "entity": "taco", "output": "contextual semantics", "neg_sample": ["taco is used for OtherScientificTerm", "currently , masked language modeling ( e.g.", ", bert ) is the prime choice to learn contextualized representations .", "due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?"], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To be specific , TACO extracts and aligns contextual semantics hidden in contextualized representations to encourage models to attend global semantics when generating contextualized representations .", "forward": true, "src_ids": "2022.acl-long.193_2800"}
+{"input": "global semantics is done by using Method| context: currently , masked language modeling ( e.g. , bert ) is the prime choice to learn contextualized representations . due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ? in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms .", "entity": "global semantics", "output": "taco", "neg_sample": ["global semantics is done by using Method", "currently , masked language modeling ( e.g.", ", bert ) is the prime choice to learn contextualized representations .", "due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?", "in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms ."], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To address these problems , we propose TACO , a simple yet effective representation learning approach to directly model global semantics .", "forward": false, "src_ids": "2022.acl-long.193_2801"}
+{"input": "contextual semantics is done by using Method| context: currently , masked language modeling ( e.g. , bert ) is the prime choice to learn contextualized representations . due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ? in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms .", "entity": "contextual semantics", "output": "taco", "neg_sample": ["contextual semantics is done by using Method", "currently , masked language modeling ( e.g.", ", bert ) is the prime choice to learn contextualized representations .", "due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?", "in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms ."], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To be specific , TACO extracts and aligns contextual semantics hidden in contextualized representations to encourage models to attend global semantics when generating contextualized representations .", "forward": false, "src_ids": "2022.acl-long.193_2802"}
+{"input": "contextualized representations is done by using Method| context: currently , masked language modeling ( e.g. , bert ) is the prime choice to learn contextualized representations . due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ? in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms .", "entity": "contextualized representations", "output": "taco", "neg_sample": ["contextualized representations is done by using Method", "currently , masked language modeling ( e.g.", ", bert ) is the prime choice to learn contextualized representations .", "due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?", "in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms ."], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To be specific , TACO extracts and aligns contextual semantics hidden in contextualized representations to encourage models to attend global semantics when generating contextualized representations .", "forward": false, "src_ids": "2022.acl-long.193_2803"}
+{"input": "taco is used for OtherScientificTerm| context: currently , masked language modeling ( e.g. , bert ) is the prime choice to learn contextualized representations . due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ? in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms .", "entity": "taco", "output": "global semantics", "neg_sample": ["taco is used for OtherScientificTerm", "currently , masked language modeling ( e.g.", ", bert ) is the prime choice to learn contextualized representations .", "due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?", "in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms ."], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To address these problems , we propose TACO , a simple yet effective representation learning approach to directly model global semantics .", "forward": true, "src_ids": "2022.acl-long.193_2804"}
+{"input": "representation learning approach is used for OtherScientificTerm| context: currently , masked language modeling ( e.g. , bert ) is the prime choice to learn contextualized representations . due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ? in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms .", "entity": "representation learning approach", "output": "global semantics", "neg_sample": ["representation learning approach is used for OtherScientificTerm", "currently , masked language modeling ( e.g.", ", bert ) is the prime choice to learn contextualized representations .", "due to the pervasiveness , it naturally raises an interesting question : how do masked language models ( mlms ) learn contextual representations ?", "in this work , we analyze the learning dynamics of mlms and find that it adopts sampled embeddings as anchors to estimate and inject contextual semantics to representations , which limits the efficiency and effectiveness of mlms ."], "relation": "used for", "id": "2022.acl-long.193", "year": 2022, "rel_sent": "To address these problems , we propose TACO , a simple yet effective representation learning approach to directly model global semantics .", "forward": true, "src_ids": "2022.acl-long.193_2805"}
+{"input": "unimodal and multimodal tasks is done by using Method| context: clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks . few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition . here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans .", "entity": "unimodal and multimodal tasks", "output": "clip 's multimodal training", "neg_sample": ["unimodal and multimodal tasks is done by using Method", "clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks .", "few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition .", "here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans ."], "relation": "used for", "id": "2022.repl4nlp-1.4", "year": 2022, "rel_sent": "We conclude that CLIP 's multimodal training is beneficial for both unimodal and multimodal tasks that require classification of human - centric concepts .", "forward": false, "src_ids": "2022.repl4nlp-1.4_2806"}
+{"input": "clip 's multimodal training is used for Task| context: clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks . however , recent work showed that in comparison to unimodal ( visual ) networks , clip 's multimodal training does not benefit generalization ( e.g. few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition . here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans .", "entity": "clip 's multimodal training", "output": "unimodal and multimodal tasks", "neg_sample": ["clip 's multimodal training is used for Task", "clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks .", "however , recent work showed that in comparison to unimodal ( visual ) networks , clip 's multimodal training does not benefit generalization ( e.g.", "few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition .", "here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans ."], "relation": "used for", "id": "2022.repl4nlp-1.4", "year": 2022, "rel_sent": "We conclude that CLIP 's multimodal training is beneficial for both unimodal and multimodal tasks that require classification of human - centric concepts .", "forward": true, "src_ids": "2022.repl4nlp-1.4_2807"}
+{"input": "movie genre classification is done by using Material| context: clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks . however , recent work showed that in comparison to unimodal ( visual ) networks , clip 's multimodal training does not benefit generalization ( e.g. few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition . here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans .", "entity": "movie genre classification", "output": "multimodal dataset", "neg_sample": ["movie genre classification is done by using Material", "clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks .", "however , recent work showed that in comparison to unimodal ( visual ) networks , clip 's multimodal training does not benefit generalization ( e.g.", "few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition .", "here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans ."], "relation": "used for", "id": "2022.repl4nlp-1.4", "year": 2022, "rel_sent": "We introduce and publicly release a new multimodal dataset for movie genre classification .", "forward": false, "src_ids": "2022.repl4nlp-1.4_2808"}
+{"input": "multimodal dataset is used for Task| context: clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks . however , recent work showed that in comparison to unimodal ( visual ) networks , clip 's multimodal training does not benefit generalization ( e.g. few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition . here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans .", "entity": "multimodal dataset", "output": "movie genre classification", "neg_sample": ["multimodal dataset is used for Task", "clip , a vision - language network trained with a multimodal contrastive learning objective on a large dataset of images and captions , has demonstrated impressive zero - shot ability in various tasks .", "however , recent work showed that in comparison to unimodal ( visual ) networks , clip 's multimodal training does not benefit generalization ( e.g.", "few - shot or transfer learning ) for standard visual classification tasks such as object , street numbers or animal recognition .", "here , we hypothesize that clip 's improved unimodal generalization abilities may be most prominent in domains that involve human - centric concepts ( cultural , social , aesthetic , affective ... ) ; this is because clip 's training dataset is mainly composed of image annotations made by humans for other humans ."], "relation": "used for", "id": "2022.repl4nlp-1.4", "year": 2022, "rel_sent": "We introduce and publicly release a new multimodal dataset for movie genre classification .", "forward": true, "src_ids": "2022.repl4nlp-1.4_2809"}
+{"input": "bert - like models is done by using Method| context: large - scale pre - trained language models have shown remarkable results in diverse nlp applications . however , these performance gains have been accompanied by a significant increase in computation time and model size , stressing the need to develop new or complementary strategies to increase the efficiency of these models .", "entity": "bert - like models", "output": "differentiable adaptive computation time strategy", "neg_sample": ["bert - like models is done by using Method", "large - scale pre - trained language models have shown remarkable results in diverse nlp applications .", "however , these performance gains have been accompanied by a significant increase in computation time and model size , stressing the need to develop new or complementary strategies to increase the efficiency of these models ."], "relation": "used for", "id": "2022.nlppower-1.10", "year": 2022, "rel_sent": "This paper proposes DACT - BERT , a differentiable adaptive computation time strategy for BERT - like models .", "forward": false, "src_ids": "2022.nlppower-1.10_2810"}
+{"input": "differentiable adaptive computation time strategy is used for Method| context: large - scale pre - trained language models have shown remarkable results in diverse nlp applications . however , these performance gains have been accompanied by a significant increase in computation time and model size , stressing the need to develop new or complementary strategies to increase the efficiency of these models .", "entity": "differentiable adaptive computation time strategy", "output": "bert - like models", "neg_sample": ["differentiable adaptive computation time strategy is used for Method", "large - scale pre - trained language models have shown remarkable results in diverse nlp applications .", "however , these performance gains have been accompanied by a significant increase in computation time and model size , stressing the need to develop new or complementary strategies to increase the efficiency of these models ."], "relation": "used for", "id": "2022.nlppower-1.10", "year": 2022, "rel_sent": "This paper proposes DACT - BERT , a differentiable adaptive computation time strategy for BERT - like models .", "forward": true, "src_ids": "2022.nlppower-1.10_2811"}
+{"input": "homophobia / transphobia detection is done by using Method| context: sexual minorities face a lot of unfair treatment and discrimination in our world . this creates enormous stress and many psychological problems for sexual minorities . there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities .", "entity": "homophobia / transphobia detection", "output": "roberta -based approach", "neg_sample": ["homophobia / transphobia detection is done by using Method", "sexual minorities face a lot of unfair treatment and discrimination in our world .", "this creates enormous stress and many psychological problems for sexual minorities .", "there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities ."], "relation": "used for", "id": "2022.ltedi-1.19", "year": 2022, "rel_sent": "ABLIMET @LT - EDI - ACL2022 : A Roberta based Approach for Homophobia / Transphobia Detection in Social Media.", "forward": false, "src_ids": "2022.ltedi-1.19_2812"}
+{"input": "homophobia / transphobia detection is done by using Method| context: sexual minorities face a lot of unfair treatment and discrimination in our world . this creates enormous stress and many psychological problems for sexual minorities . there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities .", "entity": "homophobia / transphobia detection", "output": "roberta -based approach", "neg_sample": ["homophobia / transphobia detection is done by using Method", "sexual minorities face a lot of unfair treatment and discrimination in our world .", "this creates enormous stress and many psychological problems for sexual minorities .", "there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities ."], "relation": "used for", "id": "2022.ltedi-1.19", "year": 2022, "rel_sent": "We use a Roberta -based approach to conduct Homophobia/ Transphobia detection experiments on the dataset of the competition , and get better results .", "forward": false, "src_ids": "2022.ltedi-1.19_2813"}
+{"input": "roberta -based approach is used for Task| context: sexual minorities face a lot of unfair treatment and discrimination in our world . this creates enormous stress and many psychological problems for sexual minorities . there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities .", "entity": "roberta -based approach", "output": "homophobia / transphobia detection", "neg_sample": ["roberta -based approach is used for Task", "sexual minorities face a lot of unfair treatment and discrimination in our world .", "this creates enormous stress and many psychological problems for sexual minorities .", "there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities ."], "relation": "used for", "id": "2022.ltedi-1.19", "year": 2022, "rel_sent": "ABLIMET @LT - EDI - ACL2022 : A Roberta based Approach for Homophobia / Transphobia Detection in Social Media.", "forward": true, "src_ids": "2022.ltedi-1.19_2814"}
+{"input": "roberta -based approach is used for Task| context: sexual minorities face a lot of unfair treatment and discrimination in our world . this creates enormous stress and many psychological problems for sexual minorities . there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities .", "entity": "roberta -based approach", "output": "homophobia / transphobia detection", "neg_sample": ["roberta -based approach is used for Task", "sexual minorities face a lot of unfair treatment and discrimination in our world .", "this creates enormous stress and many psychological problems for sexual minorities .", "there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities ."], "relation": "used for", "id": "2022.ltedi-1.19", "year": 2022, "rel_sent": "We use a Roberta -based approach to conduct Homophobia/ Transphobia detection experiments on the dataset of the competition , and get better results .", "forward": true, "src_ids": "2022.ltedi-1.19_2815"}
+{"input": "natural language processing technology is used for OtherScientificTerm| context: sexual minorities face a lot of unfair treatment and discrimination in our world . this creates enormous stress and many psychological problems for sexual minorities .", "entity": "natural language processing technology", "output": "homophobia / transphobia", "neg_sample": ["natural language processing technology is used for OtherScientificTerm", "sexual minorities face a lot of unfair treatment and discrimination in our world .", "this creates enormous stress and many psychological problems for sexual minorities ."], "relation": "used for", "id": "2022.ltedi-1.19", "year": 2022, "rel_sent": "Identifying and processing Homophobia/ Transphobia through natural language processing technology can improve the efficiency of processing Homophobia/ Transphobia , and can quickly screen out Homophobia / Transphobia on the Internet .", "forward": true, "src_ids": "2022.ltedi-1.19_2816"}
+{"input": "homophobia / transphobia is done by using Method| context: sexual minorities face a lot of unfair treatment and discrimination in our world . this creates enormous stress and many psychological problems for sexual minorities . there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities .", "entity": "homophobia / transphobia", "output": "natural language processing technology", "neg_sample": ["homophobia / transphobia is done by using Method", "sexual minorities face a lot of unfair treatment and discrimination in our world .", "this creates enormous stress and many psychological problems for sexual minorities .", "there is a lot of hate speech on the internet , and homophobia / transphobia is the one against sexual minorities ."], "relation": "used for", "id": "2022.ltedi-1.19", "year": 2022, "rel_sent": "Identifying and processing Homophobia/ Transphobia through natural language processing technology can improve the efficiency of processing Homophobia/ Transphobia , and can quickly screen out Homophobia / Transphobia on the Internet .", "forward": false, "src_ids": "2022.ltedi-1.19_2817"}
+{"input": "development set is used for OtherScientificTerm| context: when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models .", "entity": "development set", "output": "permutations", "neg_sample": ["development set is used for OtherScientificTerm", "when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models ."], "relation": "used for", "id": "2022.acl-long.556", "year": 2022, "rel_sent": "While one could use a development set to determine which permutations are performant , this would deviate from the true few - shot setting as it requires additional annotated data .", "forward": true, "src_ids": "2022.acl-long.556_2818"}
+{"input": "permutations is done by using Material| context: when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models . we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not .", "entity": "permutations", "output": "development set", "neg_sample": ["permutations is done by using Material", "when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models .", "we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not ."], "relation": "used for", "id": "2022.acl-long.556", "year": 2022, "rel_sent": "While one could use a development set to determine which permutations are performant , this would deviate from the true few - shot setting as it requires additional annotated data .", "forward": false, "src_ids": "2022.acl-long.556_2819"}
+{"input": "artificial development set is done by using Method| context: we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not .", "entity": "artificial development set", "output": "language models", "neg_sample": ["artificial development set is done by using Method", "we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not ."], "relation": "used for", "id": "2022.acl-long.556", "year": 2022, "rel_sent": "Instead , we use the generative nature of language models to construct an artificial development set and based on entropy statistics of the candidate permutations on this set , we identify performant prompts .", "forward": false, "src_ids": "2022.acl-long.556_2820"}
+{"input": "language models is used for Material| context: when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models . we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not .", "entity": "language models", "output": "artificial development set", "neg_sample": ["language models is used for Material", "when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models .", "we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not ."], "relation": "used for", "id": "2022.acl-long.556", "year": 2022, "rel_sent": "Instead , we use the generative nature of language models to construct an artificial development set and based on entropy statistics of the candidate permutations on this set , we identify performant prompts .", "forward": true, "src_ids": "2022.acl-long.556_2821"}
+{"input": "performant prompts is done by using OtherScientificTerm| context: when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models . we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not .", "entity": "performant prompts", "output": "entropy statistics", "neg_sample": ["performant prompts is done by using OtherScientificTerm", "when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models .", "we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not ."], "relation": "used for", "id": "2022.acl-long.556", "year": 2022, "rel_sent": "Instead , we use the generative nature of language models to construct an artificial development set and based on entropy statistics of the candidate permutations on this set , we identify performant prompts .", "forward": false, "src_ids": "2022.acl-long.556_2822"}
+{"input": "entropy statistics is used for OtherScientificTerm| context: when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models . we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not .", "entity": "entropy statistics", "output": "performant prompts", "neg_sample": ["entropy statistics is used for OtherScientificTerm", "when primed with only a handful of training samples , very large , pretrained language models such as gpt-3 have shown competitive results when compared tofully - supervised , fine - tuned , large , pretrained language models .", "we demonstrate that the order in which the samples are provided can make the difference between near state - of - the - art and random guess performance : essentially some permutations are ' fantastic ' and some not ."], "relation": "used for", "id": "2022.acl-long.556", "year": 2022, "rel_sent": "Instead , we use the generative nature of language models to construct an artificial development set and based on entropy statistics of the candidate permutations on this set , we identify performant prompts .", "forward": true, "src_ids": "2022.acl-long.556_2823"}
+{"input": "chinese pre - trained language models is done by using Method| context: chinese pre - trained language models usually exploit contextual character information to learn representations , while ignoring the linguistics knowledge , e.g. , word and sentence information .", "entity": "chinese pre - trained language models", "output": "task - free enhancement module", "neg_sample": ["chinese pre - trained language models is done by using Method", "chinese pre - trained language models usually exploit contextual character information to learn representations , while ignoring the linguistics knowledge , e.g.", ", word and sentence information ."], "relation": "used for", "id": "2022.acl-long.140", "year": 2022, "rel_sent": "Hence , we propose a task - free enhancement module termed as Heterogeneous Linguistics Graph ( HLG ) to enhance Chinese pre - trained language models by integrating linguistics knowledge .", "forward": false, "src_ids": "2022.acl-long.140_2824"}
+{"input": "chinese language is done by using OtherScientificTerm| context: chinese pre - trained language models usually exploit contextual character information to learn representations , while ignoring the linguistics knowledge , e.g. , word and sentence information .", "entity": "chinese language", "output": "hierarchical heterogeneous graph", "neg_sample": ["chinese language is done by using OtherScientificTerm", "chinese pre - trained language models usually exploit contextual character information to learn representations , while ignoring the linguistics knowledge , e.g.", ", word and sentence information ."], "relation": "used for", "id": "2022.acl-long.140", "year": 2022, "rel_sent": "Specifically , we construct a hierarchical heterogeneous graph to model the characteristics linguistics structure of Chinese language , and conduct a graph - based method to summarize and concretize information on different granularities of Chinese linguistics hierarchies . Experimental results demonstrate our model has the ability to improve the performance of vanilla BERT , BERTwwm and ERNIE 1.0 on 6 natural language processing tasks with 10 benchmark datasets .", "forward": false, "src_ids": "2022.acl-long.140_2825"}
+{"input": "hierarchical heterogeneous graph is used for Material| context: chinese pre - trained language models usually exploit contextual character information to learn representations , while ignoring the linguistics knowledge , e.g. , word and sentence information .", "entity": "hierarchical heterogeneous graph", "output": "chinese language", "neg_sample": ["hierarchical heterogeneous graph is used for Material", "chinese pre - trained language models usually exploit contextual character information to learn representations , while ignoring the linguistics knowledge , e.g.", ", word and sentence information ."], "relation": "used for", "id": "2022.acl-long.140", "year": 2022, "rel_sent": "Specifically , we construct a hierarchical heterogeneous graph to model the characteristics linguistics structure of Chinese language , and conduct a graph - based method to summarize and concretize information on different granularities of Chinese linguistics hierarchies . Experimental results demonstrate our model has the ability to improve the performance of vanilla BERT , BERTwwm and ERNIE 1.0 on 6 natural language processing tasks with 10 benchmark datasets .", "forward": true, "src_ids": "2022.acl-long.140_2826"}
+{"input": "unlabeled data is used for Method| context: recognizing causal precedence relations among the chemical interactions in biomedical literature is crucial to understanding the underlying biological mechanisms . however , detecting such causal relation can be hard because : ( 1 ) many times , such causal relations among events are not explicitly expressed by certain phrases but implicitly implied by very diverse expressions in the text , and ( 2 ) annotating such causal relation detection datasets requires considerable expert knowledge and effort .", "entity": "unlabeled data", "output": "neural models", "neg_sample": ["unlabeled data is used for Method", "recognizing causal precedence relations among the chemical interactions in biomedical literature is crucial to understanding the underlying biological mechanisms .", "however , detecting such causal relation can be hard because : ( 1 ) many times , such causal relations among events are not explicitly expressed by certain phrases but implicitly implied by very diverse expressions in the text , and ( 2 ) annotating such causal relation detection datasets requires considerable expert knowledge and effort ."], "relation": "used for", "id": "2022.bionlp-1.24", "year": 2022, "rel_sent": "We show that , by using very limited amount of labeled data , and sufficient amount of unlabeled data , the neural models outperform previous baselines on the causal precedence detection task , and are ten times faster at inference compared to the BERT base model .", "forward": true, "src_ids": "2022.bionlp-1.24_2827"}
+{"input": "labeled data is used for Method| context: recognizing causal precedence relations among the chemical interactions in biomedical literature is crucial to understanding the underlying biological mechanisms . however , detecting such causal relation can be hard because : ( 1 ) many times , such causal relations among events are not explicitly expressed by certain phrases but implicitly implied by very diverse expressions in the text , and ( 2 ) annotating such causal relation detection datasets requires considerable expert knowledge and effort .", "entity": "labeled data", "output": "neural models", "neg_sample": ["labeled data is used for Method", "recognizing causal precedence relations among the chemical interactions in biomedical literature is crucial to understanding the underlying biological mechanisms .", "however , detecting such causal relation can be hard because : ( 1 ) many times , such causal relations among events are not explicitly expressed by certain phrases but implicitly implied by very diverse expressions in the text , and ( 2 ) annotating such causal relation detection datasets requires considerable expert knowledge and effort ."], "relation": "used for", "id": "2022.bionlp-1.24", "year": 2022, "rel_sent": "We show that , by using very limited amount of labeled data , and sufficient amount of unlabeled data , the neural models outperform previous baselines on the causal precedence detection task , and are ten times faster at inference compared to the BERT base model .", "forward": true, "src_ids": "2022.bionlp-1.24_2828"}
+{"input": "neural models is done by using Material| context: recognizing causal precedence relations among the chemical interactions in biomedical literature is crucial to understanding the underlying biological mechanisms . however , detecting such causal relation can be hard because : ( 1 ) many times , such causal relations among events are not explicitly expressed by certain phrases but implicitly implied by very diverse expressions in the text , and ( 2 ) annotating such causal relation detection datasets requires considerable expert knowledge and effort .", "entity": "neural models", "output": "labeled data", "neg_sample": ["neural models is done by using Material", "recognizing causal precedence relations among the chemical interactions in biomedical literature is crucial to understanding the underlying biological mechanisms .", "however , detecting such causal relation can be hard because : ( 1 ) many times , such causal relations among events are not explicitly expressed by certain phrases but implicitly implied by very diverse expressions in the text , and ( 2 ) annotating such causal relation detection datasets requires considerable expert knowledge and effort ."], "relation": "used for", "id": "2022.bionlp-1.24", "year": 2022, "rel_sent": "We show that , by using very limited amount of labeled data , and sufficient amount of unlabeled data , the neural models outperform previous baselines on the causal precedence detection task , and are ten times faster at inference compared to the BERT base model .", "forward": false, "src_ids": "2022.bionlp-1.24_2829"}
+{"input": "monotonic change of approval prediction is done by using OtherScientificTerm| context: predicting the approval chance of a patent application is a challenging problem involving multiple facets . the most crucial facet is arguably the novelty - 35 u.s. code 102 rejects more recent applications that have very similar prior arts . such novelty evaluations differ the patent approval prediction from conventional document classification - successful patent applications may share similar writing patterns ; however , too - similar newer applications would receive the opposite label , thus confusing standard document classifiers ( e.g. , bert ) .", "entity": "monotonic change of approval prediction", "output": "classification objective", "neg_sample": ["monotonic change of approval prediction is done by using OtherScientificTerm", "predicting the approval chance of a patent application is a challenging problem involving multiple facets .", "the most crucial facet is arguably the novelty - 35 u.s. code 102 rejects more recent applications that have very similar prior arts .", "such novelty evaluations differ the patent approval prediction from conventional document classification - successful patent applications may share similar writing patterns ; however , too - similar newer applications would receive the opposite label , thus confusing standard document classifiers ( e.g.", ", bert ) ."], "relation": "used for", "id": "2022.acl-long.28", "year": 2022, "rel_sent": "Moreover , we impose a new regularization term into the classification objective to enforce the monotonic change of approval prediction w.r.t .", "forward": false, "src_ids": "2022.acl-long.28_2830"}
+{"input": "regularization term is used for Task| context: predicting the approval chance of a patent application is a challenging problem involving multiple facets . the most crucial facet is arguably the novelty - 35 u.s. code 102 rejects more recent applications that have very similar prior arts . such novelty evaluations differ the patent approval prediction from conventional document classification - successful patent applications may share similar writing patterns ; however , too - similar newer applications would receive the opposite label , thus confusing standard document classifiers ( e.g. , bert ) .", "entity": "regularization term", "output": "monotonic change of approval prediction", "neg_sample": ["regularization term is used for Task", "predicting the approval chance of a patent application is a challenging problem involving multiple facets .", "the most crucial facet is arguably the novelty - 35 u.s. code 102 rejects more recent applications that have very similar prior arts .", "such novelty evaluations differ the patent approval prediction from conventional document classification - successful patent applications may share similar writing patterns ; however , too - similar newer applications would receive the opposite label , thus confusing standard document classifiers ( e.g.", ", bert ) ."], "relation": "used for", "id": "2022.acl-long.28", "year": 2022, "rel_sent": "Moreover , we impose a new regularization term into the classification objective to enforce the monotonic change of approval prediction w.r.t .", "forward": true, "src_ids": "2022.acl-long.28_2831"}
+{"input": "classification objective is used for Task| context: predicting the approval chance of a patent application is a challenging problem involving multiple facets . the most crucial facet is arguably the novelty - 35 u.s. code 102 rejects more recent applications that have very similar prior arts . such novelty evaluations differ the patent approval prediction from conventional document classification - successful patent applications may share similar writing patterns ; however , too - similar newer applications would receive the opposite label , thus confusing standard document classifiers ( e.g. , bert ) .", "entity": "classification objective", "output": "monotonic change of approval prediction", "neg_sample": ["classification objective is used for Task", "predicting the approval chance of a patent application is a challenging problem involving multiple facets .", "the most crucial facet is arguably the novelty - 35 u.s. code 102 rejects more recent applications that have very similar prior arts .", "such novelty evaluations differ the patent approval prediction from conventional document classification - successful patent applications may share similar writing patterns ; however , too - similar newer applications would receive the opposite label , thus confusing standard document classifiers ( e.g.", ", bert ) ."], "relation": "used for", "id": "2022.acl-long.28", "year": 2022, "rel_sent": "Moreover , we impose a new regularization term into the classification objective to enforce the monotonic change of approval prediction w.r.t .", "forward": true, "src_ids": "2022.acl-long.28_2832"}
+{"input": "interactive creation is done by using Material| context: comments are widely used by users in collaborative documents every day . the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel . understanding the role of comments in communication dynamics within documents is the first step towards automating their management .", "entity": "interactive creation", "output": "intelligent collaborative document experiences", "neg_sample": ["interactive creation is done by using Material", "comments are widely used by users in collaborative documents every day .", "the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel .", "understanding the role of comments in communication dynamics within documents is the first step towards automating their management ."], "relation": "used for", "id": "2022.dialdoc-1.20", "year": 2022, "rel_sent": "We envision that the next generation of intelligent collaborative document experiences allow interactive creation and consumption of content , there We also introduce the components necessary for developing novel tools that automate the handling of comments through natural language interaction with the documents .", "forward": false, "src_ids": "2022.dialdoc-1.20_2833"}
+{"input": "intelligent collaborative document experiences is used for Task| context: comments are widely used by users in collaborative documents every day . the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel . understanding the role of comments in communication dynamics within documents is the first step towards automating their management .", "entity": "intelligent collaborative document experiences", "output": "interactive creation", "neg_sample": ["intelligent collaborative document experiences is used for Task", "comments are widely used by users in collaborative documents every day .", "the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel .", "understanding the role of comments in communication dynamics within documents is the first step towards automating their management ."], "relation": "used for", "id": "2022.dialdoc-1.20", "year": 2022, "rel_sent": "We envision that the next generation of intelligent collaborative document experiences allow interactive creation and consumption of content , there We also introduce the components necessary for developing novel tools that automate the handling of comments through natural language interaction with the documents .", "forward": true, "src_ids": "2022.dialdoc-1.20_2834"}
+{"input": "handling of comments is done by using Generic| context: comments are widely used by users in collaborative documents every day . the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel . understanding the role of comments in communication dynamics within documents is the first step towards automating their management .", "entity": "handling of comments", "output": "automatic document management tools", "neg_sample": ["handling of comments is done by using Generic", "comments are widely used by users in collaborative documents every day .", "the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel .", "understanding the role of comments in communication dynamics within documents is the first step towards automating their management ."], "relation": "used for", "id": "2022.dialdoc-1.20", "year": 2022, "rel_sent": "We envision that the next generation of intelligent collaborative document experiences allow interactive creation and consumption of content , there We also introduce the components necessary for developing novel tools that automate the handling of comments through natural language interaction with the documents .", "forward": false, "src_ids": "2022.dialdoc-1.20_2835"}
+{"input": "automatic document management tools is used for Task| context: comments are widely used by users in collaborative documents every day . the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel . understanding the role of comments in communication dynamics within documents is the first step towards automating their management .", "entity": "automatic document management tools", "output": "handling of comments", "neg_sample": ["automatic document management tools is used for Task", "comments are widely used by users in collaborative documents every day .", "the documents ' comments enable collaborative editing and review dynamics , transforming each document into a context - sensitive communication channel .", "understanding the role of comments in communication dynamics within documents is the first step towards automating their management ."], "relation": "used for", "id": "2022.dialdoc-1.20", "year": 2022, "rel_sent": "We envision that the next generation of intelligent collaborative document experiences allow interactive creation and consumption of content , there We also introduce the components necessary for developing novel tools that automate the handling of comments through natural language interaction with the documents .", "forward": true, "src_ids": "2022.dialdoc-1.20_2836"}
+{"input": "downstream tasks is done by using Method| context: user language data can contain highly sensitive personal content . as such , it is imperative to offer users a strong and interpretable privacy guarantee when learning from their data .", "entity": "downstream tasks", "output": "private document embeddings", "neg_sample": ["downstream tasks is done by using Method", "user language data can contain highly sensitive personal content .", "as such , it is imperative to offer users a strong and interpretable privacy guarantee when learning from their data ."], "relation": "used for", "id": "2022.acl-long.238", "year": 2022, "rel_sent": "Our experiments indicate that these private document embeddings are useful for downstream tasks like sentiment analysis and topic classification and even outperform baseline methods with weaker guarantees like word - level Metric DP .", "forward": false, "src_ids": "2022.acl-long.238_2837"}
+{"input": "private document embeddings is used for Task| context: user language data can contain highly sensitive personal content . as such , it is imperative to offer users a strong and interpretable privacy guarantee when learning from their data .", "entity": "private document embeddings", "output": "downstream tasks", "neg_sample": ["private document embeddings is used for Task", "user language data can contain highly sensitive personal content .", "as such , it is imperative to offer users a strong and interpretable privacy guarantee when learning from their data ."], "relation": "used for", "id": "2022.acl-long.238", "year": 2022, "rel_sent": "Our experiments indicate that these private document embeddings are useful for downstream tasks like sentiment analysis and topic classification and even outperform baseline methods with weaker guarantees like word - level Metric DP .", "forward": true, "src_ids": "2022.acl-long.238_2838"}
+{"input": "early exit is done by using Method| context: currently , pre - trained models can be considered the default choice for a wide range of nlp tasks . despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) .", "entity": "early exit", "output": "ponder albert", "neg_sample": ["early exit is done by using Method", "currently , pre - trained models can be considered the default choice for a wide range of nlp tasks .", "despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) ."], "relation": "used for", "id": "2022.repl4nlp-1.22", "year": 2022, "rel_sent": "We compared PALBERT with recent methods for performing an early exit .", "forward": false, "src_ids": "2022.repl4nlp-1.22_2839"}
+{"input": "inference speed is done by using Method| context: currently , pre - trained models can be considered the default choice for a wide range of nlp tasks . despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) .", "entity": "inference speed", "output": "adaptive computation time approaches", "neg_sample": ["inference speed is done by using Method", "currently , pre - trained models can be considered the default choice for a wide range of nlp tasks .", "despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) ."], "relation": "used for", "id": "2022.repl4nlp-1.22", "year": 2022, "rel_sent": "This problem can potentially be solved by implementing adaptive computation time approaches , which were first designed to improve inference speed . Recently proposed PonderNet may be a promising solution for performing an early exit by treating the exit layer 's index as a latent variable .", "forward": false, "src_ids": "2022.repl4nlp-1.22_2840"}
+{"input": "adaptive computation time approaches is used for Metric| context: currently , pre - trained models can be considered the default choice for a wide range of nlp tasks . despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) .", "entity": "adaptive computation time approaches", "output": "inference speed", "neg_sample": ["adaptive computation time approaches is used for Metric", "currently , pre - trained models can be considered the default choice for a wide range of nlp tasks .", "despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) ."], "relation": "used for", "id": "2022.repl4nlp-1.22", "year": 2022, "rel_sent": "This problem can potentially be solved by implementing adaptive computation time approaches , which were first designed to improve inference speed . Recently proposed PonderNet may be a promising solution for performing an early exit by treating the exit layer 's index as a latent variable .", "forward": true, "src_ids": "2022.repl4nlp-1.22_2841"}
+{"input": "early exit is done by using Method| context: currently , pre - trained models can be considered the default choice for a wide range of nlp tasks . despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) .", "entity": "early exit", "output": "pondernet", "neg_sample": ["early exit is done by using Method", "currently , pre - trained models can be considered the default choice for a wide range of nlp tasks .", "despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) ."], "relation": "used for", "id": "2022.repl4nlp-1.22", "year": 2022, "rel_sent": "This problem can potentially be solved by implementing adaptive computation time approaches , which were first designed to improve inference speed . Recently proposed PonderNet may be a promising solution for performing an early exit by treating the exit layer 's index as a latent variable .", "forward": false, "src_ids": "2022.repl4nlp-1.22_2842"}
+{"input": "pondernet is used for Task| context: currently , pre - trained models can be considered the default choice for a wide range of nlp tasks . despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) .", "entity": "pondernet", "output": "early exit", "neg_sample": ["pondernet is used for Task", "currently , pre - trained models can be considered the default choice for a wide range of nlp tasks .", "despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) ."], "relation": "used for", "id": "2022.repl4nlp-1.22", "year": 2022, "rel_sent": "This problem can potentially be solved by implementing adaptive computation time approaches , which were first designed to improve inference speed . Recently proposed PonderNet may be a promising solution for performing an early exit by treating the exit layer 's index as a latent variable .", "forward": true, "src_ids": "2022.repl4nlp-1.22_2843"}
+{"input": "ponder albert is used for Task| context: currently , pre - trained models can be considered the default choice for a wide range of nlp tasks . despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) .", "entity": "ponder albert", "output": "early exit", "neg_sample": ["ponder albert is used for Task", "currently , pre - trained models can be considered the default choice for a wide range of nlp tasks .", "despite their sota results , there is practical evidence that these models may require a different number of computing layers for different input sequences , since evaluating all layers leads to overconfidence on wrong predictions ( namely overthinking ) ."], "relation": "used for", "id": "2022.repl4nlp-1.22", "year": 2022, "rel_sent": "We compared PALBERT with recent methods for performing an early exit .", "forward": true, "src_ids": "2022.repl4nlp-1.22_2844"}
+{"input": "large language models is done by using Method| context: current state - of - the - art nlp systems use large neural networks that require extensive computational resources for training . inspired by human knowledge acquisition , researchers have proposed curriculum learning - sequencing tasks ( task - based curricula ) or ordering and sampling the datasets ( data - based curricula ) that facilitate training .", "entity": "large language models", "output": "data - based curriculum learning", "neg_sample": ["large language models is done by using Method", "current state - of - the - art nlp systems use large neural networks that require extensive computational resources for training .", "inspired by human knowledge acquisition , researchers have proposed curriculum learning - sequencing tasks ( task - based curricula ) or ordering and sampling the datasets ( data - based curricula ) that facilitate training ."], "relation": "used for", "id": "2022.insights-1.16", "year": 2022, "rel_sent": "This work investigates the benefits of data - based curriculum learning for large language models such as BERT and T5 .", "forward": false, "src_ids": "2022.insights-1.16_2845"}
+{"input": "data - based curriculum learning is used for Method| context: current state - of - the - art nlp systems use large neural networks that require extensive computational resources for training . inspired by human knowledge acquisition , researchers have proposed curriculum learning - sequencing tasks ( task - based curricula ) or ordering and sampling the datasets ( data - based curricula ) that facilitate training .", "entity": "data - based curriculum learning", "output": "large language models", "neg_sample": ["data - based curriculum learning is used for Method", "current state - of - the - art nlp systems use large neural networks that require extensive computational resources for training .", "inspired by human knowledge acquisition , researchers have proposed curriculum learning - sequencing tasks ( task - based curricula ) or ordering and sampling the datasets ( data - based curricula ) that facilitate training ."], "relation": "used for", "id": "2022.insights-1.16", "year": 2022, "rel_sent": "This work investigates the benefits of data - based curriculum learning for large language models such as BERT and T5 .", "forward": true, "src_ids": "2022.insights-1.16_2846"}
+{"input": "cross - domain sentiment analysis is done by using Method| context: cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models . as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "cross - domain sentiment analysis", "output": "adversarial soft prompt tuning method ( adspt )", "neg_sample": ["cross - domain sentiment analysis is done by using Method", "cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models .", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "Adversarial Soft Prompt Tuning for Cross - Domain Sentiment Analysis.", "forward": false, "src_ids": "2022.acl-long.174_2847"}
+{"input": "cross - domain sentiment analysis is done by using Method| context: cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models . as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "cross - domain sentiment analysis", "output": "adversarial soft prompt tuning method ( adspt )", "neg_sample": ["cross - domain sentiment analysis is done by using Method", "cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models .", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "In this paper , we propose a novel Adversarial Soft Prompt Tuning method ( AdSPT ) to better model cross - domain sentiment analysis .", "forward": false, "src_ids": "2022.acl-long.174_2848"}
+{"input": "domain discrepancy is done by using Method| context: cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models . as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "domain discrepancy", "output": "adversarial soft prompt tuning method ( adspt )", "neg_sample": ["domain discrepancy is done by using Method", "cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models .", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "On the one hand , AdSPT adopts separate soft prompts instead of hard templates to learn different vectors for different domains , thus alleviating the domain discrepancy of the [ MASK ] token in the masked language modeling task .", "forward": false, "src_ids": "2022.acl-long.174_2849"}
+{"input": "adversarial soft prompt tuning method ( adspt ) is used for Task| context: as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "adversarial soft prompt tuning method ( adspt )", "output": "cross - domain sentiment analysis", "neg_sample": ["adversarial soft prompt tuning method ( adspt ) is used for Task", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "Adversarial Soft Prompt Tuning for Cross - Domain Sentiment Analysis.", "forward": true, "src_ids": "2022.acl-long.174_2850"}
+{"input": "adversarial soft prompt tuning method ( adspt ) is used for Task| context: as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "adversarial soft prompt tuning method ( adspt )", "output": "cross - domain sentiment analysis", "neg_sample": ["adversarial soft prompt tuning method ( adspt ) is used for Task", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "In this paper , we propose a novel Adversarial Soft Prompt Tuning method ( AdSPT ) to better model cross - domain sentiment analysis .", "forward": true, "src_ids": "2022.acl-long.174_2851"}
+{"input": "adversarial soft prompt tuning method ( adspt ) is used for OtherScientificTerm| context: cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models . as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "adversarial soft prompt tuning method ( adspt )", "output": "domain discrepancy", "neg_sample": ["adversarial soft prompt tuning method ( adspt ) is used for OtherScientificTerm", "cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models .", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "On the one hand , AdSPT adopts separate soft prompts instead of hard templates to learn different vectors for different domains , thus alleviating the domain discrepancy of the [ MASK ] token in the masked language modeling task .", "forward": true, "src_ids": "2022.acl-long.174_2852"}
+{"input": "domain - invariant representations is done by using Method| context: cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models . as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "domain - invariant representations", "output": "domain adversarial training strategy", "neg_sample": ["domain - invariant representations is done by using Method", "cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models .", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "On the other hand , AdSPT uses a novel domain adversarial training strategy to learn domain - invariant representations between each source domain and the target domain .", "forward": false, "src_ids": "2022.acl-long.174_2853"}
+{"input": "domain adversarial training strategy is used for Method| context: cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models . as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks . however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique .", "entity": "domain adversarial training strategy", "output": "domain - invariant representations", "neg_sample": ["domain adversarial training strategy is used for Method", "cross - domain sentiment analysis has achieved promising results with the help of pre - trained language models .", "as gpt-3 appears , prompt tuning has been widely explored to enable better semantic modeling in many natural language processing tasks .", "however , directly using a fixed predefined template for cross - domain research can not model different distributions of the [ mask ] token in different domains , thus making underuse of the prompt tuning technique ."], "relation": "used for", "id": "2022.acl-long.174", "year": 2022, "rel_sent": "On the other hand , AdSPT uses a novel domain adversarial training strategy to learn domain - invariant representations between each source domain and the target domain .", "forward": true, "src_ids": "2022.acl-long.174_2854"}
+{"input": "unified pretraining is used for Method| context: translation quality evaluation plays a crucial role in machine translation . according to the input format , it is mainly separated into three tasks , i.e. , reference - only , source - only and source - reference - combined . recent methods , despite their promising results , are specifically designed and optimized on one of them . this limits the convenience of these methods , and overlooks the commonalities among tasks .", "entity": "unified pretraining", "output": "multi - task training", "neg_sample": ["unified pretraining is used for Method", "translation quality evaluation plays a crucial role in machine translation .", "according to the input format , it is mainly separated into three tasks , i.e.", ", reference - only , source - only and source - reference - combined .", "recent methods , despite their promising results , are specifically designed and optimized on one of them .", "this limits the convenience of these methods , and overlooks the commonalities among tasks ."], "relation": "used for", "id": "2022.acl-long.558", "year": 2022, "rel_sent": "Concretely , we propose monotonic regional attention to control the interaction among input segments , and unified pretraining to better adapt multi - task training .", "forward": true, "src_ids": "2022.acl-long.558_2855"}
+{"input": "multi - task training is done by using Method| context: translation quality evaluation plays a crucial role in machine translation . according to the input format , it is mainly separated into three tasks , i.e. , reference - only , source - only and source - reference - combined . recent methods , despite their promising results , are specifically designed and optimized on one of them . this limits the convenience of these methods , and overlooks the commonalities among tasks .", "entity": "multi - task training", "output": "unified pretraining", "neg_sample": ["multi - task training is done by using Method", "translation quality evaluation plays a crucial role in machine translation .", "according to the input format , it is mainly separated into three tasks , i.e.", ", reference - only , source - only and source - reference - combined .", "recent methods , despite their promising results , are specifically designed and optimized on one of them .", "this limits the convenience of these methods , and overlooks the commonalities among tasks ."], "relation": "used for", "id": "2022.acl-long.558", "year": 2022, "rel_sent": "Concretely , we propose monotonic regional attention to control the interaction among input segments , and unified pretraining to better adapt multi - task training .", "forward": false, "src_ids": "2022.acl-long.558_2856"}
+{"input": "multilingual tabular nli dataset is done by using Method| context: the ability to reason about tabular or semi - structured knowledge is a fundamental problem for today 's natural language processing ( nlp ) systems . while significant progress has been achieved in the direction of tabular reasoning , these advances are limited to english due to the absence of multilingual benchmark datasets for semi - structured data .", "entity": "multilingual tabular nli dataset", "output": "machine translation methods", "neg_sample": ["multilingual tabular nli dataset is done by using Method", "the ability to reason about tabular or semi - structured knowledge is a fundamental problem for today 's natural language processing ( nlp ) systems .", "while significant progress has been achieved in the direction of tabular reasoning , these advances are limited to english due to the absence of multilingual benchmark datasets for semi - structured data ."], "relation": "used for", "id": "2022.fever-1.7", "year": 2022, "rel_sent": "In this paper , we use machine translation methods to construct a multilingual tabular NLI dataset , namely XINFOTABS , which expands the English tabular NLI dataset of INFOTABS to ten diverse languages .", "forward": false, "src_ids": "2022.fever-1.7_2857"}
+{"input": "machine translation methods is used for Material| context: the ability to reason about tabular or semi - structured knowledge is a fundamental problem for today 's natural language processing ( nlp ) systems . while significant progress has been achieved in the direction of tabular reasoning , these advances are limited to english due to the absence of multilingual benchmark datasets for semi - structured data .", "entity": "machine translation methods", "output": "multilingual tabular nli dataset", "neg_sample": ["machine translation methods is used for Material", "the ability to reason about tabular or semi - structured knowledge is a fundamental problem for today 's natural language processing ( nlp ) systems .", "while significant progress has been achieved in the direction of tabular reasoning , these advances are limited to english due to the absence of multilingual benchmark datasets for semi - structured data ."], "relation": "used for", "id": "2022.fever-1.7", "year": 2022, "rel_sent": "In this paper , we use machine translation methods to construct a multilingual tabular NLI dataset , namely XINFOTABS , which expands the English tabular NLI dataset of INFOTABS to ten diverse languages .", "forward": true, "src_ids": "2022.fever-1.7_2858"}
+{"input": "coherent and diverse topical phrases is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "coherent and diverse topical phrases", "output": "uctopic", "neg_sample": ["coherent and diverse topical phrases is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "Comprehensive evaluation on topic mining shows that UCTopic can extract coherent and diverse topical phrases .", "forward": false, "src_ids": "2022.acl-long.426_2859"}
+{"input": "phrase representations is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "phrase representations", "output": "unsupervised contrastive learning framework", "neg_sample": ["phrase representations is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "UCTopic : Unsupervised Contrastive Learning for Phrase Representations and Topic Mining.", "forward": false, "src_ids": "2022.acl-long.426_2860"}
+{"input": "topic mining is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "topic mining", "output": "unsupervised contrastive learning framework", "neg_sample": ["topic mining is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "UCTopic : Unsupervised Contrastive Learning for Phrase Representations and Topic Mining.", "forward": false, "src_ids": "2022.acl-long.426_2861"}
+{"input": "context - aware phrase representations is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "context - aware phrase representations", "output": "unsupervised contrastive learning framework", "neg_sample": ["context - aware phrase representations is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "In this paper , we propose UCTopic , a novel unsupervised contrastive learning framework for context - aware phrase representations and topic mining .", "forward": false, "src_ids": "2022.acl-long.426_2862"}
+{"input": "topic mining is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "topic mining", "output": "unsupervised contrastive learning framework", "neg_sample": ["topic mining is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "In this paper , we propose UCTopic , a novel unsupervised contrastive learning framework for context - aware phrase representations and topic mining .", "forward": false, "src_ids": "2022.acl-long.426_2863"}
+{"input": "unsupervised contrastive learning framework is used for Method| context: topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "unsupervised contrastive learning framework", "output": "phrase representations", "neg_sample": ["unsupervised contrastive learning framework is used for Method", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "UCTopic : Unsupervised Contrastive Learning for Phrase Representations and Topic Mining.", "forward": true, "src_ids": "2022.acl-long.426_2864"}
+{"input": "cluster - assisted contrastive learning ( ccl ) is used for Method| context: topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "cluster - assisted contrastive learning ( ccl )", "output": "phrase representations", "neg_sample": ["cluster - assisted contrastive learning ( ccl ) is used for Method", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "Hence , we propose cluster - assisted contrastive learning ( CCL ) which largely reduces noisy negatives by selecting negatives from clusters and further improves phrase representations for topics accordingly .", "forward": true, "src_ids": "2022.acl-long.426_2865"}
+{"input": "unsupervised contrastive learning framework is used for Task| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "unsupervised contrastive learning framework", "output": "topic mining", "neg_sample": ["unsupervised contrastive learning framework is used for Task", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "UCTopic : Unsupervised Contrastive Learning for Phrase Representations and Topic Mining.", "forward": true, "src_ids": "2022.acl-long.426_2866"}
+{"input": "unsupervised contrastive learning framework is used for Task| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "unsupervised contrastive learning framework", "output": "topic mining", "neg_sample": ["unsupervised contrastive learning framework is used for Task", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "In this paper , we propose UCTopic , a novel unsupervised contrastive learning framework for context - aware phrase representations and topic mining .", "forward": true, "src_ids": "2022.acl-long.426_2867"}
+{"input": "unsupervised contrastive learning framework is used for Task| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "unsupervised contrastive learning framework", "output": "context - aware phrase representations", "neg_sample": ["unsupervised contrastive learning framework is used for Task", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "In this paper , we propose UCTopic , a novel unsupervised contrastive learning framework for context - aware phrase representations and topic mining .", "forward": true, "src_ids": "2022.acl-long.426_2868"}
+{"input": "phrase representations is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "phrase representations", "output": "cluster - assisted contrastive learning ( ccl )", "neg_sample": ["phrase representations is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "Hence , we propose cluster - assisted contrastive learning ( CCL ) which largely reduces noisy negatives by selecting negatives from clusters and further improves phrase representations for topics accordingly .", "forward": false, "src_ids": "2022.acl-long.426_2869"}
+{"input": "noisy negatives is done by using Method| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "noisy negatives", "output": "cluster - assisted contrastive learning ( ccl )", "neg_sample": ["noisy negatives is done by using Method", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "Hence , we propose cluster - assisted contrastive learning ( CCL ) which largely reduces noisy negatives by selecting negatives from clusters and further improves phrase representations for topics accordingly .", "forward": false, "src_ids": "2022.acl-long.426_2870"}
+{"input": "cluster - assisted contrastive learning ( ccl ) is used for OtherScientificTerm| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "cluster - assisted contrastive learning ( ccl )", "output": "noisy negatives", "neg_sample": ["cluster - assisted contrastive learning ( ccl ) is used for OtherScientificTerm", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "Hence , we propose cluster - assisted contrastive learning ( CCL ) which largely reduces noisy negatives by selecting negatives from clusters and further improves phrase representations for topics accordingly .", "forward": true, "src_ids": "2022.acl-long.426_2871"}
+{"input": "uctopic is used for OtherScientificTerm| context: high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a . topic mining ) . existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge .", "entity": "uctopic", "output": "coherent and diverse topical phrases", "neg_sample": ["uctopic is used for OtherScientificTerm", "high - quality phrase representations are essential tofinding topics and related terms in documents ( a.k.a .", "topic mining ) .", "existing phrase representation learning methods either simply combine unigram representations in a context - free manner or rely on extensive annotations to learn context - aware knowledge ."], "relation": "used for", "id": "2022.acl-long.426", "year": 2022, "rel_sent": "Comprehensive evaluation on topic mining shows that UCTopic can extract coherent and diverse topical phrases .", "forward": true, "src_ids": "2022.acl-long.426_2872"}
+{"input": "aspect- and sentence - based sentiment analysis is done by using OtherScientificTerm| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "aspect- and sentence - based sentiment analysis", "output": "label space", "neg_sample": ["aspect- and sentence - based sentiment analysis is done by using OtherScientificTerm", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "Towards Unifying the Label Space for Aspect- and Sentence - based Sentiment Analysis.", "forward": false, "src_ids": "2022.findings-acl.3_2873"}
+{"input": "label space is used for Task| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "label space", "output": "aspect- and sentence - based sentiment analysis", "neg_sample": ["label space is used for Task", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "Towards Unifying the Label Space for Aspect- and Sentence - based Sentiment Analysis.", "forward": true, "src_ids": "2022.findings-acl.3_2874"}
+{"input": "sentiment polarity is done by using Task| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "sentiment polarity", "output": "fine - grained task", "neg_sample": ["sentiment polarity is done by using Task", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "The aspect - based sentiment analysis ( ABSA ) is a fine - grained task that aims to determine the sentiment polarity towards targeted aspect terms occurring in the sentence .", "forward": false, "src_ids": "2022.findings-acl.3_2875"}
+{"input": "fine - grained task is used for OtherScientificTerm| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "fine - grained task", "output": "sentiment polarity", "neg_sample": ["fine - grained task is used for OtherScientificTerm", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "The aspect - based sentiment analysis ( ABSA ) is a fine - grained task that aims to determine the sentiment polarity towards targeted aspect terms occurring in the sentence .", "forward": true, "src_ids": "2022.findings-acl.3_2876"}
+{"input": "homogeneous tasks is done by using Method| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "homogeneous tasks", "output": "pseudo - label ( pl ) method", "neg_sample": ["homogeneous tasks is done by using Method", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "In this article , we follow this line , and for the first time , we manage to apply the Pseudo - Label ( PL ) method to merge the two homogeneous tasks .", "forward": false, "src_ids": "2022.findings-acl.3_2877"}
+{"input": "pseudo - label ( pl ) method is used for Generic| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "pseudo - label ( pl ) method", "output": "homogeneous tasks", "neg_sample": ["pseudo - label ( pl ) method is used for Generic", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "In this article , we follow this line , and for the first time , we manage to apply the Pseudo - Label ( PL ) method to merge the two homogeneous tasks .", "forward": true, "src_ids": "2022.findings-acl.3_2878"}
+{"input": "label granularity unification is done by using OtherScientificTerm| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "label granularity unification", "output": "pseudo labels", "neg_sample": ["label granularity unification is done by using OtherScientificTerm", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "While it seems straightforward to use generated pseudo labels to handle this case of label granularity unification for two highly related tasks , we identify its major challenge in this paper and propose a novel framework , dubbed as Dual - granularity Pseudo Labeling ( DPL ) .", "forward": false, "src_ids": "2022.findings-acl.3_2879"}
+{"input": "pseudo labels is used for Task| context: the development of the absa task is very much hindered by the lack of annotated data . to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning .", "entity": "pseudo labels", "output": "label granularity unification", "neg_sample": ["pseudo labels is used for Task", "the development of the absa task is very much hindered by the lack of annotated data .", "to tackle this , the prior works have studied the possibility of utilizing the sentiment analysis ( sa ) datasets to assist in training the absa model , primarily via pretraining or multi - task learning ."], "relation": "used for", "id": "2022.findings-acl.3", "year": 2022, "rel_sent": "While it seems straightforward to use generated pseudo labels to handle this case of label granularity unification for two highly related tasks , we identify its major challenge in this paper and propose a novel framework , dubbed as Dual - granularity Pseudo Labeling ( DPL ) .", "forward": true, "src_ids": "2022.findings-acl.3_2880"}
+{"input": "automated dementia detection is done by using OtherScientificTerm| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "automated dementia detection", "output": "verbal disfluency tags", "neg_sample": ["automated dementia detection is done by using OtherScientificTerm", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "How You Say It Matters : Measuring the Impact of Verbal Disfluency Tags on Automated Dementia Detection.", "forward": false, "src_ids": "2022.bionlp-1.4_2881"}
+{"input": "verbal disfluency tags is used for Task| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "verbal disfluency tags", "output": "automated dementia detection", "neg_sample": ["verbal disfluency tags is used for Task", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "How You Say It Matters : Measuring the Impact of Verbal Disfluency Tags on Automated Dementia Detection.", "forward": true, "src_ids": "2022.bionlp-1.4_2882"}
+{"input": "disfluencies is done by using Method| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "disfluencies", "output": "disfluency annotator", "neg_sample": ["disfluencies is done by using Method", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": false, "src_ids": "2022.bionlp-1.4_2883"}
+{"input": "speech transcripts is done by using Method| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "speech transcripts", "output": "disfluency annotator", "neg_sample": ["speech transcripts is done by using Method", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": false, "src_ids": "2022.bionlp-1.4_2884"}
+{"input": "cognitive health assessment task is done by using Method| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "cognitive health assessment task", "output": "disfluency annotator", "neg_sample": ["cognitive health assessment task is done by using Method", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": false, "src_ids": "2022.bionlp-1.4_2885"}
+{"input": "cognitive health assessment task is done by using Material| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "cognitive health assessment task", "output": "speech transcripts", "neg_sample": ["cognitive health assessment task is done by using Material", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": false, "src_ids": "2022.bionlp-1.4_2886"}
+{"input": "disfluency annotator is used for Material| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "disfluency annotator", "output": "speech transcripts", "neg_sample": ["disfluency annotator is used for Material", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": true, "src_ids": "2022.bionlp-1.4_2887"}
+{"input": "disfluency annotator is used for Task| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "disfluency annotator", "output": "cognitive health assessment task", "neg_sample": ["disfluency annotator is used for Task", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": true, "src_ids": "2022.bionlp-1.4_2888"}
+{"input": "speech transcripts is used for Task| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "speech transcripts", "output": "cognitive health assessment task", "neg_sample": ["speech transcripts is used for Task", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We experiment with an off - the - shelf disfluency annotator to tag disfluencies in speech transcripts for a well - known cognitive health assessment task .", "forward": true, "src_ids": "2022.bionlp-1.4_2889"}
+{"input": "dementia detection is done by using OtherScientificTerm| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "dementia detection", "output": "gold annotated versus automatically detected verbal disfluencies", "neg_sample": ["dementia detection is done by using OtherScientificTerm", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We evaluate the performance of this model on detecting repetitions and corrections or retracing , and measure the influence of gold annotated versus automatically detected verbal disfluencies on dementia detection through a series of experiments .", "forward": false, "src_ids": "2022.bionlp-1.4_2890"}
+{"input": "gold annotated versus automatically detected verbal disfluencies is used for Task| context: automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks . however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance .", "entity": "gold annotated versus automatically detected verbal disfluencies", "output": "dementia detection", "neg_sample": ["gold annotated versus automatically detected verbal disfluencies is used for Task", "automatic speech recognition ( asr ) systems usually incorporate postprocessing mechanisms to remove disfluencies , facilitating the generation of clear , fluent transcripts that are conducive to many downstream nlp tasks .", "however , verbal disfluencies have proved to be predictive of dementia status , although little is known about how various types of verbal disfluencies , nor automatically detected disfluencies , affect predictive performance ."], "relation": "used for", "id": "2022.bionlp-1.4", "year": 2022, "rel_sent": "We evaluate the performance of this model on detecting repetitions and corrections or retracing , and measure the influence of gold annotated versus automatically detected verbal disfluencies on dementia detection through a series of experiments .", "forward": true, "src_ids": "2022.bionlp-1.4_2891"}
+{"input": "few - shot text classification is done by using Method| context: in text classification tasks , useful information is encoded in the label names . label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction . however , use of label - semantics during pre - training has not been extensively explored .", "entity": "few - shot text classification", "output": "label semantic aware pre - training", "neg_sample": ["few - shot text classification is done by using Method", "in text classification tasks , useful information is encoded in the label names .", "label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction .", "however , use of label - semantics during pre - training has not been extensively explored ."], "relation": "used for", "id": "2022.acl-long.570", "year": 2022, "rel_sent": "Label Semantic Aware Pre - training for Few - shot Text Classification.", "forward": false, "src_ids": "2022.acl-long.570_2892"}
+{"input": "label semantic aware pre - training is used for Task| context: in text classification tasks , useful information is encoded in the label names . label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction . however , use of label - semantics during pre - training has not been extensively explored .", "entity": "label semantic aware pre - training", "output": "few - shot text classification", "neg_sample": ["label semantic aware pre - training is used for Task", "in text classification tasks , useful information is encoded in the label names .", "label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction .", "however , use of label - semantics during pre - training has not been extensively explored ."], "relation": "used for", "id": "2022.acl-long.570", "year": 2022, "rel_sent": "Label Semantic Aware Pre - training for Few - shot Text Classification.", "forward": true, "src_ids": "2022.acl-long.570_2893"}
+{"input": "sentence - label pairs is done by using Method| context: in text classification tasks , useful information is encoded in the label names . label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction . however , use of label - semantics during pre - training has not been extensively explored .", "entity": "sentence - label pairs", "output": "filtering and labeling pipeline", "neg_sample": ["sentence - label pairs is done by using Method", "in text classification tasks , useful information is encoded in the label names .", "label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction .", "however , use of label - semantics during pre - training has not been extensively explored ."], "relation": "used for", "id": "2022.acl-long.570", "year": 2022, "rel_sent": "As domain - general pre - training requires large amounts of data , we develop a filtering and labeling pipeline to automatically create sentence - label pairs from unlabeled text .", "forward": false, "src_ids": "2022.acl-long.570_2894"}
+{"input": "filtering and labeling pipeline is used for OtherScientificTerm| context: in text classification tasks , useful information is encoded in the label names . label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction . however , use of label - semantics during pre - training has not been extensively explored .", "entity": "filtering and labeling pipeline", "output": "sentence - label pairs", "neg_sample": ["filtering and labeling pipeline is used for OtherScientificTerm", "in text classification tasks , useful information is encoded in the label names .", "label semantic aware systems have leveraged this information for improved text classification performance during fine - tuning and prediction .", "however , use of label - semantics during pre - training has not been extensively explored ."], "relation": "used for", "id": "2022.acl-long.570", "year": 2022, "rel_sent": "As domain - general pre - training requires large amounts of data , we develop a filtering and labeling pipeline to automatically create sentence - label pairs from unlabeled text .", "forward": true, "src_ids": "2022.acl-long.570_2895"}
+{"input": "penalized rewards is used for OtherScientificTerm| context: dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction . generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn . thus , although being a useful metric , it can be harsh at times and underestimate the true potential of a dst model . moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations . so , using jga as the only metric for model selection may not be ideal for all scenarios .", "entity": "penalized rewards", "output": "mispredictions", "neg_sample": ["penalized rewards is used for OtherScientificTerm", "dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction .", "generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn .", "thus , although being a useful metric , it can be harsh at times and underestimate the true potential of a dst model .", "moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations .", "so , using jga as the only metric for model selection may not be ideal for all scenarios ."], "relation": "used for", "id": "2022.acl-short.35", "year": 2022, "rel_sent": "But unlike JGA , it tries to give penalized rewards to mispredictions that are locally correct i.e.", "forward": true, "src_ids": "2022.acl-short.35_2896"}
+{"input": "flexible goal accuracy is used for Method| context: dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction . generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn . due to this cumulative nature of the belief state , it is difficult to get a correct prediction once a misprediction has occurred . moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations . so , using jga as the only metric for model selection may not be ideal for all scenarios .", "entity": "flexible goal accuracy", "output": "dst model", "neg_sample": ["flexible goal accuracy is used for Method", "dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction .", "generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn .", "due to this cumulative nature of the belief state , it is difficult to get a correct prediction once a misprediction has occurred .", "moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations .", "so , using jga as the only metric for model selection may not be ideal for all scenarios ."], "relation": "used for", "id": "2022.acl-short.35", "year": 2022, "rel_sent": "We also show that FGA is a better discriminator of DST model performance .", "forward": true, "src_ids": "2022.acl-short.35_2897"}
+{"input": "dst model is done by using Method| context: dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction . generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn . due to this cumulative nature of the belief state , it is difficult to get a correct prediction once a misprediction has occurred . thus , although being a useful metric , it can be harsh at times and underestimate the true potential of a dst model . moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations . so , using jga as the only metric for model selection may not be ideal for all scenarios .", "entity": "dst model", "output": "flexible goal accuracy", "neg_sample": ["dst model is done by using Method", "dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction .", "generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn .", "due to this cumulative nature of the belief state , it is difficult to get a correct prediction once a misprediction has occurred .", "thus , although being a useful metric , it can be harsh at times and underestimate the true potential of a dst model .", "moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations .", "so , using jga as the only metric for model selection may not be ideal for all scenarios ."], "relation": "used for", "id": "2022.acl-short.35", "year": 2022, "rel_sent": "We also show that FGA is a better discriminator of DST model performance .", "forward": false, "src_ids": "2022.acl-short.35_2898"}
+{"input": "mispredictions is done by using OtherScientificTerm| context: dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction . generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn . due to this cumulative nature of the belief state , it is difficult to get a correct prediction once a misprediction has occurred . thus , although being a useful metric , it can be harsh at times and underestimate the true potential of a dst model . moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations . so , using jga as the only metric for model selection may not be ideal for all scenarios .", "entity": "mispredictions", "output": "penalized rewards", "neg_sample": ["mispredictions is done by using OtherScientificTerm", "dialogue state tracking ( dst ) is primarily evaluated using joint goal accuracy ( jga ) defined as the fraction of turns where the ground - truth dialogue state exactly matches the prediction .", "generally in dst , the dialogue state or belief state for a given turn contain all the intents shown by the user till that turn .", "due to this cumulative nature of the belief state , it is difficult to get a correct prediction once a misprediction has occurred .", "thus , although being a useful metric , it can be harsh at times and underestimate the true potential of a dst model .", "moreover , an improvement in jga can sometimes decrease the performance of turn - level or non - cumulative belief state prediction due to inconsistency in annotations .", "so , using jga as the only metric for model selection may not be ideal for all scenarios ."], "relation": "used for", "id": "2022.acl-short.35", "year": 2022, "rel_sent": "But unlike JGA , it tries to give penalized rewards to mispredictions that are locally correct i.e.", "forward": false, "src_ids": "2022.acl-short.35_2899"}
+{"input": "document - level relation extraction ( docre ) is done by using Method| context: motivated by the fact that many relations cross the sentence boundary , there has been increasing interest in document - level relation extraction ( docre ) . docre requires integrating information within and across sentences , capturing complex interactions between mentions of entities . most existing methods are pipeline - based , requiring entities as input . however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps .", "entity": "document - level relation extraction ( docre )", "output": "sequence - to - sequence approach", "neg_sample": ["document - level relation extraction ( docre ) is done by using Method", "motivated by the fact that many relations cross the sentence boundary , there has been increasing interest in document - level relation extraction ( docre ) .", "docre requires integrating information within and across sentences , capturing complex interactions between mentions of entities .", "most existing methods are pipeline - based , requiring entities as input .", "however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps ."], "relation": "used for", "id": "2022.bionlp-1.2", "year": 2022, "rel_sent": "A sequence - to - sequence approach for document - level relation extraction.", "forward": false, "src_ids": "2022.bionlp-1.2_2900"}
+{"input": "document - level relation extraction ( docre ) is done by using Method| context: motivated by the fact that many relations cross the sentence boundary , there has been increasing interest in document - level relation extraction ( docre ) . docre requires integrating information within and across sentences , capturing complex interactions between mentions of entities . most existing methods are pipeline - based , requiring entities as input . however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps .", "entity": "document - level relation extraction ( docre )", "output": "sequence - to - sequence approach", "neg_sample": ["document - level relation extraction ( docre ) is done by using Method", "motivated by the fact that many relations cross the sentence boundary , there has been increasing interest in document - level relation extraction ( docre ) .", "docre requires integrating information within and across sentences , capturing complex interactions between mentions of entities .", "most existing methods are pipeline - based , requiring entities as input .", "however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps ."], "relation": "used for", "id": "2022.bionlp-1.2", "year": 2022, "rel_sent": "In this paper , we develop a sequence - to - sequence approach , seq2rel , that can learn the subtasks of DocRE ( entity extraction , coreference resolution and relation extraction ) end - to - end , replacing a pipeline of task - specific components .", "forward": false, "src_ids": "2022.bionlp-1.2_2901"}
+{"input": "sequence - to - sequence approach is used for Task| context: most existing methods are pipeline - based , requiring entities as input . however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps .", "entity": "sequence - to - sequence approach", "output": "document - level relation extraction ( docre )", "neg_sample": ["sequence - to - sequence approach is used for Task", "most existing methods are pipeline - based , requiring entities as input .", "however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps ."], "relation": "used for", "id": "2022.bionlp-1.2", "year": 2022, "rel_sent": "A sequence - to - sequence approach for document - level relation extraction.", "forward": true, "src_ids": "2022.bionlp-1.2_2902"}
+{"input": "sequence - to - sequence approach is used for Task| context: most existing methods are pipeline - based , requiring entities as input . however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps .", "entity": "sequence - to - sequence approach", "output": "document - level relation extraction ( docre )", "neg_sample": ["sequence - to - sequence approach is used for Task", "most existing methods are pipeline - based , requiring entities as input .", "however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps ."], "relation": "used for", "id": "2022.bionlp-1.2", "year": 2022, "rel_sent": "In this paper , we develop a sequence - to - sequence approach , seq2rel , that can learn the subtasks of DocRE ( entity extraction , coreference resolution and relation extraction ) end - to - end , replacing a pipeline of task - specific components .", "forward": true, "src_ids": "2022.bionlp-1.2_2903"}
+{"input": "seq2rel is used for Task| context: most existing methods are pipeline - based , requiring entities as input . however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps .", "entity": "seq2rel", "output": "document - level relation extraction ( docre )", "neg_sample": ["seq2rel is used for Task", "most existing methods are pipeline - based , requiring entities as input .", "however , jointly learning to extract entities and relations can improve performance and be more efficient due to shared parameters and training steps ."], "relation": "used for", "id": "2022.bionlp-1.2", "year": 2022, "rel_sent": "In this paper , we develop a sequence - to - sequence approach , seq2rel , that can learn the subtasks of DocRE ( entity extraction , coreference resolution and relation extraction ) end - to - end , replacing a pipeline of task - specific components .", "forward": true, "src_ids": "2022.bionlp-1.2_2904"}
+{"input": "vision - and - language generation tasks is done by using Method| context: due to the limitations of the model structure and pre - training objectives , existing vision - and - language generation models can not utilize pair - wise images and text through bi - directional generation .", "entity": "vision - and - language generation tasks", "output": "du - vlg", "neg_sample": ["vision - and - language generation tasks is done by using Method", "due to the limitations of the model structure and pre - training objectives , existing vision - and - language generation models can not utilize pair - wise images and text through bi - directional generation ."], "relation": "used for", "id": "2022.findings-acl.201", "year": 2022, "rel_sent": "DU - VLG : Unifying Vision - and - Language Generation via Dual Sequence - to - Sequence Pre - training.", "forward": false, "src_ids": "2022.findings-acl.201_2905"}
+{"input": "du - vlg is used for Task| context: due to the limitations of the model structure and pre - training objectives , existing vision - and - language generation models can not utilize pair - wise images and text through bi - directional generation .", "entity": "du - vlg", "output": "vision - and - language generation tasks", "neg_sample": ["du - vlg is used for Task", "due to the limitations of the model structure and pre - training objectives , existing vision - and - language generation models can not utilize pair - wise images and text through bi - directional generation ."], "relation": "used for", "id": "2022.findings-acl.201", "year": 2022, "rel_sent": "DU - VLG : Unifying Vision - and - Language Generation via Dual Sequence - to - Sequence Pre - training.", "forward": true, "src_ids": "2022.findings-acl.201_2906"}
+{"input": "arabic is done by using Method| context: we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models .", "entity": "arabic", "output": "pre - trained language models", "neg_sample": ["arabic is done by using Method", "we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models ."], "relation": "used for", "id": "2022.findings-acl.135", "year": 2022, "rel_sent": "Morphosyntactic Tagging with Pre - trained Language Models for Arabic and its Dialects.", "forward": false, "src_ids": "2022.findings-acl.135_2907"}
+{"input": "low - resource scenario is done by using Material| context: we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models .", "entity": "low - resource scenario", "output": "annotated data", "neg_sample": ["low - resource scenario is done by using Material", "we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models ."], "relation": "used for", "id": "2022.findings-acl.135", "year": 2022, "rel_sent": "We explore different training setups for fine - tuning pre - trained transformer language models , including training data size , the use of external linguistic resources , and the use of annotated data from other dialects in a low - resource scenario .", "forward": false, "src_ids": "2022.findings-acl.135_2908"}
+{"input": "annotated data is used for Material| context: we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models .", "entity": "annotated data", "output": "low - resource scenario", "neg_sample": ["annotated data is used for Material", "we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models ."], "relation": "used for", "id": "2022.findings-acl.135", "year": 2022, "rel_sent": "We explore different training setups for fine - tuning pre - trained transformer language models , including training data size , the use of external linguistic resources , and the use of annotated data from other dialects in a low - resource scenario .", "forward": true, "src_ids": "2022.findings-acl.135_2909"}
+{"input": "low - resource dialect is done by using Task| context: we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models .", "entity": "low - resource dialect", "output": "strategic fine - tuning", "neg_sample": ["low - resource dialect is done by using Task", "we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models ."], "relation": "used for", "id": "2022.findings-acl.135", "year": 2022, "rel_sent": "Our results show that strategic fine - tuning using datasets from other high - resource dialects is beneficial for a low - resource dialect .", "forward": false, "src_ids": "2022.findings-acl.135_2910"}
+{"input": "strategic fine - tuning is used for Material| context: we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models .", "entity": "strategic fine - tuning", "output": "low - resource dialect", "neg_sample": ["strategic fine - tuning is used for Material", "we present state - of - the - art results on morphosyntactic tagging across different varieties of arabic using fine - tuned pre - trained transformer language models ."], "relation": "used for", "id": "2022.findings-acl.135", "year": 2022, "rel_sent": "Our results show that strategic fine - tuning using datasets from other high - resource dialects is beneficial for a low - resource dialect .", "forward": true, "src_ids": "2022.findings-acl.135_2911"}
+{"input": "translate is used for Task| context: commonsense reasoning ( csr ) requires models to be equipped with general world knowledge . while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english .", "entity": "translate", "output": "multilingual commonsense reasoning", "neg_sample": ["translate is used for Task", "commonsense reasoning ( csr ) requires models to be equipped with general world knowledge .", "while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english ."], "relation": "used for", "id": "2022.findings-acl.255", "year": 2022, "rel_sent": "Extensive results on the XCSR benchmark demonstrate that TRT with external knowledge can significantly improve multilingual commonsense reasoning in both zero - shot and translate - train settings , consistently outperforming the state - of - the - art by more than 3 % on the multilingual commonsense reasoning benchmark X - CSQA and X - CODAH .", "forward": true, "src_ids": "2022.findings-acl.255_2912"}
+{"input": "multilingual commonsense reasoning is done by using Method| context: commonsense reasoning ( csr ) requires models to be equipped with general world knowledge . while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english . thus , it remains unclear how to effectively conduct multilingual commonsense reasoning ( xcsr ) for various languages .", "entity": "multilingual commonsense reasoning", "output": "translate", "neg_sample": ["multilingual commonsense reasoning is done by using Method", "commonsense reasoning ( csr ) requires models to be equipped with general world knowledge .", "while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english .", "thus , it remains unclear how to effectively conduct multilingual commonsense reasoning ( xcsr ) for various languages ."], "relation": "used for", "id": "2022.findings-acl.255", "year": 2022, "rel_sent": "Extensive results on the XCSR benchmark demonstrate that TRT with external knowledge can significantly improve multilingual commonsense reasoning in both zero - shot and translate - train settings , consistently outperforming the state - of - the - art by more than 3 % on the multilingual commonsense reasoning benchmark X - CSQA and X - CODAH .", "forward": false, "src_ids": "2022.findings-acl.255_2913"}
+{"input": "commonsense reasoning framework is done by using Material| context: commonsense reasoning ( csr ) requires models to be equipped with general world knowledge . while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english . thus , it remains unclear how to effectively conduct multilingual commonsense reasoning ( xcsr ) for various languages .", "entity": "commonsense reasoning framework", "output": "english knowledge sources", "neg_sample": ["commonsense reasoning framework is done by using Material", "commonsense reasoning ( csr ) requires models to be equipped with general world knowledge .", "while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english .", "thus , it remains unclear how to effectively conduct multilingual commonsense reasoning ( xcsr ) for various languages ."], "relation": "used for", "id": "2022.findings-acl.255", "year": 2022, "rel_sent": "In this work , we propose to use English as a pivot language , utilizing English knowledge sources for our our commonsense reasoning framework via a translate - retrieve - translate ( TRT ) strategy .", "forward": false, "src_ids": "2022.findings-acl.255_2914"}
+{"input": "english knowledge sources is used for Method| context: commonsense reasoning ( csr ) requires models to be equipped with general world knowledge . while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english . thus , it remains unclear how to effectively conduct multilingual commonsense reasoning ( xcsr ) for various languages .", "entity": "english knowledge sources", "output": "commonsense reasoning framework", "neg_sample": ["english knowledge sources is used for Method", "commonsense reasoning ( csr ) requires models to be equipped with general world knowledge .", "while csr is a language - agnostic process , most comprehensive knowledge sources are restricted to a small number of languages , especially english .", "thus , it remains unclear how to effectively conduct multilingual commonsense reasoning ( xcsr ) for various languages ."], "relation": "used for", "id": "2022.findings-acl.255", "year": 2022, "rel_sent": "In this work , we propose to use English as a pivot language , utilizing English knowledge sources for our our commonsense reasoning framework via a translate - retrieve - translate ( TRT ) strategy .", "forward": true, "src_ids": "2022.findings-acl.255_2915"}
+{"input": "qa - based product attribute extraction is done by using Task| context: a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products . although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries .", "entity": "qa - based product attribute extraction", "output": "knowledge - driven query expansion", "neg_sample": ["qa - based product attribute extraction is done by using Task", "a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products .", "although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries ."], "relation": "used for", "id": "2022.acl-short.25", "year": 2022, "rel_sent": "Simple and Effective Knowledge - Driven Query Expansion for QA - Based Product Attribute Extraction.", "forward": false, "src_ids": "2022.acl-short.25_2916"}
+{"input": "qa - based ave is done by using Task| context: a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products . although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries .", "entity": "qa - based ave", "output": "knowledge - driven query expansion", "neg_sample": ["qa - based ave is done by using Task", "a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products .", "although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries ."], "relation": "used for", "id": "2022.acl-short.25", "year": 2022, "rel_sent": "We thus propose simple knowledge - driven query expansion based on possible answers ( values ) of a query ( attribute ) for QA - based AVE . We retrieve values of a query ( attribute ) from the training data to expand the query .", "forward": false, "src_ids": "2022.acl-short.25_2917"}
+{"input": "knowledge - driven query expansion is used for Task| context: a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products . although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries .", "entity": "knowledge - driven query expansion", "output": "qa - based product attribute extraction", "neg_sample": ["knowledge - driven query expansion is used for Task", "a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products .", "although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries ."], "relation": "used for", "id": "2022.acl-short.25", "year": 2022, "rel_sent": "Simple and Effective Knowledge - Driven Query Expansion for QA - Based Product Attribute Extraction.", "forward": true, "src_ids": "2022.acl-short.25_2918"}
+{"input": "knowledge - driven query expansion is used for Task| context: a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products . although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries .", "entity": "knowledge - driven query expansion", "output": "qa - based ave", "neg_sample": ["knowledge - driven query expansion is used for Task", "a key challenge in attribute value extraction ( ave ) from e - commerce sites is how to handle a large number of attributes for diverse products .", "although this challenge is partially addressed by a question answering ( qa ) approach which finds a value in product data for a given query ( attribute ) , it does not work effectively for rare and ambiguous queries ."], "relation": "used for", "id": "2022.acl-short.25", "year": 2022, "rel_sent": "We thus propose simple knowledge - driven query expansion based on possible answers ( values ) of a query ( attribute ) for QA - based AVE . We retrieve values of a query ( attribute ) from the training data to expand the query .", "forward": true, "src_ids": "2022.acl-short.25_2919"}
+{"input": "annotation is done by using Method| context: the automation of extracting argument structures faces a pair of challenges on ( 1 ) encoding long - term contexts tofacilitate comprehensive understanding , and ( 2 ) improving data efficiency since constructing high - quality argument structures is time - consuming .", "entity": "annotation", "output": "active learning", "neg_sample": ["annotation is done by using Method", "the automation of extracting argument structures faces a pair of challenges on ( 1 ) encoding long - term contexts tofacilitate comprehensive understanding , and ( 2 ) improving data efficiency since constructing high - quality argument structures is time - consuming ."], "relation": "used for", "id": "2022.findings-acl.36", "year": 2022, "rel_sent": "To tackle the difficulty of data annotation , we examine two complementary methods : ( i ) transfer learning to leverage existing annotated data to boost model performance in a new target domain , and ( ii ) active learning to strategically identify a small amount of samples for annotation .", "forward": false, "src_ids": "2022.findings-acl.36_2920"}
+{"input": "active learning is used for Task| context: the automation of extracting argument structures faces a pair of challenges on ( 1 ) encoding long - term contexts tofacilitate comprehensive understanding , and ( 2 ) improving data efficiency since constructing high - quality argument structures is time - consuming .", "entity": "active learning", "output": "annotation", "neg_sample": ["active learning is used for Task", "the automation of extracting argument structures faces a pair of challenges on ( 1 ) encoding long - term contexts tofacilitate comprehensive understanding , and ( 2 ) improving data efficiency since constructing high - quality argument structures is time - consuming ."], "relation": "used for", "id": "2022.findings-acl.36", "year": 2022, "rel_sent": "To tackle the difficulty of data annotation , we examine two complementary methods : ( i ) transfer learning to leverage existing annotated data to boost model performance in a new target domain , and ( ii ) active learning to strategically identify a small amount of samples for annotation .", "forward": true, "src_ids": "2022.findings-acl.36_2921"}
+{"input": "transformer is done by using Method| context: recently , a lot of research has been carried out to improve the efficiency of transformer . among them , the sparse pattern - based method is an important branch of efficient transformers . however , some existing sparse methods usually use fixed patterns to select words , without considering similarities between words . other sparse methods use clustering patterns to select words , but the clustering process is separate from the training process of the target task , which causes a decrease in effectiveness .", "entity": "transformer", "output": "self - attention mechanism", "neg_sample": ["transformer is done by using Method", "recently , a lot of research has been carried out to improve the efficiency of transformer .", "among them , the sparse pattern - based method is an important branch of efficient transformers .", "however , some existing sparse methods usually use fixed patterns to select words , without considering similarities between words .", "other sparse methods use clustering patterns to select words , but the clustering process is separate from the training process of the target task , which causes a decrease in effectiveness ."], "relation": "used for", "id": "2022.acl-long.170", "year": 2022, "rel_sent": "To address these limitations , we design a neural clustering method , which can be seamlessly integrated into the Self - Attention Mechanism in Transformer .", "forward": false, "src_ids": "2022.acl-long.170_2922"}
+{"input": "self - attention mechanism is used for Method| context: however , some existing sparse methods usually use fixed patterns to select words , without considering similarities between words . other sparse methods use clustering patterns to select words , but the clustering process is separate from the training process of the target task , which causes a decrease in effectiveness .", "entity": "self - attention mechanism", "output": "transformer", "neg_sample": ["self - attention mechanism is used for Method", "however , some existing sparse methods usually use fixed patterns to select words , without considering similarities between words .", "other sparse methods use clustering patterns to select words , but the clustering process is separate from the training process of the target task , which causes a decrease in effectiveness ."], "relation": "used for", "id": "2022.acl-long.170", "year": 2022, "rel_sent": "To address these limitations , we design a neural clustering method , which can be seamlessly integrated into the Self - Attention Mechanism in Transformer .", "forward": true, "src_ids": "2022.acl-long.170_2923"}
+{"input": "role interaction enhanced method is used for Task| context: existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles . however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles .", "entity": "role interaction enhanced method", "output": "role - oriented dialogue summarization", "neg_sample": ["role interaction enhanced method is used for Task", "existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles .", "however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles ."], "relation": "used for", "id": "2022.acl-long.182", "year": 2022, "rel_sent": "Therefore , we propose a novel role interaction enhanced method for role - oriented dialogue summarization .", "forward": true, "src_ids": "2022.acl-long.182_2924"}
+{"input": "roles ' critical dialogue utterances is done by using Method| context: role - oriented dialogue summarization is to generate summaries for different roles in the dialogue , e.g. , merchants and consumers . existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles . however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles .", "entity": "roles ' critical dialogue utterances", "output": "cross attention interaction", "neg_sample": ["roles ' critical dialogue utterances is done by using Method", "role - oriented dialogue summarization is to generate summaries for different roles in the dialogue , e.g.", ", merchants and consumers .", "existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles .", "however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles ."], "relation": "used for", "id": "2022.acl-long.182", "year": 2022, "rel_sent": "The cross attention interaction aims to select other roles ' critical dialogue utterances , while the decoder self - attention interaction aims to obtain key information from other roles ' summaries .", "forward": false, "src_ids": "2022.acl-long.182_2925"}
+{"input": "role - oriented dialogue summarization is done by using Method| context: role - oriented dialogue summarization is to generate summaries for different roles in the dialogue , e.g. , merchants and consumers . existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles . however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles .", "entity": "role - oriented dialogue summarization", "output": "role interaction enhanced method", "neg_sample": ["role - oriented dialogue summarization is done by using Method", "role - oriented dialogue summarization is to generate summaries for different roles in the dialogue , e.g.", ", merchants and consumers .", "existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles .", "however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles ."], "relation": "used for", "id": "2022.acl-long.182", "year": 2022, "rel_sent": "Therefore , we propose a novel role interaction enhanced method for role - oriented dialogue summarization .", "forward": false, "src_ids": "2022.acl-long.182_2926"}
+{"input": "cross attention interaction is used for OtherScientificTerm| context: role - oriented dialogue summarization is to generate summaries for different roles in the dialogue , e.g. , merchants and consumers . existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles . however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles .", "entity": "cross attention interaction", "output": "roles ' critical dialogue utterances", "neg_sample": ["cross attention interaction is used for OtherScientificTerm", "role - oriented dialogue summarization is to generate summaries for different roles in the dialogue , e.g.", ", merchants and consumers .", "existing methods handle this task by summarizing each role 's content separately and thus are prone to ignore the information from other roles .", "however , we believe that other roles ' content could benefit the quality of summaries , such as the omitted information mentioned by other roles ."], "relation": "used for", "id": "2022.acl-long.182", "year": 2022, "rel_sent": "The cross attention interaction aims to select other roles ' critical dialogue utterances , while the decoder self - attention interaction aims to obtain key information from other roles ' summaries .", "forward": true, "src_ids": "2022.acl-long.182_2927"}
+{"input": "grammatical error correction is done by using Method| context: modern writing assistance applications are always equipped with a grammatical error correction ( gec ) model to correct errors in user - entered sentences . different scenarios have varying requirements for correction behavior , e.g. , performing more precise corrections ( high precision ) or providing more candidates for users ( high recall ) . however , previous works adjust such trade - off only for sequence labeling approaches .", "entity": "grammatical error correction", "output": "align - and - predict decoding", "neg_sample": ["grammatical error correction is done by using Method", "modern writing assistance applications are always equipped with a grammatical error correction ( gec ) model to correct errors in user - entered sentences .", "different scenarios have varying requirements for correction behavior , e.g.", ", performing more precise corrections ( high precision ) or providing more candidates for users ( high recall ) .", "however , previous works adjust such trade - off only for sequence labeling approaches ."], "relation": "used for", "id": "2022.acl-short.77", "year": 2022, "rel_sent": "Adjusting the Precision - Recall Trade - Off with Align - and - Predict Decoding for Grammatical Error Correction.", "forward": false, "src_ids": "2022.acl-short.77_2928"}
+{"input": "align - and - predict decoding is used for Task| context: different scenarios have varying requirements for correction behavior , e.g. , performing more precise corrections ( high precision ) or providing more candidates for users ( high recall ) . however , previous works adjust such trade - off only for sequence labeling approaches .", "entity": "align - and - predict decoding", "output": "grammatical error correction", "neg_sample": ["align - and - predict decoding is used for Task", "different scenarios have varying requirements for correction behavior , e.g.", ", performing more precise corrections ( high precision ) or providing more candidates for users ( high recall ) .", "however , previous works adjust such trade - off only for sequence labeling approaches ."], "relation": "used for", "id": "2022.acl-short.77", "year": 2022, "rel_sent": "Adjusting the Precision - Recall Trade - Off with Align - and - Predict Decoding for Grammatical Error Correction.", "forward": true, "src_ids": "2022.acl-short.77_2929"}
+{"input": "context - sensitive dialogue unsafety detection is done by using Method| context: dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently . however , dialogue safety problems remain under - defined and the corresponding dataset is scarce .", "entity": "context - sensitive dialogue unsafety detection", "output": "dialogue safety classifier", "neg_sample": ["context - sensitive dialogue unsafety detection is done by using Method", "dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently .", "however , dialogue safety problems remain under - defined and the corresponding dataset is scarce ."], "relation": "used for", "id": "2022.findings-acl.308", "year": 2022, "rel_sent": "As a remedy , we train a dialogue safety classifier to provide a strong baseline for context - sensitive dialogue unsafety detection .", "forward": false, "src_ids": "2022.findings-acl.308_2930"}
+{"input": "dialogue safety classifier is used for Task| context: dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently . however , dialogue safety problems remain under - defined and the corresponding dataset is scarce .", "entity": "dialogue safety classifier", "output": "context - sensitive dialogue unsafety detection", "neg_sample": ["dialogue safety classifier is used for Task", "dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently .", "however , dialogue safety problems remain under - defined and the corresponding dataset is scarce ."], "relation": "used for", "id": "2022.findings-acl.308", "year": 2022, "rel_sent": "As a remedy , we train a dialogue safety classifier to provide a strong baseline for context - sensitive dialogue unsafety detection .", "forward": true, "src_ids": "2022.findings-acl.308_2931"}
+{"input": "safety evaluations is done by using Method| context: dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently . however , dialogue safety problems remain under - defined and the corresponding dataset is scarce .", "entity": "safety evaluations", "output": "classifier", "neg_sample": ["safety evaluations is done by using Method", "dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently .", "however , dialogue safety problems remain under - defined and the corresponding dataset is scarce ."], "relation": "used for", "id": "2022.findings-acl.308", "year": 2022, "rel_sent": "With our classifier , we perform safety evaluations on popular conversational models and show that existing dialogue systems still exhibit concerning context - sensitive safety problems .", "forward": false, "src_ids": "2022.findings-acl.308_2932"}
+{"input": "classifier is used for Task| context: dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently . however , dialogue safety problems remain under - defined and the corresponding dataset is scarce .", "entity": "classifier", "output": "safety evaluations", "neg_sample": ["classifier is used for Task", "dialogue safety problems severely limit the real - world deployment of neural conversational models and have attracted great research interests recently .", "however , dialogue safety problems remain under - defined and the corresponding dataset is scarce ."], "relation": "used for", "id": "2022.findings-acl.308", "year": 2022, "rel_sent": "With our classifier , we perform safety evaluations on popular conversational models and show that existing dialogue systems still exhibit concerning context - sensitive safety problems .", "forward": true, "src_ids": "2022.findings-acl.308_2933"}
+{"input": "multimodal troll memes is done by using Method| context: with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues . recently , memes have become a popular way of sharing information on social media . usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content . detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature .", "entity": "multimodal troll memes", "output": "deep learning techniques", "neg_sample": ["multimodal troll memes is done by using Method", "with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues .", "recently , memes have become a popular way of sharing information on social media .", "usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content .", "detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature ."], "relation": "used for", "id": "2022.dravidianlangtech-1.27", "year": 2022, "rel_sent": "CUET - NLP@DravidianLangTech - ACL2022 : Investigating Deep Learning Techniques to Detect Multimodal Troll Memes.", "forward": false, "src_ids": "2022.dravidianlangtech-1.27_2934"}
+{"input": "deep learning techniques is used for Material| context: with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues . recently , memes have become a popular way of sharing information on social media . usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content . detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature .", "entity": "deep learning techniques", "output": "multimodal troll memes", "neg_sample": ["deep learning techniques is used for Material", "with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues .", "recently , memes have become a popular way of sharing information on social media .", "usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content .", "detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature ."], "relation": "used for", "id": "2022.dravidianlangtech-1.27", "year": 2022, "rel_sent": "CUET - NLP@DravidianLangTech - ACL2022 : Investigating Deep Learning Techniques to Detect Multimodal Troll Memes.", "forward": true, "src_ids": "2022.dravidianlangtech-1.27_2935"}
+{"input": "classification task is done by using Method| context: with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues . recently , memes have become a popular way of sharing information on social media . usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content . detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature .", "entity": "classification task", "output": "computational models", "neg_sample": ["classification task is done by using Method", "with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues .", "recently , memes have become a popular way of sharing information on social media .", "usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content .", "detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature ."], "relation": "used for", "id": "2022.dravidianlangtech-1.27", "year": 2022, "rel_sent": "Several computational models have been investigated to perform the classification task .", "forward": false, "src_ids": "2022.dravidianlangtech-1.27_2936"}
+{"input": "computational models is used for Task| context: with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues . recently , memes have become a popular way of sharing information on social media . usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content . detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature .", "entity": "computational models", "output": "classification task", "neg_sample": ["computational models is used for Task", "with the substantial rise of internet usage , social media has become a powerful communication medium to convey information , opinions , and feelings on various issues .", "recently , memes have become a popular way of sharing information on social media .", "usually , memes are visuals with text incorporated into them and quickly disseminate hatred and offensive content .", "detecting or classifying memes is challenging due to their region - specific interpretation and multimodal nature ."], "relation": "used for", "id": "2022.dravidianlangtech-1.27", "year": 2022, "rel_sent": "Several computational models have been investigated to perform the classification task .", "forward": true, "src_ids": "2022.dravidianlangtech-1.27_2937"}
+{"input": "contextual link prediction is done by using Method| context: in the real world , many relational facts require context ; for instance , a politician holds a given elected position only for a particular timespan . this context ( the timespan ) is typically ignored in knowledge graph link prediction tasks , or is leveraged by models designed specifically to make use of it ( i.e. n - ary link prediction models ) .", "entity": "contextual link prediction", "output": "masked language models", "neg_sample": ["contextual link prediction is done by using Method", "in the real world , many relational facts require context ; for instance , a politician holds a given elected position only for a particular timespan .", "this context ( the timespan ) is typically ignored in knowledge graph link prediction tasks , or is leveraged by models designed specifically to make use of it ( i.e.", "n - ary link prediction models ) ."], "relation": "used for", "id": "2022.deelio-1.9", "year": 2022, "rel_sent": "On Masked Language Models for Contextual Link Prediction.", "forward": false, "src_ids": "2022.deelio-1.9_2938"}
+{"input": "masked language models is used for Task| context: in the real world , many relational facts require context ; for instance , a politician holds a given elected position only for a particular timespan . this context ( the timespan ) is typically ignored in knowledge graph link prediction tasks , or is leveraged by models designed specifically to make use of it ( i.e. n - ary link prediction models ) .", "entity": "masked language models", "output": "contextual link prediction", "neg_sample": ["masked language models is used for Task", "in the real world , many relational facts require context ; for instance , a politician holds a given elected position only for a particular timespan .", "this context ( the timespan ) is typically ignored in knowledge graph link prediction tasks , or is leveraged by models designed specifically to make use of it ( i.e.", "n - ary link prediction models ) ."], "relation": "used for", "id": "2022.deelio-1.9", "year": 2022, "rel_sent": "On Masked Language Models for Contextual Link Prediction.", "forward": true, "src_ids": "2022.deelio-1.9_2939"}
+{"input": "chinese spelling correction is done by using Method| context: recently , bert - based models have dominated the research of chinese spelling correction ( csc ) . these methods have two limitations : ( 1 ) they have poor performance on multi - typo texts . in such texts , the context of each typo contains at least one misspelled character , which brings noise information . such noisy context leads to the declining performance on multi - typo texts . ( 2 ) they tend to overcorrect valid expressions to more frequent expressions due to the masked token recovering task of bert .", "entity": "chinese spelling correction", "output": "contextual typo robust approach", "neg_sample": ["chinese spelling correction is done by using Method", "recently , bert - based models have dominated the research of chinese spelling correction ( csc ) .", "these methods have two limitations : ( 1 ) they have poor performance on multi - typo texts .", "in such texts , the context of each typo contains at least one misspelled character , which brings noise information .", "such noisy context leads to the declining performance on multi - typo texts .", "( 2 ) they tend to overcorrect valid expressions to more frequent expressions due to the masked token recovering task of bert ."], "relation": "used for", "id": "2022.findings-acl.237", "year": 2022, "rel_sent": "CRASpell : A Contextual Typo Robust Approach to Improve Chinese Spelling Correction.", "forward": false, "src_ids": "2022.findings-acl.237_2940"}
+{"input": "contextual typo robust approach is used for Task| context: these methods have two limitations : ( 1 ) they have poor performance on multi - typo texts . in such texts , the context of each typo contains at least one misspelled character , which brings noise information . such noisy context leads to the declining performance on multi - typo texts . ( 2 ) they tend to overcorrect valid expressions to more frequent expressions due to the masked token recovering task of bert .", "entity": "contextual typo robust approach", "output": "chinese spelling correction", "neg_sample": ["contextual typo robust approach is used for Task", "these methods have two limitations : ( 1 ) they have poor performance on multi - typo texts .", "in such texts , the context of each typo contains at least one misspelled character , which brings noise information .", "such noisy context leads to the declining performance on multi - typo texts .", "( 2 ) they tend to overcorrect valid expressions to more frequent expressions due to the masked token recovering task of bert ."], "relation": "used for", "id": "2022.findings-acl.237", "year": 2022, "rel_sent": "CRASpell : A Contextual Typo Robust Approach to Improve Chinese Spelling Correction.", "forward": true, "src_ids": "2022.findings-acl.237_2941"}
+{"input": "romanian biomedical domain is done by using Method| context: recognition of named entities present in text is an important step towards information extraction and natural language understanding .", "entity": "romanian biomedical domain", "output": "named entity recognition system", "neg_sample": ["romanian biomedical domain is done by using Method", "recognition of named entities present in text is an important step towards information extraction and natural language understanding ."], "relation": "used for", "id": "2022.bionlp-1.30", "year": 2022, "rel_sent": "This work presents a named entity recognition system for the Romanian biomedical domain .", "forward": false, "src_ids": "2022.bionlp-1.30_2942"}
+{"input": "named entity recognition system is used for Material| context: recognition of named entities present in text is an important step towards information extraction and natural language understanding .", "entity": "named entity recognition system", "output": "romanian biomedical domain", "neg_sample": ["named entity recognition system is used for Material", "recognition of named entities present in text is an important step towards information extraction and natural language understanding ."], "relation": "used for", "id": "2022.bionlp-1.30", "year": 2022, "rel_sent": "This work presents a named entity recognition system for the Romanian biomedical domain .", "forward": true, "src_ids": "2022.bionlp-1.30_2943"}
+{"input": "multimodal aspect - based sentiment analysis is done by using Method| context: as an important task in sentiment analysis , multimodal aspect - based sentiment analysis ( mabsa ) has attracted increasing attention inrecent years . however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodalalignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grainedaspects , opinions , and their alignments across modalities .", "entity": "multimodal aspect - based sentiment analysis", "output": "vision - language pre - training", "neg_sample": ["multimodal aspect - based sentiment analysis is done by using Method", "as an important task in sentiment analysis , multimodal aspect - based sentiment analysis ( mabsa ) has attracted increasing attention inrecent years .", "however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodalalignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grainedaspects , opinions , and their alignments across modalities ."], "relation": "used for", "id": "2022.acl-long.152", "year": 2022, "rel_sent": "Vision - Language Pre - Training for Multimodal Aspect - Based Sentiment Analysis.", "forward": false, "src_ids": "2022.acl-long.152_2944"}
+{"input": "vision - language pre - training is used for Task| context: however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodalalignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grainedaspects , opinions , and their alignments across modalities .", "entity": "vision - language pre - training", "output": "multimodal aspect - based sentiment analysis", "neg_sample": ["vision - language pre - training is used for Task", "however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodalalignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grainedaspects , opinions , and their alignments across modalities ."], "relation": "used for", "id": "2022.acl-long.152", "year": 2022, "rel_sent": "Vision - Language Pre - Training for Multimodal Aspect - Based Sentiment Analysis.", "forward": true, "src_ids": "2022.acl-long.152_2945"}
+{"input": "persona - based dialogue generation is done by using Method| context: towards building intelligent dialogue agents , there has been a growing interest in introducing explicit personas in generation models . however , with limited persona - based dialogue data at hand , it may be difficult to train a dialogue generation model well . we point out that the data challenges of this generation task lie in two aspects : first , it is expensive to scale up current persona - based dialogue datasets ; second , each data sample in this task is more complex to learn with than conventional dialogue data .", "entity": "persona - based dialogue generation", "output": "model - agnostic data manipulation method", "neg_sample": ["persona - based dialogue generation is done by using Method", "towards building intelligent dialogue agents , there has been a growing interest in introducing explicit personas in generation models .", "however , with limited persona - based dialogue data at hand , it may be difficult to train a dialogue generation model well .", "we point out that the data challenges of this generation task lie in two aspects : first , it is expensive to scale up current persona - based dialogue datasets ; second , each data sample in this task is more complex to learn with than conventional dialogue data ."], "relation": "used for", "id": "2022.acl-long.550", "year": 2022, "rel_sent": "A Model - agnostic Data Manipulation Method for Persona - based Dialogue Generation.", "forward": false, "src_ids": "2022.acl-long.550_2946"}
+{"input": "model - agnostic data manipulation method is used for Task| context: towards building intelligent dialogue agents , there has been a growing interest in introducing explicit personas in generation models . however , with limited persona - based dialogue data at hand , it may be difficult to train a dialogue generation model well . we point out that the data challenges of this generation task lie in two aspects : first , it is expensive to scale up current persona - based dialogue datasets ; second , each data sample in this task is more complex to learn with than conventional dialogue data .", "entity": "model - agnostic data manipulation method", "output": "persona - based dialogue generation", "neg_sample": ["model - agnostic data manipulation method is used for Task", "towards building intelligent dialogue agents , there has been a growing interest in introducing explicit personas in generation models .", "however , with limited persona - based dialogue data at hand , it may be difficult to train a dialogue generation model well .", "we point out that the data challenges of this generation task lie in two aspects : first , it is expensive to scale up current persona - based dialogue datasets ; second , each data sample in this task is more complex to learn with than conventional dialogue data ."], "relation": "used for", "id": "2022.acl-long.550", "year": 2022, "rel_sent": "A Model - agnostic Data Manipulation Method for Persona - based Dialogue Generation.", "forward": true, "src_ids": "2022.acl-long.550_2947"}
+{"input": "fair model is used for Method| context: however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions .", "entity": "fair model", "output": "knowledge distillation", "neg_sample": ["fair model is used for Method", "however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions ."], "relation": "used for", "id": "2022.findings-acl.55", "year": 2022, "rel_sent": "To this end , we present a novel approach to mitigate gender disparity in text generation by learning a fair model during knowledge distillation .", "forward": true, "src_ids": "2022.findings-acl.55_2948"}
+{"input": "knowledge distillation is done by using Method| context: language models excel at generating coherent text , and model compression techniques such as knowledge distillation have enabled their use in resource - constrained settings . however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions . therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model .", "entity": "knowledge distillation", "output": "fair model", "neg_sample": ["knowledge distillation is done by using Method", "language models excel at generating coherent text , and model compression techniques such as knowledge distillation have enabled their use in resource - constrained settings .", "however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions .", "therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model ."], "relation": "used for", "id": "2022.findings-acl.55", "year": 2022, "rel_sent": "To this end , we present a novel approach to mitigate gender disparity in text generation by learning a fair model during knowledge distillation .", "forward": false, "src_ids": "2022.findings-acl.55_2949"}
+{"input": "language models is used for OtherScientificTerm| context: language models excel at generating coherent text , and model compression techniques such as knowledge distillation have enabled their use in resource - constrained settings . however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions . therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model .", "entity": "language models", "output": "gender polarity", "neg_sample": ["language models is used for OtherScientificTerm", "language models excel at generating coherent text , and model compression techniques such as knowledge distillation have enabled their use in resource - constrained settings .", "however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions .", "therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model ."], "relation": "used for", "id": "2022.findings-acl.55", "year": 2022, "rel_sent": "Finally , we observe that language models that reduce gender polarity in language generation do not improve embedding fairness or downstream classification fairness .", "forward": true, "src_ids": "2022.findings-acl.55_2950"}
+{"input": "language generation is done by using Method| context: however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions . therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model .", "entity": "language generation", "output": "language models", "neg_sample": ["language generation is done by using Method", "however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions .", "therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model ."], "relation": "used for", "id": "2022.findings-acl.55", "year": 2022, "rel_sent": "Finally , we observe that language models that reduce gender polarity in language generation do not improve embedding fairness or downstream classification fairness .", "forward": false, "src_ids": "2022.findings-acl.55_2951"}
+{"input": "gender polarity is done by using Method| context: however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions . therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model .", "entity": "gender polarity", "output": "language models", "neg_sample": ["gender polarity is done by using Method", "however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions .", "therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model ."], "relation": "used for", "id": "2022.findings-acl.55", "year": 2022, "rel_sent": "Finally , we observe that language models that reduce gender polarity in language generation do not improve embedding fairness or downstream classification fairness .", "forward": false, "src_ids": "2022.findings-acl.55_2952"}
+{"input": "language models is used for Task| context: language models excel at generating coherent text , and model compression techniques such as knowledge distillation have enabled their use in resource - constrained settings . however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions . therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model .", "entity": "language models", "output": "language generation", "neg_sample": ["language models is used for Task", "language models excel at generating coherent text , and model compression techniques such as knowledge distillation have enabled their use in resource - constrained settings .", "however , these models can be biased in multiple ways , including the unfounded association of male and female genders with gender - neutral professions .", "therefore , knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model 's biases onto the distilled model ."], "relation": "used for", "id": "2022.findings-acl.55", "year": 2022, "rel_sent": "Finally , we observe that language models that reduce gender polarity in language generation do not improve embedding fairness or downstream classification fairness .", "forward": true, "src_ids": "2022.findings-acl.55_2953"}
+{"input": "fine - tuned bert is used for Task| context: public opinion in social media is increasingly becoming a critical factor in pandemic control . understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated .", "entity": "fine - tuned bert", "output": "emotion analysis", "neg_sample": ["fine - tuned bert is used for Task", "public opinion in social media is increasingly becoming a critical factor in pandemic control .", "understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated ."], "relation": "used for", "id": "2022.wassa-1.10", "year": 2022, "rel_sent": "A separate dataset of 20,254 Japanese Tweets containing COVID-19 vaccine - related keywords was also collected , on which the fine - tuned BERT was used to perform emotion analysis .", "forward": true, "src_ids": "2022.wassa-1.10_2954"}
+{"input": "emotion ratings is used for Material| context: public opinion in social media is increasingly becoming a critical factor in pandemic control . understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated .", "entity": "emotion ratings", "output": "japanese tweets", "neg_sample": ["emotion ratings is used for Material", "public opinion in social media is increasingly becoming a critical factor in pandemic control .", "understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated ."], "relation": "used for", "id": "2022.wassa-1.10", "year": 2022, "rel_sent": "Using the WRIME dataset , which provides emotion ratings for Japanese Tweets sourced from writers ( Tweet posters ) and readers , we fine - tuned a BERT model to predict levels of emotional intensity .", "forward": true, "src_ids": "2022.wassa-1.10_2955"}
+{"input": "japanese tweets is done by using OtherScientificTerm| context: public opinion in social media is increasingly becoming a critical factor in pandemic control . understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated .", "entity": "japanese tweets", "output": "emotion ratings", "neg_sample": ["japanese tweets is done by using OtherScientificTerm", "public opinion in social media is increasingly becoming a critical factor in pandemic control .", "understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated ."], "relation": "used for", "id": "2022.wassa-1.10", "year": 2022, "rel_sent": "Using the WRIME dataset , which provides emotion ratings for Japanese Tweets sourced from writers ( Tweet posters ) and readers , we fine - tuned a BERT model to predict levels of emotional intensity .", "forward": false, "src_ids": "2022.wassa-1.10_2956"}
+{"input": "emotional intensity is done by using Method| context: public opinion in social media is increasingly becoming a critical factor in pandemic control . understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated .", "entity": "emotional intensity", "output": "bert model", "neg_sample": ["emotional intensity is done by using Method", "public opinion in social media is increasingly becoming a critical factor in pandemic control .", "understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated ."], "relation": "used for", "id": "2022.wassa-1.10", "year": 2022, "rel_sent": "Using the WRIME dataset , which provides emotion ratings for Japanese Tweets sourced from writers ( Tweet posters ) and readers , we fine - tuned a BERT model to predict levels of emotional intensity .", "forward": false, "src_ids": "2022.wassa-1.10_2957"}
+{"input": "bert model is used for OtherScientificTerm| context: public opinion in social media is increasingly becoming a critical factor in pandemic control . understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated .", "entity": "bert model", "output": "emotional intensity", "neg_sample": ["bert model is used for OtherScientificTerm", "public opinion in social media is increasingly becoming a critical factor in pandemic control .", "understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated ."], "relation": "used for", "id": "2022.wassa-1.10", "year": 2022, "rel_sent": "Using the WRIME dataset , which provides emotion ratings for Japanese Tweets sourced from writers ( Tweet posters ) and readers , we fine - tuned a BERT model to predict levels of emotional intensity .", "forward": true, "src_ids": "2022.wassa-1.10_2958"}
+{"input": "emotion analysis is done by using Method| context: public opinion in social media is increasingly becoming a critical factor in pandemic control . understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated .", "entity": "emotion analysis", "output": "fine - tuned bert", "neg_sample": ["emotion analysis is done by using Method", "public opinion in social media is increasingly becoming a critical factor in pandemic control .", "understanding the emotions of a population towards vaccinations and covid-19 may be valuable in convincing members to become vaccinated ."], "relation": "used for", "id": "2022.wassa-1.10", "year": 2022, "rel_sent": "A separate dataset of 20,254 Japanese Tweets containing COVID-19 vaccine - related keywords was also collected , on which the fine - tuned BERT was used to perform emotion analysis .", "forward": false, "src_ids": "2022.wassa-1.10_2959"}
+{"input": "multi - mode word distributions is done by using Method| context: the softmax layer produces the distribution based on the dot products of a single hidden state and the embeddings of words in the vocabulary . however , we discover that this single hidden state can not produce all probability distributions regardless of the lm size or training data size because the single hidden state embedding can not be close to the embeddings of all the possible next words simultaneously when there are other interfering word embeddings between them .", "entity": "multi - mode word distributions", "output": "language models", "neg_sample": ["multi - mode word distributions is done by using Method", "the softmax layer produces the distribution based on the dot products of a single hidden state and the embeddings of words in the vocabulary .", "however , we discover that this single hidden state can not produce all probability distributions regardless of the lm size or training data size because the single hidden state embedding can not be close to the embeddings of all the possible next words simultaneously when there are other interfering word embeddings between them ."], "relation": "used for", "id": "2022.acl-long.554", "year": 2022, "rel_sent": "Softmax Bottleneck Makes Language Models Unable to Represent Multi - mode Word Distributions.", "forward": false, "src_ids": "2022.acl-long.554_2960"}
+{"input": "language models is used for OtherScientificTerm| context: neural language models ( lms ) such as gpt-2 estimate the probability distribution over the next word by a softmax over the vocabulary . the softmax layer produces the distribution based on the dot products of a single hidden state and the embeddings of words in the vocabulary . however , we discover that this single hidden state can not produce all probability distributions regardless of the lm size or training data size because the single hidden state embedding can not be close to the embeddings of all the possible next words simultaneously when there are other interfering word embeddings between them .", "entity": "language models", "output": "multi - mode word distributions", "neg_sample": ["language models is used for OtherScientificTerm", "neural language models ( lms ) such as gpt-2 estimate the probability distribution over the next word by a softmax over the vocabulary .", "the softmax layer produces the distribution based on the dot products of a single hidden state and the embeddings of words in the vocabulary .", "however , we discover that this single hidden state can not produce all probability distributions regardless of the lm size or training data size because the single hidden state embedding can not be close to the embeddings of all the possible next words simultaneously when there are other interfering word embeddings between them ."], "relation": "used for", "id": "2022.acl-long.554", "year": 2022, "rel_sent": "Softmax Bottleneck Makes Language Models Unable to Represent Multi - mode Word Distributions.", "forward": true, "src_ids": "2022.acl-long.554_2961"}
+{"input": "statistical measures is done by using OtherScientificTerm| context: eye tracking data during reading is a useful source of information to understand the cognitive processes that take place during language comprehension processes . different languages account for different cognitive triggers , however there seems to be some uniform indicatorsacross languages .", "entity": "statistical measures", "output": "regression layer", "neg_sample": ["statistical measures is done by using OtherScientificTerm", "eye tracking data during reading is a useful source of information to understand the cognitive processes that take place during language comprehension processes .", "different languages account for different cognitive triggers , however there seems to be some uniform indicatorsacross languages ."], "relation": "used for", "id": "2022.cmcl-1.11", "year": 2022, "rel_sent": "Our model uses text representations from transformers and some hand engineered features with a regression layer on top to predict statistical measures of mean and standard deviation for 2 main eye - tracking features .", "forward": false, "src_ids": "2022.cmcl-1.11_2962"}
+{"input": "regression layer is used for Metric| context: eye tracking data during reading is a useful source of information to understand the cognitive processes that take place during language comprehension processes . different languages account for different cognitive triggers , however there seems to be some uniform indicatorsacross languages .", "entity": "regression layer", "output": "statistical measures", "neg_sample": ["regression layer is used for Metric", "eye tracking data during reading is a useful source of information to understand the cognitive processes that take place during language comprehension processes .", "different languages account for different cognitive triggers , however there seems to be some uniform indicatorsacross languages ."], "relation": "used for", "id": "2022.cmcl-1.11", "year": 2022, "rel_sent": "Our model uses text representations from transformers and some hand engineered features with a regression layer on top to predict statistical measures of mean and standard deviation for 2 main eye - tracking features .", "forward": true, "src_ids": "2022.cmcl-1.11_2963"}
+{"input": "assessment of politeness strategies is used for Method| context: distinct from domain - agnostic politeness constructs , in specific domains such as online stores , booking platforms , and others , agents need to be capable of adopting highly specific vocabulary , with significant impact on lexical and grammatical aspects of utterances . then , the challenge is on improving utterances ' politeness while preserving the actual content , an utterly central requirement to achieve the task goal .", "entity": "assessment of politeness strategies", "output": "polite task - oriented dialog agents", "neg_sample": ["assessment of politeness strategies is used for Method", "distinct from domain - agnostic politeness constructs , in specific domains such as online stores , booking platforms , and others , agents need to be capable of adopting highly specific vocabulary , with significant impact on lexical and grammatical aspects of utterances .", "then , the challenge is on improving utterances ' politeness while preserving the actual content , an utterly central requirement to achieve the task goal ."], "relation": "used for", "id": "2022.wassa-1.34", "year": 2022, "rel_sent": "In this paper , we conduct a novel assessment of politeness strategies for task - oriented dialog agents under a transfer learning scenario .", "forward": true, "src_ids": "2022.wassa-1.34_2964"}
+{"input": "polite task - oriented dialog agents is done by using Method| context: for task - oriented dialog agents , the tone of voice mediates user - agent interactions , playing a central role in the flow of a conversation . distinct from domain - agnostic politeness constructs , in specific domains such as online stores , booking platforms , and others , agents need to be capable of adopting highly specific vocabulary , with significant impact on lexical and grammatical aspects of utterances . then , the challenge is on improving utterances ' politeness while preserving the actual content , an utterly central requirement to achieve the task goal .", "entity": "polite task - oriented dialog agents", "output": "assessment of politeness strategies", "neg_sample": ["polite task - oriented dialog agents is done by using Method", "for task - oriented dialog agents , the tone of voice mediates user - agent interactions , playing a central role in the flow of a conversation .", "distinct from domain - agnostic politeness constructs , in specific domains such as online stores , booking platforms , and others , agents need to be capable of adopting highly specific vocabulary , with significant impact on lexical and grammatical aspects of utterances .", "then , the challenge is on improving utterances ' politeness while preserving the actual content , an utterly central requirement to achieve the task goal ."], "relation": "used for", "id": "2022.wassa-1.34", "year": 2022, "rel_sent": "In this paper , we conduct a novel assessment of politeness strategies for task - oriented dialog agents under a transfer learning scenario .", "forward": false, "src_ids": "2022.wassa-1.34_2965"}
+{"input": "nested named entity recognition is done by using Method| context: nested named entity recognition ( ner ) is a task in which named entities may overlap with each other . span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task . however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data .", "entity": "nested named entity recognition", "output": "span selection framework", "neg_sample": ["nested named entity recognition is done by using Method", "nested named entity recognition ( ner ) is a task in which named entities may overlap with each other .", "span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task .", "however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data ."], "relation": "used for", "id": "2022.findings-acl.9", "year": 2022, "rel_sent": "Extract - Select : A Span Selection Framework for Nested Named Entity Recognition with Generative Adversarial Training.", "forward": false, "src_ids": "2022.findings-acl.9_2966"}
+{"input": "nested ner is done by using Method| context: nested named entity recognition ( ner ) is a task in which named entities may overlap with each other . span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task . however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data .", "entity": "nested ner", "output": "span selection framework", "neg_sample": ["nested ner is done by using Method", "nested named entity recognition ( ner ) is a task in which named entities may overlap with each other .", "span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task .", "however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data ."], "relation": "used for", "id": "2022.findings-acl.9", "year": 2022, "rel_sent": "In this paper , we propose Extract - Select , a span selection framework for nested NER , to tackle these problems .", "forward": false, "src_ids": "2022.findings-acl.9_2967"}
+{"input": "span selection framework is used for Task| context: span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task . however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data .", "entity": "span selection framework", "output": "nested named entity recognition", "neg_sample": ["span selection framework is used for Task", "span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task .", "however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data ."], "relation": "used for", "id": "2022.findings-acl.9", "year": 2022, "rel_sent": "Extract - Select : A Span Selection Framework for Nested Named Entity Recognition with Generative Adversarial Training.", "forward": true, "src_ids": "2022.findings-acl.9_2968"}
+{"input": "span selection framework is used for Task| context: nested named entity recognition ( ner ) is a task in which named entities may overlap with each other . however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data .", "entity": "span selection framework", "output": "nested ner", "neg_sample": ["span selection framework is used for Task", "nested named entity recognition ( ner ) is a task in which named entities may overlap with each other .", "however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data ."], "relation": "used for", "id": "2022.findings-acl.9", "year": 2022, "rel_sent": "In this paper , we propose Extract - Select , a span selection framework for nested NER , to tackle these problems .", "forward": true, "src_ids": "2022.findings-acl.9_2969"}
+{"input": "hybrid selection strategy is used for Task| context: nested named entity recognition ( ner ) is a task in which named entities may overlap with each other . span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task .", "entity": "hybrid selection strategy", "output": "long entity recognition", "neg_sample": ["hybrid selection strategy is used for Task", "nested named entity recognition ( ner ) is a task in which named entities may overlap with each other .", "span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task ."], "relation": "used for", "id": "2022.findings-acl.9", "year": 2022, "rel_sent": "Secondly , we propose a hybrid selection strategy in the extractor , which not only makes full use of span boundary but also improves the ability of long entity recognition .", "forward": true, "src_ids": "2022.findings-acl.9_2970"}
+{"input": "long entity recognition is done by using Method| context: nested named entity recognition ( ner ) is a task in which named entities may overlap with each other . span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task . however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data .", "entity": "long entity recognition", "output": "hybrid selection strategy", "neg_sample": ["long entity recognition is done by using Method", "nested named entity recognition ( ner ) is a task in which named entities may overlap with each other .", "span - based approaches regard nested ner as a two - stage span enumeration and classification task , thus having the innate ability to handle this task .", "however , they face the problems of error propagation , ignorance of span boundary , difficulty in long entity recognition and requirement on large - scale annotated data ."], "relation": "used for", "id": "2022.findings-acl.9", "year": 2022, "rel_sent": "Secondly , we propose a hybrid selection strategy in the extractor , which not only makes full use of span boundary but also improves the ability of long entity recognition .", "forward": false, "src_ids": "2022.findings-acl.9_2971"}
+{"input": "latent space is done by using Method| context: while variational autoencoders ( vaes ) have been widely applied in text generation tasks , they are troubled by two challenges : insufficient representation capacity and poor controllability . the former results from the posterior collapse and restrictive assumption , which impede better representation learning . the latter arises as continuous latent variables in traditional formulations hinder vaes from interpretability and controllability .", "entity": "latent space", "output": "contrastive learning", "neg_sample": ["latent space is done by using Method", "while variational autoencoders ( vaes ) have been widely applied in text generation tasks , they are troubled by two challenges : insufficient representation capacity and poor controllability .", "the former results from the posterior collapse and restrictive assumption , which impede better representation learning .", "the latter arises as continuous latent variables in traditional formulations hinder vaes from interpretability and controllability ."], "relation": "used for", "id": "2022.findings-acl.10", "year": 2022, "rel_sent": "Tofacilitate controlled text generation with DPrior , we propose to employ contrastive learning to separate the latent space into several parts .", "forward": false, "src_ids": "2022.findings-acl.10_2972"}
+{"input": "contrastive learning is used for OtherScientificTerm| context: while variational autoencoders ( vaes ) have been widely applied in text generation tasks , they are troubled by two challenges : insufficient representation capacity and poor controllability . the former results from the posterior collapse and restrictive assumption , which impede better representation learning . the latter arises as continuous latent variables in traditional formulations hinder vaes from interpretability and controllability .", "entity": "contrastive learning", "output": "latent space", "neg_sample": ["contrastive learning is used for OtherScientificTerm", "while variational autoencoders ( vaes ) have been widely applied in text generation tasks , they are troubled by two challenges : insufficient representation capacity and poor controllability .", "the former results from the posterior collapse and restrictive assumption , which impede better representation learning .", "the latter arises as continuous latent variables in traditional formulations hinder vaes from interpretability and controllability ."], "relation": "used for", "id": "2022.findings-acl.10", "year": 2022, "rel_sent": "Tofacilitate controlled text generation with DPrior , we propose to employ contrastive learning to separate the latent space into several parts .", "forward": true, "src_ids": "2022.findings-acl.10_2973"}
+{"input": "assistive technology is done by using Method| context: intelligent conversational assistants have become an integral part of our lives for performing simple tasks . however , such agents , for example , google bots , alexa and others are yet to have any social impact on minority population , for example , for people with neurological disorders and people with speech , language and social communication disorders , sometimes with locked - in states where speaking or typing is a challenge . language model technologies can be very powerful tools in enabling these users to carry out daily communication and social interactions .", "entity": "assistive technology", "output": "conversational agent", "neg_sample": ["assistive technology is done by using Method", "intelligent conversational assistants have become an integral part of our lives for performing simple tasks .", "however , such agents , for example , google bots , alexa and others are yet to have any social impact on minority population , for example , for people with neurological disorders and people with speech , language and social communication disorders , sometimes with locked - in states where speaking or typing is a challenge .", "language model technologies can be very powerful tools in enabling these users to carry out daily communication and social interactions ."], "relation": "used for", "id": "2022.acl-demo.19", "year": 2022, "rel_sent": "Cue - bot : A Conversational Agent for Assistive Technology.", "forward": false, "src_ids": "2022.acl-demo.19_2974"}
+{"input": "conversational agent is used for Method| context: intelligent conversational assistants have become an integral part of our lives for performing simple tasks . however , such agents , for example , google bots , alexa and others are yet to have any social impact on minority population , for example , for people with neurological disorders and people with speech , language and social communication disorders , sometimes with locked - in states where speaking or typing is a challenge . language model technologies can be very powerful tools in enabling these users to carry out daily communication and social interactions .", "entity": "conversational agent", "output": "assistive technology", "neg_sample": ["conversational agent is used for Method", "intelligent conversational assistants have become an integral part of our lives for performing simple tasks .", "however , such agents , for example , google bots , alexa and others are yet to have any social impact on minority population , for example , for people with neurological disorders and people with speech , language and social communication disorders , sometimes with locked - in states where speaking or typing is a challenge .", "language model technologies can be very powerful tools in enabling these users to carry out daily communication and social interactions ."], "relation": "used for", "id": "2022.acl-demo.19", "year": 2022, "rel_sent": "Cue - bot : A Conversational Agent for Assistive Technology.", "forward": true, "src_ids": "2022.acl-demo.19_2975"}
+{"input": "phs - bert is used for Task| context: a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence . despite its potential , the technology is still in its early stages and is not ready for widespread application . advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications .", "entity": "phs - bert", "output": "public health surveillance ( phs )", "neg_sample": ["phs - bert is used for Task", "a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence .", "despite its potential , the technology is still in its early stages and is not ready for widespread application .", "advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications ."], "relation": "used for", "id": "2022.nlppower-1.3", "year": 2022, "rel_sent": "We present and release PHS - BERT , a transformer - based PLM , to identify tasks related to public health surveillance on social media .", "forward": true, "src_ids": "2022.nlppower-1.3_2976"}
+{"input": "transformer - based plm is used for Task| context: a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence . despite its potential , the technology is still in its early stages and is not ready for widespread application . advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications .", "entity": "transformer - based plm", "output": "public health surveillance ( phs )", "neg_sample": ["transformer - based plm is used for Task", "a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence .", "despite its potential , the technology is still in its early stages and is not ready for widespread application .", "advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications ."], "relation": "used for", "id": "2022.nlppower-1.3", "year": 2022, "rel_sent": "We present and release PHS - BERT , a transformer - based PLM , to identify tasks related to public health surveillance on social media .", "forward": true, "src_ids": "2022.nlppower-1.3_2977"}
+{"input": "public health surveillance ( phs ) is done by using Method| context: a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence . this exchange of health information on social media has been regarded as an attempt to enhance public health surveillance ( phs ) . despite its potential , the technology is still in its early stages and is not ready for widespread application . advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications . however , there are no plms for social media tasks involving phs .", "entity": "public health surveillance ( phs )", "output": "phs - bert", "neg_sample": ["public health surveillance ( phs ) is done by using Method", "a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence .", "this exchange of health information on social media has been regarded as an attempt to enhance public health surveillance ( phs ) .", "despite its potential , the technology is still in its early stages and is not ready for widespread application .", "advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications .", "however , there are no plms for social media tasks involving phs ."], "relation": "used for", "id": "2022.nlppower-1.3", "year": 2022, "rel_sent": "We present and release PHS - BERT , a transformer - based PLM , to identify tasks related to public health surveillance on social media .", "forward": false, "src_ids": "2022.nlppower-1.3_2978"}
+{"input": "plm is used for Task| context: a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence . this exchange of health information on social media has been regarded as an attempt to enhance public health surveillance ( phs ) . despite its potential , the technology is still in its early stages and is not ready for widespread application . advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications . however , there are no plms for social media tasks involving phs .", "entity": "plm", "output": "phs - related tasks", "neg_sample": ["plm is used for Task", "a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence .", "this exchange of health information on social media has been regarded as an attempt to enhance public health surveillance ( phs ) .", "despite its potential , the technology is still in its early stages and is not ready for widespread application .", "advancements in pretrained language models ( plms ) have facilitated the development of several domain - specific plms and a variety of downstream applications .", "however , there are no plms for social media tasks involving phs ."], "relation": "used for", "id": "2022.nlppower-1.3", "year": 2022, "rel_sent": "Compared with existing PLMs that are mainly evaluated on limited tasks , PHS - BERT achieved state - of - the - art performance on all 25 tested datasets , showing that our PLM is robust and generalizable in the common PHS tasks .", "forward": true, "src_ids": "2022.nlppower-1.3_2979"}
+{"input": "phs - related tasks is done by using Method| context: a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence . this exchange of health information on social media has been regarded as an attempt to enhance public health surveillance ( phs ) . despite its potential , the technology is still in its early stages and is not ready for widespread application .", "entity": "phs - related tasks", "output": "plm", "neg_sample": ["phs - related tasks is done by using Method", "a user - generated text on social media enables health workers to keep track of information , identify possible outbreaks , forecast disease trends , monitor emergency cases , and ascertain disease awareness and response to official health correspondence .", "this exchange of health information on social media has been regarded as an attempt to enhance public health surveillance ( phs ) .", "despite its potential , the technology is still in its early stages and is not ready for widespread application ."], "relation": "used for", "id": "2022.nlppower-1.3", "year": 2022, "rel_sent": "Compared with existing PLMs that are mainly evaluated on limited tasks , PHS - BERT achieved state - of - the - art performance on all 25 tested datasets , showing that our PLM is robust and generalizable in the common PHS tasks .", "forward": false, "src_ids": "2022.nlppower-1.3_2980"}
+{"input": "type information is done by using OtherScientificTerm| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "type information", "output": "indirect supervision", "neg_sample": ["type information is done by using OtherScientificTerm", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": false, "src_ids": "2022.tacl-1.35_2981"}
+{"input": "natural language inference is done by using OtherScientificTerm| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "natural language inference", "output": "indirect supervision", "neg_sample": ["natural language inference is done by using OtherScientificTerm", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": false, "src_ids": "2022.tacl-1.35_2982"}
+{"input": "type information is done by using Task| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "type information", "output": "natural language inference", "neg_sample": ["type information is done by using Task", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": false, "src_ids": "2022.tacl-1.35_2983"}
+{"input": "indirect supervision is used for Task| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "indirect supervision", "output": "natural language inference", "neg_sample": ["indirect supervision is used for Task", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": true, "src_ids": "2022.tacl-1.35_2984"}
+{"input": "entity typing is done by using Method| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "entity typing", "output": "lite", "neg_sample": ["entity typing is done by using Method", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": false, "src_ids": "2022.tacl-1.35_2985"}
+{"input": "limited training data is used for Method| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "limited training data", "output": "lite", "neg_sample": ["limited training data is used for Method", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "Experiments show that , with limited training data , LITE obtains state - of - the - art performance on the UFET task .", "forward": true, "src_ids": "2022.tacl-1.35_2986"}
+{"input": "lite is used for Task| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "lite", "output": "entity typing", "neg_sample": ["lite is used for Task", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": true, "src_ids": "2022.tacl-1.35_2987"}
+{"input": "indirect supervision is used for OtherScientificTerm| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "indirect supervision", "output": "type information", "neg_sample": ["indirect supervision is used for OtherScientificTerm", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": true, "src_ids": "2022.tacl-1.35_2988"}
+{"input": "natural language inference is used for OtherScientificTerm| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "natural language inference", "output": "type information", "neg_sample": ["natural language inference is used for OtherScientificTerm", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "This work presents LITE , a new approach that formulates entity typing as a natural language inference ( NLI ) problem , making use of ( i ) the indirect supervision from NLI to infer type information meaningfully represented as textual hypotheses and alleviate the data scarcity issue , as well as ( ii ) a learning - to - rank objective to avoid the pre - defining of a type set .", "forward": true, "src_ids": "2022.tacl-1.35_2989"}
+{"input": "lite is done by using Material| context: a key challenge for this task lies in the large number of types and the scarcity of annotated data per type . existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers . this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training .", "entity": "lite", "output": "limited training data", "neg_sample": ["lite is done by using Material", "a key challenge for this task lies in the large number of types and the scarcity of annotated data per type .", "existing systems formulate the task as a multi - way classification problem and train directly or distantly supervised classifiers .", "this causes two issues : ( i ) the classifiers do not capture the type semantics because types are often converted into indices ; ( ii ) systems developed in this way are limited to predicting within a pre - defined type set , and often fall short of generalizing to types that are rarely seen or unseen in training ."], "relation": "used for", "id": "2022.tacl-1.35", "year": 2022, "rel_sent": "Experiments show that , with limited training data , LITE obtains state - of - the - art performance on the UFET task .", "forward": false, "src_ids": "2022.tacl-1.35_2990"}
+{"input": "grammars is used for OtherScientificTerm| context: magri ( 2018 ) argues that the types of implicational universals that remain valid in maxent are phonologically implausible , suggesting that maxent overgenerates noisy hg in a problematic way . however , a variety of recent work has shown that some of the possible grammars in a constraint based grammar may be unlikely to be observed because they are difficult to learn ( staubs , 2014 ; stanton , 2016 ; pater and moreton , 2012 ; hughto , 2019 ; o'hara , 2021 ) .", "entity": "grammars", "output": "harmonically bounded candidates", "neg_sample": ["grammars is used for OtherScientificTerm", "magri ( 2018 ) argues that the types of implicational universals that remain valid in maxent are phonologically implausible , suggesting that maxent overgenerates noisy hg in a problematic way .", "however , a variety of recent work has shown that some of the possible grammars in a constraint based grammar may be unlikely to be observed because they are difficult to learn ( staubs , 2014 ; stanton , 2016 ; pater and moreton , 2012 ; hughto , 2019 ; o'hara , 2021 ) ."], "relation": "used for", "id": "2022.scil-1.23", "year": 2022, "rel_sent": "Here , I show that grammars that give too much weight to harmonically bounded candidates , and violate the implicational universals that hold in Noisy HG are significantly harder to learn than those grammars that are also possible in Noisy HG .", "forward": true, "src_ids": "2022.scil-1.23_2991"}
+{"input": "harmonically bounded candidates is done by using Method| context: magri ( 2018 ) argues that the types of implicational universals that remain valid in maxent are phonologically implausible , suggesting that maxent overgenerates noisy hg in a problematic way .", "entity": "harmonically bounded candidates", "output": "grammars", "neg_sample": ["harmonically bounded candidates is done by using Method", "magri ( 2018 ) argues that the types of implicational universals that remain valid in maxent are phonologically implausible , suggesting that maxent overgenerates noisy hg in a problematic way ."], "relation": "used for", "id": "2022.scil-1.23", "year": 2022, "rel_sent": "Here , I show that grammars that give too much weight to harmonically bounded candidates , and violate the implicational universals that hold in Noisy HG are significantly harder to learn than those grammars that are also possible in Noisy HG .", "forward": false, "src_ids": "2022.scil-1.23_2992"}
+{"input": "imbalanced data is done by using Method| context: spreading positive vibes or hope content on social media may help many people to get motivated in their life .", "entity": "imbalanced data", "output": "re - sampling technique", "neg_sample": ["imbalanced data is done by using Method", "spreading positive vibes or hope content on social media may help many people to get motivated in their life ."], "relation": "used for", "id": "2022.ltedi-1.20", "year": 2022, "rel_sent": "The proposed methodology uses the re - sampling technique to deal with imbalanced data in the corpus and obtained 1st rank for English language with a macro - averaged F1 - score of 0.550 and weighted - averaged F1 - score of 0.860 .", "forward": false, "src_ids": "2022.ltedi-1.20_2993"}
+{"input": "re - sampling technique is used for Material| context: spreading positive vibes or hope content on social media may help many people to get motivated in their life .", "entity": "re - sampling technique", "output": "imbalanced data", "neg_sample": ["re - sampling technique is used for Material", "spreading positive vibes or hope content on social media may help many people to get motivated in their life ."], "relation": "used for", "id": "2022.ltedi-1.20", "year": 2022, "rel_sent": "The proposed methodology uses the re - sampling technique to deal with imbalanced data in the corpus and obtained 1st rank for English language with a macro - averaged F1 - score of 0.550 and weighted - averaged F1 - score of 0.860 .", "forward": true, "src_ids": "2022.ltedi-1.20_2994"}
+{"input": "semantic retrieval is done by using Task| context: how to learn highly compact yet effective sentence representation ? pre - trained language models have been effective in many nlp tasks . however , these models are often huge and produce large sentence embeddings . moreover , there is a big performance gap between large and small models .", "entity": "semantic retrieval", "output": "compressing sentence representation", "neg_sample": ["semantic retrieval is done by using Task", "how to learn highly compact yet effective sentence representation ?", "pre - trained language models have been effective in many nlp tasks .", "however , these models are often huge and produce large sentence embeddings .", "moreover , there is a big performance gap between large and small models ."], "relation": "used for", "id": "2022.findings-acl.64", "year": 2022, "rel_sent": "Compressing Sentence Representation for Semantic Retrieval via Homomorphic Projective Distillation.", "forward": false, "src_ids": "2022.findings-acl.64_2995"}
+{"input": "compressing sentence representation is used for Task| context: how to learn highly compact yet effective sentence representation ? pre - trained language models have been effective in many nlp tasks . however , these models are often huge and produce large sentence embeddings . moreover , there is a big performance gap between large and small models .", "entity": "compressing sentence representation", "output": "semantic retrieval", "neg_sample": ["compressing sentence representation is used for Task", "how to learn highly compact yet effective sentence representation ?", "pre - trained language models have been effective in many nlp tasks .", "however , these models are often huge and produce large sentence embeddings .", "moreover , there is a big performance gap between large and small models ."], "relation": "used for", "id": "2022.findings-acl.64", "year": 2022, "rel_sent": "Compressing Sentence Representation for Semantic Retrieval via Homomorphic Projective Distillation.", "forward": true, "src_ids": "2022.findings-acl.64_2996"}
+{"input": "compressed sentence embeddings is done by using Method| context: how to learn highly compact yet effective sentence representation ? pre - trained language models have been effective in many nlp tasks . however , these models are often huge and produce large sentence embeddings . moreover , there is a big performance gap between large and small models .", "entity": "compressed sentence embeddings", "output": "homomorphic projective distillation ( hpd )", "neg_sample": ["compressed sentence embeddings is done by using Method", "how to learn highly compact yet effective sentence representation ?", "pre - trained language models have been effective in many nlp tasks .", "however , these models are often huge and produce large sentence embeddings .", "moreover , there is a big performance gap between large and small models ."], "relation": "used for", "id": "2022.findings-acl.64", "year": 2022, "rel_sent": "In this paper , we propose Homomorphic Projective Distillation ( HPD ) to learn compressed sentence embeddings .", "forward": false, "src_ids": "2022.findings-acl.64_2997"}
+{"input": "homomorphic projective distillation ( hpd ) is used for OtherScientificTerm| context: how to learn highly compact yet effective sentence representation ? pre - trained language models have been effective in many nlp tasks . however , these models are often huge and produce large sentence embeddings . moreover , there is a big performance gap between large and small models .", "entity": "homomorphic projective distillation ( hpd )", "output": "compressed sentence embeddings", "neg_sample": ["homomorphic projective distillation ( hpd ) is used for OtherScientificTerm", "how to learn highly compact yet effective sentence representation ?", "pre - trained language models have been effective in many nlp tasks .", "however , these models are often huge and produce large sentence embeddings .", "moreover , there is a big performance gap between large and small models ."], "relation": "used for", "id": "2022.findings-acl.64", "year": 2022, "rel_sent": "In this paper , we propose Homomorphic Projective Distillation ( HPD ) to learn compressed sentence embeddings .", "forward": true, "src_ids": "2022.findings-acl.64_2998"}
+{"input": "compact representations is done by using Method| context: how to learn highly compact yet effective sentence representation ? pre - trained language models have been effective in many nlp tasks . however , these models are often huge and produce large sentence embeddings . moreover , there is a big performance gap between large and small models .", "entity": "compact representations", "output": "small transformer encoder model", "neg_sample": ["compact representations is done by using Method", "how to learn highly compact yet effective sentence representation ?", "pre - trained language models have been effective in many nlp tasks .", "however , these models are often huge and produce large sentence embeddings .", "moreover , there is a big performance gap between large and small models ."], "relation": "used for", "id": "2022.findings-acl.64", "year": 2022, "rel_sent": "Our method augments a small Transformer encoder model with learnable projection layers to produce compact representations while mimicking a large pre - trained language model to retain the sentence representation quality .", "forward": false, "src_ids": "2022.findings-acl.64_2999"}
+{"input": "small transformer encoder model is used for Generic| context: how to learn highly compact yet effective sentence representation ? pre - trained language models have been effective in many nlp tasks . however , these models are often huge and produce large sentence embeddings . moreover , there is a big performance gap between large and small models .", "entity": "small transformer encoder model", "output": "compact representations", "neg_sample": ["small transformer encoder model is used for Generic", "how to learn highly compact yet effective sentence representation ?", "pre - trained language models have been effective in many nlp tasks .", "however , these models are often huge and produce large sentence embeddings .", "moreover , there is a big performance gap between large and small models ."], "relation": "used for", "id": "2022.findings-acl.64", "year": 2022, "rel_sent": "Our method augments a small Transformer encoder model with learnable projection layers to produce compact representations while mimicking a large pre - trained language model to retain the sentence representation quality .", "forward": true, "src_ids": "2022.findings-acl.64_3000"}
+{"input": "depression assessment is done by using Method| context: many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse . at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment .", "entity": "depression assessment", "output": "contextual - embeddings", "neg_sample": ["depression assessment is done by using Method", "many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse .", "at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment ."], "relation": "used for", "id": "2022.wassa-1.9", "year": 2022, "rel_sent": "Understanding BERT 's Mood : The Role of Contextual - Embeddings as User - Representations for Depression Assessment.", "forward": false, "src_ids": "2022.wassa-1.9_3001"}
+{"input": "contextual - embeddings is used for Task| context: many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse . at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment .", "entity": "contextual - embeddings", "output": "depression assessment", "neg_sample": ["contextual - embeddings is used for Task", "many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse .", "at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment ."], "relation": "used for", "id": "2022.wassa-1.9", "year": 2022, "rel_sent": "Understanding BERT 's Mood : The Role of Contextual - Embeddings as User - Representations for Depression Assessment.", "forward": true, "src_ids": "2022.wassa-1.9_3002"}
+{"input": "human - level nlp tasks is done by using OtherScientificTerm| context: many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse . at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment .", "entity": "human - level nlp tasks", "output": "individual layers", "neg_sample": ["human - level nlp tasks is done by using OtherScientificTerm", "many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse .", "at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment ."], "relation": "used for", "id": "2022.wassa-1.9", "year": 2022, "rel_sent": "Using degree of depression as a case study , we do an empirical analysis on which off - the - shelf language model , individual layers , and combinations of layers seem most promising when applied to human - level NLP tasks .", "forward": false, "src_ids": "2022.wassa-1.9_3003"}
+{"input": "layers is used for Task| context: many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse . at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment .", "entity": "layers", "output": "human - level nlp tasks", "neg_sample": ["layers is used for Task", "many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse .", "at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment ."], "relation": "used for", "id": "2022.wassa-1.9", "year": 2022, "rel_sent": "Using degree of depression as a case study , we do an empirical analysis on which off - the - shelf language model , individual layers , and combinations of layers seem most promising when applied to human - level NLP tasks .", "forward": true, "src_ids": "2022.wassa-1.9_3004"}
+{"input": "individual layers is used for Task| context: many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse . at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment .", "entity": "individual layers", "output": "human - level nlp tasks", "neg_sample": ["individual layers is used for Task", "many recent works in natural language processing have demonstrated ability to assess aspects of mental health from personal discourse .", "at the same time , pre - trained contextual word embedding models have grown to dominate much of nlp but little is known empirically on how to best apply them for mental health assessment ."], "relation": "used for", "id": "2022.wassa-1.9", "year": 2022, "rel_sent": "Using degree of depression as a case study , we do an empirical analysis on which off - the - shelf language model , individual layers , and combinations of layers seem most promising when applied to human - level NLP tasks .", "forward": true, "src_ids": "2022.wassa-1.9_3005"}
+{"input": "boundary of the sensory word is done by using Method| context: synesthesia refers to the description of perceptions in one sensory modality through concepts from other modalities . it involves not only a linguistic phenomenon , but also a cognitive phenomenon structuring human thought and action , which makes it become a bridge between figurative linguistic phenomenon and abstract cognition , and thus be helpful to understand the deep semantics .", "entity": "boundary of the sensory word", "output": "radical - based neural network model", "neg_sample": ["boundary of the sensory word is done by using Method", "synesthesia refers to the description of perceptions in one sensory modality through concepts from other modalities .", "it involves not only a linguistic phenomenon , but also a cognitive phenomenon structuring human thought and action , which makes it become a bridge between figurative linguistic phenomenon and abstract cognition , and thus be helpful to understand the deep semantics ."], "relation": "used for", "id": "2022.findings-acl.306", "year": 2022, "rel_sent": "Upon these baselines , we further propose a radical - based neural network model to identify the boundary of the sensory word , and to jointly detect the original and synesthetic sensory modalities for the word .", "forward": false, "src_ids": "2022.findings-acl.306_3006"}
+{"input": "radical - based neural network model is used for OtherScientificTerm| context: synesthesia refers to the description of perceptions in one sensory modality through concepts from other modalities . it involves not only a linguistic phenomenon , but also a cognitive phenomenon structuring human thought and action , which makes it become a bridge between figurative linguistic phenomenon and abstract cognition , and thus be helpful to understand the deep semantics .", "entity": "radical - based neural network model", "output": "boundary of the sensory word", "neg_sample": ["radical - based neural network model is used for OtherScientificTerm", "synesthesia refers to the description of perceptions in one sensory modality through concepts from other modalities .", "it involves not only a linguistic phenomenon , but also a cognitive phenomenon structuring human thought and action , which makes it become a bridge between figurative linguistic phenomenon and abstract cognition , and thus be helpful to understand the deep semantics ."], "relation": "used for", "id": "2022.findings-acl.306", "year": 2022, "rel_sent": "Upon these baselines , we further propose a radical - based neural network model to identify the boundary of the sensory word , and to jointly detect the original and synesthetic sensory modalities for the word .", "forward": true, "src_ids": "2022.findings-acl.306_3007"}
+{"input": "derivation tree reconstruction is done by using Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "derivation tree reconstruction", "output": "semantic composition", "neg_sample": ["derivation tree reconstruction is done by using Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "Semantic Composition with PSHRG for Derivation Tree Reconstruction from Graph - Based Meaning Representations.", "forward": false, "src_ids": "2022.acl-long.372_3008"}
+{"input": "derivation tree reconstruction is done by using Method| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "derivation tree reconstruction", "output": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "neg_sample": ["derivation tree reconstruction is done by using Method", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "Semantic Composition with PSHRG for Derivation Tree Reconstruction from Graph - Based Meaning Representations.", "forward": false, "src_ids": "2022.acl-long.372_3009"}
+{"input": "syntax - semantics interface is done by using Method| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "syntax - semantics interface", "output": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "neg_sample": ["syntax - semantics interface is done by using Method", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": false, "src_ids": "2022.acl-long.372_3010"}
+{"input": "modelling compositional graph - to - tree translations is done by using Method| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "modelling compositional graph - to - tree translations", "output": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "neg_sample": ["modelling compositional graph - to - tree translations is done by using Method", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": false, "src_ids": "2022.acl-long.372_3011"}
+{"input": "channelling explainability is done by using Method| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "channelling explainability", "output": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "neg_sample": ["channelling explainability is done by using Method", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": false, "src_ids": "2022.acl-long.372_3012"}
+{"input": "probabilistic synchronous hyperedge replacement grammar ( pshrg ) is used for Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "output": "derivation tree reconstruction", "neg_sample": ["probabilistic synchronous hyperedge replacement grammar ( pshrg ) is used for Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "Semantic Composition with PSHRG for Derivation Tree Reconstruction from Graph - Based Meaning Representations.", "forward": true, "src_ids": "2022.acl-long.372_3013"}
+{"input": "semantic composition is used for Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "semantic composition", "output": "derivation tree reconstruction", "neg_sample": ["semantic composition is used for Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "Semantic Composition with PSHRG for Derivation Tree Reconstruction from Graph - Based Meaning Representations.", "forward": true, "src_ids": "2022.acl-long.372_3014"}
+{"input": "generating derivation trees is done by using Method| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "generating derivation trees", "output": "data - driven approach", "neg_sample": ["generating derivation trees is done by using Method", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "We introduce a data - driven approach to generating derivation trees from meaning representation graphs with probabilistic synchronous hyperedge replacement grammar ( PSHRG ) .", "forward": false, "src_ids": "2022.acl-long.372_3015"}
+{"input": "data - driven approach is used for Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "data - driven approach", "output": "generating derivation trees", "neg_sample": ["data - driven approach is used for Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "We introduce a data - driven approach to generating derivation trees from meaning representation graphs with probabilistic synchronous hyperedge replacement grammar ( PSHRG ) .", "forward": true, "src_ids": "2022.acl-long.372_3016"}
+{"input": "probabilistic synchronous hyperedge replacement grammar ( pshrg ) is used for OtherScientificTerm| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "output": "syntax - semantics interface", "neg_sample": ["probabilistic synchronous hyperedge replacement grammar ( pshrg ) is used for OtherScientificTerm", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": true, "src_ids": "2022.acl-long.372_3017"}
+{"input": "probabilistic synchronous hyperedge replacement grammar ( pshrg ) is used for Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "probabilistic synchronous hyperedge replacement grammar ( pshrg )", "output": "modelling compositional graph - to - tree translations", "neg_sample": ["probabilistic synchronous hyperedge replacement grammar ( pshrg ) is used for Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": true, "src_ids": "2022.acl-long.372_3018"}
+{"input": "surface realization is done by using Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "surface realization", "output": "channelling explainability", "neg_sample": ["surface realization is done by using Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": false, "src_ids": "2022.acl-long.372_3019"}
+{"input": "channelling explainability is used for Task| context: shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse .", "entity": "channelling explainability", "output": "surface realization", "neg_sample": ["channelling explainability is used for Task", "shrg has been used to produce meaning representation graphs from texts and syntax trees , but little is known about its viability on the reverse ."], "relation": "used for", "id": "2022.acl-long.372", "year": 2022, "rel_sent": "This work reveals the ability of PSHRG in formalizing a syntax - semantics interface , modelling compositional graph - to - tree translations , and channelling explainability to surface realization .", "forward": true, "src_ids": "2022.acl-long.372_3020"}
+{"input": "translation is done by using OtherScientificTerm| context: neural machine translation ( nmt ) systems exhibit problematic biases , such as stereotypical gender bias in the translation of occupation terms into languages with grammatical gender . in this paper we describe a new source of bias prevalent in nmt systems , relating to translations of sentences containing person names .", "entity": "translation", "output": "randomly switching entities", "neg_sample": ["translation is done by using OtherScientificTerm", "neural machine translation ( nmt ) systems exhibit problematic biases , such as stereotypical gender bias in the translation of occupation terms into languages with grammatical gender .", "in this paper we describe a new source of bias prevalent in nmt systems , relating to translations of sentences containing person names ."], "relation": "used for", "id": "2022.acl-long.184", "year": 2022, "rel_sent": "To mitigate these biases we propose a simple but effective data augmentation method based on randomly switching entities during translation , which effectively eliminates the problem without any effect on translation quality .", "forward": false, "src_ids": "2022.acl-long.184_3021"}
+{"input": "language technologies is used for Task| context: innu - aimun is an algonquian language spoken in eastern canada . it is the language of the innu , an indigenous people that now lives for the most part in a dozen communities across quebec and labrador . although it is alive , innu - aimun sees important preservation and revitalization challenges and issues . the state of its technology is still nascent , with very few existing applications .", "entity": "language technologies", "output": "nlp applications", "neg_sample": ["language technologies is used for Task", "innu - aimun is an algonquian language spoken in eastern canada .", "it is the language of the innu , an indigenous people that now lives for the most part in a dozen communities across quebec and labrador .", "although it is alive , innu - aimun sees important preservation and revitalization challenges and issues .", "the state of its technology is still nascent , with very few existing applications ."], "relation": "used for", "id": "2022.computel-1.13", "year": 2022, "rel_sent": "Considering the existing linguistic and textual resources , we argue that developing language technology is feasible and propose first steps towards NLP applications like machine translation .", "forward": true, "src_ids": "2022.computel-1.13_3022"}
+{"input": "nlp applications is done by using Method| context: innu - aimun is an algonquian language spoken in eastern canada . it is the language of the innu , an indigenous people that now lives for the most part in a dozen communities across quebec and labrador . although it is alive , innu - aimun sees important preservation and revitalization challenges and issues . the state of its technology is still nascent , with very few existing applications .", "entity": "nlp applications", "output": "language technologies", "neg_sample": ["nlp applications is done by using Method", "innu - aimun is an algonquian language spoken in eastern canada .", "it is the language of the innu , an indigenous people that now lives for the most part in a dozen communities across quebec and labrador .", "although it is alive , innu - aimun sees important preservation and revitalization challenges and issues .", "the state of its technology is still nascent , with very few existing applications ."], "relation": "used for", "id": "2022.computel-1.13", "year": 2022, "rel_sent": "Considering the existing linguistic and textual resources , we argue that developing language technology is feasible and propose first steps towards NLP applications like machine translation .", "forward": false, "src_ids": "2022.computel-1.13_3023"}
+{"input": "plain x is used for OtherScientificTerm| context: the plain x platform is a toolbox for multilingual adaptation , for video , audio , and text content .", "entity": "plain x", "output": "voice - over", "neg_sample": ["plain x is used for OtherScientificTerm", "the plain x platform is a toolbox for multilingual adaptation , for video , audio , and text content ."], "relation": "used for", "id": "2022.eamt-1.49", "year": 2022, "rel_sent": "As a result , plain X enables a smooth semi - automated production of subtitles or voice - over , much faster than with older , manual workflows .", "forward": true, "src_ids": "2022.eamt-1.49_3024"}
+{"input": "multilingual media production process is done by using Task| context: the plain x platform is a toolbox for multilingual adaptation , for video , audio , and text content .", "entity": "multilingual media production process", "output": "artificial intelligence ( ai )", "neg_sample": ["multilingual media production process is done by using Task", "the plain x platform is a toolbox for multilingual adaptation , for video , audio , and text content ."], "relation": "used for", "id": "2022.eamt-1.49", "year": 2022, "rel_sent": "It brings Artificial Intelligence ( AI ) into the multilingual media production process , while keeping the human in the loop .", "forward": false, "src_ids": "2022.eamt-1.49_3025"}
+{"input": "artificial intelligence ( ai ) is used for Task| context: the plain x platform is a toolbox for multilingual adaptation , for video , audio , and text content .", "entity": "artificial intelligence ( ai )", "output": "multilingual media production process", "neg_sample": ["artificial intelligence ( ai ) is used for Task", "the plain x platform is a toolbox for multilingual adaptation , for video , audio , and text content ."], "relation": "used for", "id": "2022.eamt-1.49", "year": 2022, "rel_sent": "It brings Artificial Intelligence ( AI ) into the multilingual media production process , while keeping the human in the loop .", "forward": true, "src_ids": "2022.eamt-1.49_3026"}
+{"input": "contrastive visual semantic pretraining is used for OtherScientificTerm| context: we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions .", "entity": "contrastive visual semantic pretraining", "output": "semantics of natural language representations", "neg_sample": ["contrastive visual semantic pretraining is used for OtherScientificTerm", "we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions ."], "relation": "used for", "id": "2022.acl-long.217", "year": 2022, "rel_sent": "Contrastive Visual Semantic Pretraining Magnifies the Semantics of Natural Language Representations.", "forward": true, "src_ids": "2022.acl-long.217_3027"}
+{"input": "contrastive visual semantic pretraining is used for OtherScientificTerm| context: we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions .", "entity": "contrastive visual semantic pretraining", "output": "anisotropy", "neg_sample": ["contrastive visual semantic pretraining is used for OtherScientificTerm", "we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions ."], "relation": "used for", "id": "2022.acl-long.217", "year": 2022, "rel_sent": "We find that contrastive visual semantic pretraining significantly mitigates the anisotropy found in contextualized word embeddings from GPT-2 , such that the intra - layer self - similarity ( mean pairwise cosine similarity ) of CLIP word embeddings is under .25 in all layers , compared to greater than .95 in the top layer of GPT-2 .", "forward": true, "src_ids": "2022.acl-long.217_3028"}
+{"input": "contrastive visual semantic pretraining is used for Method| context: we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions .", "entity": "contrastive visual semantic pretraining", "output": "contextualized word embeddings", "neg_sample": ["contrastive visual semantic pretraining is used for Method", "we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions ."], "relation": "used for", "id": "2022.acl-long.217", "year": 2022, "rel_sent": "We find that contrastive visual semantic pretraining significantly mitigates the anisotropy found in contextualized word embeddings from GPT-2 , such that the intra - layer self - similarity ( mean pairwise cosine similarity ) of CLIP word embeddings is under .25 in all layers , compared to greater than .95 in the top layer of GPT-2 .", "forward": true, "src_ids": "2022.acl-long.217_3029"}
+{"input": "reasoning is done by using Method| context: the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem . it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g. , count , superlative , comparative ) .", "entity": "reasoning", "output": "mixture - of - experts neural network", "neg_sample": ["reasoning is done by using Method", "the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem .", "it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g.", ", count , superlative , comparative ) ."], "relation": "used for", "id": "2022.findings-acl.13", "year": 2022, "rel_sent": "Specifically , we have developed a mixture - of - experts neural network to recognize and execute different types of reasoning - the network is composed of multiple experts , each handling a specific part of the semantics for reasoning , whereas a management module is applied to decide the contribution of each expert network to the verification result .", "forward": false, "src_ids": "2022.findings-acl.13_3030"}
+{"input": "mixture - of - experts neural network is used for Task| context: the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem .", "entity": "mixture - of - experts neural network", "output": "reasoning", "neg_sample": ["mixture - of - experts neural network is used for Task", "the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem ."], "relation": "used for", "id": "2022.findings-acl.13", "year": 2022, "rel_sent": "Specifically , we have developed a mixture - of - experts neural network to recognize and execute different types of reasoning - the network is composed of multiple experts , each handling a specific part of the semantics for reasoning , whereas a management module is applied to decide the contribution of each expert network to the verification result .", "forward": true, "src_ids": "2022.findings-acl.13_3031"}
+{"input": "expert network is done by using Method| context: the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem . it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g. , count , superlative , comparative ) .", "entity": "expert network", "output": "management module", "neg_sample": ["expert network is done by using Method", "the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem .", "it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g.", ", count , superlative , comparative ) ."], "relation": "used for", "id": "2022.findings-acl.13", "year": 2022, "rel_sent": "Specifically , we have developed a mixture - of - experts neural network to recognize and execute different types of reasoning - the network is composed of multiple experts , each handling a specific part of the semantics for reasoning , whereas a management module is applied to decide the contribution of each expert network to the verification result .", "forward": false, "src_ids": "2022.findings-acl.13_3032"}
+{"input": "self - adaptive method is used for Method| context: the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem . it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g. , count , superlative , comparative ) .", "entity": "self - adaptive method", "output": "management module", "neg_sample": ["self - adaptive method is used for Method", "the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem .", "it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g.", ", count , superlative , comparative ) ."], "relation": "used for", "id": "2022.findings-acl.13", "year": 2022, "rel_sent": "A self - adaptive method is developed to teach the management module combining results of different experts more efficiently without external knowledge .", "forward": true, "src_ids": "2022.findings-acl.13_3033"}
+{"input": "management module is used for Method| context: the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem . it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g. , count , superlative , comparative ) .", "entity": "management module", "output": "expert network", "neg_sample": ["management module is used for Method", "the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem .", "it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g.", ", count , superlative , comparative ) ."], "relation": "used for", "id": "2022.findings-acl.13", "year": 2022, "rel_sent": "Specifically , we have developed a mixture - of - experts neural network to recognize and execute different types of reasoning - the network is composed of multiple experts , each handling a specific part of the semantics for reasoning , whereas a management module is applied to decide the contribution of each expert network to the verification result .", "forward": true, "src_ids": "2022.findings-acl.13_3034"}
+{"input": "management module is done by using Method| context: the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem . it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g. , count , superlative , comparative ) .", "entity": "management module", "output": "self - adaptive method", "neg_sample": ["management module is done by using Method", "the table - based fact verification task has recently gained widespread attention and yet remains to be a very challenging problem .", "it inherently requires informative reasoning over natural language together with different numerical and logical reasoning on tables ( e.g.", ", count , superlative , comparative ) ."], "relation": "used for", "id": "2022.findings-acl.13", "year": 2022, "rel_sent": "A self - adaptive method is developed to teach the management module combining results of different experts more efficiently without external knowledge .", "forward": false, "src_ids": "2022.findings-acl.13_3035"}
+{"input": "knowledge base question answering is done by using Method| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "knowledge base question answering", "output": "generation augmented iterative ranking", "neg_sample": ["knowledge base question answering is done by using Method", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "RNG - KBQA : Generation Augmented Iterative Ranking for Knowledge Base Question Answering.", "forward": false, "src_ids": "2022.acl-long.417_3036"}
+{"input": "generation augmented iterative ranking is used for Task| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "generation augmented iterative ranking", "output": "knowledge base question answering", "neg_sample": ["generation augmented iterative ranking is used for Task", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "RNG - KBQA : Generation Augmented Iterative Ranking for Knowledge Base Question Answering.", "forward": true, "src_ids": "2022.acl-long.417_3037"}
+{"input": "kbqa is done by using Method| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "kbqa", "output": "rank - and - generate approach", "neg_sample": ["kbqa is done by using Method", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "We present RnG - KBQA , a Rank - and - Generate approach for KBQA , which remedies the coverage issue with a generation model while preserving a strong generalization capability .", "forward": false, "src_ids": "2022.acl-long.417_3038"}
+{"input": "rank - and - generate approach is used for Task| context: test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "rank - and - generate approach", "output": "kbqa", "neg_sample": ["rank - and - generate approach is used for Task", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "We present RnG - KBQA , a Rank - and - Generate approach for KBQA , which remedies the coverage issue with a generation model while preserving a strong generalization capability .", "forward": true, "src_ids": "2022.acl-long.417_3039"}
+{"input": "candidate logical forms is done by using Method| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "candidate logical forms", "output": "contrastive ranker", "neg_sample": ["candidate logical forms is done by using Method", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "Our approach first uses a contrastive ranker to rank a set of candidate logical forms obtained by searching over the knowledge graph .", "forward": false, "src_ids": "2022.acl-long.417_3040"}
+{"input": "contrastive ranker is used for OtherScientificTerm| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "contrastive ranker", "output": "candidate logical forms", "neg_sample": ["contrastive ranker is used for OtherScientificTerm", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "Our approach first uses a contrastive ranker to rank a set of candidate logical forms obtained by searching over the knowledge graph .", "forward": true, "src_ids": "2022.acl-long.417_3041"}
+{"input": "logical form is done by using Method| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "logical form", "output": "tailored generation model", "neg_sample": ["logical form is done by using Method", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "It then introduces a tailored generation model conditioned on the question and the top - ranked candidates to compose the final logical form .", "forward": false, "src_ids": "2022.acl-long.417_3042"}
+{"input": "tailored generation model is used for OtherScientificTerm| context: existing kbqa approaches , despite achieving strong performance on i.i.d . test data , often struggle in generalizing to questions involving unseen kb schema items . prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue .", "entity": "tailored generation model", "output": "logical form", "neg_sample": ["tailored generation model is used for OtherScientificTerm", "existing kbqa approaches , despite achieving strong performance on i.i.d .", "test data , often struggle in generalizing to questions involving unseen kb schema items .", "prior ranking - based approaches have shown some success in generalization , but suffer from the coverage issue ."], "relation": "used for", "id": "2022.acl-long.417", "year": 2022, "rel_sent": "It then introduces a tailored generation model conditioned on the question and the top - ranked candidates to compose the final logical form .", "forward": true, "src_ids": "2022.acl-long.417_3043"}
+{"input": "pretrained language models is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "pretrained language models", "output": "bert2bert", "neg_sample": ["pretrained language models is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "bert2BERT : Towards Reusable Pretrained Language Models.", "forward": false, "src_ids": "2022.acl-long.151_3044"}
+{"input": "gpt_{base } is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "gpt_{base }", "output": "bert2bert", "neg_sample": ["gpt_{base } is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "In particular , bert2BERT saves about 45 % and 47 % computational cost of pre - training BERT_{BASE } and GPT_{BASE } by reusing the models of almost their half sizes .", "forward": false, "src_ids": "2022.acl-long.151_3045"}
+{"input": "bert2bert is used for Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "bert2bert", "output": "pretrained language models", "neg_sample": ["bert2bert is used for Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "bert2BERT : Towards Reusable Pretrained Language Models.", "forward": true, "src_ids": "2022.acl-long.151_3046"}
+{"input": "large model is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "large model", "output": "smaller pre - trained model", "neg_sample": ["large model is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "In this paper , we propose bert2BERT , which can effectively transfer the knowledge of an existing smaller pre - trained model to a large model through parameter initialization and significantly improve the pre - training efficiency of the large model .", "forward": false, "src_ids": "2022.acl-long.151_3047"}
+{"input": "smaller pre - trained model is used for Generic| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "smaller pre - trained model", "output": "large model", "neg_sample": ["smaller pre - trained model is used for Generic", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "In this paper , we propose bert2BERT , which can effectively transfer the knowledge of an existing smaller pre - trained model to a large model through parameter initialization and significantly improve the pre - training efficiency of the large model .", "forward": true, "src_ids": "2022.acl-long.151_3048"}
+{"input": "computer vision is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "computer vision", "output": "function - preserving method", "neg_sample": ["computer vision is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "Specifically , we extend the previous function - preserving method proposed in computer vision on the Transformer - based language model , and further improve it by proposing a novel method , advanced knowledge for large model 's initialization .", "forward": false, "src_ids": "2022.acl-long.151_3049"}
+{"input": "transformer - based language model is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "transformer - based language model", "output": "function - preserving method", "neg_sample": ["transformer - based language model is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "Specifically , we extend the previous function - preserving method proposed in computer vision on the Transformer - based language model , and further improve it by proposing a novel method , advanced knowledge for large model 's initialization .", "forward": false, "src_ids": "2022.acl-long.151_3050"}
+{"input": "function - preserving method is used for Task| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "function - preserving method", "output": "computer vision", "neg_sample": ["function - preserving method is used for Task", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "Specifically , we extend the previous function - preserving method proposed in computer vision on the Transformer - based language model , and further improve it by proposing a novel method , advanced knowledge for large model 's initialization .", "forward": true, "src_ids": "2022.acl-long.151_3051"}
+{"input": "function - preserving method is used for Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "function - preserving method", "output": "transformer - based language model", "neg_sample": ["function - preserving method is used for Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "Specifically , we extend the previous function - preserving method proposed in computer vision on the Transformer - based language model , and further improve it by proposing a novel method , advanced knowledge for large model 's initialization .", "forward": true, "src_ids": "2022.acl-long.151_3052"}
+{"input": "bert2bert is used for Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "bert2bert", "output": "gpt_{base }", "neg_sample": ["bert2bert is used for Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.151", "year": 2022, "rel_sent": "In particular , bert2BERT saves about 45 % and 47 % computational cost of pre - training BERT_{BASE } and GPT_{BASE } by reusing the models of almost their half sizes .", "forward": true, "src_ids": "2022.acl-long.151_3053"}
+{"input": "layer normalization is used for Metric| context: although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with .", "entity": "layer normalization", "output": "cross - entropy", "neg_sample": ["layer normalization is used for Metric", "although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with ."], "relation": "used for", "id": "2022.acl-long.527", "year": 2022, "rel_sent": "Second , we use layer normalization to bring the cross - entropy of both models arbitrarily close to zero .", "forward": true, "src_ids": "2022.acl-long.527_3054"}
+{"input": "transformer is used for OtherScientificTerm| context: although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with . hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer .", "entity": "transformer", "output": "parity", "neg_sample": ["transformer is used for OtherScientificTerm", "although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with .", "hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer ."], "relation": "used for", "id": "2022.acl-long.527", "year": 2022, "rel_sent": "First , we settle an open question by constructing a transformer that recognizes PARITY with perfect accuracy , and similarly for FIRST .", "forward": true, "src_ids": "2022.acl-long.527_3055"}
+{"input": "cross - entropy is done by using Method| context: although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with . hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer .", "entity": "cross - entropy", "output": "layer normalization", "neg_sample": ["cross - entropy is done by using Method", "although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with .", "hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer ."], "relation": "used for", "id": "2022.acl-long.527", "year": 2022, "rel_sent": "Second , we use layer normalization to bring the cross - entropy of both models arbitrarily close to zero .", "forward": false, "src_ids": "2022.acl-long.527_3056"}
+{"input": "disability biases is done by using Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "disability biases", "output": "stereotype content model", "neg_sample": ["disability biases is done by using Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "Applying the Stereotype Content Model to assess disability bias in popular pre - trained NLP models underlying AI - based assistive technologies.", "forward": false, "src_ids": "2022.slpat-1.8_3057"}
+{"input": "pre - trained nlp models is done by using Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "pre - trained nlp models", "output": "stereotype content model", "neg_sample": ["pre - trained nlp models is done by using Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "Applying the Stereotype Content Model to assess disability bias in popular pre - trained NLP models underlying AI - based assistive technologies.", "forward": false, "src_ids": "2022.slpat-1.8_3058"}
+{"input": "stereotype content model is used for OtherScientificTerm| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "entity": "stereotype content model", "output": "disability biases", "neg_sample": ["stereotype content model is used for OtherScientificTerm", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "Applying the Stereotype Content Model to assess disability bias in popular pre - trained NLP models underlying AI - based assistive technologies.", "forward": true, "src_ids": "2022.slpat-1.8_3059"}
+{"input": "stereotype content model is used for Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "stereotype content model", "output": "pre - trained nlp models", "neg_sample": ["stereotype content model is used for Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "Applying the Stereotype Content Model to assess disability bias in popular pre - trained NLP models underlying AI - based assistive technologies.", "forward": true, "src_ids": "2022.slpat-1.8_3060"}
+{"input": "disability is done by using Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "disability", "output": "psychology - based stereotype assessment", "neg_sample": ["disability is done by using Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": false, "src_ids": "2022.slpat-1.8_3061"}
+{"input": "bert is done by using Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "bert", "output": "psychology - based stereotype assessment", "neg_sample": ["bert is done by using Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": false, "src_ids": "2022.slpat-1.8_3062"}
+{"input": "blindness is done by using Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "blindness", "output": "psychology - based stereotype assessment", "neg_sample": ["blindness is done by using Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": false, "src_ids": "2022.slpat-1.8_3063"}
+{"input": "deafness is done by using Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "deafness", "output": "psychology - based stereotype assessment", "neg_sample": ["deafness is done by using Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": false, "src_ids": "2022.slpat-1.8_3064"}
+{"input": "psychology - based stereotype assessment is used for OtherScientificTerm| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "entity": "psychology - based stereotype assessment", "output": "disability", "neg_sample": ["psychology - based stereotype assessment is used for OtherScientificTerm", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": true, "src_ids": "2022.slpat-1.8_3065"}
+{"input": "psychology - based stereotype assessment is used for OtherScientificTerm| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "psychology - based stereotype assessment", "output": "deafness", "neg_sample": ["psychology - based stereotype assessment is used for OtherScientificTerm", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": true, "src_ids": "2022.slpat-1.8_3066"}
+{"input": "psychology - based stereotype assessment is used for Method| context: stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities . if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population . ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases .", "entity": "psychology - based stereotype assessment", "output": "bert", "neg_sample": ["psychology - based stereotype assessment is used for Method", "stereotypes are a positive or negative , generalized , and often widely shared belief about the attributes of certain groups of people , such as people with sensory disabilities .", "if stereotypes manifest in assistive technologies used by deaf or blind people , they can harm the user in a number of ways , especially considering the vulnerable nature of the target population .", "ai models underlying assistive technologies have been shown to contain biased stereotypes , including racial , gender , and disability biases ."], "relation": "used for", "id": "2022.slpat-1.8", "year": 2022, "rel_sent": "We build on this work to present a psychology - based stereotype assessment of the representation of disability , deafness , and blindness in BERT using the Stereotype Content Model .", "forward": true, "src_ids": "2022.slpat-1.8_3067"}
+{"input": "linguistic relations is used for Task| context: domain adaptation methods often exploit domain - transferable input features , a.k.a . pivots . the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g. from restaurants to laptops ) .", "entity": "linguistic relations", "output": "cross - domain aspect term extraction", "neg_sample": ["linguistic relations is used for Task", "domain adaptation methods often exploit domain - transferable input features , a.k.a .", "pivots .", "the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g.", "from restaurants to laptops ) ."], "relation": "used for", "id": "2022.wassa-1.11", "year": 2022, "rel_sent": "In this paper , we investigate and establish empirically a prior conjecture , which suggests that the linguistic relations connecting opinion terms to their aspects transfer well across domains and therefore can be leveraged for cross - domain aspect term extraction .", "forward": true, "src_ids": "2022.wassa-1.11_3068"}
+{"input": "relation patterns is done by using Method| context: domain adaptation methods often exploit domain - transferable input features , a.k.a . pivots . the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g. from restaurants to laptops ) .", "entity": "relation patterns", "output": "linguistic dependency formalisms", "neg_sample": ["relation patterns is done by using Method", "domain adaptation methods often exploit domain - transferable input features , a.k.a .", "pivots .", "the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g.", "from restaurants to laptops ) ."], "relation": "used for", "id": "2022.wassa-1.11", "year": 2022, "rel_sent": "We present several analyses supporting this conjecture , via experiments with four linguistic dependency formalisms to represent relation patterns .", "forward": false, "src_ids": "2022.wassa-1.11_3069"}
+{"input": "linguistic dependency formalisms is used for OtherScientificTerm| context: domain adaptation methods often exploit domain - transferable input features , a.k.a . pivots . the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g. from restaurants to laptops ) .", "entity": "linguistic dependency formalisms", "output": "relation patterns", "neg_sample": ["linguistic dependency formalisms is used for OtherScientificTerm", "domain adaptation methods often exploit domain - transferable input features , a.k.a .", "pivots .", "the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g.", "from restaurants to laptops ) ."], "relation": "used for", "id": "2022.wassa-1.11", "year": 2022, "rel_sent": "We present several analyses supporting this conjecture , via experiments with four linguistic dependency formalisms to represent relation patterns .", "forward": true, "src_ids": "2022.wassa-1.11_3070"}
+{"input": "cross - domain aspect term extraction is done by using OtherScientificTerm| context: domain adaptation methods often exploit domain - transferable input features , a.k.a . pivots . the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g. from restaurants to laptops ) .", "entity": "cross - domain aspect term extraction", "output": "linguistic relations", "neg_sample": ["cross - domain aspect term extraction is done by using OtherScientificTerm", "domain adaptation methods often exploit domain - transferable input features , a.k.a .", "pivots .", "the task of aspect and opinion term extraction presents a special challenge for domain transfer : while opinion terms largely transfer across domains , aspects change drastically from one domain to another ( e.g.", "from restaurants to laptops ) ."], "relation": "used for", "id": "2022.wassa-1.11", "year": 2022, "rel_sent": "In this paper , we investigate and establish empirically a prior conjecture , which suggests that the linguistic relations connecting opinion terms to their aspects transfer well across domains and therefore can be leveraged for cross - domain aspect term extraction .", "forward": false, "src_ids": "2022.wassa-1.11_3071"}
+{"input": "pre - trained models is used for Task| context: while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models .", "entity": "pre - trained models", "output": "symmetric classification tasks", "neg_sample": ["pre - trained models is used for Task", "while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models ."], "relation": "used for", "id": "2022.findings-acl.148", "year": 2022, "rel_sent": "Striking a Balance : Alleviating Inconsistency in Pre - trained Models for Symmetric Classification Tasks.", "forward": true, "src_ids": "2022.findings-acl.148_3072"}
+{"input": "symmetric classification is done by using Method| context: while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models .", "entity": "symmetric classification", "output": "consistency loss function", "neg_sample": ["symmetric classification is done by using Method", "while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models ."], "relation": "used for", "id": "2022.findings-acl.148", "year": 2022, "rel_sent": "Specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores . We highlight this model shortcoming and apply a consistency loss function to alleviate inconsistency in symmetric classification .", "forward": false, "src_ids": "2022.findings-acl.148_3073"}
+{"input": "consistency loss function is used for Task| context: while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models .", "entity": "consistency loss function", "output": "symmetric classification", "neg_sample": ["consistency loss function is used for Task", "while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models ."], "relation": "used for", "id": "2022.findings-acl.148", "year": 2022, "rel_sent": "Specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores . We highlight this model shortcoming and apply a consistency loss function to alleviate inconsistency in symmetric classification .", "forward": true, "src_ids": "2022.findings-acl.148_3074"}
+{"input": "multi - label text classification ( mltc ) is done by using Method| context: multi - label text classification ( mltc ) is a fundamental and challenging task in natural language processing . previous studies mainly focus on learning text representation and modeling label correlation but neglect the rich knowledge from the existing similar instances when predicting labels of a specific text .", "entity": "multi - label text classification ( mltc )", "output": "contrastive learning - enhanced nearest neighbor mechanism", "neg_sample": ["multi - label text classification ( mltc ) is done by using Method", "multi - label text classification ( mltc ) is a fundamental and challenging task in natural language processing .", "previous studies mainly focus on learning text representation and modeling label correlation but neglect the rich knowledge from the existing similar instances when predicting labels of a specific text ."], "relation": "used for", "id": "2022.acl-short.75", "year": 2022, "rel_sent": "Contrastive Learning - Enhanced Nearest Neighbor Mechanism for Multi - Label Text Classification.", "forward": false, "src_ids": "2022.acl-short.75_3075"}
+{"input": "contrastive learning - enhanced nearest neighbor mechanism is used for Task| context: previous studies mainly focus on learning text representation and modeling label correlation but neglect the rich knowledge from the existing similar instances when predicting labels of a specific text .", "entity": "contrastive learning - enhanced nearest neighbor mechanism", "output": "multi - label text classification ( mltc )", "neg_sample": ["contrastive learning - enhanced nearest neighbor mechanism is used for Task", "previous studies mainly focus on learning text representation and modeling label correlation but neglect the rich knowledge from the existing similar instances when predicting labels of a specific text ."], "relation": "used for", "id": "2022.acl-short.75", "year": 2022, "rel_sent": "Contrastive Learning - Enhanced Nearest Neighbor Mechanism for Multi - Label Text Classification.", "forward": true, "src_ids": "2022.acl-short.75_3076"}
+{"input": "idioms is done by using Method| context: idioms are unlike most phrases in two important ways . first , words in an idiom have non - canonical meanings . second , the non - canonical meanings of words in an idiom are contingent on the presence of other words in the idiom . linguistic theories differ on whether these properties depend on one another , as well as whether special theoretical machinery is needed to accommodate idioms .", "entity": "idioms", "output": "special machinery", "neg_sample": ["idioms is done by using Method", "idioms are unlike most phrases in two important ways .", "first , words in an idiom have non - canonical meanings .", "second , the non - canonical meanings of words in an idiom are contingent on the presence of other words in the idiom .", "linguistic theories differ on whether these properties depend on one another , as well as whether special theoretical machinery is needed to accommodate idioms ."], "relation": "used for", "id": "2022.acl-long.278", "year": 2022, "rel_sent": "Our results suggest that introducing special machinery to handle idioms may not be warranted .", "forward": false, "src_ids": "2022.acl-long.278_3077"}
+{"input": "special machinery is used for OtherScientificTerm| context: first , words in an idiom have non - canonical meanings . second , the non - canonical meanings of words in an idiom are contingent on the presence of other words in the idiom .", "entity": "special machinery", "output": "idioms", "neg_sample": ["special machinery is used for OtherScientificTerm", "first , words in an idiom have non - canonical meanings .", "second , the non - canonical meanings of words in an idiom are contingent on the presence of other words in the idiom ."], "relation": "used for", "id": "2022.acl-long.278", "year": 2022, "rel_sent": "Our results suggest that introducing special machinery to handle idioms may not be warranted .", "forward": true, "src_ids": "2022.acl-long.278_3078"}
+{"input": "social media text classification is done by using Method| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "social media text classification", "output": "bi - directional recurrent neural ordinary differential equations", "neg_sample": ["social media text classification is done by using Method", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "Bi - Directional Recurrent Neural Ordinary Differential Equations for Social Media Text Classification.", "forward": false, "src_ids": "2022.wit-1.3_3079"}
+{"input": "bi - directional recurrent neural ordinary differential equations is used for Task| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "bi - directional recurrent neural ordinary differential equations", "output": "social media text classification", "neg_sample": ["bi - directional recurrent neural ordinary differential equations is used for Task", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "Bi - Directional Recurrent Neural Ordinary Differential Equations for Social Media Text Classification.", "forward": true, "src_ids": "2022.wit-1.3_3080"}
+{"input": "social media post classification is done by using Method| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "social media post classification", "output": "recurrent neural ordinary differential equations ( rnode )", "neg_sample": ["social media post classification is done by using Method", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "In this work , we propose to use recurrent neural ordinary differential equations ( RNODE ) for social media post classification which consider the time of posting and allow the computation of hidden representation to evolve in a time - sensitive continuous manner .", "forward": false, "src_ids": "2022.wit-1.3_3081"}
+{"input": "recurrent neural ordinary differential equations ( rnode ) is used for Task| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "recurrent neural ordinary differential equations ( rnode )", "output": "social media post classification", "neg_sample": ["recurrent neural ordinary differential equations ( rnode ) is used for Task", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "In this work , we propose to use recurrent neural ordinary differential equations ( RNODE ) for social media post classification which consider the time of posting and allow the computation of hidden representation to evolve in a time - sensitive continuous manner .", "forward": true, "src_ids": "2022.wit-1.3_3082"}
+{"input": "stance classification of rumours is done by using Method| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "stance classification of rumours", "output": "recurrent neural ordinary differential equations", "neg_sample": ["stance classification of rumours is done by using Method", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "Our experiments demonstrate that RNODE and Bi - RNODE are effective for the problem of stance classification of rumours in social media .", "forward": false, "src_ids": "2022.wit-1.3_3083"}
+{"input": "bi - rnode is used for Task| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "bi - rnode", "output": "stance classification of rumours", "neg_sample": ["bi - rnode is used for Task", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "Our experiments demonstrate that RNODE and Bi - RNODE are effective for the problem of stance classification of rumours in social media .", "forward": true, "src_ids": "2022.wit-1.3_3084"}
+{"input": "recurrent neural ordinary differential equations is used for Task| context: classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts . sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature . rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting .", "entity": "recurrent neural ordinary differential equations", "output": "stance classification of rumours", "neg_sample": ["recurrent neural ordinary differential equations is used for Task", "classification of posts in social media such as twitter is difficult due to the noisy and short nature of texts .", "sequence classification models based on recurrent neural networks ( rnn ) are popular for classifying posts that are sequential in nature .", "rnns assume the hidden representation dynamics to evolve in a discrete manner and do not consider the exact time of the posting ."], "relation": "used for", "id": "2022.wit-1.3", "year": 2022, "rel_sent": "Our experiments demonstrate that RNODE and Bi - RNODE are effective for the problem of stance classification of rumours in social media .", "forward": true, "src_ids": "2022.wit-1.3_3085"}
+{"input": "multimodal sentiment analysis is done by using Method| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity . through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly .", "entity": "multimodal sentiment analysis", "output": "sentiment word aware multimodal refinement", "neg_sample": ["multimodal sentiment analysis is done by using Method", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "Sentiment Word Aware Multimodal Refinement for Multimodal Sentiment Analysis with ASR Errors.", "forward": false, "src_ids": "2022.findings-acl.109_3086"}
+{"input": "sentiment word aware multimodal refinement is used for Task| context: however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "entity": "sentiment word aware multimodal refinement", "output": "multimodal sentiment analysis", "neg_sample": ["sentiment word aware multimodal refinement is used for Task", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "Sentiment Word Aware Multimodal Refinement for Multimodal Sentiment Analysis with ASR Errors.", "forward": true, "src_ids": "2022.findings-acl.109_3087"}
+{"input": "sentiment word position detection module is used for OtherScientificTerm| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "entity": "sentiment word position detection module", "output": "sentiment words", "neg_sample": ["sentiment word position detection module is used for OtherScientificTerm", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "Specifically , we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings .", "forward": true, "src_ids": "2022.findings-acl.109_3088"}
+{"input": "sentiment words is done by using Method| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity . through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly .", "entity": "sentiment words", "output": "sentiment word position detection module", "neg_sample": ["sentiment words is done by using Method", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "Specifically , we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings .", "forward": false, "src_ids": "2022.findings-acl.109_3089"}
+{"input": "sentiment word embeddings is done by using Method| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity . through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly .", "entity": "sentiment word embeddings", "output": "multimodal sentiment word refinement module", "neg_sample": ["sentiment word embeddings is done by using Method", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "Specifically , we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings .", "forward": false, "src_ids": "2022.findings-acl.109_3090"}
+{"input": "multimodal sentiment word refinement module is used for OtherScientificTerm| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity . through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly .", "entity": "multimodal sentiment word refinement module", "output": "sentiment word embeddings", "neg_sample": ["multimodal sentiment word refinement module is used for OtherScientificTerm", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "Specifically , we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings .", "forward": true, "src_ids": "2022.findings-acl.109_3091"}
+{"input": "sentiment labels is done by using Method| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity . through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly .", "entity": "sentiment labels", "output": "multimodal feature fusion module", "neg_sample": ["sentiment labels is done by using Method", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "The refined embeddings are taken as the textual inputs of the multimodal feature fusion module to predict the sentiment labels .", "forward": false, "src_ids": "2022.findings-acl.109_3092"}
+{"input": "multimodal feature fusion module is used for OtherScientificTerm| context: multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed . however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world . we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity . through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly .", "entity": "multimodal feature fusion module", "output": "sentiment labels", "neg_sample": ["multimodal feature fusion module is used for OtherScientificTerm", "multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed .", "however , the performance of the state - of - the - art models decreases sharply when they are deployed in the real world .", "we find that the main reason is that real - world applications can only access the text outputs by the automatic speech recognition ( asr ) models , which may be with errors because of the limitation of model capacity .", "through further analysis of the asr outputs , we find that in some cases the sentiment words , the key sentiment elements in the textual modality , are recognized as other words , which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly ."], "relation": "used for", "id": "2022.findings-acl.109", "year": 2022, "rel_sent": "The refined embeddings are taken as the textual inputs of the multimodal feature fusion module to predict the sentiment labels .", "forward": true, "src_ids": "2022.findings-acl.109_3093"}
+{"input": "transfer learning ( tl ) model is used for Task| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "transfer learning ( tl ) model", "output": "detecting signs of depression", "neg_sample": ["transfer learning ( tl ) model is used for Task", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "To address the issue of detecting the sign of depression content on social media , in this paper , we - team MUCS , describe an Ensemble of Machine Learning ( ML ) models and a Transfer Learning ( TL ) model submitted to ' Detecting Signs of Depression from Social Media Text - LT - EDI@ACL 2022 ' ( DepSign - LT - EDI@ACL-2022 ) shared task at Association for Computational Linguistics ( ACL ) 2022 .", "forward": true, "src_ids": "2022.ltedi-1.47_3094"}
+{"input": "detecting signs of depression is done by using Method| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "detecting signs of depression", "output": "transfer learning ( tl ) model", "neg_sample": ["detecting signs of depression is done by using Method", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "To address the issue of detecting the sign of depression content on social media , in this paper , we - team MUCS , describe an Ensemble of Machine Learning ( ML ) models and a Transfer Learning ( TL ) model submitted to ' Detecting Signs of Depression from Social Media Text - LT - EDI@ACL 2022 ' ( DepSign - LT - EDI@ACL-2022 ) shared task at Association for Computational Linguistics ( ACL ) 2022 .", "forward": false, "src_ids": "2022.ltedi-1.47_3095"}
+{"input": "ensemble model is done by using OtherScientificTerm| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "ensemble model", "output": "frequency and text based features", "neg_sample": ["ensemble model is done by using OtherScientificTerm", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "Both frequency and text based features are used to train an Ensemble model and Bidirectional Encoder Representations from Transformers ( BERT ) fine - tuned with raw text is used to train the TL model .", "forward": false, "src_ids": "2022.ltedi-1.47_3096"}
+{"input": "frequency and text based features is used for Method| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "frequency and text based features", "output": "ensemble model", "neg_sample": ["frequency and text based features is used for Method", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "Both frequency and text based features are used to train an Ensemble model and Bidirectional Encoder Representations from Transformers ( BERT ) fine - tuned with raw text is used to train the TL model .", "forward": true, "src_ids": "2022.ltedi-1.47_3097"}
+{"input": "tl model is done by using Method| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "tl model", "output": "bidirectional encoder representations", "neg_sample": ["tl model is done by using Method", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "Both frequency and text based features are used to train an Ensemble model and Bidirectional Encoder Representations from Transformers ( BERT ) fine - tuned with raw text is used to train the TL model .", "forward": false, "src_ids": "2022.ltedi-1.47_3098"}
+{"input": "tl model is done by using Material| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "tl model", "output": "raw text", "neg_sample": ["tl model is done by using Material", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "Both frequency and text based features are used to train an Ensemble model and Bidirectional Encoder Representations from Transformers ( BERT ) fine - tuned with raw text is used to train the TL model .", "forward": false, "src_ids": "2022.ltedi-1.47_3099"}
+{"input": "bidirectional encoder representations is used for Method| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "bidirectional encoder representations", "output": "tl model", "neg_sample": ["bidirectional encoder representations is used for Method", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "Both frequency and text based features are used to train an Ensemble model and Bidirectional Encoder Representations from Transformers ( BERT ) fine - tuned with raw text is used to train the TL model .", "forward": true, "src_ids": "2022.ltedi-1.47_3100"}
+{"input": "raw text is used for Method| context: social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media . users having the sign of depression may post negative or disturbing content seeking the attention of other users . hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required . however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same .", "entity": "raw text", "output": "tl model", "neg_sample": ["raw text is used for Method", "social media has seen enormous growth in its users recently and knowingly or unknowingly the behavior of a person will be reflected in the comments she / he posts on social media .", "users having the sign of depression may post negative or disturbing content seeking the attention of other users .", "hence , social media data can be analysed to check whether the users ' have the sign of depression and help them to get through the situation if required .", "however , as analyzing the increasing amount of social media data manually in laborious and error - prone , automated tools have to be developed for the same ."], "relation": "used for", "id": "2022.ltedi-1.47", "year": 2022, "rel_sent": "Both frequency and text based features are used to train an Ensemble model and Bidirectional Encoder Representations from Transformers ( BERT ) fine - tuned with raw text is used to train the TL model .", "forward": true, "src_ids": "2022.ltedi-1.47_3101"}
+{"input": "dialogue systems is done by using Method| context: users interacting with voice assistants today need to phrase their requests in a very specific manner to elicit an appropriate response . this limits the user experience , and is partly due to the lack of reasoning capabilities of dialogue platforms and the hand - crafted rules that require extensive labor . one possible solution to improve user experience and relieve the manual efforts of designers is to build an end - to - end dialogue system that can do reasoning itself while perceiving user 's utterances .", "entity": "dialogue systems", "output": "large - scale interpretable knowledge graph reasoning", "neg_sample": ["dialogue systems is done by using Method", "users interacting with voice assistants today need to phrase their requests in a very specific manner to elicit an appropriate response .", "this limits the user experience , and is partly due to the lack of reasoning capabilities of dialogue platforms and the hand - crafted rules that require extensive labor .", "one possible solution to improve user experience and relieve the manual efforts of designers is to build an end - to - end dialogue system that can do reasoning itself while perceiving user 's utterances ."], "relation": "used for", "id": "2022.findings-acl.33", "year": 2022, "rel_sent": "Towards Large - Scale Interpretable Knowledge Graph Reasoning for Dialogue Systems.", "forward": false, "src_ids": "2022.findings-acl.33_3102"}
+{"input": "large - scale interpretable knowledge graph reasoning is used for Task| context: users interacting with voice assistants today need to phrase their requests in a very specific manner to elicit an appropriate response . this limits the user experience , and is partly due to the lack of reasoning capabilities of dialogue platforms and the hand - crafted rules that require extensive labor . one possible solution to improve user experience and relieve the manual efforts of designers is to build an end - to - end dialogue system that can do reasoning itself while perceiving user 's utterances .", "entity": "large - scale interpretable knowledge graph reasoning", "output": "dialogue systems", "neg_sample": ["large - scale interpretable knowledge graph reasoning is used for Task", "users interacting with voice assistants today need to phrase their requests in a very specific manner to elicit an appropriate response .", "this limits the user experience , and is partly due to the lack of reasoning capabilities of dialogue platforms and the hand - crafted rules that require extensive labor .", "one possible solution to improve user experience and relieve the manual efforts of designers is to build an end - to - end dialogue system that can do reasoning itself while perceiving user 's utterances ."], "relation": "used for", "id": "2022.findings-acl.33", "year": 2022, "rel_sent": "Towards Large - Scale Interpretable Knowledge Graph Reasoning for Dialogue Systems.", "forward": true, "src_ids": "2022.findings-acl.33_3103"}
+{"input": "multimodal dialog systems is done by using Method| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved . nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses .", "entity": "multimodal dialog systems", "output": "unitranser", "neg_sample": ["multimodal dialog systems is done by using Method", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "To address these issues , we propose UniTranSeR , a Unified Transformer Semantic Representation framework with feature alignment and intention reasoning for multimodal dialog systems .", "forward": false, "src_ids": "2022.acl-long.9_3104"}
+{"input": "multimodal task - oriented dialog system is done by using Method| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved . nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses .", "entity": "multimodal task - oriented dialog system", "output": "unified transformer semantic representation framework", "neg_sample": ["multimodal task - oriented dialog system is done by using Method", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "UniTranSeR : A Unified Transformer Semantic Representation Framework for Multimodal Task - Oriented Dialog System.", "forward": false, "src_ids": "2022.acl-long.9_3105"}
+{"input": "unified transformer semantic representation framework is used for Task| context: nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses .", "entity": "unified transformer semantic representation framework", "output": "multimodal task - oriented dialog system", "neg_sample": ["unified transformer semantic representation framework is used for Task", "nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "UniTranSeR : A Unified Transformer Semantic Representation Framework for Multimodal Task - Oriented Dialog System.", "forward": true, "src_ids": "2022.acl-long.9_3106"}
+{"input": "unified transformer semantic space is used for OtherScientificTerm| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "entity": "unified transformer semantic space", "output": "inter - modal interactions", "neg_sample": ["unified transformer semantic space is used for OtherScientificTerm", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "Specifically , we first embed the multimodal features into a unified Transformer semantic space to prompt inter - modal interactions , and then devise a feature alignment and intention reasoning ( FAIR ) layer to perform cross - modal entity alignment and fine - grained key - value reasoning , so as to effectively identify user 's intention for generating more accurate responses .", "forward": true, "src_ids": "2022.acl-long.9_3107"}
+{"input": "multimodal features is used for OtherScientificTerm| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "entity": "multimodal features", "output": "inter - modal interactions", "neg_sample": ["multimodal features is used for OtherScientificTerm", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "Specifically , we first embed the multimodal features into a unified Transformer semantic space to prompt inter - modal interactions , and then devise a feature alignment and intention reasoning ( FAIR ) layer to perform cross - modal entity alignment and fine - grained key - value reasoning , so as to effectively identify user 's intention for generating more accurate responses .", "forward": true, "src_ids": "2022.acl-long.9_3108"}
+{"input": "unified transformer semantic representation framework is used for Task| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved . nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses .", "entity": "unified transformer semantic representation framework", "output": "multimodal dialog systems", "neg_sample": ["unified transformer semantic representation framework is used for Task", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "To address these issues , we propose UniTranSeR , a Unified Transformer Semantic Representation framework with feature alignment and intention reasoning for multimodal dialog systems .", "forward": true, "src_ids": "2022.acl-long.9_3109"}
+{"input": "unitranser is used for Task| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved . nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses .", "entity": "unitranser", "output": "multimodal dialog systems", "neg_sample": ["unitranser is used for Task", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "To address these issues , we propose UniTranSeR , a Unified Transformer Semantic Representation framework with feature alignment and intention reasoning for multimodal dialog systems .", "forward": true, "src_ids": "2022.acl-long.9_3110"}
+{"input": "inter - modal interactions is done by using OtherScientificTerm| context: as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved . nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses .", "entity": "inter - modal interactions", "output": "multimodal features", "neg_sample": ["inter - modal interactions is done by using OtherScientificTerm", "as a more natural and intelligent interaction manner , multimodal task - oriented dialog system recently has received great attention and many remarkable progresses have been achieved .", "nevertheless , almost all existing studies follow the pipeline tofirst learn intra - modal features separately and then conduct simple feature concatenation or attention - based feature fusion to generate responses , which hampers them from learning inter - modal interactions and conducting cross - modal feature alignment for generating more intention - aware responses ."], "relation": "used for", "id": "2022.acl-long.9", "year": 2022, "rel_sent": "Specifically , we first embed the multimodal features into a unified Transformer semantic space to prompt inter - modal interactions , and then devise a feature alignment and intention reasoning ( FAIR ) layer to perform cross - modal entity alignment and fine - grained key - value reasoning , so as to effectively identify user 's intention for generating more accurate responses .", "forward": false, "src_ids": "2022.acl-long.9_3111"}
+{"input": "initialization is done by using OtherScientificTerm| context: prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks . however , prompt tuning is yet to be fully explored . in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning .", "entity": "initialization", "output": "soft prompts", "neg_sample": ["initialization is done by using OtherScientificTerm", "prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks .", "however , prompt tuning is yet to be fully explored .", "in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning ."], "relation": "used for", "id": "2022.acl-long.576", "year": 2022, "rel_sent": "Therefore , in this work , we propose to pre - train prompts by adding soft prompts into the pre - training stage to obtain a better initialization .", "forward": false, "src_ids": "2022.acl-long.576_3112"}
+{"input": "unified task is done by using OtherScientificTerm| context: prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks . however , prompt tuning is yet to be fully explored . in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning .", "entity": "unified task", "output": "soft prompts", "neg_sample": ["unified task is done by using OtherScientificTerm", "prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks .", "however , prompt tuning is yet to be fully explored .", "in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning ."], "relation": "used for", "id": "2022.acl-long.576", "year": 2022, "rel_sent": "To ensure the generalization of PPT , we formulate similar classification tasks into a unified task form and pre - train soft prompts for this unified task .", "forward": false, "src_ids": "2022.acl-long.576_3113"}
+{"input": "soft prompts is used for OtherScientificTerm| context: prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks . among these methods , prompt tuning , which freezes plms and only tunes soft prompts , provides an efficient and effective solution for adapting large - scale plms to downstream tasks . however , prompt tuning is yet to be fully explored . in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning . we attribute this low performance to the manner of initializing soft prompts .", "entity": "soft prompts", "output": "initialization", "neg_sample": ["soft prompts is used for OtherScientificTerm", "prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks .", "among these methods , prompt tuning , which freezes plms and only tunes soft prompts , provides an efficient and effective solution for adapting large - scale plms to downstream tasks .", "however , prompt tuning is yet to be fully explored .", "in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning .", "we attribute this low performance to the manner of initializing soft prompts ."], "relation": "used for", "id": "2022.acl-long.576", "year": 2022, "rel_sent": "Therefore , in this work , we propose to pre - train prompts by adding soft prompts into the pre - training stage to obtain a better initialization .", "forward": true, "src_ids": "2022.acl-long.576_3114"}
+{"input": "soft prompts is used for Generic| context: prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks . among these methods , prompt tuning , which freezes plms and only tunes soft prompts , provides an efficient and effective solution for adapting large - scale plms to downstream tasks . however , prompt tuning is yet to be fully explored . in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning . we attribute this low performance to the manner of initializing soft prompts .", "entity": "soft prompts", "output": "unified task", "neg_sample": ["soft prompts is used for Generic", "prompts for pre - trained language models ( plms ) have shown remarkable performance by bridging the gap between pre - training tasks and various downstream tasks .", "among these methods , prompt tuning , which freezes plms and only tunes soft prompts , provides an efficient and effective solution for adapting large - scale plms to downstream tasks .", "however , prompt tuning is yet to be fully explored .", "in our pilot experiments , we find that prompt tuning performs comparably with conventional full - model tuning when downstream data are sufficient , whereas it is much worse under few - shot learning settings , which may hinder the application of prompt tuning .", "we attribute this low performance to the manner of initializing soft prompts ."], "relation": "used for", "id": "2022.acl-long.576", "year": 2022, "rel_sent": "To ensure the generalization of PPT , we formulate similar classification tasks into a unified task form and pre - train soft prompts for this unified task .", "forward": true, "src_ids": "2022.acl-long.576_3115"}
+{"input": "language model rationale extraction is done by using Method| context: an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction . ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e. , task model 's ) task performance . although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata .", "entity": "language model rationale extraction", "output": "unified learning framework", "neg_sample": ["language model rationale extraction is done by using Method", "an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction .", "ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e.", ", task model 's ) task performance .", "although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata ."], "relation": "used for", "id": "2022.bigscience-1.5", "year": 2022, "rel_sent": "UNIREX : A Unified Learning Framework for Language Model Rationale Extraction.", "forward": false, "src_ids": "2022.bigscience-1.5_3116"}
+{"input": "unified learning framework is used for Task| context: an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction . ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e. , task model 's ) task performance . although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata .", "entity": "unified learning framework", "output": "language model rationale extraction", "neg_sample": ["unified learning framework is used for Task", "an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction .", "ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e.", ", task model 's ) task performance .", "although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata ."], "relation": "used for", "id": "2022.bigscience-1.5", "year": 2022, "rel_sent": "UNIREX : A Unified Learning Framework for Language Model Rationale Extraction.", "forward": true, "src_ids": "2022.bigscience-1.5_3117"}
+{"input": "rationale extractor optimization is done by using Method| context: an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction . ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e. , task model 's ) task performance . although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata .", "entity": "rationale extractor optimization", "output": "flexible learning framework", "neg_sample": ["rationale extractor optimization is done by using Method", "an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction .", "ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e.", ", task model 's ) task performance .", "although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata ."], "relation": "used for", "id": "2022.bigscience-1.5", "year": 2022, "rel_sent": "In light of this , we propose UNIREX , a flexible learning framework which generalizes rationale extractor optimization as follows : ( 1 ) specify architecture for a learned rationale extractor ; ( 2 ) select explainability objectives ( i.e. , faithfulness and plausibility criteria ) ; and ( 3 ) jointly the train task model and rationale extractor on the task using selected objectives .", "forward": false, "src_ids": "2022.bigscience-1.5_3118"}
+{"input": "flexible learning framework is used for Task| context: an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction . ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e. , task model 's ) task performance . although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata .", "entity": "flexible learning framework", "output": "rationale extractor optimization", "neg_sample": ["flexible learning framework is used for Task", "an extractive rationale explains a language model 's ( lm 's ) prediction on a given task instance by highlighting the text inputs that most influenced the prediction .", "ideally , rationale extraction should be faithful ( reflective of lm 's actual behavior ) and plausible ( convincing to humans ) , without compromising the lm 's ( i.e.", ", task model 's ) task performance .", "although attribution algorithms and select - predict pipelines are commonly used in rationale extraction , they both rely on certain heuristics that hinder them from satisfying all three desiderata ."], "relation": "used for", "id": "2022.bigscience-1.5", "year": 2022, "rel_sent": "In light of this , we propose UNIREX , a flexible learning framework which generalizes rationale extractor optimization as follows : ( 1 ) specify architecture for a learned rationale extractor ; ( 2 ) select explainability objectives ( i.e. , faithfulness and plausibility criteria ) ; and ( 3 ) jointly the train task model and rationale extractor on the task using selected objectives .", "forward": true, "src_ids": "2022.bigscience-1.5_3119"}
+{"input": "umuteam is used for Task| context: depression is a mental condition related to sadness and the lack of interest in common daily tasks . this task is somehow related to other relevant natural language processing tasks such as emotion analysis .", "entity": "umuteam", "output": "identification of signs of depression", "neg_sample": ["umuteam is used for Task", "depression is a mental condition related to sadness and the lack of interest in common daily tasks .", "this task is somehow related to other relevant natural language processing tasks such as emotion analysis ."], "relation": "used for", "id": "2022.ltedi-1.17", "year": 2022, "rel_sent": "In this working - notes , we describe the proposal of the UMUTeam in the LT - EDI shared task ( ACL 2022 ) concerning the identification of signs of depression in social network posts .", "forward": true, "src_ids": "2022.ltedi-1.17_3120"}
+{"input": "identification of signs of depression is done by using Method| context: depression is a mental condition related to sadness and the lack of interest in common daily tasks . this task is somehow related to other relevant natural language processing tasks such as emotion analysis .", "entity": "identification of signs of depression", "output": "umuteam", "neg_sample": ["identification of signs of depression is done by using Method", "depression is a mental condition related to sadness and the lack of interest in common daily tasks .", "this task is somehow related to other relevant natural language processing tasks such as emotion analysis ."], "relation": "used for", "id": "2022.ltedi-1.17", "year": 2022, "rel_sent": "In this working - notes , we describe the proposal of the UMUTeam in the LT - EDI shared task ( ACL 2022 ) concerning the identification of signs of depression in social network posts .", "forward": false, "src_ids": "2022.ltedi-1.17_3121"}
+{"input": "pre - training is done by using Method| context: availability of the user reviews in vernacular languages is helpful for the users to get information regarding the products . since most of the e - commerce websites allow the reviews in english language only , it is important to provide the translated versions of the reviews to the non - english speaking users . translation of the user reviews from english to vernacular languages is a challenging task , predominantly due to the lack of sufficient in - domain datasets .", "entity": "pre - training", "output": "decoder", "neg_sample": ["pre - training is done by using Method", "availability of the user reviews in vernacular languages is helpful for the users to get information regarding the products .", "since most of the e - commerce websites allow the reviews in english language only , it is important to provide the translated versions of the reviews to the non - english speaking users .", "translation of the user reviews from english to vernacular languages is a challenging task , predominantly due to the lack of sufficient in - domain datasets ."], "relation": "used for", "id": "2022.eamt-1.27", "year": 2022, "rel_sent": "The decoder for the pre - training is trained over the cross - lingual target samples where the phrases are replaced with their translated counterparts .", "forward": false, "src_ids": "2022.eamt-1.27_3122"}
+{"input": "decoder is used for Method| context: availability of the user reviews in vernacular languages is helpful for the users to get information regarding the products . since most of the e - commerce websites allow the reviews in english language only , it is important to provide the translated versions of the reviews to the non - english speaking users . translation of the user reviews from english to vernacular languages is a challenging task , predominantly due to the lack of sufficient in - domain datasets .", "entity": "decoder", "output": "pre - training", "neg_sample": ["decoder is used for Method", "availability of the user reviews in vernacular languages is helpful for the users to get information regarding the products .", "since most of the e - commerce websites allow the reviews in english language only , it is important to provide the translated versions of the reviews to the non - english speaking users .", "translation of the user reviews from english to vernacular languages is a challenging task , predominantly due to the lack of sufficient in - domain datasets ."], "relation": "used for", "id": "2022.eamt-1.27", "year": 2022, "rel_sent": "The decoder for the pre - training is trained over the cross - lingual target samples where the phrases are replaced with their translated counterparts .", "forward": true, "src_ids": "2022.eamt-1.27_3123"}
+{"input": "transfer is done by using Method| context: multilingual language models were shown to allow for nontrivial transfer across scripts and languages .", "entity": "transfer", "output": "internal representations", "neg_sample": ["transfer is done by using Method", "multilingual language models were shown to allow for nontrivial transfer across scripts and languages ."], "relation": "used for", "id": "2022.repl4nlp-1.8", "year": 2022, "rel_sent": "In this work , we study the structure of the internal representations that enable this transfer .", "forward": false, "src_ids": "2022.repl4nlp-1.8_3124"}
+{"input": "loss function is used for Task| context: our system is an end - to - end speech translation model that leverages pretrained models and cross modality transfer learning .", "entity": "loss function", "output": "translation task", "neg_sample": ["loss function is used for Task", "our system is an end - to - end speech translation model that leverages pretrained models and cross modality transfer learning ."], "relation": "used for", "id": "2022.iwslt-1.12", "year": 2022, "rel_sent": "First , we implemented a new loss function that reduces knowledge gap between audio and text modalities in translation task effectively .", "forward": true, "src_ids": "2022.iwslt-1.12_3125"}
+{"input": "translation task is done by using OtherScientificTerm| context: this paper describes amazon alexa ai 's submission to the iwslt 2022 offline speech translation task . our system is an end - to - end speech translation model that leverages pretrained models and cross modality transfer learning .", "entity": "translation task", "output": "loss function", "neg_sample": ["translation task is done by using OtherScientificTerm", "this paper describes amazon alexa ai 's submission to the iwslt 2022 offline speech translation task .", "our system is an end - to - end speech translation model that leverages pretrained models and cross modality transfer learning ."], "relation": "used for", "id": "2022.iwslt-1.12", "year": 2022, "rel_sent": "First , we implemented a new loss function that reduces knowledge gap between audio and text modalities in translation task effectively .", "forward": false, "src_ids": "2022.iwslt-1.12_3126"}
+{"input": "unsupervised setting is used for Method| context: current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps .", "entity": "unsupervised setting", "output": "pmctg", "neg_sample": ["unsupervised setting is used for Method", "current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps ."], "relation": "used for", "id": "2022.findings-acl.111", "year": 2022, "rel_sent": "We show that under the unsupervised setting , PMCTG achieves new state - of - the - art results in two representative tasks , namely keywords- to - sentence generation and paraphrasing .", "forward": true, "src_ids": "2022.findings-acl.111_3127"}
+{"input": "pmctg is done by using Task| context: current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps .", "entity": "pmctg", "output": "unsupervised setting", "neg_sample": ["pmctg is done by using Task", "current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps ."], "relation": "used for", "id": "2022.findings-acl.111", "year": 2022, "rel_sent": "We show that under the unsupervised setting , PMCTG achieves new state - of - the - art results in two representative tasks , namely keywords- to - sentence generation and paraphrasing .", "forward": false, "src_ids": "2022.findings-acl.111_3128"}
+{"input": "kernel based extreme learning machines(elm ) is used for Task| context: code - switching refers to the textual or spoken data containing multiple languages .", "entity": "kernel based extreme learning machines(elm )", "output": "sentiment analysis", "neg_sample": ["kernel based extreme learning machines(elm ) is used for Task", "code - switching refers to the textual or spoken data containing multiple languages ."], "relation": "used for", "id": "2022.dravidianlangtech-1.29", "year": 2022, "rel_sent": "This paper shows the experiment results of building a Kernel based Extreme Learning Machines(ELM ) for sentiment analysis for code - switched Dravidian languages with English .", "forward": true, "src_ids": "2022.dravidianlangtech-1.29_3129"}
+{"input": "sentiment analysis is done by using Method| context: code - switching refers to the textual or spoken data containing multiple languages . application of natural language processing ( nlp ) tasks like sentiment analysis is a harder problem on code - switched languages due to the irregularities in the sentence structuring and ordering .", "entity": "sentiment analysis", "output": "kernel based extreme learning machines(elm )", "neg_sample": ["sentiment analysis is done by using Method", "code - switching refers to the textual or spoken data containing multiple languages .", "application of natural language processing ( nlp ) tasks like sentiment analysis is a harder problem on code - switched languages due to the irregularities in the sentence structuring and ordering ."], "relation": "used for", "id": "2022.dravidianlangtech-1.29", "year": 2022, "rel_sent": "This paper shows the experiment results of building a Kernel based Extreme Learning Machines(ELM ) for sentiment analysis for code - switched Dravidian languages with English .", "forward": false, "src_ids": "2022.dravidianlangtech-1.29_3130"}
+{"input": "elm architecture is done by using Method| context: code - switching refers to the textual or spoken data containing multiple languages . application of natural language processing ( nlp ) tasks like sentiment analysis is a harder problem on code - switched languages due to the irregularities in the sentence structuring and ordering .", "entity": "elm architecture", "output": "polynomial kernels", "neg_sample": ["elm architecture is done by using Method", "code - switching refers to the textual or spoken data containing multiple languages .", "application of natural language processing ( nlp ) tasks like sentiment analysis is a harder problem on code - switched languages due to the irregularities in the sentence structuring and ordering ."], "relation": "used for", "id": "2022.dravidianlangtech-1.29", "year": 2022, "rel_sent": "We also show that Polynomial kernels perform better than others in the ELM architecture .", "forward": false, "src_ids": "2022.dravidianlangtech-1.29_3131"}
+{"input": "polynomial kernels is used for Method| context: code - switching refers to the textual or spoken data containing multiple languages . application of natural language processing ( nlp ) tasks like sentiment analysis is a harder problem on code - switched languages due to the irregularities in the sentence structuring and ordering .", "entity": "polynomial kernels", "output": "elm architecture", "neg_sample": ["polynomial kernels is used for Method", "code - switching refers to the textual or spoken data containing multiple languages .", "application of natural language processing ( nlp ) tasks like sentiment analysis is a harder problem on code - switched languages due to the irregularities in the sentence structuring and ordering ."], "relation": "used for", "id": "2022.dravidianlangtech-1.29", "year": 2022, "rel_sent": "We also show that Polynomial kernels perform better than others in the ELM architecture .", "forward": true, "src_ids": "2022.dravidianlangtech-1.29_3132"}
+{"input": "hallucinating synthetic new word forms is done by using OtherScientificTerm| context: deep learning sequence models have been successful with morphological inflection generation . the sigmorphon shared task results in the past several years indicate that such models can perform well , but only if the training data covers a good amount of different lemmata , or if the lemmata to be inflected at test time have also been seen in training , as has indeed been largely the case in these tasks . surprisingly , we find that standard models such as the transformer almost completely fail at generalizing inflection patterns when trained on a limited number of lemmata and asked to inflect previously unseen lemmata - i.e.", "entity": "hallucinating synthetic new word forms", "output": "copying bias", "neg_sample": ["hallucinating synthetic new word forms is done by using OtherScientificTerm", "deep learning sequence models have been successful with morphological inflection generation .", "the sigmorphon shared task results in the past several years indicate that such models can perform well , but only if the training data covers a good amount of different lemmata , or if the lemmata to be inflected at test time have also been seen in training , as has indeed been largely the case in these tasks .", "surprisingly , we find that standard models such as the transformer almost completely fail at generalizing inflection patterns when trained on a limited number of lemmata and asked to inflect previously unseen lemmata - i.e."], "relation": "used for", "id": "2022.acl-short.84", "year": 2022, "rel_sent": "While established data augmentation techniques can be employed to alleviate this shortcoming by introducing a copying bias through hallucinating synthetic new word forms using the alphabet in the language at hand , our experiment results show that , to be more effective , the hallucination process needs to pay attention to substrings of syllable - like length rather than individual characters .", "forward": false, "src_ids": "2022.acl-short.84_3133"}
+{"input": "copying bias is used for Task| context: deep learning sequence models have been successful with morphological inflection generation . the sigmorphon shared task results in the past several years indicate that such models can perform well , but only if the training data covers a good amount of different lemmata , or if the lemmata to be inflected at test time have also been seen in training , as has indeed been largely the case in these tasks . surprisingly , we find that standard models such as the transformer almost completely fail at generalizing inflection patterns when trained on a limited number of lemmata and asked to inflect previously unseen lemmata - i.e.", "entity": "copying bias", "output": "hallucinating synthetic new word forms", "neg_sample": ["copying bias is used for Task", "deep learning sequence models have been successful with morphological inflection generation .", "the sigmorphon shared task results in the past several years indicate that such models can perform well , but only if the training data covers a good amount of different lemmata , or if the lemmata to be inflected at test time have also been seen in training , as has indeed been largely the case in these tasks .", "surprisingly , we find that standard models such as the transformer almost completely fail at generalizing inflection patterns when trained on a limited number of lemmata and asked to inflect previously unseen lemmata - i.e."], "relation": "used for", "id": "2022.acl-short.84", "year": 2022, "rel_sent": "While established data augmentation techniques can be employed to alleviate this shortcoming by introducing a copying bias through hallucinating synthetic new word forms using the alphabet in the language at hand , our experiment results show that , to be more effective , the hallucination process needs to pay attention to substrings of syllable - like length rather than individual characters .", "forward": true, "src_ids": "2022.acl-short.84_3134"}
+{"input": "question answering is done by using Method| context: characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner .", "entity": "question answering", "output": "extractive and generative readers", "neg_sample": ["question answering is done by using Method", "characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner ."], "relation": "used for", "id": "2022.spanlp-1.2", "year": 2022, "rel_sent": "Choose Your QA Model Wisely : A Systematic Study of Generative and Extractive Readers for Question Answering.", "forward": false, "src_ids": "2022.spanlp-1.2_3135"}
+{"input": "question answering is done by using Method| context: characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner .", "entity": "question answering", "output": "extractive and generative readers", "neg_sample": ["question answering is done by using Method", "characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner ."], "relation": "used for", "id": "2022.spanlp-1.2", "year": 2022, "rel_sent": "Motivated by this goal , we make the first attempt to systematically study the comparison of extractive and generative readers for question answering .", "forward": false, "src_ids": "2022.spanlp-1.2_3136"}
+{"input": "extractive and generative readers is used for Task| context: characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner .", "entity": "extractive and generative readers", "output": "question answering", "neg_sample": ["extractive and generative readers is used for Task", "characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner ."], "relation": "used for", "id": "2022.spanlp-1.2", "year": 2022, "rel_sent": "Choose Your QA Model Wisely : A Systematic Study of Generative and Extractive Readers for Question Answering.", "forward": true, "src_ids": "2022.spanlp-1.2_3137"}
+{"input": "extractive and generative readers is used for Task| context: characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner .", "entity": "extractive and generative readers", "output": "question answering", "neg_sample": ["extractive and generative readers is used for Task", "characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner ."], "relation": "used for", "id": "2022.spanlp-1.2", "year": 2022, "rel_sent": "Motivated by this goal , we make the first attempt to systematically study the comparison of extractive and generative readers for question answering .", "forward": true, "src_ids": "2022.spanlp-1.2_3138"}
+{"input": "long context qa is done by using Method| context: characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner .", "entity": "long context qa", "output": "generative readers", "neg_sample": ["long context qa is done by using Method", "characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner ."], "relation": "used for", "id": "2022.spanlp-1.2", "year": 2022, "rel_sent": "Among several interesting findings , it is important to highlight that ( 1 ) the generative readers perform better in long context QA , ( 2 ) the extractive readers perform better in short context while also showing better out - of - domain generalization , and ( 3 ) the encoder of encoder - decoder PrLMs ( e.g. , T5 ) turns out to be a strong extractive reader and outperforms the standard choice of encoder - only PrLMs ( e.g. , RoBERTa ) .", "forward": false, "src_ids": "2022.spanlp-1.2_3139"}
+{"input": "generative readers is used for Task| context: while both extractive and generative readers have been successfully applied to the question answering ( qa ) task , little attention has been paid toward the systematic comparison of them . characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner .", "entity": "generative readers", "output": "long context qa", "neg_sample": ["generative readers is used for Task", "while both extractive and generative readers have been successfully applied to the question answering ( qa ) task , little attention has been paid toward the systematic comparison of them .", "characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but alsofor developing a deeper understanding tofoster further research on improving readers in a principled manner ."], "relation": "used for", "id": "2022.spanlp-1.2", "year": 2022, "rel_sent": "Among several interesting findings , it is important to highlight that ( 1 ) the generative readers perform better in long context QA , ( 2 ) the extractive readers perform better in short context while also showing better out - of - domain generalization , and ( 3 ) the encoder of encoder - decoder PrLMs ( e.g. , T5 ) turns out to be a strong extractive reader and outperforms the standard choice of encoder - only PrLMs ( e.g. , RoBERTa ) .", "forward": true, "src_ids": "2022.spanlp-1.2_3140"}
+{"input": "multimodal machine translation is done by using Method| context: multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs .", "entity": "multimodal machine translation", "output": "phrase - level retrieval - based method", "neg_sample": ["multimodal machine translation is done by using Method", "multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs ."], "relation": "used for", "id": "2022.acl-long.390", "year": 2022, "rel_sent": "In this paper , we propose a phrase - level retrieval - based method for MMT to get visual information for the source input from existing sentence - image data sets so that MMT can break the limitation of paired sentence - image input .", "forward": false, "src_ids": "2022.acl-long.390_3141"}
+{"input": "visual information is done by using Method| context: multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs .", "entity": "visual information", "output": "phrase - level retrieval - based method", "neg_sample": ["visual information is done by using Method", "multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs ."], "relation": "used for", "id": "2022.acl-long.390", "year": 2022, "rel_sent": "In this paper , we propose a phrase - level retrieval - based method for MMT to get visual information for the source input from existing sentence - image data sets so that MMT can break the limitation of paired sentence - image input .", "forward": false, "src_ids": "2022.acl-long.390_3142"}
+{"input": "visual representations is done by using Method| context: multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs .", "entity": "visual representations", "output": "conditional variational auto - encoder", "neg_sample": ["visual representations is done by using Method", "multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs ."], "relation": "used for", "id": "2022.acl-long.390", "year": 2022, "rel_sent": "Furthermore , our method employs the conditional variational auto - encoder to learn visual representations which can filter redundant visual information and only retain visual information related to the phrase .", "forward": false, "src_ids": "2022.acl-long.390_3143"}
+{"input": "redundant visual information is done by using Method| context: multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs .", "entity": "redundant visual information", "output": "visual representations", "neg_sample": ["redundant visual information is done by using Method", "multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs ."], "relation": "used for", "id": "2022.acl-long.390", "year": 2022, "rel_sent": "Furthermore , our method employs the conditional variational auto - encoder to learn visual representations which can filter redundant visual information and only retain visual information related to the phrase .", "forward": false, "src_ids": "2022.acl-long.390_3144"}
+{"input": "conditional variational auto - encoder is used for Method| context: multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs .", "entity": "conditional variational auto - encoder", "output": "visual representations", "neg_sample": ["conditional variational auto - encoder is used for Method", "multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs ."], "relation": "used for", "id": "2022.acl-long.390", "year": 2022, "rel_sent": "Furthermore , our method employs the conditional variational auto - encoder to learn visual representations which can filter redundant visual information and only retain visual information related to the phrase .", "forward": true, "src_ids": "2022.acl-long.390_3145"}
+{"input": "visual representations is used for OtherScientificTerm| context: multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs .", "entity": "visual representations", "output": "redundant visual information", "neg_sample": ["visual representations is used for OtherScientificTerm", "multimodal machine translation ( mmt ) aims to improve neural machine translation ( nmt ) with additional visual information , but most existing mmt methods require paired input of source sentence and image , which makes them suffer from shortage of sentence - image pairs ."], "relation": "used for", "id": "2022.acl-long.390", "year": 2022, "rel_sent": "Furthermore , our method employs the conditional variational auto - encoder to learn visual representations which can filter redundant visual information and only retain visual information related to the phrase .", "forward": true, "src_ids": "2022.acl-long.390_3146"}
+{"input": "fairness notions is used for Method| context: the textual representations in english can be desirably transferred to multilingualism and support downstream multimodal tasks for different languages . nevertheless , the principle of multilingual fairness is rarely scrutinized : do multilingual multimodal models treat languages equally ? are their performances biased towards particular languages ?", "entity": "fairness notions", "output": "pre - trained multimodal representations", "neg_sample": ["fairness notions is used for Method", "the textual representations in english can be desirably transferred to multilingualism and support downstream multimodal tasks for different languages .", "nevertheless , the principle of multilingual fairness is rarely scrutinized : do multilingual multimodal models treat languages equally ?", "are their performances biased towards particular languages ?"], "relation": "used for", "id": "2022.findings-acl.211", "year": 2022, "rel_sent": "To answer these questions , we view language as the fairness recipient and introduce two new fairness notions , multilingual individual fairness and multilingual group fairness , for pre - trained multimodal models .", "forward": true, "src_ids": "2022.findings-acl.211_3147"}
+{"input": "pre - trained multimodal representations is done by using Generic| context: recently pre - trained multimodal models , such as clip , have shown exceptional capabilities towards connecting images and natural language . the textual representations in english can be desirably transferred to multilingualism and support downstream multimodal tasks for different languages . nevertheless , the principle of multilingual fairness is rarely scrutinized : do multilingual multimodal models treat languages equally ? are their performances biased towards particular languages ?", "entity": "pre - trained multimodal representations", "output": "fairness notions", "neg_sample": ["pre - trained multimodal representations is done by using Generic", "recently pre - trained multimodal models , such as clip , have shown exceptional capabilities towards connecting images and natural language .", "the textual representations in english can be desirably transferred to multilingualism and support downstream multimodal tasks for different languages .", "nevertheless , the principle of multilingual fairness is rarely scrutinized : do multilingual multimodal models treat languages equally ?", "are their performances biased towards particular languages ?"], "relation": "used for", "id": "2022.findings-acl.211", "year": 2022, "rel_sent": "To answer these questions , we view language as the fairness recipient and introduce two new fairness notions , multilingual individual fairness and multilingual group fairness , for pre - trained multimodal models .", "forward": false, "src_ids": "2022.findings-acl.211_3148"}
+{"input": "morphologically complex languages is done by using Method| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "morphologically complex languages", "output": "phoneme recognition setup", "neg_sample": ["morphologically complex languages is done by using Method", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "To automate data preparation , training and evaluation steps , we also developed a phoneme recognition setup which handles morphologically complex languages and writing systems for which no pronunciation dictionary exists . We find that fine - tuning a multilingual pretrained model yields an average phoneme error rate ( PER ) of 15 % for 6 languages with 99 minutes or less of transcribed data for training .", "forward": false, "src_ids": "2022.findings-acl.180_3149"}
+{"input": "phoneme recognition setup is used for Material| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "phoneme recognition setup", "output": "morphologically complex languages", "neg_sample": ["phoneme recognition setup is used for Material", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "To automate data preparation , training and evaluation steps , we also developed a phoneme recognition setup which handles morphologically complex languages and writing systems for which no pronunciation dictionary exists . We find that fine - tuning a multilingual pretrained model yields an average phoneme error rate ( PER ) of 15 % for 6 languages with 99 minutes or less of transcribed data for training .", "forward": true, "src_ids": "2022.findings-acl.180_3150"}
+{"input": "speaker - dependent situation is done by using Task| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "speaker - dependent situation", "output": "automatic speech recognition", "neg_sample": ["speaker - dependent situation is done by using Task", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "These results on a number of varied languages suggest that ASR can now significantly reduce transcription efforts in the speaker - dependent situation common in endangered language work .", "forward": false, "src_ids": "2022.findings-acl.180_3151"}
+{"input": "transcription efforts is done by using Task| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "transcription efforts", "output": "automatic speech recognition", "neg_sample": ["transcription efforts is done by using Task", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "These results on a number of varied languages suggest that ASR can now significantly reduce transcription efforts in the speaker - dependent situation common in endangered language work .", "forward": false, "src_ids": "2022.findings-acl.180_3152"}
+{"input": "speaker - dependent situation is done by using Task| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "speaker - dependent situation", "output": "transcription efforts", "neg_sample": ["speaker - dependent situation is done by using Task", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "These results on a number of varied languages suggest that ASR can now significantly reduce transcription efforts in the speaker - dependent situation common in endangered language work .", "forward": false, "src_ids": "2022.findings-acl.180_3153"}
+{"input": "automatic speech recognition is used for Task| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "automatic speech recognition", "output": "transcription efforts", "neg_sample": ["automatic speech recognition is used for Task", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "These results on a number of varied languages suggest that ASR can now significantly reduce transcription efforts in the speaker - dependent situation common in endangered language work .", "forward": true, "src_ids": "2022.findings-acl.180_3154"}
+{"input": "automatic speech recognition is used for OtherScientificTerm| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "automatic speech recognition", "output": "speaker - dependent situation", "neg_sample": ["automatic speech recognition is used for OtherScientificTerm", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "These results on a number of varied languages suggest that ASR can now significantly reduce transcription efforts in the speaker - dependent situation common in endangered language work .", "forward": true, "src_ids": "2022.findings-acl.180_3155"}
+{"input": "transcription efforts is used for OtherScientificTerm| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "transcription efforts", "output": "speaker - dependent situation", "neg_sample": ["transcription efforts is used for OtherScientificTerm", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.180", "year": 2022, "rel_sent": "These results on a number of varied languages suggest that ASR can now significantly reduce transcription efforts in the speaker - dependent situation common in endangered language work .", "forward": true, "src_ids": "2022.findings-acl.180_3156"}
+{"input": "aspect - based sentiment classification is done by using Method| context: dependency trees have been intensively used with graph neural networks for aspect - based sentiment classification . though being effective , such methods rely on external dependency parsers , which can be unavailable for low - resource languages or perform worse in low - resource domains . in addition , dependency trees are also not optimized for aspect - based sentiment classification .", "entity": "aspect - based sentiment classification", "output": "discrete opinion tree induction", "neg_sample": ["aspect - based sentiment classification is done by using Method", "dependency trees have been intensively used with graph neural networks for aspect - based sentiment classification .", "though being effective , such methods rely on external dependency parsers , which can be unavailable for low - resource languages or perform worse in low - resource domains .", "in addition , dependency trees are also not optimized for aspect - based sentiment classification ."], "relation": "used for", "id": "2022.acl-long.145", "year": 2022, "rel_sent": "Discrete Opinion Tree Induction for Aspect - based Sentiment Analysis.", "forward": false, "src_ids": "2022.acl-long.145_3157"}
+{"input": "discrete opinion tree induction is used for Task| context: though being effective , such methods rely on external dependency parsers , which can be unavailable for low - resource languages or perform worse in low - resource domains .", "entity": "discrete opinion tree induction", "output": "aspect - based sentiment classification", "neg_sample": ["discrete opinion tree induction is used for Task", "though being effective , such methods rely on external dependency parsers , which can be unavailable for low - resource languages or perform worse in low - resource domains ."], "relation": "used for", "id": "2022.acl-long.145", "year": 2022, "rel_sent": "Discrete Opinion Tree Induction for Aspect - based Sentiment Analysis.", "forward": true, "src_ids": "2022.acl-long.145_3158"}
+{"input": "open - domain dialogue generation is done by using Method| context: a major issue in open - domain dialogue generation is the agent 's tendency to generate repetitive and generic responses . the lack in response diversity has been addressed in recent years via the use of latent variable models , such as the conditional variational auto - encoder ( cvae ) , which typically involve learning a latent gaussian distribution over potential response intents . however , due to latent variable collapse , training latent variable dialogue models are notoriously complex , requiring substantial modification to the standard training process and loss function . other approaches proposed to improve response diversity also largely entail a significant increase in training complexity .", "entity": "open - domain dialogue generation", "output": "randomized link ( rl ) transformer", "neg_sample": ["open - domain dialogue generation is done by using Method", "a major issue in open - domain dialogue generation is the agent 's tendency to generate repetitive and generic responses .", "the lack in response diversity has been addressed in recent years via the use of latent variable models , such as the conditional variational auto - encoder ( cvae ) , which typically involve learning a latent gaussian distribution over potential response intents .", "however , due to latent variable collapse , training latent variable dialogue models are notoriously complex , requiring substantial modification to the standard training process and loss function .", "other approaches proposed to improve response diversity also largely entail a significant increase in training complexity ."], "relation": "used for", "id": "2022.nlp4convai-1.1", "year": 2022, "rel_sent": "A Randomized Link Transformer for Diverse Open - Domain Dialogue Generation.", "forward": false, "src_ids": "2022.nlp4convai-1.1_3159"}
+{"input": "randomized link ( rl ) transformer is used for Task| context: the lack in response diversity has been addressed in recent years via the use of latent variable models , such as the conditional variational auto - encoder ( cvae ) , which typically involve learning a latent gaussian distribution over potential response intents . however , due to latent variable collapse , training latent variable dialogue models are notoriously complex , requiring substantial modification to the standard training process and loss function . other approaches proposed to improve response diversity also largely entail a significant increase in training complexity .", "entity": "randomized link ( rl ) transformer", "output": "open - domain dialogue generation", "neg_sample": ["randomized link ( rl ) transformer is used for Task", "the lack in response diversity has been addressed in recent years via the use of latent variable models , such as the conditional variational auto - encoder ( cvae ) , which typically involve learning a latent gaussian distribution over potential response intents .", "however , due to latent variable collapse , training latent variable dialogue models are notoriously complex , requiring substantial modification to the standard training process and loss function .", "other approaches proposed to improve response diversity also largely entail a significant increase in training complexity ."], "relation": "used for", "id": "2022.nlp4convai-1.1", "year": 2022, "rel_sent": "A Randomized Link Transformer for Diverse Open - Domain Dialogue Generation.", "forward": true, "src_ids": "2022.nlp4convai-1.1_3160"}
+{"input": "inferences is done by using Method| context: transformers have been shown to be able to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language . recent works show that such models can also produce the reasoning steps ( i.e. , the proof graph ) that emulate the model 's logical reasoning process . currently , these black - box models generate both the proof graph and intermediate inferences within the same model and thus may be unfaithful .", "entity": "inferences", "output": "knowledge composition", "neg_sample": ["inferences is done by using Method", "transformers have been shown to be able to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language .", "recent works show that such models can also produce the reasoning steps ( i.e.", ", the proof graph ) that emulate the model 's logical reasoning process .", "currently , these black - box models generate both the proof graph and intermediate inferences within the same model and thus may be unfaithful ."], "relation": "used for", "id": "2022.acl-long.77", "year": 2022, "rel_sent": "The rule and fact selection steps select the candidate rule and facts to be used and then the knowledge composition combines them to generate new inferences .", "forward": false, "src_ids": "2022.acl-long.77_3161"}
+{"input": "knowledge composition is used for Task| context: transformers have been shown to be able to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language . recent works show that such models can also produce the reasoning steps ( i.e. , the proof graph ) that emulate the model 's logical reasoning process .", "entity": "knowledge composition", "output": "inferences", "neg_sample": ["knowledge composition is used for Task", "transformers have been shown to be able to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language .", "recent works show that such models can also produce the reasoning steps ( i.e.", ", the proof graph ) that emulate the model 's logical reasoning process ."], "relation": "used for", "id": "2022.acl-long.77", "year": 2022, "rel_sent": "The rule and fact selection steps select the candidate rule and facts to be used and then the knowledge composition combines them to generate new inferences .", "forward": true, "src_ids": "2022.acl-long.77_3162"}
+{"input": "link prediction is done by using Method| context: knowledge graphs ( kgs ) are symbolically structured storages of facts . the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world . furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues .", "entity": "link prediction", "output": "knowledge embedding representation", "neg_sample": ["link prediction is done by using Method", "knowledge graphs ( kgs ) are symbolically structured storages of facts .", "the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world .", "furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues ."], "relation": "used for", "id": "2022.acl-srw.27", "year": 2022, "rel_sent": "MEKER : Memory Efficient Knowledge Embedding Representation for Link Prediction and Question Answering.", "forward": false, "src_ids": "2022.acl-srw.27_3163"}
+{"input": "knowledge embedding representation is used for Task| context: knowledge graphs ( kgs ) are symbolically structured storages of facts . the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world . furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues .", "entity": "knowledge embedding representation", "output": "link prediction", "neg_sample": ["knowledge embedding representation is used for Task", "knowledge graphs ( kgs ) are symbolically structured storages of facts .", "the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world .", "furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues ."], "relation": "used for", "id": "2022.acl-srw.27", "year": 2022, "rel_sent": "MEKER : Memory Efficient Knowledge Embedding Representation for Link Prediction and Question Answering.", "forward": true, "src_ids": "2022.acl-srw.27_3164"}
+{"input": "optimization gradients is done by using Method| context: knowledge graphs ( kgs ) are symbolically structured storages of facts . the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world . furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues .", "entity": "optimization gradients", "output": "cp - als algorithm", "neg_sample": ["optimization gradients is done by using Method", "knowledge graphs ( kgs ) are symbolically structured storages of facts .", "the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world .", "furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues ."], "relation": "used for", "id": "2022.acl-srw.27", "year": 2022, "rel_sent": "The generalization of the standard CP - ALS algorithm allows obtaining optimization gradients without a backpropagation mechanism .", "forward": false, "src_ids": "2022.acl-srw.27_3165"}
+{"input": "cp - als algorithm is used for OtherScientificTerm| context: knowledge graphs ( kgs ) are symbolically structured storages of facts . the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world . furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues .", "entity": "cp - als algorithm", "output": "optimization gradients", "neg_sample": ["cp - als algorithm is used for OtherScientificTerm", "knowledge graphs ( kgs ) are symbolically structured storages of facts .", "the kg embedding contains concise data used in nlp tasks requiring implicit information about the real world .", "furthermore , the size of kgs that may be useful in actual nlp assignments is enormous , and creating embedding over it has memory cost issues ."], "relation": "used for", "id": "2022.acl-srw.27", "year": 2022, "rel_sent": "The generalization of the standard CP - ALS algorithm allows obtaining optimization gradients without a backpropagation mechanism .", "forward": true, "src_ids": "2022.acl-srw.27_3166"}
+{"input": "multilingual models is done by using OtherScientificTerm| context: when tasked with supporting multiple languages for a given problem , two approaches have arisen : training a model for each language with the annotation budget divided equally among them , and training on a high - resource language followed by zero - shot transfer to the remaining languages .", "entity": "multilingual models", "output": "annotations", "neg_sample": ["multilingual models is done by using OtherScientificTerm", "when tasked with supporting multiple languages for a given problem , two approaches have arisen : training a model for each language with the annotation budget divided equally among them , and training on a high - resource language followed by zero - shot transfer to the remaining languages ."], "relation": "used for", "id": "2022.acl-short.9", "year": 2022, "rel_sent": "On Efficiently Acquiring Annotations for Multilingual Models.", "forward": false, "src_ids": "2022.acl-short.9_3167"}
+{"input": "annotations is used for Method| context: when tasked with supporting multiple languages for a given problem , two approaches have arisen : training a model for each language with the annotation budget divided equally among them , and training on a high - resource language followed by zero - shot transfer to the remaining languages .", "entity": "annotations", "output": "multilingual models", "neg_sample": ["annotations is used for Method", "when tasked with supporting multiple languages for a given problem , two approaches have arisen : training a model for each language with the annotation budget divided equally among them , and training on a high - resource language followed by zero - shot transfer to the remaining languages ."], "relation": "used for", "id": "2022.acl-short.9", "year": 2022, "rel_sent": "On Efficiently Acquiring Annotations for Multilingual Models.", "forward": true, "src_ids": "2022.acl-short.9_3168"}
+{"input": "natural - language answers is done by using Task| context: existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history . it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations .", "entity": "natural - language answers", "output": "conversational question answering", "neg_sample": ["natural - language answers is done by using Task", "existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history .", "it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations ."], "relation": "used for", "id": "2022.acl-long.555", "year": 2022, "rel_sent": "Conversational question answering aims to provide natural - language answers to users in information - seeking conversations .", "forward": false, "src_ids": "2022.acl-long.555_3169"}
+{"input": "information - seeking conversations is done by using OtherScientificTerm| context: existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history . it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations .", "entity": "information - seeking conversations", "output": "natural - language answers", "neg_sample": ["information - seeking conversations is done by using OtherScientificTerm", "existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history .", "it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations ."], "relation": "used for", "id": "2022.acl-long.555", "year": 2022, "rel_sent": "Conversational question answering aims to provide natural - language answers to users in information - seeking conversations .", "forward": false, "src_ids": "2022.acl-long.555_3170"}
+{"input": "conversational question answering is used for OtherScientificTerm| context: existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history . it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations .", "entity": "conversational question answering", "output": "natural - language answers", "neg_sample": ["conversational question answering is used for OtherScientificTerm", "existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history .", "it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations ."], "relation": "used for", "id": "2022.acl-long.555", "year": 2022, "rel_sent": "Conversational question answering aims to provide natural - language answers to users in information - seeking conversations .", "forward": true, "src_ids": "2022.acl-long.555_3171"}
+{"input": "natural - language answers is used for Material| context: existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history . it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations .", "entity": "natural - language answers", "output": "information - seeking conversations", "neg_sample": ["natural - language answers is used for Material", "existing conversational qa benchmarks compare models with pre - collected human - human conversations , using ground - truth answers provided in conversational history .", "it remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real - world human - machine conversations ."], "relation": "used for", "id": "2022.acl-long.555", "year": 2022, "rel_sent": "Conversational question answering aims to provide natural - language answers to users in information - seeking conversations .", "forward": true, "src_ids": "2022.acl-long.555_3172"}
+{"input": "abstractive news summarization is done by using Task| context: recent improvements in automatic news summarization fundamentally rely on large corpora of news articles and their summaries . these corpora are often constructed by scraping news websites , which results in including not only summaries but also other kinds of texts . apart from more generic noise , we identify straplines as a form of text scraped from news websites that commonly turn out not to be summaries . the presence of these non - summaries threatens the validity of scraped corpora as benchmarks for news summarization .", "entity": "abstractive news summarization", "output": "automatically discarding straplines", "neg_sample": ["abstractive news summarization is done by using Task", "recent improvements in automatic news summarization fundamentally rely on large corpora of news articles and their summaries .", "these corpora are often constructed by scraping news websites , which results in including not only summaries but also other kinds of texts .", "apart from more generic noise , we identify straplines as a form of text scraped from news websites that commonly turn out not to be summaries .", "the presence of these non - summaries threatens the validity of scraped corpora as benchmarks for news summarization ."], "relation": "used for", "id": "2022.nlppower-1.5", "year": 2022, "rel_sent": "Automatically Discarding Straplines to Improve Data Quality for Abstractive News Summarization.", "forward": false, "src_ids": "2022.nlppower-1.5_3173"}
+{"input": "automatically discarding straplines is used for Task| context: recent improvements in automatic news summarization fundamentally rely on large corpora of news articles and their summaries . these corpora are often constructed by scraping news websites , which results in including not only summaries but also other kinds of texts . apart from more generic noise , we identify straplines as a form of text scraped from news websites that commonly turn out not to be summaries . the presence of these non - summaries threatens the validity of scraped corpora as benchmarks for news summarization .", "entity": "automatically discarding straplines", "output": "abstractive news summarization", "neg_sample": ["automatically discarding straplines is used for Task", "recent improvements in automatic news summarization fundamentally rely on large corpora of news articles and their summaries .", "these corpora are often constructed by scraping news websites , which results in including not only summaries but also other kinds of texts .", "apart from more generic noise , we identify straplines as a form of text scraped from news websites that commonly turn out not to be summaries .", "the presence of these non - summaries threatens the validity of scraped corpora as benchmarks for news summarization ."], "relation": "used for", "id": "2022.nlppower-1.5", "year": 2022, "rel_sent": "Automatically Discarding Straplines to Improve Data Quality for Abstractive News Summarization.", "forward": true, "src_ids": "2022.nlppower-1.5_3174"}
+{"input": "unified plug - and - play model is used for Task| context: despite their success , existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub - tasks and greater data annotation overhead .", "entity": "unified plug - and - play model", "output": "task - oriented dialogue", "neg_sample": ["unified plug - and - play model is used for Task", "despite their success , existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub - tasks and greater data annotation overhead ."], "relation": "used for", "id": "2022.acl-long.319", "year": 2022, "rel_sent": "In this study , we present PPTOD , a unified plug - and - play model for task - oriented dialogue .", "forward": true, "src_ids": "2022.acl-long.319_3175"}
+{"input": "task - oriented dialogue is done by using Method| context: pre - trained language models have been recently shown to benefit task - oriented dialogue ( tod ) systems . despite their success , existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub - tasks and greater data annotation overhead .", "entity": "task - oriented dialogue", "output": "unified plug - and - play model", "neg_sample": ["task - oriented dialogue is done by using Method", "pre - trained language models have been recently shown to benefit task - oriented dialogue ( tod ) systems .", "despite their success , existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub - tasks and greater data annotation overhead ."], "relation": "used for", "id": "2022.acl-long.319", "year": 2022, "rel_sent": "In this study , we present PPTOD , a unified plug - and - play model for task - oriented dialogue .", "forward": false, "src_ids": "2022.acl-long.319_3176"}
+{"input": "multi - hop knowledge base question answering is done by using Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "multi - hop knowledge base question answering", "output": "subgraph retrieval enhanced model", "neg_sample": ["multi - hop knowledge base question answering is done by using Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Subgraph Retrieval Enhanced Model for Multi - hop Knowledge Base Question Answering.", "forward": false, "src_ids": "2022.acl-long.396_3177"}
+{"input": "subgraph retrieval enhanced model is used for Task| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "subgraph retrieval enhanced model", "output": "multi - hop knowledge base question answering", "neg_sample": ["subgraph retrieval enhanced model is used for Task", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Subgraph Retrieval Enhanced Model for Multi - hop Knowledge Base Question Answering.", "forward": true, "src_ids": "2022.acl-long.396_3178"}
+{"input": "subgraph - oriented kbqa model is done by using Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "subgraph - oriented kbqa model", "output": "trainable subgraph retriever ( sr )", "neg_sample": ["subgraph - oriented kbqa model is done by using Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "This paper proposes a trainable subgraph retriever ( SR ) decoupled from the subsequent reasoning process , which enables a plug - and - play framework to enhance any subgraph - oriented KBQA model .", "forward": false, "src_ids": "2022.acl-long.396_3179"}
+{"input": "plug - and - play framework is used for Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "plug - and - play framework", "output": "subgraph - oriented kbqa model", "neg_sample": ["plug - and - play framework is used for Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "This paper proposes a trainable subgraph retriever ( SR ) decoupled from the subsequent reasoning process , which enables a plug - and - play framework to enhance any subgraph - oriented KBQA model .", "forward": true, "src_ids": "2022.acl-long.396_3180"}
+{"input": "trainable subgraph retriever ( sr ) is used for Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "trainable subgraph retriever ( sr )", "output": "subgraph - oriented kbqa model", "neg_sample": ["trainable subgraph retriever ( sr ) is used for Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "This paper proposes a trainable subgraph retriever ( SR ) decoupled from the subsequent reasoning process , which enables a plug - and - play framework to enhance any subgraph - oriented KBQA model .", "forward": true, "src_ids": "2022.acl-long.396_3181"}
+{"input": "weakly supervised pre - training is used for Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "weakly supervised pre - training", "output": "subgraph retriever", "neg_sample": ["weakly supervised pre - training is used for Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Via weakly supervised pre - training as well as the end - to - end fine - tuning , SR achieves new state - of - the - art performance when combined with NSM ( He et al . , 2021 ) , a subgraph - oriented reasoner , for embedding - based KBQA methods .", "forward": true, "src_ids": "2022.acl-long.396_3182"}
+{"input": "end - to - end fine - tuning is used for Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "end - to - end fine - tuning", "output": "subgraph retriever", "neg_sample": ["end - to - end fine - tuning is used for Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Via weakly supervised pre - training as well as the end - to - end fine - tuning , SR achieves new state - of - the - art performance when combined with NSM ( He et al . , 2021 ) , a subgraph - oriented reasoner , for embedding - based KBQA methods .", "forward": true, "src_ids": "2022.acl-long.396_3183"}
+{"input": "subgraph retriever is done by using Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "subgraph retriever", "output": "weakly supervised pre - training", "neg_sample": ["subgraph retriever is done by using Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Via weakly supervised pre - training as well as the end - to - end fine - tuning , SR achieves new state - of - the - art performance when combined with NSM ( He et al . , 2021 ) , a subgraph - oriented reasoner , for embedding - based KBQA methods .", "forward": false, "src_ids": "2022.acl-long.396_3184"}
+{"input": "embedding - based kbqa methods is done by using Method| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "embedding - based kbqa methods", "output": "subgraph - oriented reasoner", "neg_sample": ["embedding - based kbqa methods is done by using Method", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Via weakly supervised pre - training as well as the end - to - end fine - tuning , SR achieves new state - of - the - art performance when combined with NSM ( He et al . , 2021 ) , a subgraph - oriented reasoner , for embedding - based KBQA methods .", "forward": false, "src_ids": "2022.acl-long.396_3185"}
+{"input": "subgraph - oriented reasoner is used for Task| context: recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning . the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises . however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing .", "entity": "subgraph - oriented reasoner", "output": "embedding - based kbqa methods", "neg_sample": ["subgraph - oriented reasoner is used for Task", "recent works on knowledge base question answering ( kbqa ) retrieve subgraphs for easier reasoning .", "the desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises .", "however , the existing retrieval is either heuristic or interwoven with the reasoning , causing reasoning on the partial subgraphs , which increases the reasoning bias when the intermediate supervision is missing ."], "relation": "used for", "id": "2022.acl-long.396", "year": 2022, "rel_sent": "Via weakly supervised pre - training as well as the end - to - end fine - tuning , SR achieves new state - of - the - art performance when combined with NSM ( He et al . , 2021 ) , a subgraph - oriented reasoner , for embedding - based KBQA methods .", "forward": true, "src_ids": "2022.acl-long.396_3186"}
+{"input": "sequence modeling is done by using OtherScientificTerm| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "sequence modeling", "output": "relaxed equilibrium", "neg_sample": ["sequence modeling is done by using OtherScientificTerm", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Leveraging Relaxed Equilibrium by Lazy Transition for Sequence Modeling.", "forward": false, "src_ids": "2022.acl-long.208_3187"}
+{"input": "sequence modeling is done by using Method| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "sequence modeling", "output": "lazy transition", "neg_sample": ["sequence modeling is done by using Method", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Leveraging Relaxed Equilibrium by Lazy Transition for Sequence Modeling.", "forward": false, "src_ids": "2022.acl-long.208_3188"}
+{"input": "lt ( lazy transformer ) is done by using Method| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "lt ( lazy transformer )", "output": "lazy transition", "neg_sample": ["lt ( lazy transformer ) is done by using Method", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Our lazy transition is deployed on top of UT to build LT ( lazy transformer ) , where all tokens are processed unequally towards depth .", "forward": false, "src_ids": "2022.acl-long.208_3189"}
+{"input": "lazy transition is used for Task| context: however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "lazy transition", "output": "sequence modeling", "neg_sample": ["lazy transition is used for Task", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Leveraging Relaxed Equilibrium by Lazy Transition for Sequence Modeling.", "forward": true, "src_ids": "2022.acl-long.208_3190"}
+{"input": "relaxed equilibrium is used for Task| context: however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "relaxed equilibrium", "output": "sequence modeling", "neg_sample": ["relaxed equilibrium is used for Task", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Leveraging Relaxed Equilibrium by Lazy Transition for Sequence Modeling.", "forward": true, "src_ids": "2022.acl-long.208_3191"}
+{"input": "token representation is done by using OtherScientificTerm| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "token representation", "output": "iterative refinements", "neg_sample": ["token representation is done by using OtherScientificTerm", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Inspired by the equilibrium phenomenon , we present a lazy transition , a mechanism to adjust the significance of iterative refinements for each token representation .", "forward": false, "src_ids": "2022.acl-long.208_3192"}
+{"input": "iterative refinements is used for Method| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "iterative refinements", "output": "token representation", "neg_sample": ["iterative refinements is used for Method", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Inspired by the equilibrium phenomenon , we present a lazy transition , a mechanism to adjust the significance of iterative refinements for each token representation .", "forward": true, "src_ids": "2022.acl-long.208_3193"}
+{"input": "lt ( lazy transformer ) is done by using Method| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "entity": "lt ( lazy transformer )", "output": "ut", "neg_sample": ["lt ( lazy transformer ) is done by using Method", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Our lazy transition is deployed on top of UT to build LT ( lazy transformer ) , where all tokens are processed unequally towards depth .", "forward": false, "src_ids": "2022.acl-long.208_3194"}
+{"input": "lazy transition is used for Method| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "lazy transition", "output": "lt ( lazy transformer )", "neg_sample": ["lazy transition is used for Method", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Our lazy transition is deployed on top of UT to build LT ( lazy transformer ) , where all tokens are processed unequally towards depth .", "forward": true, "src_ids": "2022.acl-long.208_3195"}
+{"input": "ut is used for Method| context: in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation . however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth .", "entity": "ut", "output": "lt ( lazy transformer )", "neg_sample": ["ut is used for Method", "in sequence modeling , certain tokens are usually less ambiguous than others , and representations of these tokens require fewer refinements for disambiguation .", "however , given the nature of attention - based models like transformer and ut ( universal transformer ) , all tokens are equally processed towards depth ."], "relation": "used for", "id": "2022.acl-long.208", "year": 2022, "rel_sent": "Our lazy transition is deployed on top of UT to build LT ( lazy transformer ) , where all tokens are processed unequally towards depth .", "forward": true, "src_ids": "2022.acl-long.208_3196"}
+{"input": "rule based systems is used for OtherScientificTerm| context: open information extraction ( openie ) is the task of extracting ( subject , predicate , object ) triples from natural language sentences . current openie systems extract all triple slots independently .", "entity": "rule based systems", "output": "extraction slots", "neg_sample": ["rule based systems is used for OtherScientificTerm", "open information extraction ( openie ) is the task of extracting ( subject , predicate , object ) triples from natural language sentences .", "current openie systems extract all triple slots independently ."], "relation": "used for", "id": "2022.acl-long.478", "year": 2022, "rel_sent": "Due to the iterative nature , the system is also modularit is possible to seamlessly integrate rule based extraction systems with a neural end - to - end system , thereby allowing rule based systems to supply extraction slots which MILIE can leverage for extracting the remaining slots .", "forward": true, "src_ids": "2022.acl-long.478_3197"}
+{"input": "extraction slots is done by using Method| context: open information extraction ( openie ) is the task of extracting ( subject , predicate , object ) triples from natural language sentences . current openie systems extract all triple slots independently .", "entity": "extraction slots", "output": "rule based systems", "neg_sample": ["extraction slots is done by using Method", "open information extraction ( openie ) is the task of extracting ( subject , predicate , object ) triples from natural language sentences .", "current openie systems extract all triple slots independently ."], "relation": "used for", "id": "2022.acl-long.478", "year": 2022, "rel_sent": "Due to the iterative nature , the system is also modularit is possible to seamlessly integrate rule based extraction systems with a neural end - to - end system , thereby allowing rule based systems to supply extraction slots which MILIE can leverage for extracting the remaining slots .", "forward": false, "src_ids": "2022.acl-long.478_3198"}
+{"input": "multilingual language understanding is done by using Method| context: scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations . the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g. , web - scale corpora ) .", "entity": "multilingual language understanding", "output": "data augmentation", "neg_sample": ["multilingual language understanding is done by using Method", "scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations .", "the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g.", ", web - scale corpora ) ."], "relation": "used for", "id": "2022.findings-acl.160", "year": 2022, "rel_sent": "Data Augmentation and Learned Layer Aggregation for Improved Multilingual Language Understanding in Dialogue.", "forward": false, "src_ids": "2022.findings-acl.160_3199"}
+{"input": "layer aggregation is used for Task| context: scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations . the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g. , web - scale corpora ) .", "entity": "layer aggregation", "output": "multilingual language understanding", "neg_sample": ["layer aggregation is used for Task", "scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations .", "the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g.", ", web - scale corpora ) ."], "relation": "used for", "id": "2022.findings-acl.160", "year": 2022, "rel_sent": "Data Augmentation and Learned Layer Aggregation for Improved Multilingual Language Understanding in Dialogue.", "forward": true, "src_ids": "2022.findings-acl.160_3200"}
+{"input": "data augmentation is used for Task| context: scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations . the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g. , web - scale corpora ) .", "entity": "data augmentation", "output": "multilingual language understanding", "neg_sample": ["data augmentation is used for Task", "scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations .", "the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g.", ", web - scale corpora ) ."], "relation": "used for", "id": "2022.findings-acl.160", "year": 2022, "rel_sent": "Data Augmentation and Learned Layer Aggregation for Improved Multilingual Language Understanding in Dialogue.", "forward": true, "src_ids": "2022.findings-acl.160_3201"}
+{"input": "semantic information is done by using Method| context: scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations . the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g. , web - scale corpora ) .", "entity": "semantic information", "output": "layeragg", "neg_sample": ["semantic information is done by using Method", "scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations .", "the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g.", ", web - scale corpora ) ."], "relation": "used for", "id": "2022.findings-acl.160", "year": 2022, "rel_sent": "LayerAgg learns to select and combine useful semantic information scattered across different layers of a Transformer model ( e.g. , mBERT ) ; it is especially suited for zero - shot scenarios as semantically richer representations should strengthen the model 's cross - lingual capabilities .", "forward": false, "src_ids": "2022.findings-acl.160_3202"}
+{"input": "layeragg is used for OtherScientificTerm| context: scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations . the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g. , web - scale corpora ) .", "entity": "layeragg", "output": "semantic information", "neg_sample": ["layeragg is used for OtherScientificTerm", "scaling dialogue systems to a multitude of domains , tasks and languages relies on costly and time - consuming data annotation for different domain - task - language configurations .", "the annotation efforts might be substantially reduced by the methods that generalise well in zero- and few - shot scenarios , and also effectively leverage external unannotated data sources ( e.g.", ", web - scale corpora ) ."], "relation": "used for", "id": "2022.findings-acl.160", "year": 2022, "rel_sent": "LayerAgg learns to select and combine useful semantic information scattered across different layers of a Transformer model ( e.g. , mBERT ) ; it is especially suited for zero - shot scenarios as semantically richer representations should strengthen the model 's cross - lingual capabilities .", "forward": true, "src_ids": "2022.findings-acl.160_3203"}
+{"input": "text simplification corpora is done by using Method| context: ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g. , w.r.t .", "entity": "text simplification corpora", "output": "annotation tool", "neg_sample": ["text simplification corpora is done by using Method", "ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g.", ", w.r.t ."], "relation": "used for", "id": "2022.acl-demo.14", "year": 2022, "rel_sent": "TS - ANNO : An Annotation Tool to Build , Annotate and Evaluate Text Simplification Corpora.", "forward": false, "src_ids": "2022.acl-demo.14_3204"}
+{"input": "annotation tool is used for Material| context: ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g. , w.r.t .", "entity": "annotation tool", "output": "text simplification corpora", "neg_sample": ["annotation tool is used for Material", "ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g.", ", w.r.t ."], "relation": "used for", "id": "2022.acl-demo.14", "year": 2022, "rel_sent": "TS - ANNO : An Annotation Tool to Build , Annotate and Evaluate Text Simplification Corpora.", "forward": true, "src_ids": "2022.acl-demo.14_3205"}
+{"input": "text simplification is done by using Material| context: ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g. , w.r.t .", "entity": "text simplification", "output": "parallel corpora", "neg_sample": ["text simplification is done by using Material", "ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g.", ", w.r.t ."], "relation": "used for", "id": "2022.acl-demo.14", "year": 2022, "rel_sent": "We introduce TS - ANNO , an open - source web application for manual creation and for evaluation of parallel corpora for text simplification .", "forward": false, "src_ids": "2022.acl-demo.14_3206"}
+{"input": "parallel corpora is used for Task| context: ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g. , w.r.t .", "entity": "parallel corpora", "output": "text simplification", "neg_sample": ["parallel corpora is used for Task", "ts - anno can be used for i ) sentence - wise alignment , ii ) rating alignment pairs ( e.g.", ", w.r.t ."], "relation": "used for", "id": "2022.acl-demo.14", "year": 2022, "rel_sent": "We introduce TS - ANNO , an open - source web application for manual creation and for evaluation of parallel corpora for text simplification .", "forward": true, "src_ids": "2022.acl-demo.14_3207"}
+{"input": "pcvg task is done by using Material| context: given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image . naturally , models trained on these biased data lead to over - estimation of performance on the benchmark .", "entity": "pcvg task", "output": "debiased dataset", "neg_sample": ["pcvg task is done by using Material", "given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image .", "naturally , models trained on these biased data lead to over - estimation of performance on the benchmark ."], "relation": "used for", "id": "2022.acl-short.39", "year": 2022, "rel_sent": "We argue our debiased dataset offers the PCVG task a more practical baseline for reliable benchmarking and future improvements .", "forward": false, "src_ids": "2022.acl-short.39_3208"}
+{"input": "benchmarking is done by using Material| context: given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image . naturally , models trained on these biased data lead to over - estimation of performance on the benchmark .", "entity": "benchmarking", "output": "debiased dataset", "neg_sample": ["benchmarking is done by using Material", "given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image .", "naturally , models trained on these biased data lead to over - estimation of performance on the benchmark ."], "relation": "used for", "id": "2022.acl-short.39", "year": 2022, "rel_sent": "We argue our debiased dataset offers the PCVG task a more practical baseline for reliable benchmarking and future improvements .", "forward": false, "src_ids": "2022.acl-short.39_3209"}
+{"input": "debiased dataset is used for Task| context: we present a debiased dataset for the person - centric visual grounding ( pcvg ) task first proposed by cui et al . given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image . naturally , models trained on these biased data lead to over - estimation of performance on the benchmark .", "entity": "debiased dataset", "output": "pcvg task", "neg_sample": ["debiased dataset is used for Task", "we present a debiased dataset for the person - centric visual grounding ( pcvg ) task first proposed by cui et al .", "given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image .", "naturally , models trained on these biased data lead to over - estimation of performance on the benchmark ."], "relation": "used for", "id": "2022.acl-short.39", "year": 2022, "rel_sent": "We argue our debiased dataset offers the PCVG task a more practical baseline for reliable benchmarking and future improvements .", "forward": true, "src_ids": "2022.acl-short.39_3210"}
+{"input": "end - to - end simultaneous translation is done by using Method| context: a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate . thus the policy is crucial to balance translation quality and latency .", "entity": "end - to - end simultaneous translation", "output": "learning adaptive segmentation policy", "neg_sample": ["end - to - end simultaneous translation is done by using Method", "a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate .", "thus the policy is crucial to balance translation quality and latency ."], "relation": "used for", "id": "2022.acl-long.542", "year": 2022, "rel_sent": "Learning Adaptive Segmentation Policy for End - to - End Simultaneous Translation.", "forward": false, "src_ids": "2022.acl-long.542_3211"}
+{"input": "end - to - end st is done by using Method| context: a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate . thus the policy is crucial to balance translation quality and latency .", "entity": "end - to - end st", "output": "learning adaptive segmentation policy", "neg_sample": ["end - to - end st is done by using Method", "a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate .", "thus the policy is crucial to balance translation quality and latency ."], "relation": "used for", "id": "2022.acl-long.542", "year": 2022, "rel_sent": "This paper proposes an adaptive segmentation policy for end - to - end ST .", "forward": false, "src_ids": "2022.acl-long.542_3212"}
+{"input": "learning adaptive segmentation policy is used for Task| context: a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate . thus the policy is crucial to balance translation quality and latency .", "entity": "learning adaptive segmentation policy", "output": "end - to - end simultaneous translation", "neg_sample": ["learning adaptive segmentation policy is used for Task", "a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate .", "thus the policy is crucial to balance translation quality and latency ."], "relation": "used for", "id": "2022.acl-long.542", "year": 2022, "rel_sent": "Learning Adaptive Segmentation Policy for End - to - End Simultaneous Translation.", "forward": true, "src_ids": "2022.acl-long.542_3213"}
+{"input": "learning adaptive segmentation policy is used for Task| context: a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate . thus the policy is crucial to balance translation quality and latency .", "entity": "learning adaptive segmentation policy", "output": "end - to - end st", "neg_sample": ["learning adaptive segmentation policy is used for Task", "a typical simultaneous translation ( st ) system consists of a speech translation model and a policy module , which determines when to wait and when to translate .", "thus the policy is crucial to balance translation quality and latency ."], "relation": "used for", "id": "2022.acl-long.542", "year": 2022, "rel_sent": "This paper proposes an adaptive segmentation policy for end - to - end ST .", "forward": true, "src_ids": "2022.acl-long.542_3214"}
+{"input": "graphs is done by using Method| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks . however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs . unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g. , generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts .", "entity": "graphs", "output": "pre - trained language models", "neg_sample": ["graphs is done by using Method", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs .", "unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g.", ", generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "We first show that with limited supervision , pre - trained language models often generate graphs that either violate these constraints or are semantically incoherent .", "forward": false, "src_ids": "2022.acl-long.85_3215"}
+{"input": "limited supervision is used for Method| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks . however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs . unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g. , generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts .", "entity": "limited supervision", "output": "pre - trained language models", "neg_sample": ["limited supervision is used for Method", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs .", "unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g.", ", generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "We first show that with limited supervision , pre - trained language models often generate graphs that either violate these constraints or are semantically incoherent .", "forward": true, "src_ids": "2022.acl-long.85_3216"}
+{"input": "contrastive learning models is done by using OtherScientificTerm| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "entity": "contrastive learning models", "output": "graphs", "neg_sample": ["contrastive learning models is done by using OtherScientificTerm", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "Next , we leverage these graphs in different contrastive learning models with Max - Margin and InfoNCE losses .", "forward": false, "src_ids": "2022.acl-long.85_3217"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "entity": "pre - trained language models", "output": "graphs", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "We first show that with limited supervision , pre - trained language models often generate graphs that either violate these constraints or are semantically incoherent .", "forward": true, "src_ids": "2022.acl-long.85_3218"}
+{"input": "pre - trained language models is done by using OtherScientificTerm| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks . however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs . unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g. , generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts . in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs .", "entity": "pre - trained language models", "output": "limited supervision", "neg_sample": ["pre - trained language models is done by using OtherScientificTerm", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs .", "unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g.", ", generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts .", "in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "We first show that with limited supervision , pre - trained language models often generate graphs that either violate these constraints or are semantically incoherent .", "forward": false, "src_ids": "2022.acl-long.85_3219"}
+{"input": "graphs is used for Method| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks . however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs . unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g. , generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts . in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs .", "entity": "graphs", "output": "contrastive learning models", "neg_sample": ["graphs is used for Method", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs .", "unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g.", ", generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts .", "in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "Next , we leverage these graphs in different contrastive learning models with Max - Margin and InfoNCE losses .", "forward": true, "src_ids": "2022.acl-long.85_3220"}
+{"input": "human errors is used for Method| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks . however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs . unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g. , generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts . in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs .", "entity": "human errors", "output": "contrastive learning models", "neg_sample": ["human errors is used for Method", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs .", "unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g.", ", generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts .", "in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "Lastly , we show that human errors are the best negatives for contrastive learning and also that automatically generating more such human - like negative graphs can lead tofurther improvements .", "forward": true, "src_ids": "2022.acl-long.85_3221"}
+{"input": "contrastive learning models is done by using OtherScientificTerm| context: pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks . however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs . unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g. , generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts . in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs .", "entity": "contrastive learning models", "output": "human errors", "neg_sample": ["contrastive learning models is done by using OtherScientificTerm", "pre - trained sequence - to - sequence language models have led to widespread success in many natural language generation tasks .", "however , there has been relatively less work on analyzing their ability to generate structured outputs such as graphs .", "unlike natural language , graphs have distinct structural and semantic properties in the context of a downstream nlp task , e.g.", ", generating a graph that is connected and acyclic can be attributed to its structural constraints , while the semantics of a graph can refer to how meaningfully an edge represents the relation between two node concepts .", "in this work , we study pre - trained language models that generate explanation graphs in an end - to - end manner and analyze their ability to learn the structural constraints and semantics of such graphs ."], "relation": "used for", "id": "2022.acl-long.85", "year": 2022, "rel_sent": "Lastly , we show that human errors are the best negatives for contrastive learning and also that automatically generating more such human - like negative graphs can lead tofurther improvements .", "forward": false, "src_ids": "2022.acl-long.85_3222"}
+{"input": "referring expression comprehension is done by using OtherScientificTerm| context: training a referring expression comprehension ( rec ) model for a new visual domain requires collecting referring expressions , and potentially corresponding bounding boxes , for images in the domain . while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec.", "entity": "referring expression comprehension", "output": "zero - shot baselines", "neg_sample": ["referring expression comprehension is done by using OtherScientificTerm", "training a referring expression comprehension ( rec ) model for a new visual domain requires collecting referring expressions , and potentially corresponding bounding boxes , for images in the domain .", "while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec."], "relation": "used for", "id": "2022.acl-long.357", "year": 2022, "rel_sent": "ReCLIP : A Strong Zero - Shot Baseline for Referring Expression Comprehension.", "forward": false, "src_ids": "2022.acl-long.357_3223"}
+{"input": "clip is done by using OtherScientificTerm| context: training a referring expression comprehension ( rec ) model for a new visual domain requires collecting referring expressions , and potentially corresponding bounding boxes , for images in the domain . while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec.", "entity": "clip", "output": "zero - shot baselines", "neg_sample": ["clip is done by using OtherScientificTerm", "training a referring expression comprehension ( rec ) model for a new visual domain requires collecting referring expressions , and potentially corresponding bounding boxes , for images in the domain .", "while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec."], "relation": "used for", "id": "2022.acl-long.357", "year": 2022, "rel_sent": "We present ReCLIP , a simple but strong zero - shot baseline that repurposes CLIP , a state - of - the - art large - scale model , for ReC.", "forward": false, "src_ids": "2022.acl-long.357_3224"}
+{"input": "zero - shot baselines is used for Task| context: while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec.", "entity": "zero - shot baselines", "output": "referring expression comprehension", "neg_sample": ["zero - shot baselines is used for Task", "while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec."], "relation": "used for", "id": "2022.acl-long.357", "year": 2022, "rel_sent": "ReCLIP : A Strong Zero - Shot Baseline for Referring Expression Comprehension.", "forward": true, "src_ids": "2022.acl-long.357_3225"}
+{"input": "zero - shot baselines is used for Method| context: training a referring expression comprehension ( rec ) model for a new visual domain requires collecting referring expressions , and potentially corresponding bounding boxes , for images in the domain . while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec.", "entity": "zero - shot baselines", "output": "clip", "neg_sample": ["zero - shot baselines is used for Method", "training a referring expression comprehension ( rec ) model for a new visual domain requires collecting referring expressions , and potentially corresponding bounding boxes , for images in the domain .", "while large - scale pre - trained models are useful for image classification across domains , it remains unclear if they can be applied in a zero - shot manner to more complex tasks like rec."], "relation": "used for", "id": "2022.acl-long.357", "year": 2022, "rel_sent": "We present ReCLIP , a simple but strong zero - shot baseline that repurposes CLIP , a state - of - the - art large - scale model , for ReC.", "forward": true, "src_ids": "2022.acl-long.357_3226"}
+{"input": "education is done by using Task| context: we show that a significant portion of errors in such systems arise from asking irrelevant or un - interpretable questions and that such errors can be ameliorated by providing summarized input .", "entity": "education", "output": "answer - unaware question generation", "neg_sample": ["education is done by using Task", "we show that a significant portion of errors in such systems arise from asking irrelevant or un - interpretable questions and that such errors can be ameliorated by providing summarized input ."], "relation": "used for", "id": "2022.findings-acl.151", "year": 2022, "rel_sent": "A Feasibility Study of Answer - Unaware Question Generation for Education.", "forward": false, "src_ids": "2022.findings-acl.151_3227"}
+{"input": "answer - unaware question generation is used for Material| context: we conduct a feasibility study into the applicability of answer - unaware question generation models to textbook passages . we show that a significant portion of errors in such systems arise from asking irrelevant or un - interpretable questions and that such errors can be ameliorated by providing summarized input .", "entity": "answer - unaware question generation", "output": "education", "neg_sample": ["answer - unaware question generation is used for Material", "we conduct a feasibility study into the applicability of answer - unaware question generation models to textbook passages .", "we show that a significant portion of errors in such systems arise from asking irrelevant or un - interpretable questions and that such errors can be ameliorated by providing summarized input ."], "relation": "used for", "id": "2022.findings-acl.151", "year": 2022, "rel_sent": "A Feasibility Study of Answer - Unaware Question Generation for Education.", "forward": true, "src_ids": "2022.findings-acl.151_3228"}
+{"input": "open - domain question answering is done by using Method| context: open - domain question answering is a challenging task with a wide variety of practical applications . existing modern approaches mostly follow a standard two - stage paradigm : retriever then reader .", "entity": "open - domain question answering", "output": "copy - augmented generative approach", "neg_sample": ["open - domain question answering is done by using Method", "open - domain question answering is a challenging task with a wide variety of practical applications .", "existing modern approaches mostly follow a standard two - stage paradigm : retriever then reader ."], "relation": "used for", "id": "2022.acl-short.47", "year": 2022, "rel_sent": "A Copy - Augmented Generative Model for Open - Domain Question Answering.", "forward": false, "src_ids": "2022.acl-short.47_3229"}
+{"input": "copy - augmented generative approach is used for Task| context: existing modern approaches mostly follow a standard two - stage paradigm : retriever then reader .", "entity": "copy - augmented generative approach", "output": "open - domain question answering", "neg_sample": ["copy - augmented generative approach is used for Task", "existing modern approaches mostly follow a standard two - stage paradigm : retriever then reader ."], "relation": "used for", "id": "2022.acl-short.47", "year": 2022, "rel_sent": "A Copy - Augmented Generative Model for Open - Domain Question Answering.", "forward": true, "src_ids": "2022.acl-short.47_3230"}
+{"input": "pruning criteria is done by using Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "pruning criteria", "output": "distillation", "neg_sample": ["pruning criteria is done by using Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "Unlike previous approaches that treat distillation and pruning separately , we use distillation to inform the pruning criteria , without requiring a separate student network as in knowledge distillation .", "forward": false, "src_ids": "2022.findings-acl.267_3231"}
+{"input": "distillation is used for OtherScientificTerm| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "distillation", "output": "pruning criteria", "neg_sample": ["distillation is used for OtherScientificTerm", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "Unlike previous approaches that treat distillation and pruning separately , we use distillation to inform the pruning criteria , without requiring a separate student network as in knowledge distillation .", "forward": true, "src_ids": "2022.findings-acl.267_3232"}
+{"input": "knowledge distillation is done by using Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "knowledge distillation", "output": "student network", "neg_sample": ["knowledge distillation is done by using Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "Unlike previous approaches that treat distillation and pruning separately , we use distillation to inform the pruning criteria , without requiring a separate student network as in knowledge distillation .", "forward": false, "src_ids": "2022.findings-acl.267_3233"}
+{"input": "student network is used for Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "student network", "output": "knowledge distillation", "neg_sample": ["student network is used for Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "Unlike previous approaches that treat distillation and pruning separately , we use distillation to inform the pruning criteria , without requiring a separate student network as in knowledge distillation .", "forward": true, "src_ids": "2022.findings-acl.267_3234"}
+{"input": "sparse solutions is done by using Metric| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "sparse solutions", "output": "cross - correlation objective", "neg_sample": ["sparse solutions is done by using Metric", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "We show that the proposed cross - correlation objective for self - distilled pruning implicitly encourages sparse solutions , naturally complementing magnitude - based pruning criteria .", "forward": false, "src_ids": "2022.findings-acl.267_3235"}
+{"input": "self - distilled pruning is done by using Metric| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "self - distilled pruning", "output": "cross - correlation objective", "neg_sample": ["self - distilled pruning is done by using Metric", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "We show that the proposed cross - correlation objective for self - distilled pruning implicitly encourages sparse solutions , naturally complementing magnitude - based pruning criteria .", "forward": false, "src_ids": "2022.findings-acl.267_3236"}
+{"input": "mono- and cross - lingual language model is done by using Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "mono- and cross - lingual language model", "output": "self - distilled pruning", "neg_sample": ["mono- and cross - lingual language model is done by using Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "Experiments on the GLUE and XGLUE benchmarks show that self - distilled pruning increases mono- and cross - lingual language model performance .", "forward": false, "src_ids": "2022.findings-acl.267_3237"}
+{"input": "cross - correlation objective is used for Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "cross - correlation objective", "output": "self - distilled pruning", "neg_sample": ["cross - correlation objective is used for Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "We show that the proposed cross - correlation objective for self - distilled pruning implicitly encourages sparse solutions , naturally complementing magnitude - based pruning criteria .", "forward": true, "src_ids": "2022.findings-acl.267_3238"}
+{"input": "cross - correlation objective is used for OtherScientificTerm| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "cross - correlation objective", "output": "sparse solutions", "neg_sample": ["cross - correlation objective is used for OtherScientificTerm", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "We show that the proposed cross - correlation objective for self - distilled pruning implicitly encourages sparse solutions , naturally complementing magnitude - based pruning criteria .", "forward": true, "src_ids": "2022.findings-acl.267_3239"}
+{"input": "self - distilled pruning is used for Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "self - distilled pruning", "output": "mono- and cross - lingual language model", "neg_sample": ["self - distilled pruning is used for Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "Experiments on the GLUE and XGLUE benchmarks show that self - distilled pruning increases mono- and cross - lingual language model performance .", "forward": true, "src_ids": "2022.findings-acl.267_3240"}
+{"input": "class separability is done by using Method| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "class separability", "output": "self - distillation", "neg_sample": ["class separability is done by using Method", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "We also observe that self - distillation ( 1 ) maximizes class separability , ( 2 ) increases the signal - to - noise ratio , and ( 3 ) converges faster after pruning steps , providing further insights into why self - distilled pruning improves generalization .", "forward": false, "src_ids": "2022.findings-acl.267_3241"}
+{"input": "self - distillation is used for OtherScientificTerm| context: pruning aims to reduce the number of parameters while maintaining performance close to the original network .", "entity": "self - distillation", "output": "class separability", "neg_sample": ["self - distillation is used for OtherScientificTerm", "pruning aims to reduce the number of parameters while maintaining performance close to the original network ."], "relation": "used for", "id": "2022.findings-acl.267", "year": 2022, "rel_sent": "We also observe that self - distillation ( 1 ) maximizes class separability , ( 2 ) increases the signal - to - noise ratio , and ( 3 ) converges faster after pruning steps , providing further insights into why self - distilled pruning improves generalization .", "forward": true, "src_ids": "2022.findings-acl.267_3242"}
+{"input": "- x is used for Task| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "- x", "output": "privacy - preserving inference", "neg_sample": ["- x is used for Task", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "In this work , we introduce THE - X , an approximation approach for transformers , which enables privacy - preserving inference of pre - trained models developed by popular frameworks .", "forward": true, "src_ids": "2022.findings-acl.277_3243"}
+{"input": "approximation approach is used for Task| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "approximation approach", "output": "privacy - preserving inference", "neg_sample": ["approximation approach is used for Task", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "In this work , we introduce THE - X , an approximation approach for transformers , which enables privacy - preserving inference of pre - trained models developed by popular frameworks .", "forward": true, "src_ids": "2022.findings-acl.277_3244"}
+{"input": "privacy - preserving inference is done by using Method| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "privacy - preserving inference", "output": "- x", "neg_sample": ["privacy - preserving inference is done by using Method", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "In this work , we introduce THE - X , an approximation approach for transformers , which enables privacy - preserving inference of pre - trained models developed by popular frameworks .", "forward": false, "src_ids": "2022.findings-acl.277_3245"}
+{"input": "transformer inference is done by using Method| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "transformer inference", "output": "- x", "neg_sample": ["transformer inference is done by using Method", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "Experiments reveal our proposed THE - X can enable transformer inference on encrypted data for different downstream tasks , all with negligible performance drop but enjoying the theory - guaranteed privacy - preserving advantage .", "forward": false, "src_ids": "2022.findings-acl.277_3246"}
+{"input": "downstream tasks is done by using Method| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "downstream tasks", "output": "- x", "neg_sample": ["downstream tasks is done by using Method", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "Experiments reveal our proposed THE - X can enable transformer inference on encrypted data for different downstream tasks , all with negligible performance drop but enjoying the theory - guaranteed privacy - preserving advantage .", "forward": false, "src_ids": "2022.findings-acl.277_3247"}
+{"input": "transformers is done by using Method| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "transformers", "output": "approximation approach", "neg_sample": ["transformers is done by using Method", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "In this work , we introduce THE - X , an approximation approach for transformers , which enables privacy - preserving inference of pre - trained models developed by popular frameworks .", "forward": false, "src_ids": "2022.findings-acl.277_3248"}
+{"input": "approximation approach is used for Method| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "approximation approach", "output": "transformers", "neg_sample": ["approximation approach is used for Method", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "In this work , we introduce THE - X , an approximation approach for transformers , which enables privacy - preserving inference of pre - trained models developed by popular frameworks .", "forward": true, "src_ids": "2022.findings-acl.277_3249"}
+{"input": "downstream tasks is done by using Task| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "downstream tasks", "output": "transformer inference", "neg_sample": ["downstream tasks is done by using Task", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "Experiments reveal our proposed THE - X can enable transformer inference on encrypted data for different downstream tasks , all with negligible performance drop but enjoying the theory - guaranteed privacy - preserving advantage .", "forward": false, "src_ids": "2022.findings-acl.277_3250"}
+{"input": "- x is used for Task| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "- x", "output": "transformer inference", "neg_sample": ["- x is used for Task", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "Experiments reveal our proposed THE - X can enable transformer inference on encrypted data for different downstream tasks , all with negligible performance drop but enjoying the theory - guaranteed privacy - preserving advantage .", "forward": true, "src_ids": "2022.findings-acl.277_3251"}
+{"input": "transformer inference is used for Task| context: as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g. , search history , medical record , bank account ) . privacy - preserving inference of transformer models is on the demand of cloud service users . to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) . however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet .", "entity": "transformer inference", "output": "downstream tasks", "neg_sample": ["transformer inference is used for Task", "as more and more pre - trained language models adopt on - cloud deployment , the privacy issues grow quickly , mainly for the exposure of plain - text user data ( e.g.", ", search history , medical record , bank account ) .", "privacy - preserving inference of transformer models is on the demand of cloud service users .", "to protect privacy , it is an attractive choice to compute only with ciphertext in homomorphic encryption ( he ) .", "however , enabling pre - trained models inference on ciphertext data is difficult due to the complex computations in transformer blocks , which are not supported by current he tools yet ."], "relation": "used for", "id": "2022.findings-acl.277", "year": 2022, "rel_sent": "Experiments reveal our proposed THE - X can enable transformer inference on encrypted data for different downstream tasks , all with negligible performance drop but enjoying the theory - guaranteed privacy - preserving advantage .", "forward": true, "src_ids": "2022.findings-acl.277_3252"}
+{"input": "sentences representation is done by using Method| context: neural networks are widely used in various nlp tasks for their remarkable performance . however , the complexity makes them difficult to interpret , i.e. , they are not guaranteed right for the right reason . besides the complexity , we reveal that the model pathology - the inconsistency between word saliency and model confidence , further hurts the interpretability .", "entity": "sentences representation", "output": "pathological contrastive training ( pct ) framework", "neg_sample": ["sentences representation is done by using Method", "neural networks are widely used in various nlp tasks for their remarkable performance .", "however , the complexity makes them difficult to interpret , i.e.", ", they are not guaranteed right for the right reason .", "besides the complexity , we reveal that the model pathology - the inconsistency between word saliency and model confidence , further hurts the interpretability ."], "relation": "used for", "id": "2022.findings-acl.175", "year": 2022, "rel_sent": "In this paper , to mitigate the pathology and obtain more interpretable models , we propose Pathological Contrastive Training ( PCT ) framework , which adopts contrastive learning and saliency - based samples augmentation to calibrate the sentences representation .", "forward": false, "src_ids": "2022.findings-acl.175_3253"}
+{"input": "pathological contrastive training ( pct ) framework is used for Task| context: neural networks are widely used in various nlp tasks for their remarkable performance . however , the complexity makes them difficult to interpret , i.e. , they are not guaranteed right for the right reason . besides the complexity , we reveal that the model pathology - the inconsistency between word saliency and model confidence , further hurts the interpretability .", "entity": "pathological contrastive training ( pct ) framework", "output": "sentences representation", "neg_sample": ["pathological contrastive training ( pct ) framework is used for Task", "neural networks are widely used in various nlp tasks for their remarkable performance .", "however , the complexity makes them difficult to interpret , i.e.", ", they are not guaranteed right for the right reason .", "besides the complexity , we reveal that the model pathology - the inconsistency between word saliency and model confidence , further hurts the interpretability ."], "relation": "used for", "id": "2022.findings-acl.175", "year": 2022, "rel_sent": "In this paper , to mitigate the pathology and obtain more interpretable models , we propose Pathological Contrastive Training ( PCT ) framework , which adopts contrastive learning and saliency - based samples augmentation to calibrate the sentences representation .", "forward": true, "src_ids": "2022.findings-acl.175_3254"}
+{"input": "saliency - based samples augmentation is used for Task| context: neural networks are widely used in various nlp tasks for their remarkable performance . however , the complexity makes them difficult to interpret , i.e. , they are not guaranteed right for the right reason . besides the complexity , we reveal that the model pathology - the inconsistency between word saliency and model confidence , further hurts the interpretability .", "entity": "saliency - based samples augmentation", "output": "sentences representation", "neg_sample": ["saliency - based samples augmentation is used for Task", "neural networks are widely used in various nlp tasks for their remarkable performance .", "however , the complexity makes them difficult to interpret , i.e.", ", they are not guaranteed right for the right reason .", "besides the complexity , we reveal that the model pathology - the inconsistency between word saliency and model confidence , further hurts the interpretability ."], "relation": "used for", "id": "2022.findings-acl.175", "year": 2022, "rel_sent": "In this paper , to mitigate the pathology and obtain more interpretable models , we propose Pathological Contrastive Training ( PCT ) framework , which adopts contrastive learning and saliency - based samples augmentation to calibrate the sentences representation .", "forward": true, "src_ids": "2022.findings-acl.175_3255"}
+{"input": "simultaneous speech translation task is done by using Method| context: in this paper , we describe our submission to the simultaneous speech translation at iwslt 2022 .", "entity": "simultaneous speech translation task", "output": "cuni - kit system", "neg_sample": ["simultaneous speech translation task is done by using Method", "in this paper , we describe our submission to the simultaneous speech translation at iwslt 2022 ."], "relation": "used for", "id": "2022.iwslt-1.24", "year": 2022, "rel_sent": "CUNI - KIT System for Simultaneous Speech Translation Task at IWSLT 2022.", "forward": false, "src_ids": "2022.iwslt-1.24_3256"}
+{"input": "extractive opinion summarization is done by using Method| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "extractive opinion summarization", "output": "semantic autoencoder ( semae )", "neg_sample": ["extractive opinion summarization is done by using Method", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "We present Semantic Autoencoder ( SemAE ) to perform extractive opinion summarization in an unsupervised manner .", "forward": false, "src_ids": "2022.acl-long.86_3257"}
+{"input": "semantic autoencoder ( semae ) is used for Task| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "semantic autoencoder ( semae )", "output": "extractive opinion summarization", "neg_sample": ["semantic autoencoder ( semae ) is used for Task", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "We present Semantic Autoencoder ( SemAE ) to perform extractive opinion summarization in an unsupervised manner .", "forward": true, "src_ids": "2022.acl-long.86_3258"}
+{"input": "controllable summarization is done by using Method| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "controllable summarization", "output": "semantic autoencoder", "neg_sample": ["controllable summarization is done by using Method", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE is also able to perform controllable summarization to generate aspect - specific summaries using only a few samples .", "forward": false, "src_ids": "2022.acl-long.86_3259"}
+{"input": "aspect - specific summaries is done by using Method| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "aspect - specific summaries", "output": "semantic autoencoder", "neg_sample": ["aspect - specific summaries is done by using Method", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE is also able to perform controllable summarization to generate aspect - specific summaries using only a few samples .", "forward": false, "src_ids": "2022.acl-long.86_3260"}
+{"input": "semantic information is done by using Method| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "semantic information", "output": "dictionary learning", "neg_sample": ["semantic information is done by using Method", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE uses dictionary learning to implicitly capture semantic information from the review text and learns a latent representation of each sentence over semantic units .", "forward": false, "src_ids": "2022.acl-long.86_3261"}
+{"input": "dictionary learning is used for OtherScientificTerm| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "dictionary learning", "output": "semantic information", "neg_sample": ["dictionary learning is used for OtherScientificTerm", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE uses dictionary learning to implicitly capture semantic information from the review text and learns a latent representation of each sentence over semantic units .", "forward": true, "src_ids": "2022.acl-long.86_3262"}
+{"input": "representative opinions is done by using Method| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "representative opinions", "output": "extractive summarization algorithm", "neg_sample": ["representative opinions is done by using Method", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "Our extractive summarization algorithm leverages the representations to identify representative opinions among hundreds of reviews .", "forward": false, "src_ids": "2022.acl-long.86_3263"}
+{"input": "extractive summarization algorithm is used for OtherScientificTerm| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "extractive summarization algorithm", "output": "representative opinions", "neg_sample": ["extractive summarization algorithm is used for OtherScientificTerm", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "Our extractive summarization algorithm leverages the representations to identify representative opinions among hundreds of reviews .", "forward": true, "src_ids": "2022.acl-long.86_3264"}
+{"input": "aspect - specific summaries is done by using Task| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "aspect - specific summaries", "output": "controllable summarization", "neg_sample": ["aspect - specific summaries is done by using Task", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE is also able to perform controllable summarization to generate aspect - specific summaries using only a few samples .", "forward": false, "src_ids": "2022.acl-long.86_3265"}
+{"input": "semantic autoencoder is used for Task| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "semantic autoencoder", "output": "controllable summarization", "neg_sample": ["semantic autoencoder is used for Task", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE is also able to perform controllable summarization to generate aspect - specific summaries using only a few samples .", "forward": true, "src_ids": "2022.acl-long.86_3266"}
+{"input": "controllable summarization is used for OtherScientificTerm| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "controllable summarization", "output": "aspect - specific summaries", "neg_sample": ["controllable summarization is used for OtherScientificTerm", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE is also able to perform controllable summarization to generate aspect - specific summaries using only a few samples .", "forward": true, "src_ids": "2022.acl-long.86_3267"}
+{"input": "semantic autoencoder is used for OtherScientificTerm| context: opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews .", "entity": "semantic autoencoder", "output": "aspect - specific summaries", "neg_sample": ["semantic autoencoder is used for OtherScientificTerm", "opinion summarization is the task of automatically generating summaries that encapsulate information expressed in multiple user reviews ."], "relation": "used for", "id": "2022.acl-long.86", "year": 2022, "rel_sent": "SemAE is also able to perform controllable summarization to generate aspect - specific summaries using only a few samples .", "forward": true, "src_ids": "2022.acl-long.86_3268"}
+{"input": "writing support tools is done by using OtherScientificTerm| context: improvements in language technology have led to an increasing interest in writing support tools .", "entity": "writing support tools", "output": "design space", "neg_sample": ["writing support tools is done by using OtherScientificTerm", "improvements in language technology have led to an increasing interest in writing support tools ."], "relation": "used for", "id": "2022.in2writing-1.2", "year": 2022, "rel_sent": "A Design Space for Writing Support Tools Using a Cognitive Process Model of Writing.", "forward": false, "src_ids": "2022.in2writing-1.2_3269"}
+{"input": "source code summarization is done by using OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "source code summarization", "output": "hierarchical syntax structure of code", "neg_sample": ["source code summarization is done by using OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Modeling Hierarchical Syntax Structure with Triplet Position for Source Code Summarization.", "forward": false, "src_ids": "2022.acl-long.37_3270"}
+{"input": "codescribe is used for OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "codescribe", "output": "hierarchical syntax structure of code", "neg_sample": ["codescribe is used for OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In this paper , we propose CODESCRIBE to model the hierarchical syntax structure of code by introducing a novel triplet position for code summarization .", "forward": true, "src_ids": "2022.acl-long.37_3271"}
+{"input": "automatic code summarization is done by using OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "automatic code summarization", "output": "triplet position", "neg_sample": ["automatic code summarization is done by using OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In this paper , we propose CODESCRIBE to model the hierarchical syntax structure of code by introducing a novel triplet position for code summarization .", "forward": false, "src_ids": "2022.acl-long.37_3272"}
+{"input": "hierarchical syntax structure of code is used for Task| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "hierarchical syntax structure of code", "output": "source code summarization", "neg_sample": ["hierarchical syntax structure of code is used for Task", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Modeling Hierarchical Syntax Structure with Triplet Position for Source Code Summarization.", "forward": true, "src_ids": "2022.acl-long.37_3273"}
+{"input": "triplet position is used for Task| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "triplet position", "output": "source code summarization", "neg_sample": ["triplet position is used for Task", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Modeling Hierarchical Syntax Structure with Triplet Position for Source Code Summarization.", "forward": true, "src_ids": "2022.acl-long.37_3274"}
+{"input": "triplet position is used for Task| context: our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "triplet position", "output": "automatic code summarization", "neg_sample": ["triplet position is used for Task", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In this paper , we propose CODESCRIBE to model the hierarchical syntax structure of code by introducing a novel triplet position for code summarization .", "forward": true, "src_ids": "2022.acl-long.37_3275"}
+{"input": "hierarchical syntax structure of code is done by using Method| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "hierarchical syntax structure of code", "output": "codescribe", "neg_sample": ["hierarchical syntax structure of code is done by using Method", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In this paper , we propose CODESCRIBE to model the hierarchical syntax structure of code by introducing a novel triplet position for code summarization .", "forward": false, "src_ids": "2022.acl-long.37_3276"}
+{"input": "structural and sequential information of code is done by using Method| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "structural and sequential information of code", "output": "codescribe", "neg_sample": ["structural and sequential information of code is done by using Method", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Specifically , CODESCRIBE leverages the graph neural network and Transformer to preserve the structural and sequential information of code , respectively .", "forward": false, "src_ids": "2022.acl-long.37_3277"}
+{"input": "transformer is used for OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "transformer", "output": "structural and sequential information of code", "neg_sample": ["transformer is used for OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Specifically , CODESCRIBE leverages the graph neural network and Transformer to preserve the structural and sequential information of code , respectively .", "forward": true, "src_ids": "2022.acl-long.37_3278"}
+{"input": "graph neural network is used for OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "graph neural network", "output": "structural and sequential information of code", "neg_sample": ["graph neural network is used for OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Specifically , CODESCRIBE leverages the graph neural network and Transformer to preserve the structural and sequential information of code , respectively .", "forward": true, "src_ids": "2022.acl-long.37_3279"}
+{"input": "codescribe is used for OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "codescribe", "output": "structural and sequential information of code", "neg_sample": ["codescribe is used for OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "Specifically , CODESCRIBE leverages the graph neural network and Transformer to preserve the structural and sequential information of code , respectively .", "forward": true, "src_ids": "2022.acl-long.37_3280"}
+{"input": "summary generation is done by using Method| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "summary generation", "output": "pointer - generator network", "neg_sample": ["summary generation is done by using Method", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In addition , we propose a pointer - generator network that pays attention to both the structure and sequential tokens of code for a better summary generation .", "forward": false, "src_ids": "2022.acl-long.37_3281"}
+{"input": "structure and sequential tokens of code is done by using Method| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "structure and sequential tokens of code", "output": "pointer - generator network", "neg_sample": ["structure and sequential tokens of code is done by using Method", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In addition , we propose a pointer - generator network that pays attention to both the structure and sequential tokens of code for a better summary generation .", "forward": false, "src_ids": "2022.acl-long.37_3282"}
+{"input": "pointer - generator network is used for OtherScientificTerm| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "pointer - generator network", "output": "structure and sequential tokens of code", "neg_sample": ["pointer - generator network is used for OtherScientificTerm", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In addition , we propose a pointer - generator network that pays attention to both the structure and sequential tokens of code for a better summary generation .", "forward": true, "src_ids": "2022.acl-long.37_3283"}
+{"input": "pointer - generator network is used for Task| context: automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance . our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches . one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art . however , the hierarchical structures of asts have not been well explored .", "entity": "pointer - generator network", "output": "summary generation", "neg_sample": ["pointer - generator network is used for Task", "automatic code summarization , which aims to describe the source code in natural language , has become an essential task in software maintenance .", "our fellow researchers have attempted to achieve such a purpose through various machine learning - based approaches .", "one key challenge keeping these approaches from being practical lies in the lacking of retaining the semantic structure of source code , which has unfortunately been overlooked by the state - of - the - art .", "however , the hierarchical structures of asts have not been well explored ."], "relation": "used for", "id": "2022.acl-long.37", "year": 2022, "rel_sent": "In addition , we propose a pointer - generator network that pays attention to both the structure and sequential tokens of code for a better summary generation .", "forward": true, "src_ids": "2022.acl-long.37_3284"}
+{"input": "bli task is done by using Method| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "bli task", "output": "two - stage contrastive learning framework", "neg_sample": ["bli task is done by using Method", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "In this work , we propose a robust and effective two - stage contrastive learning framework for the BLI task .", "forward": false, "src_ids": "2022.acl-long.299_3285"}
+{"input": "two - stage contrastive learning framework is used for Task| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "two - stage contrastive learning framework", "output": "bli task", "neg_sample": ["two - stage contrastive learning framework is used for Task", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "In this work , we propose a robust and effective two - stage contrastive learning framework for the BLI task .", "forward": true, "src_ids": "2022.acl-long.299_3286"}
+{"input": "self - learning procedure is used for OtherScientificTerm| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "self - learning procedure", "output": "cross - lingual linear maps", "neg_sample": ["self - learning procedure is used for OtherScientificTerm", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "At Stage C1 , we propose to refine standard cross - lingual linear maps between static word embeddings ( WEs ) via a contrastive learning objective ; we also show how to integrate it into the self - learning procedure for even more refined cross - lingual maps .", "forward": true, "src_ids": "2022.acl-long.299_3287"}
+{"input": "cross - lingual linear maps is done by using Method| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "cross - lingual linear maps", "output": "self - learning procedure", "neg_sample": ["cross - lingual linear maps is done by using Method", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "At Stage C1 , we propose to refine standard cross - lingual linear maps between static word embeddings ( WEs ) via a contrastive learning objective ; we also show how to integrate it into the self - learning procedure for even more refined cross - lingual maps .", "forward": false, "src_ids": "2022.acl-long.299_3288"}
+{"input": "word translation capability is done by using Method| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "word translation capability", "output": "bli - oriented contrastive fine - tuning", "neg_sample": ["word translation capability is done by using Method", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "In Stage C2 , we conduct BLI - oriented contrastive fine - tuning of mBERT , unlocking its word translation capability .", "forward": false, "src_ids": "2022.acl-long.299_3289"}
+{"input": "mbert is done by using Method| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "mbert", "output": "bli - oriented contrastive fine - tuning", "neg_sample": ["mbert is done by using Method", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "In Stage C2 , we conduct BLI - oriented contrastive fine - tuning of mBERT , unlocking its word translation capability .", "forward": false, "src_ids": "2022.acl-long.299_3290"}
+{"input": "bli - oriented contrastive fine - tuning is used for Method| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "bli - oriented contrastive fine - tuning", "output": "mbert", "neg_sample": ["bli - oriented contrastive fine - tuning is used for Method", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "In Stage C2 , we conduct BLI - oriented contrastive fine - tuning of mBERT , unlocking its word translation capability .", "forward": true, "src_ids": "2022.acl-long.299_3291"}
+{"input": "bli - oriented contrastive fine - tuning is used for OtherScientificTerm| context: word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages .", "entity": "bli - oriented contrastive fine - tuning", "output": "word translation capability", "neg_sample": ["bli - oriented contrastive fine - tuning is used for OtherScientificTerm", "word translation or bilingual lexicon induction ( bli ) is a key cross - lingual task , aiming to bridge the lexical gap between different languages ."], "relation": "used for", "id": "2022.acl-long.299", "year": 2022, "rel_sent": "In Stage C2 , we conduct BLI - oriented contrastive fine - tuning of mBERT , unlocking its word translation capability .", "forward": true, "src_ids": "2022.acl-long.299_3292"}
+{"input": "educational experts is used for Material| context: question answering ( qa ) is a fundamental means tofacilitate assessment and training of narrative comprehension skills for both machines and young children , yet there is scarcity of high - quality qa datasets carefully designed to serve this purpose . in particular , existing datasets rarely distinguish fine - grained reading skills , such as the understanding of varying narrative elements .", "entity": "educational experts", "output": "fairytaleqa", "neg_sample": ["educational experts is used for Material", "question answering ( qa ) is a fundamental means tofacilitate assessment and training of narrative comprehension skills for both machines and young children , yet there is scarcity of high - quality qa datasets carefully designed to serve this purpose .", "in particular , existing datasets rarely distinguish fine - grained reading skills , such as the understanding of varying narrative elements ."], "relation": "used for", "id": "2022.acl-long.34", "year": 2022, "rel_sent": "Generated by educational experts based on an evidence - based theoretical framework , FairytaleQA consists of 10,580 explicit and implicit questions derived from 278 children - friendly stories , covering seven types of narrative elements or relations .", "forward": true, "src_ids": "2022.acl-long.34_3293"}
+{"input": "fairytaleqa is done by using OtherScientificTerm| context: question answering ( qa ) is a fundamental means tofacilitate assessment and training of narrative comprehension skills for both machines and young children , yet there is scarcity of high - quality qa datasets carefully designed to serve this purpose . in particular , existing datasets rarely distinguish fine - grained reading skills , such as the understanding of varying narrative elements .", "entity": "fairytaleqa", "output": "educational experts", "neg_sample": ["fairytaleqa is done by using OtherScientificTerm", "question answering ( qa ) is a fundamental means tofacilitate assessment and training of narrative comprehension skills for both machines and young children , yet there is scarcity of high - quality qa datasets carefully designed to serve this purpose .", "in particular , existing datasets rarely distinguish fine - grained reading skills , such as the understanding of varying narrative elements ."], "relation": "used for", "id": "2022.acl-long.34", "year": 2022, "rel_sent": "Generated by educational experts based on an evidence - based theoretical framework , FairytaleQA consists of 10,580 explicit and implicit questions derived from 278 children - friendly stories , covering seven types of narrative elements or relations .", "forward": false, "src_ids": "2022.acl-long.34_3294"}
+{"input": "edu - level pre - trained edu representation is done by using Method| context: pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing . current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e.", "entity": "edu - level pre - trained edu representation", "output": "light bi - gram edu modification", "neg_sample": ["edu - level pre - trained edu representation is done by using Method", "pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing .", "current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e."], "relation": "used for", "id": "2022.acl-long.294", "year": 2022, "rel_sent": "element discourse unit ( EDU).To this end , we propose a second - stage EDU - level pre - training approach in this work , which presents two novel tasks to learn effective EDU representations continually based on well pre - trained language models . Concretely , the two tasks are ( 1 ) next EDU prediction ( NEP ) and ( 2 ) discourse marker prediction ( DMP).We take a state - of - the - art transition - based neural parser as baseline , and adopt it with a light bi - gram EDU modification to effectively explore the EDU - level pre - trained EDU representation . Experimental results on a benckmark dataset show that our method is highly effective , leading a 2.1 - point improvement in F1 - score . All codes and pre - trained models will be released publicly tofacilitate future studies .", "forward": false, "src_ids": "2022.acl-long.294_3295"}
+{"input": "light bi - gram edu modification is used for Method| context: pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing . current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e.", "entity": "light bi - gram edu modification", "output": "edu - level pre - trained edu representation", "neg_sample": ["light bi - gram edu modification is used for Method", "pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing .", "current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e."], "relation": "used for", "id": "2022.acl-long.294", "year": 2022, "rel_sent": "element discourse unit ( EDU).To this end , we propose a second - stage EDU - level pre - training approach in this work , which presents two novel tasks to learn effective EDU representations continually based on well pre - trained language models . Concretely , the two tasks are ( 1 ) next EDU prediction ( NEP ) and ( 2 ) discourse marker prediction ( DMP).We take a state - of - the - art transition - based neural parser as baseline , and adopt it with a light bi - gram EDU modification to effectively explore the EDU - level pre - trained EDU representation . Experimental results on a benckmark dataset show that our method is highly effective , leading a 2.1 - point improvement in F1 - score . All codes and pre - trained models will be released publicly tofacilitate future studies .", "forward": true, "src_ids": "2022.acl-long.294_3296"}
+{"input": "dependency parsing is done by using OtherScientificTerm| context: the experiments include both zero - shot settings as well as multilingual models . previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "dependency parsing", "output": "zero - shot transfers", "neg_sample": ["dependency parsing is done by using OtherScientificTerm", "the experiments include both zero - shot settings as well as multilingual models .", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "Evaluating zero - shot transfers and multilingual models for dependency parsing and POS tagging within the low - resource language family Tupian.", "forward": false, "src_ids": "2022.acl-srw.1_3297"}
+{"input": "pos tagging is done by using OtherScientificTerm| context: the experiments include both zero - shot settings as well as multilingual models . previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "pos tagging", "output": "zero - shot transfers", "neg_sample": ["pos tagging is done by using OtherScientificTerm", "the experiments include both zero - shot settings as well as multilingual models .", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "Evaluating zero - shot transfers and multilingual models for dependency parsing and POS tagging within the low - resource language family Tupian.", "forward": false, "src_ids": "2022.acl-srw.1_3298"}
+{"input": "dependency parsing is done by using Method| context: previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "dependency parsing", "output": "multilingual models", "neg_sample": ["dependency parsing is done by using Method", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "Evaluating zero - shot transfers and multilingual models for dependency parsing and POS tagging within the low - resource language family Tupian.", "forward": false, "src_ids": "2022.acl-srw.1_3299"}
+{"input": "pos tagging is done by using Method| context: previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "pos tagging", "output": "multilingual models", "neg_sample": ["pos tagging is done by using Method", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "Evaluating zero - shot transfers and multilingual models for dependency parsing and POS tagging within the low - resource language family Tupian.", "forward": false, "src_ids": "2022.acl-srw.1_3300"}
+{"input": "multilingual models is used for Task| context: the experiments include both zero - shot settings as well as multilingual models . previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "multilingual models", "output": "dependency parsing", "neg_sample": ["multilingual models is used for Task", "the experiments include both zero - shot settings as well as multilingual models .", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "Evaluating zero - shot transfers and multilingual models for dependency parsing and POS tagging within the low - resource language family Tupian.", "forward": true, "src_ids": "2022.acl-srw.1_3301"}
+{"input": "zero - shot transfers is used for Task| context: the experiments include both zero - shot settings as well as multilingual models . previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "zero - shot transfers", "output": "dependency parsing", "neg_sample": ["zero - shot transfers is used for Task", "the experiments include both zero - shot settings as well as multilingual models .", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "Evaluating zero - shot transfers and multilingual models for dependency parsing and POS tagging within the low - resource language family Tupian.", "forward": true, "src_ids": "2022.acl-srw.1_3302"}
+{"input": "nlp applications is done by using Method| context: the experiments include both zero - shot settings as well as multilingual models . previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "nlp applications", "output": "transfer learning", "neg_sample": ["nlp applications is done by using Method", "the experiments include both zero - shot settings as well as multilingual models .", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "This suggests that in addition to leveraging similarity between two related languages , the incorporation of multiple languages of the same family might lead to better results in transfer learning for NLP applications .", "forward": false, "src_ids": "2022.acl-srw.1_3303"}
+{"input": "transfer learning is used for Task| context: the experiments include both zero - shot settings as well as multilingual models . previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases .", "entity": "transfer learning", "output": "nlp applications", "neg_sample": ["transfer learning is used for Task", "the experiments include both zero - shot settings as well as multilingual models .", "previous studies have found that even a comparably small treebank from a closely related language will improve sequence labelling considerably in such cases ."], "relation": "used for", "id": "2022.acl-srw.1", "year": 2022, "rel_sent": "This suggests that in addition to leveraging similarity between two related languages , the incorporation of multiple languages of the same family might lead to better results in transfer learning for NLP applications .", "forward": true, "src_ids": "2022.acl-srw.1_3304"}
+{"input": "human si rates is done by using OtherScientificTerm| context: scalar implicature ( si ) arises when a speaker uses an expression ( e.g. , ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g. , ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning . prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale .", "entity": "human si rates", "output": "scale uncertainty", "neg_sample": ["human si rates is done by using OtherScientificTerm", "scalar implicature ( si ) arises when a speaker uses an expression ( e.g.", ", ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g.", ", ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning .", "prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale ."], "relation": "used for", "id": "2022.cmcl-1.8", "year": 2022, "rel_sent": "We find that scale uncertainty predicts human SI rates , measured as entropy over the sampled alternatives and over latent classes among alternatives in sentence embedding space .", "forward": false, "src_ids": "2022.cmcl-1.8_3305"}
+{"input": "scale uncertainty is used for Metric| context: scalar implicature ( si ) arises when a speaker uses an expression ( e.g. , ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g. , ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning . prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale .", "entity": "scale uncertainty", "output": "human si rates", "neg_sample": ["scale uncertainty is used for Metric", "scalar implicature ( si ) arises when a speaker uses an expression ( e.g.", ", ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g.", ", ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning .", "prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale ."], "relation": "used for", "id": "2022.cmcl-1.8", "year": 2022, "rel_sent": "We find that scale uncertainty predicts human SI rates , measured as entropy over the sampled alternatives and over latent classes among alternatives in sentence embedding space .", "forward": true, "src_ids": "2022.cmcl-1.8_3306"}
+{"input": "text infilling task is done by using Method| context: scalar implicature ( si ) arises when a speaker uses an expression ( e.g. , ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g. , ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning . prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale .", "entity": "text infilling task", "output": "t5 model", "neg_sample": ["text infilling task is done by using Method", "scalar implicature ( si ) arises when a speaker uses an expression ( e.g.", ", ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g.", ", ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning .", "prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale ."], "relation": "used for", "id": "2022.cmcl-1.8", "year": 2022, "rel_sent": "We use a T5 model fine - tuned on a text infilling task to estimate this distribution .", "forward": false, "src_ids": "2022.cmcl-1.8_3307"}
+{"input": "t5 model is used for Task| context: scalar implicature ( si ) arises when a speaker uses an expression ( e.g. , ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g. , ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning . prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale .", "entity": "t5 model", "output": "text infilling task", "neg_sample": ["t5 model is used for Task", "scalar implicature ( si ) arises when a speaker uses an expression ( e.g.", ", ' some ' ) that is semantically compatible with a logically stronger alternative on the same scale ( e.g.", ", ' all ' ) , leading the listener to infer that they did not intend to convey the stronger meaning .", "prior work has demonstrated that si rates are highly variable across scales , raising the question of what factors determine the si strength for a particular scale ."], "relation": "used for", "id": "2022.cmcl-1.8", "year": 2022, "rel_sent": "We use a T5 model fine - tuned on a text infilling task to estimate this distribution .", "forward": true, "src_ids": "2022.cmcl-1.8_3308"}
+{"input": "multimodal audio - visual speech recognition is done by using Method| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "multimodal audio - visual speech recognition", "output": "unimodal self - supervised learning", "neg_sample": ["multimodal audio - visual speech recognition is done by using Method", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "Leveraging Unimodal Self - Supervised Learning for Multimodal Audio - Visual Speech Recognition.", "forward": false, "src_ids": "2022.acl-long.308_3309"}
+{"input": "multimodal avsr is done by using Method| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "multimodal avsr", "output": "unimodal self - supervised learning", "neg_sample": ["multimodal avsr is done by using Method", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "In this work , we successfully leverage unimodal self - supervised learning to promote the multimodal AVSR .", "forward": false, "src_ids": "2022.acl-long.308_3310"}
+{"input": "unimodal self - supervised learning is used for Task| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "unimodal self - supervised learning", "output": "multimodal audio - visual speech recognition", "neg_sample": ["unimodal self - supervised learning is used for Task", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "Leveraging Unimodal Self - Supervised Learning for Multimodal Audio - Visual Speech Recognition.", "forward": true, "src_ids": "2022.acl-long.308_3311"}
+{"input": "unimodal self - supervised learning is used for Method| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "unimodal self - supervised learning", "output": "multimodal avsr", "neg_sample": ["unimodal self - supervised learning is used for Method", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "In this work , we successfully leverage unimodal self - supervised learning to promote the multimodal AVSR .", "forward": true, "src_ids": "2022.acl-long.308_3312"}
+{"input": "characters is done by using Method| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "characters", "output": "multimodal framework", "neg_sample": ["characters is done by using Method", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "In particular , audio and visual front - ends are trained on large - scale unimodal datasets , then we integrate components of both front - ends into a larger multimodal framework which learns to recognize parallel audio - visual data into characters through a combination of CTC and seq2seq decoding .", "forward": false, "src_ids": "2022.acl-long.308_3313"}
+{"input": "characters is done by using Material| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "characters", "output": "parallel audio - visual data", "neg_sample": ["characters is done by using Material", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "In particular , audio and visual front - ends are trained on large - scale unimodal datasets , then we integrate components of both front - ends into a larger multimodal framework which learns to recognize parallel audio - visual data into characters through a combination of CTC and seq2seq decoding .", "forward": false, "src_ids": "2022.acl-long.308_3314"}
+{"input": "multimodal framework is used for OtherScientificTerm| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "multimodal framework", "output": "characters", "neg_sample": ["multimodal framework is used for OtherScientificTerm", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "In particular , audio and visual front - ends are trained on large - scale unimodal datasets , then we integrate components of both front - ends into a larger multimodal framework which learns to recognize parallel audio - visual data into characters through a combination of CTC and seq2seq decoding .", "forward": true, "src_ids": "2022.acl-long.308_3315"}
+{"input": "parallel audio - visual data is used for OtherScientificTerm| context: training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) . thus it makes a lot of sense to make use of unlabelled unimodal data . on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored .", "entity": "parallel audio - visual data", "output": "characters", "neg_sample": ["parallel audio - visual data is used for OtherScientificTerm", "training transformer - based models demands a large amount of data , while obtaining aligned and labelled data in multimodality is rather cost - demanding , especially for audio - visual speech recognition ( avsr ) .", "thus it makes a lot of sense to make use of unlabelled unimodal data .", "on the other side , although the effectiveness of large - scale self - supervised learning is well established in both audio and visual modalities , how to integrate those pre - trained models into a multimodal scenario remains underexplored ."], "relation": "used for", "id": "2022.acl-long.308", "year": 2022, "rel_sent": "In particular , audio and visual front - ends are trained on large - scale unimodal datasets , then we integrate components of both front - ends into a larger multimodal framework which learns to recognize parallel audio - visual data into characters through a combination of CTC and seq2seq decoding .", "forward": true, "src_ids": "2022.acl-long.308_3316"}
+{"input": "computer vision is done by using Method| context: previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models .", "entity": "computer vision", "output": "transformer", "neg_sample": ["computer vision is done by using Method", "previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models ."], "relation": "used for", "id": "2022.acl-long.438", "year": 2022, "rel_sent": "Given the fact that Transformer is becoming popular in computer vision , we experiment with various strong models ( such as Vision Transformer ) and enhanced features ( such as object - detection and image captioning ) .", "forward": false, "src_ids": "2022.acl-long.438_3317"}
+{"input": "transformer is used for Task| context: previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models .", "entity": "transformer", "output": "computer vision", "neg_sample": ["transformer is used for Task", "previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models ."], "relation": "used for", "id": "2022.acl-long.438", "year": 2022, "rel_sent": "Given the fact that Transformer is becoming popular in computer vision , we experiment with various strong models ( such as Vision Transformer ) and enhanced features ( such as object - detection and image captioning ) .", "forward": true, "src_ids": "2022.acl-long.438_3318"}
+{"input": "patch - level contribution is done by using Method| context: previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models .", "entity": "patch - level contribution", "output": "selective attention model", "neg_sample": ["patch - level contribution is done by using Method", "previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models ."], "relation": "used for", "id": "2022.acl-long.438", "year": 2022, "rel_sent": "We develop a selective attention model to study the patch - level contribution of an image in MMT .", "forward": false, "src_ids": "2022.acl-long.438_3319"}
+{"input": "multimodal machine translation ( mmt ) is done by using Method| context: previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models .", "entity": "multimodal machine translation ( mmt )", "output": "selective attention model", "neg_sample": ["multimodal machine translation ( mmt ) is done by using Method", "previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models ."], "relation": "used for", "id": "2022.acl-long.438", "year": 2022, "rel_sent": "We develop a selective attention model to study the patch - level contribution of an image in MMT .", "forward": false, "src_ids": "2022.acl-long.438_3320"}
+{"input": "selective attention model is used for OtherScientificTerm| context: previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models .", "entity": "selective attention model", "output": "patch - level contribution", "neg_sample": ["selective attention model is used for OtherScientificTerm", "previous work on multimodal machine translation ( mmt ) has focused on the way of incorporating vision features into translation but little attention is on the quality of vision models ."], "relation": "used for", "id": "2022.acl-long.438", "year": 2022, "rel_sent": "We develop a selective attention model to study the patch - level contribution of an image in MMT .", "forward": true, "src_ids": "2022.acl-long.438_3321"}
+{"input": "grammar is used for OtherScientificTerm| context: the concurrent learning of both unseen structures and grammar is an enduring problem in phonological acquisition .", "entity": "grammar", "output": "rich base analysis", "neg_sample": ["grammar is used for OtherScientificTerm", "the concurrent learning of both unseen structures and grammar is an enduring problem in phonological acquisition ."], "relation": "used for", "id": "2022.scil-1.4", "year": 2022, "rel_sent": "When given an option between acquiring a grammar that supported a rich base analysis and one that did n't , the learner always acquired the grammar that supported rich bases .", "forward": true, "src_ids": "2022.scil-1.4_3322"}
+{"input": "english llms sentence completion is done by using Method| context: current language technology is ubiquitous and directly influences individuals ' lives worldwide . given the recent trend in ai on training and constantly releasing new and powerful large language models ( llms ) , there is a need to assess their biases and potential concrete consequences . while some studies have highlighted the shortcomings of these models , there is only little on the negative impact of llms on lgbtqia+ individuals .", "entity": "english llms sentence completion", "output": "template - based approach", "neg_sample": ["english llms sentence completion is done by using Method", "current language technology is ubiquitous and directly influences individuals ' lives worldwide .", "given the recent trend in ai on training and constantly releasing new and powerful large language models ( llms ) , there is a need to assess their biases and potential concrete consequences .", "while some studies have highlighted the shortcomings of these models , there is only little on the negative impact of llms on lgbtqia+ individuals ."], "relation": "used for", "id": "2022.ltedi-1.4", "year": 2022, "rel_sent": "In this paper , we investigated a state - of - the - art template - based approach for measuring the harmfulness of English LLMs sentence completion when the subjects belong to the LGBTQIA+ community .", "forward": false, "src_ids": "2022.ltedi-1.4_3323"}
+{"input": "template - based approach is used for Task| context: current language technology is ubiquitous and directly influences individuals ' lives worldwide . given the recent trend in ai on training and constantly releasing new and powerful large language models ( llms ) , there is a need to assess their biases and potential concrete consequences . while some studies have highlighted the shortcomings of these models , there is only little on the negative impact of llms on lgbtqia+ individuals .", "entity": "template - based approach", "output": "english llms sentence completion", "neg_sample": ["template - based approach is used for Task", "current language technology is ubiquitous and directly influences individuals ' lives worldwide .", "given the recent trend in ai on training and constantly releasing new and powerful large language models ( llms ) , there is a need to assess their biases and potential concrete consequences .", "while some studies have highlighted the shortcomings of these models , there is only little on the negative impact of llms on lgbtqia+ individuals ."], "relation": "used for", "id": "2022.ltedi-1.4", "year": 2022, "rel_sent": "In this paper , we investigated a state - of - the - art template - based approach for measuring the harmfulness of English LLMs sentence completion when the subjects belong to the LGBTQIA+ community .", "forward": true, "src_ids": "2022.ltedi-1.4_3324"}
+{"input": "neural machine translation is done by using Method| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus . earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage .", "entity": "neural machine translation", "output": "denoising entity pre - training method", "neg_sample": ["neural machine translation is done by using Method", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "DEEP : DEnoising Entity Pre - training for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.123_3325"}
+{"input": "denoising entity pre - training method is used for Task| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus . earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage .", "entity": "denoising entity pre - training method", "output": "neural machine translation", "neg_sample": ["denoising entity pre - training method is used for Task", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "DEEP : DEnoising Entity Pre - training for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.123_3326"}
+{"input": "pre - trained neural machine translation model is done by using Method| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus . earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage .", "entity": "pre - trained neural machine translation model", "output": "multi - task learning strategy", "neg_sample": ["pre - trained neural machine translation model is done by using Method", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "Besides , we investigate a multi - task learning strategy that finetunes a pre - trained neural machine translation model on both entity - augmented monolingual data and parallel data tofurther improve entity translation .", "forward": false, "src_ids": "2022.acl-long.123_3327"}
+{"input": "entity translation is done by using Method| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus . earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage .", "entity": "entity translation", "output": "multi - task learning strategy", "neg_sample": ["entity translation is done by using Method", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "Besides , we investigate a multi - task learning strategy that finetunes a pre - trained neural machine translation model on both entity - augmented monolingual data and parallel data tofurther improve entity translation .", "forward": false, "src_ids": "2022.acl-long.123_3328"}
+{"input": "multi - task learning strategy is used for Method| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus . earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage .", "entity": "multi - task learning strategy", "output": "pre - trained neural machine translation model", "neg_sample": ["multi - task learning strategy is used for Method", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "Besides , we investigate a multi - task learning strategy that finetunes a pre - trained neural machine translation model on both entity - augmented monolingual data and parallel data tofurther improve entity translation .", "forward": true, "src_ids": "2022.acl-long.123_3329"}
+{"input": "entity translation is done by using Material| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus . earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage .", "entity": "entity translation", "output": "parallel data", "neg_sample": ["entity translation is done by using Material", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "earlier named entity translation methods mainly focus on phonetic transliteration , which ignores the sentence context for translation and is limited in domain and language coverage ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "Besides , we investigate a multi - task learning strategy that finetunes a pre - trained neural machine translation model on both entity - augmented monolingual data and parallel data tofurther improve entity translation .", "forward": false, "src_ids": "2022.acl-long.123_3330"}
+{"input": "parallel data is used for Task| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "entity": "parallel data", "output": "entity translation", "neg_sample": ["parallel data is used for Task", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "Besides , we investigate a multi - task learning strategy that finetunes a pre - trained neural machine translation model on both entity - augmented monolingual data and parallel data tofurther improve entity translation .", "forward": true, "src_ids": "2022.acl-long.123_3331"}
+{"input": "multi - task learning strategy is used for Task| context: it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus .", "entity": "multi - task learning strategy", "output": "entity translation", "neg_sample": ["multi - task learning strategy is used for Task", "it has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus ."], "relation": "used for", "id": "2022.acl-long.123", "year": 2022, "rel_sent": "Besides , we investigate a multi - task learning strategy that finetunes a pre - trained neural machine translation model on both entity - augmented monolingual data and parallel data tofurther improve entity translation .", "forward": true, "src_ids": "2022.acl-long.123_3332"}
+{"input": "compressed intermediate document representations is done by using Method| context: bert based ranking models have achieved superior performance on various information retrieval tasks . however , the large number of parameters and complex self - attention operations come at a significant latency overhead . to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency .", "entity": "compressed intermediate document representations", "output": "succinct document representation ( sdr ) scheme", "neg_sample": ["compressed intermediate document representations is done by using Method", "bert based ranking models have achieved superior performance on various information retrieval tasks .", "however , the large number of parameters and complex self - attention operations come at a significant latency overhead .", "to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency ."], "relation": "used for", "id": "2022.acl-long.457", "year": 2022, "rel_sent": "Nonetheless , having solved the immediate latency issue , these methods now introduce storage costs and network fetching latency , which limit their adoption in real - life production systems . In this work , we propose the Succinct Document Representation ( SDR ) scheme that computes highly compressed intermediate document representations , mitigating the storage / network issue .", "forward": false, "src_ids": "2022.acl-long.457_3333"}
+{"input": "succinct document representation ( sdr ) scheme is used for Method| context: bert based ranking models have achieved superior performance on various information retrieval tasks . however , the large number of parameters and complex self - attention operations come at a significant latency overhead . to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency .", "entity": "succinct document representation ( sdr ) scheme", "output": "compressed intermediate document representations", "neg_sample": ["succinct document representation ( sdr ) scheme is used for Method", "bert based ranking models have achieved superior performance on various information retrieval tasks .", "however , the large number of parameters and complex self - attention operations come at a significant latency overhead .", "to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency ."], "relation": "used for", "id": "2022.acl-long.457", "year": 2022, "rel_sent": "Nonetheless , having solved the immediate latency issue , these methods now introduce storage costs and network fetching latency , which limit their adoption in real - life production systems . In this work , we propose the Succinct Document Representation ( SDR ) scheme that computes highly compressed intermediate document representations , mitigating the storage / network issue .", "forward": true, "src_ids": "2022.acl-long.457_3334"}
+{"input": "encoding and decoding phases is done by using OtherScientificTerm| context: bert based ranking models have achieved superior performance on various information retrieval tasks . however , the large number of parameters and complex self - attention operations come at a significant latency overhead . to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency .", "entity": "encoding and decoding phases", "output": "document 's textual content", "neg_sample": ["encoding and decoding phases is done by using OtherScientificTerm", "bert based ranking models have achieved superior performance on various information retrieval tasks .", "however , the large number of parameters and complex self - attention operations come at a significant latency overhead .", "to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency ."], "relation": "used for", "id": "2022.acl-long.457", "year": 2022, "rel_sent": "Our approach first reduces the dimension of token representations by encoding them using a novel autoencoder architecture that uses the document 's textual content in both the encoding and decoding phases .", "forward": false, "src_ids": "2022.acl-long.457_3335"}
+{"input": "document 's textual content is used for Task| context: bert based ranking models have achieved superior performance on various information retrieval tasks . however , the large number of parameters and complex self - attention operations come at a significant latency overhead . to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency .", "entity": "document 's textual content", "output": "encoding and decoding phases", "neg_sample": ["document 's textual content is used for Task", "bert based ranking models have achieved superior performance on various information retrieval tasks .", "however , the large number of parameters and complex self - attention operations come at a significant latency overhead .", "to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency ."], "relation": "used for", "id": "2022.acl-long.457", "year": 2022, "rel_sent": "Our approach first reduces the dimension of token representations by encoding them using a novel autoencoder architecture that uses the document 's textual content in both the encoding and decoding phases .", "forward": true, "src_ids": "2022.acl-long.457_3336"}
+{"input": "implicitly toxic text is done by using Method| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "entity": "implicitly toxic text", "output": "toxigen", "neg_sample": ["implicitly toxic text is done by using Method", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate ."], "relation": "used for", "id": "2022.acl-long.234", "year": 2022, "rel_sent": "Controlling machine generation in this way allows ToxiGen to cover implicitly toxic text at a larger scale , and about more demographic groups , than previous resources of human - written text .", "forward": false, "src_ids": "2022.acl-long.234_3337"}
+{"input": "adversarial and implicit hate speech detection is done by using Material| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "entity": "adversarial and implicit hate speech detection", "output": "large - scale and machine - generated dataset", "neg_sample": ["adversarial and implicit hate speech detection is done by using Material", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate ."], "relation": "used for", "id": "2022.acl-long.234", "year": 2022, "rel_sent": "ToxiGen : A Large - Scale Machine - Generated Dataset for Adversarial and Implicit Hate Speech Detection.", "forward": false, "src_ids": "2022.acl-long.234_3338"}
+{"input": "large - scale and machine - generated dataset is used for Task| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "entity": "large - scale and machine - generated dataset", "output": "adversarial and implicit hate speech detection", "neg_sample": ["large - scale and machine - generated dataset is used for Task", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate ."], "relation": "used for", "id": "2022.acl-long.234", "year": 2022, "rel_sent": "ToxiGen : A Large - Scale Machine - Generated Dataset for Adversarial and Implicit Hate Speech Detection.", "forward": true, "src_ids": "2022.acl-long.234_3339"}
+{"input": "subtly toxic and benign text is done by using Method| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "entity": "subtly toxic and benign text", "output": "demonstration - based prompting framework", "neg_sample": ["subtly toxic and benign text is done by using Method", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate ."], "relation": "used for", "id": "2022.acl-long.234", "year": 2022, "rel_sent": "We develop a demonstration - based prompting framework and an adversarial classifier - in - the - loop decoding method to generate subtly toxic and benign text with a massive pretrained language model .", "forward": false, "src_ids": "2022.acl-long.234_3340"}
+{"input": "demonstration - based prompting framework is used for Material| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "entity": "demonstration - based prompting framework", "output": "subtly toxic and benign text", "neg_sample": ["demonstration - based prompting framework is used for Material", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate ."], "relation": "used for", "id": "2022.acl-long.234", "year": 2022, "rel_sent": "We develop a demonstration - based prompting framework and an adversarial classifier - in - the - loop decoding method to generate subtly toxic and benign text with a massive pretrained language model .", "forward": true, "src_ids": "2022.acl-long.234_3341"}
+{"input": "toxigen is used for Material| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "entity": "toxigen", "output": "implicitly toxic text", "neg_sample": ["toxigen is used for Material", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate ."], "relation": "used for", "id": "2022.acl-long.234", "year": 2022, "rel_sent": "Controlling machine generation in this way allows ToxiGen to cover implicitly toxic text at a larger scale , and about more demographic groups , than previous resources of human - written text .", "forward": true, "src_ids": "2022.acl-long.234_3342"}
+{"input": "data augmentation is done by using Generic| context: we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track .", "entity": "data augmentation", "output": "dominating techniques", "neg_sample": ["data augmentation is done by using Generic", "we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track ."], "relation": "used for", "id": "2022.iwslt-1.17", "year": 2022, "rel_sent": "We also strengthen our system with some dominating techniques related to data augmentation , such as knowledge distillation , tagged back - translation , and iterative back - translation .", "forward": false, "src_ids": "2022.iwslt-1.17_3343"}
+{"input": "dominating techniques is used for Method| context: we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track .", "entity": "dominating techniques", "output": "data augmentation", "neg_sample": ["dominating techniques is used for Method", "we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track ."], "relation": "used for", "id": "2022.iwslt-1.17", "year": 2022, "rel_sent": "We also strengthen our system with some dominating techniques related to data augmentation , such as knowledge distillation , tagged back - translation , and iterative back - translation .", "forward": true, "src_ids": "2022.iwslt-1.17_3344"}
+{"input": "naive transformer model is done by using Method| context: we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track .", "entity": "naive transformer model", "output": "training techniques", "neg_sample": ["naive transformer model is done by using Method", "we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track ."], "relation": "used for", "id": "2022.iwslt-1.17", "year": 2022, "rel_sent": "We also incorporate novel training techniques such as R - drop , deep model , and large batch training which have been shown to be beneficial to the naive Transformer model .", "forward": false, "src_ids": "2022.iwslt-1.17_3345"}
+{"input": "large batch training is used for Method| context: we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track .", "entity": "large batch training", "output": "naive transformer model", "neg_sample": ["large batch training is used for Method", "we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track ."], "relation": "used for", "id": "2022.iwslt-1.17", "year": 2022, "rel_sent": "We also incorporate novel training techniques such as R - drop , deep model , and large batch training which have been shown to be beneficial to the naive Transformer model .", "forward": true, "src_ids": "2022.iwslt-1.17_3346"}
+{"input": "training techniques is used for Method| context: we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track .", "entity": "training techniques", "output": "naive transformer model", "neg_sample": ["training techniques is used for Method", "we participate in the english - to - mandarin chinese text - to - text ( noted as t2 t ) track ."], "relation": "used for", "id": "2022.iwslt-1.17", "year": 2022, "rel_sent": "We also incorporate novel training techniques such as R - drop , deep model , and large batch training which have been shown to be beneficial to the naive Transformer model .", "forward": true, "src_ids": "2022.iwslt-1.17_3347"}
+{"input": "certainty calibration is done by using Method| context: certainty calibration is an important goal on the path to interpretability and trustworthy ai . particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential . in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest .", "entity": "certainty calibration", "output": "region - dependent temperature scaling", "neg_sample": ["certainty calibration is done by using Method", "certainty calibration is an important goal on the path to interpretability and trustworthy ai .", "particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential .", "in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest ."], "relation": "used for", "id": "2022.acl-short.59", "year": 2022, "rel_sent": "Region - dependent temperature scaling for certainty calibration and application to class - imbalanced token classification.", "forward": false, "src_ids": "2022.acl-short.59_3348"}
+{"input": "region - dependent temperature scaling is used for Task| context: particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential . in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest .", "entity": "region - dependent temperature scaling", "output": "certainty calibration", "neg_sample": ["region - dependent temperature scaling is used for Task", "particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential .", "in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest ."], "relation": "used for", "id": "2022.acl-short.59", "year": 2022, "rel_sent": "Region - dependent temperature scaling for certainty calibration and application to class - imbalanced token classification.", "forward": true, "src_ids": "2022.acl-short.59_3349"}
+{"input": "region - balanced calibration error metric is used for OtherScientificTerm| context: certainty calibration is an important goal on the path to interpretability and trustworthy ai . particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential . in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest .", "entity": "region - balanced calibration error metric", "output": "certainty regions", "neg_sample": ["region - balanced calibration error metric is used for OtherScientificTerm", "certainty calibration is an important goal on the path to interpretability and trustworthy ai .", "particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential .", "in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest ."], "relation": "used for", "id": "2022.acl-short.59", "year": 2022, "rel_sent": "We introduce a region - balanced calibration error metric that weights all certainty regions equally .", "forward": true, "src_ids": "2022.acl-short.59_3350"}
+{"input": "certainty regions is done by using Method| context: certainty calibration is an important goal on the path to interpretability and trustworthy ai . particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential . in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest .", "entity": "certainty regions", "output": "region - balanced calibration error metric", "neg_sample": ["certainty regions is done by using Method", "certainty calibration is an important goal on the path to interpretability and trustworthy ai .", "particularly in the context of human - in - the - loop systems , high - quality low to mid - range certainty estimates are essential .", "in the presence of a dominant high - certainty class , for instance the non - entity class in ner problems , existing calibration error measures are completely insensitive to potentially large errors in this certainty region of interest ."], "relation": "used for", "id": "2022.acl-short.59", "year": 2022, "rel_sent": "We introduce a region - balanced calibration error metric that weights all certainty regions equally .", "forward": false, "src_ids": "2022.acl-short.59_3351"}
+{"input": "prediction bias is done by using Task| context: although adapting pre - trained language models with few examples has shown promising performance on text classification , there is a lack of understanding of where the performance gain comes from .", "entity": "prediction bias", "output": "few - shot fine - tuning", "neg_sample": ["prediction bias is done by using Task", "although adapting pre - trained language models with few examples has shown promising performance on text classification , there is a lack of understanding of where the performance gain comes from ."], "relation": "used for", "id": "2022.insights-1.20", "year": 2022, "rel_sent": "BERT and RoBERTa ) show strong prediction bias across labels ; ( 2 ) although few - shot fine - tuning can mitigate the prediction bias and demonstrate promising prediction performance , our analysis shows models gain performance improvement by capturing non - task - related features ( e.g.", "forward": false, "src_ids": "2022.insights-1.20_3352"}
+{"input": "few - shot fine - tuning is used for OtherScientificTerm| context: although adapting pre - trained language models with few examples has shown promising performance on text classification , there is a lack of understanding of where the performance gain comes from .", "entity": "few - shot fine - tuning", "output": "prediction bias", "neg_sample": ["few - shot fine - tuning is used for OtherScientificTerm", "although adapting pre - trained language models with few examples has shown promising performance on text classification , there is a lack of understanding of where the performance gain comes from ."], "relation": "used for", "id": "2022.insights-1.20", "year": 2022, "rel_sent": "BERT and RoBERTa ) show strong prediction bias across labels ; ( 2 ) although few - shot fine - tuning can mitigate the prediction bias and demonstrate promising prediction performance , our analysis shows models gain performance improvement by capturing non - task - related features ( e.g.", "forward": true, "src_ids": "2022.insights-1.20_3353"}
+{"input": "real - world scenarios is done by using Material| context: large multilingual transformer - based machine translation models have had a pivotal role in making translation systems available for hundreds of languages with good zero - shot translation performance . one such example is the universal model with shared encoder - decoder architecture . additionally , jointly trained language - specific encoder - decoder systems have been proposed for multilingual neural machine translation ( nmt ) models .", "entity": "real - world scenarios", "output": "multilingual dataset", "neg_sample": ["real - world scenarios is done by using Material", "large multilingual transformer - based machine translation models have had a pivotal role in making translation systems available for hundreds of languages with good zero - shot translation performance .", "one such example is the universal model with shared encoder - decoder architecture .", "additionally , jointly trained language - specific encoder - decoder systems have been proposed for multilingual neural machine translation ( nmt ) models ."], "relation": "used for", "id": "2022.eamt-1.12", "year": 2022, "rel_sent": "Experiments on a multilingual dataset set up to model real - world scenarios , including zero - shot and low - resource translation , show that our proposed models achieve higher translation quality compared to purely universal and language - specific approaches .", "forward": false, "src_ids": "2022.eamt-1.12_3354"}
+{"input": "multilingual dataset is used for OtherScientificTerm| context: large multilingual transformer - based machine translation models have had a pivotal role in making translation systems available for hundreds of languages with good zero - shot translation performance . one such example is the universal model with shared encoder - decoder architecture . additionally , jointly trained language - specific encoder - decoder systems have been proposed for multilingual neural machine translation ( nmt ) models .", "entity": "multilingual dataset", "output": "real - world scenarios", "neg_sample": ["multilingual dataset is used for OtherScientificTerm", "large multilingual transformer - based machine translation models have had a pivotal role in making translation systems available for hundreds of languages with good zero - shot translation performance .", "one such example is the universal model with shared encoder - decoder architecture .", "additionally , jointly trained language - specific encoder - decoder systems have been proposed for multilingual neural machine translation ( nmt ) models ."], "relation": "used for", "id": "2022.eamt-1.12", "year": 2022, "rel_sent": "Experiments on a multilingual dataset set up to model real - world scenarios , including zero - shot and low - resource translation , show that our proposed models achieve higher translation quality compared to purely universal and language - specific approaches .", "forward": true, "src_ids": "2022.eamt-1.12_3355"}
+{"input": "nlp is done by using Task| context: as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets .", "entity": "nlp", "output": "benchmark dataset sharing", "neg_sample": ["nlp is done by using Task", "as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets ."], "relation": "used for", "id": "2022.nlppower-1.1", "year": 2022, "rel_sent": "This paper critically examines the current practices of benchmark dataset sharing in NLP and suggests a better way to inform reusers of the benchmark dataset .", "forward": false, "src_ids": "2022.nlppower-1.1_3356"}
+{"input": "benchmark dataset sharing is used for Task| context: as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets .", "entity": "benchmark dataset sharing", "output": "nlp", "neg_sample": ["benchmark dataset sharing is used for Task", "as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets ."], "relation": "used for", "id": "2022.nlppower-1.1", "year": 2022, "rel_sent": "This paper critically examines the current practices of benchmark dataset sharing in NLP and suggests a better way to inform reusers of the benchmark dataset .", "forward": true, "src_ids": "2022.nlppower-1.1_3357"}
+{"input": "data curator is used for OtherScientificTerm| context: as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets .", "entity": "data curator", "output": "social impact metadata", "neg_sample": ["data curator is used for OtherScientificTerm", "as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets ."], "relation": "used for", "id": "2022.nlppower-1.1", "year": 2022, "rel_sent": "We propose that the benchmark dataset should develop social impact metadata and data curator should take a role in managing the social impact metadata .", "forward": true, "src_ids": "2022.nlppower-1.1_3358"}
+{"input": "social impact metadata is done by using Method| context: as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets .", "entity": "social impact metadata", "output": "data curator", "neg_sample": ["social impact metadata is done by using Method", "as the dataset sharing platform plays a key role not only in distributing the dataset but also in informing the potential reusers about the dataset , we believe data - sharing platforms should provide a comprehensive context of the datasets ."], "relation": "used for", "id": "2022.nlppower-1.1", "year": 2022, "rel_sent": "We propose that the benchmark dataset should develop social impact metadata and data curator should take a role in managing the social impact metadata .", "forward": false, "src_ids": "2022.nlppower-1.1_3359"}
+{"input": "endangered languages is done by using Material| context: we describe recent extensions to the open source learning and reading assistant ( lara ) supporting image - based and phonetically annotated texts .", "entity": "endangered languages", "output": "image - based and phonetically annotated multimodal texts", "neg_sample": ["endangered languages is done by using Material", "we describe recent extensions to the open source learning and reading assistant ( lara ) supporting image - based and phonetically annotated texts ."], "relation": "used for", "id": "2022.computel-1.9", "year": 2022, "rel_sent": "Using LARA to create image - based and phonetically annotated multimodal texts for endangered languages.", "forward": false, "src_ids": "2022.computel-1.9_3360"}
+{"input": "learning and reading assistant is used for Material| context: we describe recent extensions to the open source learning and reading assistant ( lara ) supporting image - based and phonetically annotated texts .", "entity": "learning and reading assistant", "output": "image - based and phonetically annotated multimodal texts", "neg_sample": ["learning and reading assistant is used for Material", "we describe recent extensions to the open source learning and reading assistant ( lara ) supporting image - based and phonetically annotated texts ."], "relation": "used for", "id": "2022.computel-1.9", "year": 2022, "rel_sent": "Using LARA to create image - based and phonetically annotated multimodal texts for endangered languages.", "forward": true, "src_ids": "2022.computel-1.9_3361"}
+{"input": "image - based and phonetically annotated multimodal texts is used for Material| context: we describe recent extensions to the open source learning and reading assistant ( lara ) supporting image - based and phonetically annotated texts .", "entity": "image - based and phonetically annotated multimodal texts", "output": "endangered languages", "neg_sample": ["image - based and phonetically annotated multimodal texts is used for Material", "we describe recent extensions to the open source learning and reading assistant ( lara ) supporting image - based and phonetically annotated texts ."], "relation": "used for", "id": "2022.computel-1.9", "year": 2022, "rel_sent": "Using LARA to create image - based and phonetically annotated multimodal texts for endangered languages.", "forward": true, "src_ids": "2022.computel-1.9_3362"}
+{"input": "zero - shot cross - lingual transfer is done by using Method| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "zero - shot cross - lingual transfer", "output": "learning disentangled semantic representations", "neg_sample": ["zero - shot cross - lingual transfer is done by using Method", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "Learning Disentangled Semantic Representations for Zero - Shot Cross - Lingual Transfer in Multilingual Machine Reading Comprehension.", "forward": false, "src_ids": "2022.acl-long.70_3363"}
+{"input": "multilingual machine reading comprehension is done by using Method| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "multilingual machine reading comprehension", "output": "learning disentangled semantic representations", "neg_sample": ["multilingual machine reading comprehension is done by using Method", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "Learning Disentangled Semantic Representations for Zero - Shot Cross - Lingual Transfer in Multilingual Machine Reading Comprehension.", "forward": false, "src_ids": "2022.acl-long.70_3364"}
+{"input": "multilingual machine reading comprehension is done by using Task| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "multilingual machine reading comprehension", "output": "zero - shot cross - lingual transfer", "neg_sample": ["multilingual machine reading comprehension is done by using Task", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "Learning Disentangled Semantic Representations for Zero - Shot Cross - Lingual Transfer in Multilingual Machine Reading Comprehension.", "forward": false, "src_ids": "2022.acl-long.70_3365"}
+{"input": "learning disentangled semantic representations is used for Task| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "learning disentangled semantic representations", "output": "zero - shot cross - lingual transfer", "neg_sample": ["learning disentangled semantic representations is used for Task", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "Learning Disentangled Semantic Representations for Zero - Shot Cross - Lingual Transfer in Multilingual Machine Reading Comprehension.", "forward": true, "src_ids": "2022.acl-long.70_3366"}
+{"input": "zero - shot cross - lingual transfer is used for Task| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "zero - shot cross - lingual transfer", "output": "multilingual machine reading comprehension", "neg_sample": ["zero - shot cross - lingual transfer is used for Task", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "Learning Disentangled Semantic Representations for Zero - Shot Cross - Lingual Transfer in Multilingual Machine Reading Comprehension.", "forward": true, "src_ids": "2022.acl-long.70_3367"}
+{"input": "semantics is done by using Method| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "semantics", "output": "siamese semantic disentanglement model ( s2dm )", "neg_sample": ["semantics is done by using Method", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "In this paper , we propose a novel multilingual MRC framework equipped with a Siamese Semantic Disentanglement Model ( S2DM ) to disassociate semantics from syntax in representations learned by multilingual pre - trained models .", "forward": false, "src_ids": "2022.acl-long.70_3368"}
+{"input": "siamese semantic disentanglement model ( s2dm ) is used for OtherScientificTerm| context: multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) . however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language .", "entity": "siamese semantic disentanglement model ( s2dm )", "output": "semantics", "neg_sample": ["siamese semantic disentanglement model ( s2dm ) is used for OtherScientificTerm", "multilingual pre - trained models are able to zero - shot transfer knowledge from rich - resource to low - resource languages in machine reading comprehension ( mrc ) .", "however , inherent linguistic discrepancies in different languages could make answer spans predicted by zero - shot transfer violate syntactic constraints of the target language ."], "relation": "used for", "id": "2022.acl-long.70", "year": 2022, "rel_sent": "In this paper , we propose a novel multilingual MRC framework equipped with a Siamese Semantic Disentanglement Model ( S2DM ) to disassociate semantics from syntax in representations learned by multilingual pre - trained models .", "forward": true, "src_ids": "2022.acl-long.70_3369"}
+{"input": "counting is done by using Method| context: recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format . although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting .", "entity": "counting", "output": "logical inference", "neg_sample": ["counting is done by using Method", "recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format .", "although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting ."], "relation": "used for", "id": "2022.acl-srw.8", "year": 2022, "rel_sent": "Logical Inference for Counting on Semi - structured Tables.", "forward": false, "src_ids": "2022.acl-srw.8_3370"}
+{"input": "logical inference is used for OtherScientificTerm| context: recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format .", "entity": "logical inference", "output": "counting", "neg_sample": ["logical inference is used for OtherScientificTerm", "recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format ."], "relation": "used for", "id": "2022.acl-srw.8", "year": 2022, "rel_sent": "Logical Inference for Counting on Semi - structured Tables.", "forward": true, "src_ids": "2022.acl-srw.8_3371"}
+{"input": "model checking is used for Task| context: recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format .", "entity": "model checking", "output": "numerical type of inference", "neg_sample": ["model checking is used for Task", "recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format ."], "relation": "used for", "id": "2022.acl-srw.8", "year": 2022, "rel_sent": "We use logical representations as meaning representations for tables and texts and use model checking to handle a numerical type of inference between texts and tables .", "forward": true, "src_ids": "2022.acl-srw.8_3372"}
+{"input": "meaning representations is done by using Method| context: recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format . although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting .", "entity": "meaning representations", "output": "logical representations", "neg_sample": ["meaning representations is done by using Method", "recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format .", "although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting ."], "relation": "used for", "id": "2022.acl-srw.8", "year": 2022, "rel_sent": "We use logical representations as meaning representations for tables and texts and use model checking to handle a numerical type of inference between texts and tables .", "forward": false, "src_ids": "2022.acl-srw.8_3373"}
+{"input": "logical representations is used for Method| context: recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format . although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting .", "entity": "logical representations", "output": "meaning representations", "neg_sample": ["logical representations is used for Method", "recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format .", "although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting ."], "relation": "used for", "id": "2022.acl-srw.8", "year": 2022, "rel_sent": "We use logical representations as meaning representations for tables and texts and use model checking to handle a numerical type of inference between texts and tables .", "forward": true, "src_ids": "2022.acl-srw.8_3374"}
+{"input": "numerical type of inference is done by using Method| context: recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format . although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting .", "entity": "numerical type of inference", "output": "model checking", "neg_sample": ["numerical type of inference is done by using Method", "recently , the natural language inference ( nli ) task has been studied for semi - structured tables that do not have a strict format .", "although neural approaches have achieved high performance in various types of nli , including nli between semi - structured tables and texts , they still have difficulty in performing a numerical type of inference , such as counting ."], "relation": "used for", "id": "2022.acl-srw.8", "year": 2022, "rel_sent": "We use logical representations as meaning representations for tables and texts and use model checking to handle a numerical type of inference between texts and tables .", "forward": false, "src_ids": "2022.acl-srw.8_3375"}
+{"input": "fidelity - diversity trade - off is done by using OtherScientificTerm| context: round - trip machine translation ( mt ) is a popular choice for paraphrase generation , which leverages readily available parallel corpora for supervision . in this paper , we formalize the implicit similarity function induced by this approach , and show that it is susceptible to non - paraphrase pairs sharing a single ambiguous translation .", "entity": "fidelity - diversity trade - off", "output": "adjustable parameter", "neg_sample": ["fidelity - diversity trade - off is done by using OtherScientificTerm", "round - trip machine translation ( mt ) is a popular choice for paraphrase generation , which leverages readily available parallel corpora for supervision .", "in this paper , we formalize the implicit similarity function induced by this approach , and show that it is susceptible to non - paraphrase pairs sharing a single ambiguous translation ."], "relation": "used for", "id": "2022.acl-long.114", "year": 2022, "rel_sent": "In addition to being more principled and efficient than round - trip MT , our approach offers an adjustable parameter to control the fidelity - diversity trade - off , and obtains better results in our experiments .", "forward": false, "src_ids": "2022.acl-long.114_3376"}
+{"input": "adjustable parameter is used for OtherScientificTerm| context: round - trip machine translation ( mt ) is a popular choice for paraphrase generation , which leverages readily available parallel corpora for supervision . in this paper , we formalize the implicit similarity function induced by this approach , and show that it is susceptible to non - paraphrase pairs sharing a single ambiguous translation .", "entity": "adjustable parameter", "output": "fidelity - diversity trade - off", "neg_sample": ["adjustable parameter is used for OtherScientificTerm", "round - trip machine translation ( mt ) is a popular choice for paraphrase generation , which leverages readily available parallel corpora for supervision .", "in this paper , we formalize the implicit similarity function induced by this approach , and show that it is susceptible to non - paraphrase pairs sharing a single ambiguous translation ."], "relation": "used for", "id": "2022.acl-long.114", "year": 2022, "rel_sent": "In addition to being more principled and efficient than round - trip MT , our approach offers an adjustable parameter to control the fidelity - diversity trade - off , and obtains better results in our experiments .", "forward": true, "src_ids": "2022.acl-long.114_3377"}
+{"input": "explainable conversation reasoning is done by using Method| context: knowledge - grounded dialogue systems utilise external knowledge such as knowledge graphs to generate informative and appropriate responses . a crucial challenge of such systems is to select facts from a knowledge graph pertinent to the dialogue context for response generation . this fact selection can be formulated as path traversal over a knowledge graph conditioned on the dialogue context . such paths can originate from facts mentioned in the dialogue history and terminate at the facts to be mentioned in the response . these walks , in turn , provide an explanation of the flow of the conversation .", "entity": "explainable conversation reasoning", "output": "kg - cruse", "neg_sample": ["explainable conversation reasoning is done by using Method", "knowledge - grounded dialogue systems utilise external knowledge such as knowledge graphs to generate informative and appropriate responses .", "a crucial challenge of such systems is to select facts from a knowledge graph pertinent to the dialogue context for response generation .", "this fact selection can be formulated as path traversal over a knowledge graph conditioned on the dialogue context .", "such paths can originate from facts mentioned in the dialogue history and terminate at the facts to be mentioned in the response .", "these walks , in turn , provide an explanation of the flow of the conversation ."], "relation": "used for", "id": "2022.nlp4convai-1.9", "year": 2022, "rel_sent": "KG - CRuSE : Recurrent Walks over Knowledge Graph for Explainable Conversation Reasoning using Semantic Embeddings.", "forward": false, "src_ids": "2022.nlp4convai-1.9_3378"}
+{"input": "kg - cruse is used for Task| context: knowledge - grounded dialogue systems utilise external knowledge such as knowledge graphs to generate informative and appropriate responses . a crucial challenge of such systems is to select facts from a knowledge graph pertinent to the dialogue context for response generation . this fact selection can be formulated as path traversal over a knowledge graph conditioned on the dialogue context . such paths can originate from facts mentioned in the dialogue history and terminate at the facts to be mentioned in the response . these walks , in turn , provide an explanation of the flow of the conversation .", "entity": "kg - cruse", "output": "explainable conversation reasoning", "neg_sample": ["kg - cruse is used for Task", "knowledge - grounded dialogue systems utilise external knowledge such as knowledge graphs to generate informative and appropriate responses .", "a crucial challenge of such systems is to select facts from a knowledge graph pertinent to the dialogue context for response generation .", "this fact selection can be formulated as path traversal over a knowledge graph conditioned on the dialogue context .", "such paths can originate from facts mentioned in the dialogue history and terminate at the facts to be mentioned in the response .", "these walks , in turn , provide an explanation of the flow of the conversation ."], "relation": "used for", "id": "2022.nlp4convai-1.9", "year": 2022, "rel_sent": "KG - CRuSE : Recurrent Walks over Knowledge Graph for Explainable Conversation Reasoning using Semantic Embeddings.", "forward": true, "src_ids": "2022.nlp4convai-1.9_3379"}
+{"input": "coreference resolution is done by using Task| context: the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions .", "entity": "coreference resolution", "output": "graph refinement", "neg_sample": ["coreference resolution is done by using Task", "the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions ."], "relation": "used for", "id": "2022.findings-acl.215", "year": 2022, "rel_sent": "Graph Refinement for Coreference Resolution.", "forward": false, "src_ids": "2022.findings-acl.215_3380"}
+{"input": "coreference is done by using Method| context: the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions .", "entity": "coreference", "output": "modelling approach", "neg_sample": ["coreference is done by using Method", "the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions ."], "relation": "used for", "id": "2022.findings-acl.215", "year": 2022, "rel_sent": "We propose a modelling approach that learns coreference at the document - level and takes global decisions .", "forward": false, "src_ids": "2022.findings-acl.215_3381"}
+{"input": "conference resolution is done by using OtherScientificTerm| context: the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions .", "entity": "conference resolution", "output": "document - level information", "neg_sample": ["conference resolution is done by using OtherScientificTerm", "the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions ."], "relation": "used for", "id": "2022.findings-acl.215", "year": 2022, "rel_sent": "The experimental results show improvements over various baselines , reinforcing the hypothesis that document - level information improves conference resolution .", "forward": false, "src_ids": "2022.findings-acl.215_3382"}
+{"input": "document - level information is used for Task| context: the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions .", "entity": "document - level information", "output": "conference resolution", "neg_sample": ["document - level information is used for Task", "the state - of - the - art models for coreference resolution are based on independent mention pair - wise decisions ."], "relation": "used for", "id": "2022.findings-acl.215", "year": 2022, "rel_sent": "The experimental results show improvements over various baselines , reinforcing the hypothesis that document - level information improves conference resolution .", "forward": true, "src_ids": "2022.findings-acl.215_3383"}
+{"input": "icd code classification is done by using Method| context: the international classification of diseases ( icd ) system is the international standard for classifying diseases and procedures during a healthcare encounter and is widely used for healthcare reporting and management purposes . assigning correct codes for clinical procedures is important for clinical , operational and financial decision - making in healthcare . contextual word embedding models have achieved state - of - the - art results in multiple nlp tasks . however , these models have yet to achieve state - of - the - art results in the icd classification task since one of their main disadvantages is that they can only process documents that contain a small number of tokens which is rarely the case with real patient notes .", "entity": "icd code classification", "output": "contextual embedding model", "neg_sample": ["icd code classification is done by using Method", "the international classification of diseases ( icd ) system is the international standard for classifying diseases and procedures during a healthcare encounter and is widely used for healthcare reporting and management purposes .", "assigning correct codes for clinical procedures is important for clinical , operational and financial decision - making in healthcare .", "contextual word embedding models have achieved state - of - the - art results in multiple nlp tasks .", "however , these models have yet to achieve state - of - the - art results in the icd classification task since one of their main disadvantages is that they can only process documents that contain a small number of tokens which is rarely the case with real patient notes ."], "relation": "used for", "id": "2022.bionlp-1.32", "year": 2022, "rel_sent": "ICDBigBird : A Contextual Embedding Model for ICD Code Classification.", "forward": false, "src_ids": "2022.bionlp-1.32_3384"}
+{"input": "contextual embedding model is used for Task| context: the international classification of diseases ( icd ) system is the international standard for classifying diseases and procedures during a healthcare encounter and is widely used for healthcare reporting and management purposes . assigning correct codes for clinical procedures is important for clinical , operational and financial decision - making in healthcare . contextual word embedding models have achieved state - of - the - art results in multiple nlp tasks . however , these models have yet to achieve state - of - the - art results in the icd classification task since one of their main disadvantages is that they can only process documents that contain a small number of tokens which is rarely the case with real patient notes .", "entity": "contextual embedding model", "output": "icd code classification", "neg_sample": ["contextual embedding model is used for Task", "the international classification of diseases ( icd ) system is the international standard for classifying diseases and procedures during a healthcare encounter and is widely used for healthcare reporting and management purposes .", "assigning correct codes for clinical procedures is important for clinical , operational and financial decision - making in healthcare .", "contextual word embedding models have achieved state - of - the - art results in multiple nlp tasks .", "however , these models have yet to achieve state - of - the - art results in the icd classification task since one of their main disadvantages is that they can only process documents that contain a small number of tokens which is rarely the case with real patient notes ."], "relation": "used for", "id": "2022.bionlp-1.32", "year": 2022, "rel_sent": "ICDBigBird : A Contextual Embedding Model for ICD Code Classification.", "forward": true, "src_ids": "2022.bionlp-1.32_3385"}
+{"input": "spoken language modeling is done by using Material| context: we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows .", "entity": "spoken language modeling", "output": "asr - generated text", "neg_sample": ["spoken language modeling is done by using Material", "we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows ."], "relation": "used for", "id": "2022.bigscience-1.2", "year": 2022, "rel_sent": "Using ASR - Generated Text for Spoken Language Modeling.", "forward": false, "src_ids": "2022.bigscience-1.2_3386"}
+{"input": "spoken language modeling is done by using Material| context: we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows .", "entity": "spoken language modeling", "output": "asr - generated text", "neg_sample": ["spoken language modeling is done by using Material", "we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows ."], "relation": "used for", "id": "2022.bigscience-1.2", "year": 2022, "rel_sent": "Experimental results show that FlauBERT - Oral is better than its initial FlauBERT version demonstrating that , despite its inherent noisy nature , ASR - Generated text can be useful to improve spoken language modeling .", "forward": false, "src_ids": "2022.bigscience-1.2_3387"}
+{"input": "asr - generated text is used for Task| context: we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows .", "entity": "asr - generated text", "output": "spoken language modeling", "neg_sample": ["asr - generated text is used for Task", "we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows ."], "relation": "used for", "id": "2022.bigscience-1.2", "year": 2022, "rel_sent": "Using ASR - Generated Text for Spoken Language Modeling.", "forward": true, "src_ids": "2022.bigscience-1.2_3388"}
+{"input": "asr - generated text is used for Task| context: we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows .", "entity": "asr - generated text", "output": "spoken language modeling", "neg_sample": ["asr - generated text is used for Task", "we leverage the ina ( french national audiovisual institute ) collection and obtain 19 gb of text after applying asr on 350,000 hours of diverse tv shows ."], "relation": "used for", "id": "2022.bigscience-1.2", "year": 2022, "rel_sent": "Experimental results show that FlauBERT - Oral is better than its initial FlauBERT version demonstrating that , despite its inherent noisy nature , ASR - Generated text can be useful to improve spoken language modeling .", "forward": true, "src_ids": "2022.bigscience-1.2_3389"}
+{"input": "semantic shift detection is done by using Method| context: contextual word embedding techniques for semantic shift detection are receiving more and more attention .", "entity": "semantic shift detection", "output": "incremental approach", "neg_sample": ["semantic shift detection is done by using Method", "contextual word embedding techniques for semantic shift detection are receiving more and more attention ."], "relation": "used for", "id": "2022.lchange-1.4", "year": 2022, "rel_sent": "What is Done is Done : an Incremental Approach to Semantic Shift Detection.", "forward": false, "src_ids": "2022.lchange-1.4_3390"}
+{"input": "semantic shift detection is done by using Method| context: contextual word embedding techniques for semantic shift detection are receiving more and more attention .", "entity": "semantic shift detection", "output": "incremental approach", "neg_sample": ["semantic shift detection is done by using Method", "contextual word embedding techniques for semantic shift detection are receiving more and more attention ."], "relation": "used for", "id": "2022.lchange-1.4", "year": 2022, "rel_sent": "In this paper , we present What is Done is Done ( WiDiD ) , an incremental approach to semantic shift detection based on incremental clustering techniques and contextual embedding methods to capture the changes over the meanings of a target word along a diachronic corpus .", "forward": false, "src_ids": "2022.lchange-1.4_3391"}
+{"input": "translation is done by using Method| context: computational fact - checking aims at supporting the verification process of textual claims by exploiting trustworthy sources . however , there are large classes of complex claims that can not be automatically verified , for instance those related to temporal reasoning .", "entity": "translation", "output": "deep neural approach", "neg_sample": ["translation is done by using Method", "computational fact - checking aims at supporting the verification process of textual claims by exploiting trustworthy sources .", "however , there are large classes of complex claims that can not be automatically verified , for instance those related to temporal reasoning ."], "relation": "used for", "id": "2022.fever-1.8", "year": 2022, "rel_sent": "The adopted deep neural approach shows promising preliminary results for the translation of 10 categories of claims extracted from real use cases .", "forward": false, "src_ids": "2022.fever-1.8_3392"}
+{"input": "deep neural approach is used for Task| context: computational fact - checking aims at supporting the verification process of textual claims by exploiting trustworthy sources . however , there are large classes of complex claims that can not be automatically verified , for instance those related to temporal reasoning .", "entity": "deep neural approach", "output": "translation", "neg_sample": ["deep neural approach is used for Task", "computational fact - checking aims at supporting the verification process of textual claims by exploiting trustworthy sources .", "however , there are large classes of complex claims that can not be automatically verified , for instance those related to temporal reasoning ."], "relation": "used for", "id": "2022.fever-1.8", "year": 2022, "rel_sent": "The adopted deep neural approach shows promising preliminary results for the translation of 10 categories of claims extracted from real use cases .", "forward": true, "src_ids": "2022.fever-1.8_3393"}
+{"input": "self - attention is used for Method| context: learned self - attention functions in state - of - the - art nlp models often correlate with human attention .", "entity": "self - attention", "output": "large - scale pre - trained language models", "neg_sample": ["self - attention is used for Method", "learned self - attention functions in state - of - the - art nlp models often correlate with human attention ."], "relation": "used for", "id": "2022.acl-long.296", "year": 2022, "rel_sent": "We investigate whether self - attention in large - scale pre - trained language models is as predictive of human eye fixation patterns during task - reading as classical cognitive models of human attention .", "forward": true, "src_ids": "2022.acl-long.296_3394"}
+{"input": "self - attention is used for OtherScientificTerm| context: learned self - attention functions in state - of - the - art nlp models often correlate with human attention .", "entity": "self - attention", "output": "human eye fixation patterns", "neg_sample": ["self - attention is used for OtherScientificTerm", "learned self - attention functions in state - of - the - art nlp models often correlate with human attention ."], "relation": "used for", "id": "2022.acl-long.296", "year": 2022, "rel_sent": "We investigate whether self - attention in large - scale pre - trained language models is as predictive of human eye fixation patterns during task - reading as classical cognitive models of human attention .", "forward": true, "src_ids": "2022.acl-long.296_3395"}
+{"input": "sentiment analysis is done by using Material| context: learned self - attention functions in state - of - the - art nlp models often correlate with human attention .", "entity": "sentiment analysis", "output": "task - specific reading datasets", "neg_sample": ["sentiment analysis is done by using Material", "learned self - attention functions in state - of - the - art nlp models often correlate with human attention ."], "relation": "used for", "id": "2022.acl-long.296", "year": 2022, "rel_sent": "We compare attention functions across two task - specific reading datasets for sentiment analysis and relation extraction .", "forward": false, "src_ids": "2022.acl-long.296_3396"}
+{"input": "relation extraction is done by using Material| context: learned self - attention functions in state - of - the - art nlp models often correlate with human attention .", "entity": "relation extraction", "output": "task - specific reading datasets", "neg_sample": ["relation extraction is done by using Material", "learned self - attention functions in state - of - the - art nlp models often correlate with human attention ."], "relation": "used for", "id": "2022.acl-long.296", "year": 2022, "rel_sent": "We compare attention functions across two task - specific reading datasets for sentiment analysis and relation extraction .", "forward": false, "src_ids": "2022.acl-long.296_3397"}
+{"input": "task - specific reading datasets is used for Task| context: learned self - attention functions in state - of - the - art nlp models often correlate with human attention .", "entity": "task - specific reading datasets", "output": "sentiment analysis", "neg_sample": ["task - specific reading datasets is used for Task", "learned self - attention functions in state - of - the - art nlp models often correlate with human attention ."], "relation": "used for", "id": "2022.acl-long.296", "year": 2022, "rel_sent": "We compare attention functions across two task - specific reading datasets for sentiment analysis and relation extraction .", "forward": true, "src_ids": "2022.acl-long.296_3398"}
+{"input": "attention functions is used for Task| context: learned self - attention functions in state - of - the - art nlp models often correlate with human attention .", "entity": "attention functions", "output": "sentiment analysis", "neg_sample": ["attention functions is used for Task", "learned self - attention functions in state - of - the - art nlp models often correlate with human attention ."], "relation": "used for", "id": "2022.acl-long.296", "year": 2022, "rel_sent": "We compare attention functions across two task - specific reading datasets for sentiment analysis and relation extraction .", "forward": true, "src_ids": "2022.acl-long.296_3399"}
+{"input": "clients ' heterogeneity is done by using Method| context: in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned .", "entity": "clients ' heterogeneity", "output": "actperfl", "neg_sample": ["clients ' heterogeneity is done by using Method", "in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned ."], "relation": "used for", "id": "2022.fl4nlp-1.1", "year": 2022, "rel_sent": "Consequently , ActPerFL can adapt to the underlying clients ' heterogeneity with uncertainty - driven local training and model aggregation .", "forward": false, "src_ids": "2022.fl4nlp-1.1_3400"}
+{"input": "bayesian hierarchical models is used for Method| context: in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned .", "entity": "bayesian hierarchical models", "output": "actperfl", "neg_sample": ["bayesian hierarchical models is used for Method", "in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned ."], "relation": "used for", "id": "2022.fl4nlp-1.1", "year": 2022, "rel_sent": "Inspired by Bayesian hierarchical models , we develop ActPerFL , a self - aware personalized FL method where each client can automatically balance the training of its local personal model and the global model that implicitly contributes to other clients ' training .", "forward": true, "src_ids": "2022.fl4nlp-1.1_3401"}
+{"input": "actperfl is used for OtherScientificTerm| context: in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned .", "entity": "actperfl", "output": "clients ' heterogeneity", "neg_sample": ["actperfl is used for OtherScientificTerm", "in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned ."], "relation": "used for", "id": "2022.fl4nlp-1.1", "year": 2022, "rel_sent": "Consequently , ActPerFL can adapt to the underlying clients ' heterogeneity with uncertainty - driven local training and model aggregation .", "forward": true, "src_ids": "2022.fl4nlp-1.1_3402"}
+{"input": "actperfl is done by using Method| context: in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned .", "entity": "actperfl", "output": "bayesian hierarchical models", "neg_sample": ["actperfl is done by using Method", "in the context of personalized federated learning ( fl ) , the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned ."], "relation": "used for", "id": "2022.fl4nlp-1.1", "year": 2022, "rel_sent": "Inspired by Bayesian hierarchical models , we develop ActPerFL , a self - aware personalized FL method where each client can automatically balance the training of its local personal model and the global model that implicitly contributes to other clients ' training .", "forward": false, "src_ids": "2022.fl4nlp-1.1_3403"}
+{"input": "reward signal is used for Method| context: despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations . including these factual hallucinations in a summary can be beneficial because they provide useful background information .", "entity": "reward signal", "output": "summarization system", "neg_sample": ["reward signal is used for Method", "despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations .", "including these factual hallucinations in a summary can be beneficial because they provide useful background information ."], "relation": "used for", "id": "2022.acl-long.236", "year": 2022, "rel_sent": "Empirical results suggest that our method vastly outperforms two baselines in both accuracy and F1 scores and has a strong correlation with human judgments on factuality classification tasks . Furthermore , we use our method as a reward signal to train a summarization system using an off - line reinforcement learning ( RL ) algorithm that can significantly improve the factuality of generated summaries while maintaining the level of abstractiveness .", "forward": true, "src_ids": "2022.acl-long.236_3404"}
+{"input": "summarization system is done by using OtherScientificTerm| context: state - of - the - art abstractive summarization systems often generate hallucinations ; i.e. , content that is not directly inferable from the source text . despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations . including these factual hallucinations in a summary can be beneficial because they provide useful background information .", "entity": "summarization system", "output": "reward signal", "neg_sample": ["summarization system is done by using OtherScientificTerm", "state - of - the - art abstractive summarization systems often generate hallucinations ; i.e.", ", content that is not directly inferable from the source text .", "despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations .", "including these factual hallucinations in a summary can be beneficial because they provide useful background information ."], "relation": "used for", "id": "2022.acl-long.236", "year": 2022, "rel_sent": "Empirical results suggest that our method vastly outperforms two baselines in both accuracy and F1 scores and has a strong correlation with human judgments on factuality classification tasks . Furthermore , we use our method as a reward signal to train a summarization system using an off - line reinforcement learning ( RL ) algorithm that can significantly improve the factuality of generated summaries while maintaining the level of abstractiveness .", "forward": false, "src_ids": "2022.acl-long.236_3405"}
+{"input": "structured document understanding is done by using Method| context: structured document understanding has attracted considerable attention and made significant progress recently , owing to its crucial role in intelligent document processing . however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited .", "entity": "structured document understanding", "output": "language - independent layout transformer ( lilt )", "neg_sample": ["structured document understanding is done by using Method", "structured document understanding has attracted considerable attention and made significant progress recently , owing to its crucial role in intelligent document processing .", "however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited ."], "relation": "used for", "id": "2022.acl-long.534", "year": 2022, "rel_sent": "LiLT : A Simple yet Effective Language - Independent Layout Transformer for Structured Document Understanding.", "forward": false, "src_ids": "2022.acl-long.534_3406"}
+{"input": "structured document understanding is done by using Method| context: structured document understanding has attracted considerable attention and made significant progress recently , owing to its crucial role in intelligent document processing . however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited .", "entity": "structured document understanding", "output": "language - independent layout transformer ( lilt )", "neg_sample": ["structured document understanding is done by using Method", "structured document understanding has attracted considerable attention and made significant progress recently , owing to its crucial role in intelligent document processing .", "however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited ."], "relation": "used for", "id": "2022.acl-long.534", "year": 2022, "rel_sent": "To address this issue , we propose a simple yet effective Language - independent Layout Transformer ( LiLT ) for structured document understanding .", "forward": false, "src_ids": "2022.acl-long.534_3407"}
+{"input": "language - independent layout transformer ( lilt ) is used for Task| context: however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited .", "entity": "language - independent layout transformer ( lilt )", "output": "structured document understanding", "neg_sample": ["language - independent layout transformer ( lilt ) is used for Task", "however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited ."], "relation": "used for", "id": "2022.acl-long.534", "year": 2022, "rel_sent": "LiLT : A Simple yet Effective Language - Independent Layout Transformer for Structured Document Understanding.", "forward": true, "src_ids": "2022.acl-long.534_3408"}
+{"input": "language - independent layout transformer ( lilt ) is used for Task| context: however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited .", "entity": "language - independent layout transformer ( lilt )", "output": "structured document understanding", "neg_sample": ["language - independent layout transformer ( lilt ) is used for Task", "however , most existing related models can only deal with the document data of specific language(s ) ( typically english ) included in the pre - training collection , which is extremely limited ."], "relation": "used for", "id": "2022.acl-long.534", "year": 2022, "rel_sent": "To address this issue , we propose a simple yet effective Language - independent Layout Transformer ( LiLT ) for structured document understanding .", "forward": true, "src_ids": "2022.acl-long.534_3409"}
+{"input": "nmt is done by using Task| context: generalising to unseen domains is under - explored and remains a challenge in neural machine translation .", "entity": "nmt", "output": "domain generalisation", "neg_sample": ["nmt is done by using Task", "generalising to unseen domains is under - explored and remains a challenge in neural machine translation ."], "relation": "used for", "id": "2022.findings-acl.49", "year": 2022, "rel_sent": "Domain Generalisation of NMT : Fusing Adapters with Leave - One - Domain - Out Training.", "forward": false, "src_ids": "2022.findings-acl.49_3410"}
+{"input": "domain generalisation is used for Method| context: generalising to unseen domains is under - explored and remains a challenge in neural machine translation .", "entity": "domain generalisation", "output": "nmt", "neg_sample": ["domain generalisation is used for Method", "generalising to unseen domains is under - explored and remains a challenge in neural machine translation ."], "relation": "used for", "id": "2022.findings-acl.49", "year": 2022, "rel_sent": "Domain Generalisation of NMT : Fusing Adapters with Leave - One - Domain - Out Training.", "forward": true, "src_ids": "2022.findings-acl.49_3411"}
+{"input": "fusion - based generalisation method is used for OtherScientificTerm| context: generalising to unseen domains is under - explored and remains a challenge in neural machine translation .", "entity": "fusion - based generalisation method", "output": "domain - specific parameters", "neg_sample": ["fusion - based generalisation method is used for OtherScientificTerm", "generalising to unseen domains is under - explored and remains a challenge in neural machine translation ."], "relation": "used for", "id": "2022.findings-acl.49", "year": 2022, "rel_sent": "Inspired by recent research in parameter - efficient transfer learning from pretrained models , this paper proposes a fusion - based generalisation method that learns to combine domain - specific parameters .", "forward": true, "src_ids": "2022.findings-acl.49_3412"}
+{"input": "domain - specific parameters is done by using Method| context: generalising to unseen domains is under - explored and remains a challenge in neural machine translation .", "entity": "domain - specific parameters", "output": "fusion - based generalisation method", "neg_sample": ["domain - specific parameters is done by using Method", "generalising to unseen domains is under - explored and remains a challenge in neural machine translation ."], "relation": "used for", "id": "2022.findings-acl.49", "year": 2022, "rel_sent": "Inspired by recent research in parameter - efficient transfer learning from pretrained models , this paper proposes a fusion - based generalisation method that learns to combine domain - specific parameters .", "forward": false, "src_ids": "2022.findings-acl.49_3413"}
+{"input": "commonsense inference is done by using Task| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "commonsense inference", "output": "learning from missing relations", "neg_sample": ["commonsense inference is done by using Task", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Learning from Missing Relations : Contrastive Learning with Commonsense Knowledge Graphs for Commonsense Inference.", "forward": false, "src_ids": "2022.findings-acl.119_3414"}
+{"input": "learning commonsense knowledge graphs is done by using OtherScientificTerm| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "learning commonsense knowledge graphs", "output": "missing relations", "neg_sample": ["learning commonsense knowledge graphs is done by using OtherScientificTerm", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "In - depth analysis of SOLAR sheds light on the effects of the missing relations utilized in learning commonsense knowledge graphs .", "forward": false, "src_ids": "2022.findings-acl.119_3415"}
+{"input": "commonsense inference is done by using Method| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "commonsense inference", "output": "contrastive learning", "neg_sample": ["commonsense inference is done by using Method", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Learning from Missing Relations : Contrastive Learning with Commonsense Knowledge Graphs for Commonsense Inference.", "forward": false, "src_ids": "2022.findings-acl.119_3416"}
+{"input": "commonsense inference is done by using OtherScientificTerm| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "commonsense inference", "output": "learning commonsense knowledge graphs", "neg_sample": ["commonsense inference is done by using OtherScientificTerm", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Learning from Missing Relations : Contrastive Learning with Commonsense Knowledge Graphs for Commonsense Inference.", "forward": false, "src_ids": "2022.findings-acl.119_3417"}
+{"input": "missing relations is used for OtherScientificTerm| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "missing relations", "output": "learning commonsense knowledge graphs", "neg_sample": ["missing relations is used for OtherScientificTerm", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "In - depth analysis of SOLAR sheds light on the effects of the missing relations utilized in learning commonsense knowledge graphs .", "forward": true, "src_ids": "2022.findings-acl.119_3418"}
+{"input": "contrastive learning is used for Task| context: however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "contrastive learning", "output": "commonsense inference", "neg_sample": ["contrastive learning is used for Task", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Learning from Missing Relations : Contrastive Learning with Commonsense Knowledge Graphs for Commonsense Inference.", "forward": true, "src_ids": "2022.findings-acl.119_3419"}
+{"input": "learning from missing relations is used for Task| context: however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "learning from missing relations", "output": "commonsense inference", "neg_sample": ["learning from missing relations is used for Task", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Learning from Missing Relations : Contrastive Learning with Commonsense Knowledge Graphs for Commonsense Inference.", "forward": true, "src_ids": "2022.findings-acl.119_3420"}
+{"input": "learning commonsense knowledge graphs is used for Task| context: however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "learning commonsense knowledge graphs", "output": "commonsense inference", "neg_sample": ["learning commonsense knowledge graphs is used for Task", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Learning from Missing Relations : Contrastive Learning with Commonsense Knowledge Graphs for Commonsense Inference.", "forward": true, "src_ids": "2022.findings-acl.119_3421"}
+{"input": "solar is used for Task| context: however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "solar", "output": "commonsense inference", "neg_sample": ["solar is used for Task", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Empirical results demonstrate the efficacy of SOLAR in commonsense inference of diverse commonsense knowledge graphs .", "forward": true, "src_ids": "2022.findings-acl.119_3422"}
+{"input": "commonsense transformers is used for Task| context: however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "commonsense transformers", "output": "commonsense inference", "neg_sample": ["commonsense transformers is used for Task", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Specifically , SOLAR outperforms the state - of - the - art commonsense transformer on commonsense inference with ConceptNet by 1.84 % on average among 8 automatic evaluation metrics .", "forward": true, "src_ids": "2022.findings-acl.119_3423"}
+{"input": "commonsense inference is done by using Method| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "commonsense inference", "output": "commonsense transformers", "neg_sample": ["commonsense inference is done by using Method", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Specifically , SOLAR outperforms the state - of - the - art commonsense transformer on commonsense inference with ConceptNet by 1.84 % on average among 8 automatic evaluation metrics .", "forward": false, "src_ids": "2022.findings-acl.119_3424"}
+{"input": "commonsense inference is done by using Method| context: commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event . existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs . however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality .", "entity": "commonsense inference", "output": "solar", "neg_sample": ["commonsense inference is done by using Method", "commonsense inference poses a unique challenge to reason and generate the physical , social , and causal conditions of a given event .", "existing approaches to commonsense inference utilize commonsense transformers , which are large - scale language models that learn commonsense knowledge graphs .", "however , they suffer from a lack of coverage and expressive diversity of the graphs , resulting in a degradation of the representation quality ."], "relation": "used for", "id": "2022.findings-acl.119", "year": 2022, "rel_sent": "Empirical results demonstrate the efficacy of SOLAR in commonsense inference of diverse commonsense knowledge graphs .", "forward": false, "src_ids": "2022.findings-acl.119_3425"}
+{"input": "concordancer is used for Material| context: this paper presents udconcord , a concordancer web application .", "entity": "concordancer", "output": "universal dependencies treebanks", "neg_sample": ["concordancer is used for Material", "this paper presents udconcord , a concordancer web application ."], "relation": "used for", "id": "2022.udfestbr-1.6", "year": 2022, "rel_sent": "UDConcord : A Concordancer for Universal Dependencies Treebanks.", "forward": true, "src_ids": "2022.udfestbr-1.6_3426"}
+{"input": "udconcord is used for Task| context: this paper presents udconcord , a concordancer web application .", "entity": "udconcord", "output": "sentence annotation", "neg_sample": ["udconcord is used for Task", "this paper presents udconcord , a concordancer web application ."], "relation": "used for", "id": "2022.udfestbr-1.6", "year": 2022, "rel_sent": "UDConcord also allows users to easily modify sentence annotation .", "forward": true, "src_ids": "2022.udfestbr-1.6_3427"}
+{"input": "dravidian languages is done by using Task| context: there were two multi - class classification sub - tasks as a part of this shared task .", "entity": "dravidian languages", "output": "speech and language technologies", "neg_sample": ["dravidian languages is done by using Task", "there were two multi - class classification sub - tasks as a part of this shared task ."], "relation": "used for", "id": "2022.dravidianlangtech-1.14", "year": 2022, "rel_sent": "This paper describes the systems built by our team for the ' Emotion Analysis in Tamil ' shared task at the Second Workshop on Speech and Language Technologies for Dravidian Languages at ACL 2022 .", "forward": false, "src_ids": "2022.dravidianlangtech-1.14_3428"}
+{"input": "speech and language technologies is used for Material| context: there were two multi - class classification sub - tasks as a part of this shared task .", "entity": "speech and language technologies", "output": "dravidian languages", "neg_sample": ["speech and language technologies is used for Material", "there were two multi - class classification sub - tasks as a part of this shared task ."], "relation": "used for", "id": "2022.dravidianlangtech-1.14", "year": 2022, "rel_sent": "This paper describes the systems built by our team for the ' Emotion Analysis in Tamil ' shared task at the Second Workshop on Speech and Language Technologies for Dravidian Languages at ACL 2022 .", "forward": true, "src_ids": "2022.dravidianlangtech-1.14_3429"}
+{"input": "transformer - based language models is done by using Task| context: reasoning using negation is known to be difficult for transformer - based language models . while previous studies have used the tools of psycholinguistics to probe a transformer 's ability to reason over negation , none have focused on the types of negation studied in developmental psychology .", "entity": "transformer - based language models", "output": "developmental negation processing", "neg_sample": ["transformer - based language models is done by using Task", "reasoning using negation is known to be difficult for transformer - based language models .", "while previous studies have used the tools of psycholinguistics to probe a transformer 's ability to reason over negation , none have focused on the types of negation studied in developmental psychology ."], "relation": "used for", "id": "2022.acl-short.60", "year": 2022, "rel_sent": "Developmental Negation Processing in Transformer Language Models.", "forward": false, "src_ids": "2022.acl-short.60_3430"}
+{"input": "developmental negation processing is used for Method| context: while previous studies have used the tools of psycholinguistics to probe a transformer 's ability to reason over negation , none have focused on the types of negation studied in developmental psychology .", "entity": "developmental negation processing", "output": "transformer - based language models", "neg_sample": ["developmental negation processing is used for Method", "while previous studies have used the tools of psycholinguistics to probe a transformer 's ability to reason over negation , none have focused on the types of negation studied in developmental psychology ."], "relation": "used for", "id": "2022.acl-short.60", "year": 2022, "rel_sent": "Developmental Negation Processing in Transformer Language Models.", "forward": true, "src_ids": "2022.acl-short.60_3431"}
+{"input": "negation is done by using Method| context: reasoning using negation is known to be difficult for transformer - based language models . while previous studies have used the tools of psycholinguistics to probe a transformer 's ability to reason over negation , none have focused on the types of negation studied in developmental psychology .", "entity": "negation", "output": "transformers", "neg_sample": ["negation is done by using Method", "reasoning using negation is known to be difficult for transformer - based language models .", "while previous studies have used the tools of psycholinguistics to probe a transformer 's ability to reason over negation , none have focused on the types of negation studied in developmental psychology ."], "relation": "used for", "id": "2022.acl-short.60", "year": 2022, "rel_sent": "We explore how well transformers can process such categories of negation , by framing the problem as a natural language inference ( NLI ) task .", "forward": false, "src_ids": "2022.acl-short.60_3432"}
+{"input": "snippets is done by using Method| context: we model products ' reviews to generate comparative responses consisting of positive and negative experiences regarding the product .", "entity": "snippets", "output": "bert model", "neg_sample": ["snippets is done by using Method", "we model products ' reviews to generate comparative responses consisting of positive and negative experiences regarding the product ."], "relation": "used for", "id": "2022.ecnlp-1.7", "year": 2022, "rel_sent": "We contribute the first dataset for this task of Comparative Snippet Generation from contrasting opinions regarding a product , and an analysis of performance of a pre - trained BERT model to generate such snippets .", "forward": false, "src_ids": "2022.ecnlp-1.7_3433"}
+{"input": "bert model is used for OtherScientificTerm| context: we model products ' reviews to generate comparative responses consisting of positive and negative experiences regarding the product .", "entity": "bert model", "output": "snippets", "neg_sample": ["bert model is used for OtherScientificTerm", "we model products ' reviews to generate comparative responses consisting of positive and negative experiences regarding the product ."], "relation": "used for", "id": "2022.ecnlp-1.7", "year": 2022, "rel_sent": "We contribute the first dataset for this task of Comparative Snippet Generation from contrasting opinions regarding a product , and an analysis of performance of a pre - trained BERT model to generate such snippets .", "forward": true, "src_ids": "2022.ecnlp-1.7_3434"}
+{"input": "prediction is done by using OtherScientificTerm| context: transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive .", "entity": "prediction", "output": "single token embedding", "neg_sample": ["prediction is done by using OtherScientificTerm", "transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive ."], "relation": "used for", "id": "2022.acl-long.602", "year": 2022, "rel_sent": "A recent line of works use various heuristics to successively shorten sequence length while transforming tokens through encoders , in tasks such as classification and ranking that require a single token embedding for prediction . We present a novel solution to this problem , called Pyramid - BERT where we replace previously used heuristics with a core - set based token selection method justified by theoretical results .", "forward": false, "src_ids": "2022.acl-long.602_3435"}
+{"input": "single token embedding is used for Task| context: transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive .", "entity": "single token embedding", "output": "prediction", "neg_sample": ["single token embedding is used for Task", "transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive ."], "relation": "used for", "id": "2022.acl-long.602", "year": 2022, "rel_sent": "A recent line of works use various heuristics to successively shorten sequence length while transforming tokens through encoders , in tasks such as classification and ranking that require a single token embedding for prediction . We present a novel solution to this problem , called Pyramid - BERT where we replace previously used heuristics with a core - set based token selection method justified by theoretical results .", "forward": true, "src_ids": "2022.acl-long.602_3436"}
+{"input": "sequence length is done by using Method| context: transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive .", "entity": "sequence length", "output": "heuristics", "neg_sample": ["sequence length is done by using Method", "transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive ."], "relation": "used for", "id": "2022.acl-long.602", "year": 2022, "rel_sent": "A recent line of works use various heuristics to successively shorten sequence length while transforming tokens through encoders , in tasks such as classification and ranking that require a single token embedding for prediction . We present a novel solution to this problem , called Pyramid - BERT where we replace previously used heuristics with a core - set based token selection method justified by theoretical results .", "forward": false, "src_ids": "2022.acl-long.602_3437"}
+{"input": "heuristics is used for OtherScientificTerm| context: transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive .", "entity": "heuristics", "output": "sequence length", "neg_sample": ["heuristics is used for OtherScientificTerm", "transformer - based language models such as bert ( citation ) have achieved the state - of - the - art performance on various nlp tasks , but are computationally prohibitive ."], "relation": "used for", "id": "2022.acl-long.602", "year": 2022, "rel_sent": "A recent line of works use various heuristics to successively shorten sequence length while transforming tokens through encoders , in tasks such as classification and ranking that require a single token embedding for prediction . We present a novel solution to this problem , called Pyramid - BERT where we replace previously used heuristics with a core - set based token selection method justified by theoretical results .", "forward": true, "src_ids": "2022.acl-long.602_3438"}
+{"input": "english and spanish sentences is done by using OtherScientificTerm| context: whilst it is possible to automatically translate this information , the resulting text may contain specialised terminology , or may be written in a style that is difficult for lay readers to understand .", "entity": "english and spanish sentences", "output": "manual simplifications", "neg_sample": ["english and spanish sentences is done by using OtherScientificTerm", "whilst it is possible to automatically translate this information , the resulting text may contain specialised terminology , or may be written in a style that is difficult for lay readers to understand ."], "relation": "used for", "id": "2022.eamt-1.33", "year": 2022, "rel_sent": "Sofar , we have collected a new dataset with manual simplifications for English and Spanish sentences in the TICO-19 dataset , as well as implemented baseline pipelines combining Machine Translation and Text Simplification models .", "forward": false, "src_ids": "2022.eamt-1.33_3439"}
+{"input": "manual simplifications is used for Material| context: whilst it is possible to automatically translate this information , the resulting text may contain specialised terminology , or may be written in a style that is difficult for lay readers to understand .", "entity": "manual simplifications", "output": "english and spanish sentences", "neg_sample": ["manual simplifications is used for Material", "whilst it is possible to automatically translate this information , the resulting text may contain specialised terminology , or may be written in a style that is difficult for lay readers to understand ."], "relation": "used for", "id": "2022.eamt-1.33", "year": 2022, "rel_sent": "Sofar , we have collected a new dataset with manual simplifications for English and Spanish sentences in the TICO-19 dataset , as well as implemented baseline pipelines combining Machine Translation and Text Simplification models .", "forward": true, "src_ids": "2022.eamt-1.33_3440"}
+{"input": "labeling scheme is used for OtherScientificTerm| context: entity recognition is a fundamental task in understanding document images . traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice .", "entity": "labeling scheme", "output": "label surface names", "neg_sample": ["labeling scheme is used for OtherScientificTerm", "entity recognition is a fundamental task in understanding document images .", "traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice ."], "relation": "used for", "id": "2022.findings-acl.329", "year": 2022, "rel_sent": "The proposed model follows a new labeling scheme that generates the label surface names word - by - word explicitly after generating the entities .", "forward": true, "src_ids": "2022.findings-acl.329_3441"}
+{"input": "label surface names is done by using Method| context: entity recognition is a fundamental task in understanding document images . traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice .", "entity": "label surface names", "output": "labeling scheme", "neg_sample": ["label surface names is done by using Method", "entity recognition is a fundamental task in understanding document images .", "traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice ."], "relation": "used for", "id": "2022.findings-acl.329", "year": 2022, "rel_sent": "The proposed model follows a new labeling scheme that generates the label surface names word - by - word explicitly after generating the entities .", "forward": false, "src_ids": "2022.findings-acl.329_3442"}
+{"input": "laser is used for OtherScientificTerm| context: entity recognition is a fundamental task in understanding document images . traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice .", "entity": "laser", "output": "label semantics", "neg_sample": ["laser is used for OtherScientificTerm", "entity recognition is a fundamental task in understanding document images .", "traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice ."], "relation": "used for", "id": "2022.findings-acl.329", "year": 2022, "rel_sent": "During training , LASER refines the label semantics by updating the label surface name representations and also strengthens the label - region correlation .", "forward": true, "src_ids": "2022.findings-acl.329_3443"}
+{"input": "label semantics is done by using Method| context: entity recognition is a fundamental task in understanding document images . traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice .", "entity": "label semantics", "output": "laser", "neg_sample": ["label semantics is done by using Method", "entity recognition is a fundamental task in understanding document images .", "traditional sequence labeling frameworks treat the entity types as class ids and rely on extensive data and high - quality annotations to learn semantics which are typically expensive in practice ."], "relation": "used for", "id": "2022.findings-acl.329", "year": 2022, "rel_sent": "During training , LASER refines the label semantics by updating the label surface name representations and also strengthens the label - region correlation .", "forward": false, "src_ids": "2022.findings-acl.329_3444"}
+{"input": "antibiotic resistance genes is done by using Method| context: antibiotic resistance has become a growing worldwide concern as new resistance mechanisms are emerging and spreading globally , and thus detecting and collecting the cause - antibiotic resistance genes ( args ) , have been more critical than ever .", "entity": "antibiotic resistance genes", "output": "natural language processing techniques", "neg_sample": ["antibiotic resistance genes is done by using Method", "antibiotic resistance has become a growing worldwide concern as new resistance mechanisms are emerging and spreading globally , and thus detecting and collecting the cause - antibiotic resistance genes ( args ) , have been more critical than ever ."], "relation": "used for", "id": "2022.bionlp-1.40", "year": 2022, "rel_sent": "To the best of our knowledge , we are the first to leverage natural language processing techniques to curate all validated ARGs from scientific papers .", "forward": false, "src_ids": "2022.bionlp-1.40_3445"}
+{"input": "inference is done by using Method| context: building huge and highly capable language models has been a trend in the past years . despite their great performance , they incur high computational cost . a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression .", "entity": "inference", "output": "dynamic inference approach", "neg_sample": ["inference is done by using Method", "building huge and highly capable language models has been a trend in the past years .", "despite their great performance , they incur high computational cost .", "a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression ."], "relation": "used for", "id": "2022.acl-long.359", "year": 2022, "rel_sent": "This paper proposes an effective dynamic inference approach , called E - LANG , which distributes the inference between large accurate Super - models and light - weight Swift models .", "forward": false, "src_ids": "2022.acl-long.359_3446"}
+{"input": "dynamic inference approach is used for Task| context: building huge and highly capable language models has been a trend in the past years . despite their great performance , they incur high computational cost . a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression .", "entity": "dynamic inference approach", "output": "inference", "neg_sample": ["dynamic inference approach is used for Task", "building huge and highly capable language models has been a trend in the past years .", "despite their great performance , they incur high computational cost .", "a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression ."], "relation": "used for", "id": "2022.acl-long.359", "year": 2022, "rel_sent": "This paper proposes an effective dynamic inference approach , called E - LANG , which distributes the inference between large accurate Super - models and light - weight Swift models .", "forward": true, "src_ids": "2022.acl-long.359_3447"}
+{"input": "super or swift models is done by using Method| context: building huge and highly capable language models has been a trend in the past years . despite their great performance , they incur high computational cost . a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression .", "entity": "super or swift models", "output": "decision making module", "neg_sample": ["super or swift models is done by using Method", "building huge and highly capable language models has been a trend in the past years .", "despite their great performance , they incur high computational cost .", "a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression ."], "relation": "used for", "id": "2022.acl-long.359", "year": 2022, "rel_sent": "To this end , a decision making module routes the inputs to Super or Swift models based on the energy characteristics of the representations in the latent space .", "forward": false, "src_ids": "2022.acl-long.359_3448"}
+{"input": "decision making module is used for Method| context: building huge and highly capable language models has been a trend in the past years . despite their great performance , they incur high computational cost . a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression .", "entity": "decision making module", "output": "super or swift models", "neg_sample": ["decision making module is used for Method", "building huge and highly capable language models has been a trend in the past years .", "despite their great performance , they incur high computational cost .", "a common solution is to apply model compression or choose light - weight architectures , which often need a separate fixed - size model for each desirable computational budget , and may lose performance in case of heavy compression ."], "relation": "used for", "id": "2022.acl-long.359", "year": 2022, "rel_sent": "To this end , a decision making module routes the inputs to Super or Swift models based on the energy characteristics of the representations in the latent space .", "forward": true, "src_ids": "2022.acl-long.359_3449"}
+{"input": "self - supervised learning of phoneme inventory is done by using Method| context: phonemes are defined by their relationship to words : changing a phoneme changes the word . learning a phoneme inventory with little supervision has been a longstanding challenge with important applications to under - resourced speech technology .", "entity": "self - supervised learning of phoneme inventory", "output": "neural discrete representation learning model", "neg_sample": ["self - supervised learning of phoneme inventory is done by using Method", "phonemes are defined by their relationship to words : changing a phoneme changes the word .", "learning a phoneme inventory with little supervision has been a longstanding challenge with important applications to under - resourced speech technology ."], "relation": "used for", "id": "2022.acl-long.553", "year": 2022, "rel_sent": "In this paper , we bridge the gap between the linguistic and statistical definition of phonemes and propose a novel neural discrete representation learning model for self - supervised learning of phoneme inventory with raw speech and word labels .", "forward": false, "src_ids": "2022.acl-long.553_3450"}
+{"input": "neural discrete representation learning model is used for Task| context: phonemes are defined by their relationship to words : changing a phoneme changes the word . learning a phoneme inventory with little supervision has been a longstanding challenge with important applications to under - resourced speech technology .", "entity": "neural discrete representation learning model", "output": "self - supervised learning of phoneme inventory", "neg_sample": ["neural discrete representation learning model is used for Task", "phonemes are defined by their relationship to words : changing a phoneme changes the word .", "learning a phoneme inventory with little supervision has been a longstanding challenge with important applications to under - resourced speech technology ."], "relation": "used for", "id": "2022.acl-long.553", "year": 2022, "rel_sent": "In this paper , we bridge the gap between the linguistic and statistical definition of phonemes and propose a novel neural discrete representation learning model for self - supervised learning of phoneme inventory with raw speech and word labels .", "forward": true, "src_ids": "2022.acl-long.553_3451"}
+{"input": "pointwise mutual information ( pmi ) is used for OtherScientificTerm| context: bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary . his ) .", "entity": "pointwise mutual information ( pmi )", "output": "mwes", "neg_sample": ["pointwise mutual information ( pmi ) is used for OtherScientificTerm", "bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary .", "his ) ."], "relation": "used for", "id": "2022.insights-1.24", "year": 2022, "rel_sent": "Moreover , we find that Pointwise Mutual Information ( PMI ) instead of frequency finds better MWEs ( e.g. , New_York , Statue_of_Liberty , neither .", "forward": true, "src_ids": "2022.insights-1.24_3452"}
+{"input": "neural machine translation is done by using OtherScientificTerm| context: bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary . an intuitive relaxation would be to extend a bpe vocabulary with multi - word expressions ( mwes ): bigrams ( in_a ) , trigrams ( out_of_the ) , and skip - grams ( he . his ) .", "entity": "neural machine translation", "output": "multi word expressions", "neg_sample": ["neural machine translation is done by using OtherScientificTerm", "bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary .", "an intuitive relaxation would be to extend a bpe vocabulary with multi - word expressions ( mwes ): bigrams ( in_a ) , trigrams ( out_of_the ) , and skip - grams ( he .", "his ) ."], "relation": "used for", "id": "2022.insights-1.24", "year": 2022, "rel_sent": "BPE beyond Word Boundary : How NOT to use Multi Word Expressions in Neural Machine Translation.", "forward": false, "src_ids": "2022.insights-1.24_3453"}
+{"input": "multi word expressions is used for Task| context: bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary . an intuitive relaxation would be to extend a bpe vocabulary with multi - word expressions ( mwes ): bigrams ( in_a ) , trigrams ( out_of_the ) , and skip - grams ( he . his ) .", "entity": "multi word expressions", "output": "neural machine translation", "neg_sample": ["multi word expressions is used for Task", "bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary .", "an intuitive relaxation would be to extend a bpe vocabulary with multi - word expressions ( mwes ): bigrams ( in_a ) , trigrams ( out_of_the ) , and skip - grams ( he .", "his ) ."], "relation": "used for", "id": "2022.insights-1.24", "year": 2022, "rel_sent": "BPE beyond Word Boundary : How NOT to use Multi Word Expressions in Neural Machine Translation.", "forward": true, "src_ids": "2022.insights-1.24_3454"}
+{"input": "mwes is done by using OtherScientificTerm| context: bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary . an intuitive relaxation would be to extend a bpe vocabulary with multi - word expressions ( mwes ): bigrams ( in_a ) , trigrams ( out_of_the ) , and skip - grams ( he . his ) .", "entity": "mwes", "output": "pointwise mutual information ( pmi )", "neg_sample": ["mwes is done by using OtherScientificTerm", "bpe tokenization merges characters into longer tokens by finding frequently occurring contiguous patterns within the word boundary .", "an intuitive relaxation would be to extend a bpe vocabulary with multi - word expressions ( mwes ): bigrams ( in_a ) , trigrams ( out_of_the ) , and skip - grams ( he .", "his ) ."], "relation": "used for", "id": "2022.insights-1.24", "year": 2022, "rel_sent": "Moreover , we find that Pointwise Mutual Information ( PMI ) instead of frequency finds better MWEs ( e.g. , New_York , Statue_of_Liberty , neither .", "forward": false, "src_ids": "2022.insights-1.24_3455"}
+{"input": "real - world scenarios is done by using Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "entity": "real - world scenarios", "output": "big models", "neg_sample": ["real - world scenarios is done by using Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "To address the computation bottleneck encountered in deploying big models in real - world scenarios , we introduce an open - source toolkit for big model inference and tuning ( BMInf ) , which can support big model inference and tuning at extremely low computation cost .", "forward": false, "src_ids": "2022.acl-demo.22_3456"}
+{"input": "cpu - gpu scheduling optimization is used for Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "entity": "cpu - gpu scheduling optimization", "output": "big models", "neg_sample": ["cpu - gpu scheduling optimization is used for Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "At the implementation level , we apply model offloading , model checkpointing , and CPU - GPU scheduling optimization tofurther reduce the computation and memory cost of big models .", "forward": true, "src_ids": "2022.acl-demo.22_3457"}
+{"input": "model checkpointing is used for Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "entity": "model checkpointing", "output": "big models", "neg_sample": ["model checkpointing is used for Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "At the implementation level , we apply model offloading , model checkpointing , and CPU - GPU scheduling optimization tofurther reduce the computation and memory cost of big models .", "forward": true, "src_ids": "2022.acl-demo.22_3458"}
+{"input": "big models is used for OtherScientificTerm| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "big models", "output": "real - world scenarios", "neg_sample": ["big models is used for OtherScientificTerm", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "To address the computation bottleneck encountered in deploying big models in real - world scenarios , we introduce an open - source toolkit for big model inference and tuning ( BMInf ) , which can support big model inference and tuning at extremely low computation cost .", "forward": true, "src_ids": "2022.acl-demo.22_3459"}
+{"input": "model inference and tuning is done by using Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "model inference and tuning", "output": "model quantization", "neg_sample": ["model inference and tuning is done by using Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "More specifically , at the algorithm level , we introduce model quantization and parameter - efficient tuning for efficient model inference and tuning .", "forward": false, "src_ids": "2022.acl-demo.22_3460"}
+{"input": "parameter - efficient tuning is used for Task| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "parameter - efficient tuning", "output": "model inference and tuning", "neg_sample": ["parameter - efficient tuning is used for Task", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "More specifically , at the algorithm level , we introduce model quantization and parameter - efficient tuning for efficient model inference and tuning .", "forward": true, "src_ids": "2022.acl-demo.22_3461"}
+{"input": "model quantization is used for Task| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "model quantization", "output": "model inference and tuning", "neg_sample": ["model quantization is used for Task", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "More specifically , at the algorithm level , we introduce model quantization and parameter - efficient tuning for efficient model inference and tuning .", "forward": true, "src_ids": "2022.acl-demo.22_3462"}
+{"input": "big models is done by using Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "big models", "output": "model checkpointing", "neg_sample": ["big models is done by using Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "At the implementation level , we apply model offloading , model checkpointing , and CPU - GPU scheduling optimization tofurther reduce the computation and memory cost of big models .", "forward": false, "src_ids": "2022.acl-demo.22_3463"}
+{"input": "computation and memory cost is done by using Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "computation and memory cost", "output": "cpu - gpu scheduling optimization", "neg_sample": ["computation and memory cost is done by using Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "At the implementation level , we apply model offloading , model checkpointing , and CPU - GPU scheduling optimization tofurther reduce the computation and memory cost of big models .", "forward": false, "src_ids": "2022.acl-demo.22_3464"}
+{"input": "cpu - gpu scheduling optimization is used for Metric| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "cpu - gpu scheduling optimization", "output": "computation and memory cost", "neg_sample": ["cpu - gpu scheduling optimization is used for Metric", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "At the implementation level , we apply model offloading , model checkpointing , and CPU - GPU scheduling optimization tofurther reduce the computation and memory cost of big models .", "forward": true, "src_ids": "2022.acl-demo.22_3465"}
+{"input": "pre - trained language models is done by using Method| context: in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks . although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "pre - trained language models", "output": "distributed learning toolkits", "neg_sample": ["pre - trained language models is done by using Method", "in recent years , large - scale pre - trained language models ( plms ) containing billions of parameters have achieved promising results on various nlp tasks .", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "Based on above efforts , we can efficiently perform big model inference and tuning with a single GPU ( even a consumer - level GPU like GTX 1060 ) instead of computing clusters , which is difficult for existing distributed learning toolkits for PLMs .", "forward": false, "src_ids": "2022.acl-demo.22_3466"}
+{"input": "distributed learning toolkits is used for Method| context: although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task .", "entity": "distributed learning toolkits", "output": "pre - trained language models", "neg_sample": ["distributed learning toolkits is used for Method", "although we can pre - train these big models by stacking computing clusters at any cost , it is impractical to use such huge computing resources to apply big models for each downstream task ."], "relation": "used for", "id": "2022.acl-demo.22", "year": 2022, "rel_sent": "Based on above efforts , we can efficiently perform big model inference and tuning with a single GPU ( even a consumer - level GPU like GTX 1060 ) instead of computing clusters , which is difficult for existing distributed learning toolkits for PLMs .", "forward": true, "src_ids": "2022.acl-demo.22_3467"}
+{"input": "few - shot methods is used for Task| context: the few - shot natural language understanding ( nlu ) task has attracted much recent attention . however , prior methods have been evaluated under a disparate set of protocols , which hinders fair comparison and measuring the progress of the field .", "entity": "few - shot methods", "output": "nlu tasks", "neg_sample": ["few - shot methods is used for Task", "the few - shot natural language understanding ( nlu ) task has attracted much recent attention .", "however , prior methods have been evaluated under a disparate set of protocols , which hinders fair comparison and measuring the progress of the field ."], "relation": "used for", "id": "2022.acl-long.38", "year": 2022, "rel_sent": "Under this new evaluation framework , we re - evaluate several state - of - the - art few - shot methods for NLU tasks .", "forward": true, "src_ids": "2022.acl-long.38_3468"}
+{"input": "nlu tasks is done by using Method| context: the few - shot natural language understanding ( nlu ) task has attracted much recent attention . however , prior methods have been evaluated under a disparate set of protocols , which hinders fair comparison and measuring the progress of the field .", "entity": "nlu tasks", "output": "few - shot methods", "neg_sample": ["nlu tasks is done by using Method", "the few - shot natural language understanding ( nlu ) task has attracted much recent attention .", "however , prior methods have been evaluated under a disparate set of protocols , which hinders fair comparison and measuring the progress of the field ."], "relation": "used for", "id": "2022.acl-long.38", "year": 2022, "rel_sent": "Under this new evaluation framework , we re - evaluate several state - of - the - art few - shot methods for NLU tasks .", "forward": false, "src_ids": "2022.acl-long.38_3469"}
+{"input": "routing fluctuation problem is done by using Method| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "routing fluctuation problem", "output": "stablemoe", "neg_sample": ["routing fluctuation problem is done by using Method", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In this paper , we propose StableMoE with two training stages to address the routing fluctuation problem .", "forward": false, "src_ids": "2022.acl-long.489_3470"}
+{"input": "mixture of experts is done by using Method| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "mixture of experts", "output": "stable routing strategy", "neg_sample": ["mixture of experts is done by using Method", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "StableMoE : Stable Routing Strategy for Mixture of Experts.", "forward": false, "src_ids": "2022.acl-long.489_3471"}
+{"input": "stable routing strategy is used for Method| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "stable routing strategy", "output": "mixture of experts", "neg_sample": ["stable routing strategy is used for Method", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "StableMoE : Stable Routing Strategy for Mixture of Experts.", "forward": true, "src_ids": "2022.acl-long.489_3472"}
+{"input": "stablemoe is used for Task| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "entity": "stablemoe", "output": "routing fluctuation problem", "neg_sample": ["stablemoe is used for Task", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In this paper , we propose StableMoE with two training stages to address the routing fluctuation problem .", "forward": true, "src_ids": "2022.acl-long.489_3473"}
+{"input": "training stages is used for Task| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "entity": "training stages", "output": "routing fluctuation problem", "neg_sample": ["training stages is used for Task", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In this paper , we propose StableMoE with two training stages to address the routing fluctuation problem .", "forward": true, "src_ids": "2022.acl-long.489_3474"}
+{"input": "routing fluctuation problem is done by using Generic| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "routing fluctuation problem", "output": "training stages", "neg_sample": ["routing fluctuation problem is done by using Generic", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In this paper , we propose StableMoE with two training stages to address the routing fluctuation problem .", "forward": false, "src_ids": "2022.acl-long.489_3475"}
+{"input": "balanced and cohesive routing strategy is done by using Generic| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "balanced and cohesive routing strategy", "output": "training stages", "neg_sample": ["balanced and cohesive routing strategy is done by using Generic", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In the first training stage , we learn a balanced and cohesive routing strategy and distill it into a lightweight router decoupled from the backbone model .", "forward": false, "src_ids": "2022.acl-long.489_3476"}
+{"input": "training stages is used for Method| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "training stages", "output": "balanced and cohesive routing strategy", "neg_sample": ["training stages is used for Method", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In the first training stage , we learn a balanced and cohesive routing strategy and distill it into a lightweight router decoupled from the backbone model .", "forward": true, "src_ids": "2022.acl-long.489_3477"}
+{"input": "token - to - expert assignment is done by using OtherScientificTerm| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "token - to - expert assignment", "output": "distilled router", "neg_sample": ["token - to - expert assignment is done by using OtherScientificTerm", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In the second training stage , we utilize the distilled router to determine the token - to - expert assignment and freeze it for a stable routing strategy .", "forward": false, "src_ids": "2022.acl-long.489_3478"}
+{"input": "distilled router is used for OtherScientificTerm| context: the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead . we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e. , the target expert of the same input may change along with training , but only one expert will be activated for the input during inference . the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used .", "entity": "distilled router", "output": "token - to - expert assignment", "neg_sample": ["distilled router is used for OtherScientificTerm", "the mixture - of - experts ( moe ) technique can scale up the model size of transformers with an affordable computational overhead .", "we point out that existing learning - to - route moe methods suffer from the routing fluctuation issue , i.e.", ", the target expert of the same input may change along with training , but only one expert will be activated for the input during inference .", "the routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used ."], "relation": "used for", "id": "2022.acl-long.489", "year": 2022, "rel_sent": "In the second training stage , we utilize the distilled router to determine the token - to - expert assignment and freeze it for a stable routing strategy .", "forward": true, "src_ids": "2022.acl-long.489_3479"}
+{"input": "part - of - speech model is used for OtherScientificTerm| context: the tree model is well known for expressing the historic evolution of languages . this model has been considered as a method of describing genetic relationships between languages . nevertheless , some researchers question the model 's ability to predict the proximity between two languages , since it represents genetic relatedness rather than linguistic resemblance . defining other language proximity models has been an active research area for many years .", "entity": "part - of - speech model", "output": "proximity between languages", "neg_sample": ["part - of - speech model is used for OtherScientificTerm", "the tree model is well known for expressing the historic evolution of languages .", "this model has been considered as a method of describing genetic relationships between languages .", "nevertheless , some researchers question the model 's ability to predict the proximity between two languages , since it represents genetic relatedness rather than linguistic resemblance .", "defining other language proximity models has been an active research area for many years ."], "relation": "used for", "id": "2022.lchange-1.8", "year": 2022, "rel_sent": "In this paper we explore a part - of - speech model for defining proximity between languages using a multilingual language model that was fine - tuned on the task of cross - lingual part - of - speech tagging .", "forward": true, "src_ids": "2022.lchange-1.8_3480"}
+{"input": "proximity between languages is done by using Method| context: the tree model is well known for expressing the historic evolution of languages . this model has been considered as a method of describing genetic relationships between languages . nevertheless , some researchers question the model 's ability to predict the proximity between two languages , since it represents genetic relatedness rather than linguistic resemblance . defining other language proximity models has been an active research area for many years .", "entity": "proximity between languages", "output": "part - of - speech model", "neg_sample": ["proximity between languages is done by using Method", "the tree model is well known for expressing the historic evolution of languages .", "this model has been considered as a method of describing genetic relationships between languages .", "nevertheless , some researchers question the model 's ability to predict the proximity between two languages , since it represents genetic relatedness rather than linguistic resemblance .", "defining other language proximity models has been an active research area for many years ."], "relation": "used for", "id": "2022.lchange-1.8", "year": 2022, "rel_sent": "In this paper we explore a part - of - speech model for defining proximity between languages using a multilingual language model that was fine - tuned on the task of cross - lingual part - of - speech tagging .", "forward": false, "src_ids": "2022.lchange-1.8_3481"}
+{"input": "ss - aga is used for Method| context: due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages . however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner .", "entity": "ss - aga", "output": "kgs", "neg_sample": ["ss - aga is used for Method", "due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages .", "however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner ."], "relation": "used for", "id": "2022.acl-long.36", "year": 2022, "rel_sent": "Specifically , SS - AGA fuses all KGs as a whole graph by regarding alignment as a new edge type .", "forward": true, "src_ids": "2022.acl-long.36_3482"}
+{"input": "collective knowledge is done by using Task| context: predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete . due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages . however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner .", "entity": "collective knowledge", "output": "multilingual kg completion", "neg_sample": ["collective knowledge is done by using Task", "predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete .", "due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages .", "however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner ."], "relation": "used for", "id": "2022.acl-long.36", "year": 2022, "rel_sent": "In this paper , we explore multilingual KG completion , which leverages limited seed alignment as a bridge , to embrace the collective knowledge from multiple languages .", "forward": false, "src_ids": "2022.acl-long.36_3483"}
+{"input": "multilingual kg completion is used for OtherScientificTerm| context: predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete . due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages . however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner .", "entity": "multilingual kg completion", "output": "collective knowledge", "neg_sample": ["multilingual kg completion is used for OtherScientificTerm", "predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete .", "due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages .", "however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner ."], "relation": "used for", "id": "2022.acl-long.36", "year": 2022, "rel_sent": "In this paper , we explore multilingual KG completion , which leverages limited seed alignment as a bridge , to embrace the collective knowledge from multiple languages .", "forward": true, "src_ids": "2022.acl-long.36_3484"}
+{"input": "kgs is done by using Method| context: predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete . due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages . however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner .", "entity": "kgs", "output": "ss - aga", "neg_sample": ["kgs is done by using Method", "predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete .", "due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages .", "however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner ."], "relation": "used for", "id": "2022.acl-long.36", "year": 2022, "rel_sent": "Specifically , SS - AGA fuses all KGs as a whole graph by regarding alignment as a new edge type .", "forward": false, "src_ids": "2022.acl-long.36_3485"}
+{"input": "alignment pairs is done by using Method| context: predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete . due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages . however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner .", "entity": "alignment pairs", "output": "pair generator", "neg_sample": ["alignment pairs is done by using Method", "predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete .", "due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages .", "however , language alignment used in prior works is still not fully exploited : ( 1 ) alignment pairs are treated equally to maximally push parallel entities to be close , which ignores kg capacity inconsistency ; ( 2 ) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner ."], "relation": "used for", "id": "2022.acl-long.36", "year": 2022, "rel_sent": "Meanwhile , SS - AGA features a new pair generator that dynamically captures potential alignment pairs in a self - supervised paradigm .", "forward": false, "src_ids": "2022.acl-long.36_3486"}
+{"input": "pair generator is used for OtherScientificTerm| context: predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete . due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages .", "entity": "pair generator", "output": "alignment pairs", "neg_sample": ["pair generator is used for OtherScientificTerm", "predicting missing facts in a knowledge graph ( kg ) is crucial as modern kgs are far from complete .", "due to labor - intensive human labeling , this phenomenon deteriorates when handling knowledge represented in various languages ."], "relation": "used for", "id": "2022.acl-long.36", "year": 2022, "rel_sent": "Meanwhile , SS - AGA features a new pair generator that dynamically captures potential alignment pairs in a self - supervised paradigm .", "forward": true, "src_ids": "2022.acl-long.36_3487"}
+{"input": "readability assessment is done by using Method| context: automatic readability assessment ( ara ) , the task of assigning a reading level to a text , is traditionally treated as a classification problem in nlp research .", "entity": "readability assessment", "output": "neural , pairwise ranking approach", "neg_sample": ["readability assessment is done by using Method", "automatic readability assessment ( ara ) , the task of assigning a reading level to a text , is traditionally treated as a classification problem in nlp research ."], "relation": "used for", "id": "2022.findings-acl.300", "year": 2022, "rel_sent": "A Neural Pairwise Ranking Model for Readability Assessment.", "forward": false, "src_ids": "2022.findings-acl.300_3488"}
+{"input": "automatic readability assessment ( ara ) is done by using Method| context: automatic readability assessment ( ara ) , the task of assigning a reading level to a text , is traditionally treated as a classification problem in nlp research .", "entity": "automatic readability assessment ( ara )", "output": "neural , pairwise ranking approach", "neg_sample": ["automatic readability assessment ( ara ) is done by using Method", "automatic readability assessment ( ara ) , the task of assigning a reading level to a text , is traditionally treated as a classification problem in nlp research ."], "relation": "used for", "id": "2022.findings-acl.300", "year": 2022, "rel_sent": "In this paper , we propose the first neural , pairwise ranking approach to ARA and compare it with existing classification , regression , and ( non - neural ) ranking methods .", "forward": false, "src_ids": "2022.findings-acl.300_3489"}
+{"input": "automatic readability assessment ( ara ) is done by using Method| context: automatic readability assessment ( ara ) , the task of assigning a reading level to a text , is traditionally treated as a classification problem in nlp research .", "entity": "automatic readability assessment ( ara )", "output": "neural , pairwise ranking approach", "neg_sample": ["automatic readability assessment ( ara ) is done by using Method", "automatic readability assessment ( ara ) , the task of assigning a reading level to a text , is traditionally treated as a classification problem in nlp research ."], "relation": "used for", "id": "2022.findings-acl.300", "year": 2022, "rel_sent": "To our knowledge , this paper proposes the first neural pairwise ranking model for ARA , and shows the first results of cross - lingual , zero - shot evaluation of ARA with neural models .", "forward": false, "src_ids": "2022.findings-acl.300_3490"}
+{"input": "stance detection is done by using Material| context: stance detection infers a text author 's attitude towards a target . this is challenging when the model lacks background knowledge about the target .", "entity": "stance detection", "output": "wikipedia", "neg_sample": ["stance detection is done by using Material", "stance detection infers a text author 's attitude towards a target .", "this is challenging when the model lacks background knowledge about the target ."], "relation": "used for", "id": "2022.wassa-1.7", "year": 2022, "rel_sent": "Infusing Knowledge from Wikipedia to Enhance Stance Detection.", "forward": false, "src_ids": "2022.wassa-1.7_3491"}
+{"input": "wikipedia is used for Task| context: this is challenging when the model lacks background knowledge about the target .", "entity": "wikipedia", "output": "stance detection", "neg_sample": ["wikipedia is used for Task", "this is challenging when the model lacks background knowledge about the target ."], "relation": "used for", "id": "2022.wassa-1.7", "year": 2022, "rel_sent": "Infusing Knowledge from Wikipedia to Enhance Stance Detection.", "forward": true, "src_ids": "2022.wassa-1.7_3492"}
+{"input": "background knowledge is used for Task| context: this is challenging when the model lacks background knowledge about the target .", "entity": "background knowledge", "output": "stance detection", "neg_sample": ["background knowledge is used for Task", "this is challenging when the model lacks background knowledge about the target ."], "relation": "used for", "id": "2022.wassa-1.7", "year": 2022, "rel_sent": "Here , we show how background knowledge from Wikipedia can help enhance the performance on stance detection .", "forward": true, "src_ids": "2022.wassa-1.7_3493"}
+{"input": "stance detection is done by using OtherScientificTerm| context: stance detection infers a text author 's attitude towards a target .", "entity": "stance detection", "output": "background knowledge", "neg_sample": ["stance detection is done by using OtherScientificTerm", "stance detection infers a text author 's attitude towards a target ."], "relation": "used for", "id": "2022.wassa-1.7", "year": 2022, "rel_sent": "Here , we show how background knowledge from Wikipedia can help enhance the performance on stance detection .", "forward": false, "src_ids": "2022.wassa-1.7_3494"}
+{"input": "annotations is used for Task| context: it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g. , semi - structured attributes , text descriptions , user - provided contents , etc .", "entity": "annotations", "output": "answer generation", "neg_sample": ["annotations is used for Task", "it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g.", ", semi - structured attributes , text descriptions , user - provided contents , etc ."], "relation": "used for", "id": "2022.ecnlp-1.13", "year": 2022, "rel_sent": "In this paper , we build a benchmark with annotations for both evidence selection and answer generation covering 6 information sources .", "forward": true, "src_ids": "2022.ecnlp-1.13_3495"}
+{"input": "data augmentation method is used for Task| context: it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g. , semi - structured attributes , text descriptions , user - provided contents , etc .", "entity": "data augmentation method", "output": "answer generation", "neg_sample": ["data augmentation method is used for Task", "it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g.", ", semi - structured attributes , text descriptions , user - provided contents , etc ."], "relation": "used for", "id": "2022.ecnlp-1.13", "year": 2022, "rel_sent": "We further propose a novel data augmentation method to iteratively create training samples for answer generation , which achieves close - to - human performance with only a few thousandannotations .", "forward": true, "src_ids": "2022.ecnlp-1.13_3496"}
+{"input": "answer generation is done by using Method| context: it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g. , semi - structured attributes , text descriptions , user - provided contents , etc . however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation .", "entity": "answer generation", "output": "data augmentation method", "neg_sample": ["answer generation is done by using Method", "it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g.", ", semi - structured attributes , text descriptions , user - provided contents , etc .", "however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation ."], "relation": "used for", "id": "2022.ecnlp-1.13", "year": 2022, "rel_sent": "We further propose a novel data augmentation method to iteratively create training samples for answer generation , which achieves close - to - human performance with only a few thousandannotations .", "forward": false, "src_ids": "2022.ecnlp-1.13_3497"}
+{"input": "evidence selection is done by using OtherScientificTerm| context: it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g. , semi - structured attributes , text descriptions , user - provided contents , etc . however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation .", "entity": "evidence selection", "output": "annotations", "neg_sample": ["evidence selection is done by using OtherScientificTerm", "it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g.", ", semi - structured attributes , text descriptions , user - provided contents , etc .", "however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation ."], "relation": "used for", "id": "2022.ecnlp-1.13", "year": 2022, "rel_sent": "In this paper , we build a benchmark with annotations for both evidence selection and answer generation covering 6 information sources .", "forward": false, "src_ids": "2022.ecnlp-1.13_3498"}
+{"input": "answer generation is done by using OtherScientificTerm| context: it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g. , semi - structured attributes , text descriptions , user - provided contents , etc . however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation .", "entity": "answer generation", "output": "annotations", "neg_sample": ["answer generation is done by using OtherScientificTerm", "it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g.", ", semi - structured attributes , text descriptions , user - provided contents , etc .", "however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation ."], "relation": "used for", "id": "2022.ecnlp-1.13", "year": 2022, "rel_sent": "In this paper , we build a benchmark with annotations for both evidence selection and answer generation covering 6 information sources .", "forward": false, "src_ids": "2022.ecnlp-1.13_3499"}
+{"input": "annotations is used for Task| context: it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g. , semi - structured attributes , text descriptions , user - provided contents , etc . however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation .", "entity": "annotations", "output": "evidence selection", "neg_sample": ["annotations is used for Task", "it is of great value to answer product questions based on heterogeneous information sources available on web product pages , e.g.", ", semi - structured attributes , text descriptions , user - provided contents , etc .", "however , these sources have different structures and writing styles , which poses challenges for ( 1 ) evidence ranking , ( 2 ) source selection , and ( 3 ) answer generation ."], "relation": "used for", "id": "2022.ecnlp-1.13", "year": 2022, "rel_sent": "In this paper , we build a benchmark with annotations for both evidence selection and answer generation covering 6 information sources .", "forward": true, "src_ids": "2022.ecnlp-1.13_3500"}
+{"input": "suicidality assessment is done by using Method| context: recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings . with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk . however , such systems may generate uncertain predictions , leading to severe consequences .", "entity": "suicidality assessment", "output": "risk - averse mechanism", "neg_sample": ["suicidality assessment is done by using Method", "recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings .", "with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk .", "however , such systems may generate uncertain predictions , leading to severe consequences ."], "relation": "used for", "id": "2022.acl-short.70", "year": 2022, "rel_sent": "A Risk - Averse Mechanism for Suicidality Assessment on Social Media.", "forward": false, "src_ids": "2022.acl-short.70_3501"}
+{"input": "risk - averse mechanism is used for Task| context: recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings . with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk . however , such systems may generate uncertain predictions , leading to severe consequences .", "entity": "risk - averse mechanism", "output": "suicidality assessment", "neg_sample": ["risk - averse mechanism is used for Task", "recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings .", "with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk .", "however , such systems may generate uncertain predictions , leading to severe consequences ."], "relation": "used for", "id": "2022.acl-short.70", "year": 2022, "rel_sent": "A Risk - Averse Mechanism for Suicidality Assessment on Social Media.", "forward": true, "src_ids": "2022.acl-short.70_3502"}
+{"input": "sasi is used for OtherScientificTerm| context: recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings . with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk .", "entity": "sasi", "output": "uncertain predictions", "neg_sample": ["sasi is used for OtherScientificTerm", "recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings .", "with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk ."], "relation": "used for", "id": "2022.acl-short.70", "year": 2022, "rel_sent": "We propose SASI , a risk - averse and self - aware transformer - based hierarchical attention classifier , augmented to refrain from making uncertain predictions .", "forward": true, "src_ids": "2022.acl-short.70_3503"}
+{"input": "sasi is used for Task| context: recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings . with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk . however , such systems may generate uncertain predictions , leading to severe consequences .", "entity": "sasi", "output": "suicide risk assessment", "neg_sample": ["sasi is used for Task", "recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings .", "with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk .", "however , such systems may generate uncertain predictions , leading to severe consequences ."], "relation": "used for", "id": "2022.acl-short.70", "year": 2022, "rel_sent": "Furthermore , we discuss the qualitative , practical , and ethical aspects of SASI for suicide risk assessment as a human - in - the - loop framework .", "forward": true, "src_ids": "2022.acl-short.70_3504"}
+{"input": "uncertain predictions is done by using Method| context: recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings . with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk . however , such systems may generate uncertain predictions , leading to severe consequences .", "entity": "uncertain predictions", "output": "sasi", "neg_sample": ["uncertain predictions is done by using Method", "recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings .", "with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk .", "however , such systems may generate uncertain predictions , leading to severe consequences ."], "relation": "used for", "id": "2022.acl-short.70", "year": 2022, "rel_sent": "We propose SASI , a risk - averse and self - aware transformer - based hierarchical attention classifier , augmented to refrain from making uncertain predictions .", "forward": false, "src_ids": "2022.acl-short.70_3505"}
+{"input": "suicide risk assessment is done by using Method| context: recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings . with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk . however , such systems may generate uncertain predictions , leading to severe consequences .", "entity": "suicide risk assessment", "output": "sasi", "neg_sample": ["suicide risk assessment is done by using Method", "recent studies have shown that social media has increasingly become a platform for users to express suicidal thoughts outside traditional clinical settings .", "with advances in natural language processing strategies , it is now possible to design automated systems to assess suicide risk .", "however , such systems may generate uncertain predictions , leading to severe consequences ."], "relation": "used for", "id": "2022.acl-short.70", "year": 2022, "rel_sent": "Furthermore , we discuss the qualitative , practical , and ethical aspects of SASI for suicide risk assessment as a human - in - the - loop framework .", "forward": false, "src_ids": "2022.acl-short.70_3506"}
+{"input": "knowledge - based visual question answering is done by using Method| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "knowledge - based visual question answering", "output": "weakly - supervised multi - hop reasoning", "neg_sample": ["knowledge - based visual question answering is done by using Method", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "Hypergraph Transformer : Weakly - Supervised Multi - hop Reasoning for Knowledge - based Visual Question Answering.", "forward": false, "src_ids": "2022.acl-long.29_3507"}
+{"input": "weakly - supervised multi - hop reasoning is used for Task| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "weakly - supervised multi - hop reasoning", "output": "knowledge - based visual question answering", "neg_sample": ["weakly - supervised multi - hop reasoning is used for Task", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "Hypergraph Transformer : Weakly - Supervised Multi - hop Reasoning for Knowledge - based Visual Question Answering.", "forward": true, "src_ids": "2022.acl-long.29_3508"}
+{"input": "high - level semantics is done by using OtherScientificTerm| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "high - level semantics", "output": "hypergraphs", "neg_sample": ["high - level semantics is done by using OtherScientificTerm", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "In this paper , we introduce a concept of hypergraph to encode high - level semantics of a question and a knowledge base , and to learn high - order associations between them .", "forward": false, "src_ids": "2022.acl-long.29_3509"}
+{"input": "knowledge base is done by using OtherScientificTerm| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "knowledge base", "output": "hypergraphs", "neg_sample": ["knowledge base is done by using OtherScientificTerm", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "In this paper , we introduce a concept of hypergraph to encode high - level semantics of a question and a knowledge base , and to learn high - order associations between them .", "forward": false, "src_ids": "2022.acl-long.29_3510"}
+{"input": "high - order associations is done by using OtherScientificTerm| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "high - order associations", "output": "hypergraphs", "neg_sample": ["high - order associations is done by using OtherScientificTerm", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "In this paper , we introduce a concept of hypergraph to encode high - level semantics of a question and a knowledge base , and to learn high - order associations between them .", "forward": false, "src_ids": "2022.acl-long.29_3511"}
+{"input": "hypergraphs is used for OtherScientificTerm| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "hypergraphs", "output": "high - level semantics", "neg_sample": ["hypergraphs is used for OtherScientificTerm", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "In this paper , we introduce a concept of hypergraph to encode high - level semantics of a question and a knowledge base , and to learn high - order associations between them .", "forward": true, "src_ids": "2022.acl-long.29_3512"}
+{"input": "hypergraphs is used for Material| context: answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured .", "entity": "hypergraphs", "output": "knowledge base", "neg_sample": ["hypergraphs is used for Material", "answering complex questions that require multi - hop reasoning under weak supervision is considered as a challenging problem since i ) no supervision is given to the reasoning process and ii ) high - order semantics of multi - hop knowledge facts need to be captured ."], "relation": "used for", "id": "2022.acl-long.29", "year": 2022, "rel_sent": "In this paper , we introduce a concept of hypergraph to encode high - level semantics of a question and a knowledge base , and to learn high - order associations between them .", "forward": true, "src_ids": "2022.acl-long.29_3513"}
+{"input": "nlu is done by using Method| context: prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training . however , in the context of nlu , prior work reveals that prompt tuning does not perform well for normal - sized pretrained models . we alsofind that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality .", "entity": "nlu", "output": "deep prompt tuning ( citation )", "neg_sample": ["nlu is done by using Method", "prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training .", "however , in the context of nlu , prior work reveals that prompt tuning does not perform well for normal - sized pretrained models .", "we alsofind that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality ."], "relation": "used for", "id": "2022.acl-short.8", "year": 2022, "rel_sent": "Our method P - Tuning v2 is an implementation of Deep Prompt Tuning ( CITATION ) optimized and adapted for NLU .", "forward": false, "src_ids": "2022.acl-short.8_3514"}
+{"input": "deep prompt tuning ( citation ) is used for Task| context: prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training . we alsofind that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality .", "entity": "deep prompt tuning ( citation )", "output": "nlu", "neg_sample": ["deep prompt tuning ( citation ) is used for Task", "prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training .", "we alsofind that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality ."], "relation": "used for", "id": "2022.acl-short.8", "year": 2022, "rel_sent": "Our method P - Tuning v2 is an implementation of Deep Prompt Tuning ( CITATION ) optimized and adapted for NLU .", "forward": true, "src_ids": "2022.acl-short.8_3515"}
+{"input": "semantic similarity is done by using Method| context: ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term .", "entity": "semantic similarity", "output": "high - precision cross - encoder bert model", "neg_sample": ["semantic similarity is done by using Method", "ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term ."], "relation": "used for", "id": "2022.ecnlp-1.6", "year": 2022, "rel_sent": "In this work we leverage a high - precision cross - encoder BERT model for semantic similarity between customer query and products and survey its effectiveness for three ranking applications where offline - generated scores could be used : ( 1 ) as an offline metric for estimating relevance quality impact , ( 2 ) as a re - ranking feature covering head / torso queries , and ( 3 ) as a training objective for optimization .", "forward": false, "src_ids": "2022.ecnlp-1.6_3516"}
+{"input": "high - precision cross - encoder bert model is used for OtherScientificTerm| context: ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term .", "entity": "high - precision cross - encoder bert model", "output": "semantic similarity", "neg_sample": ["high - precision cross - encoder bert model is used for OtherScientificTerm", "ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term ."], "relation": "used for", "id": "2022.ecnlp-1.6", "year": 2022, "rel_sent": "In this work we leverage a high - precision cross - encoder BERT model for semantic similarity between customer query and products and survey its effectiveness for three ranking applications where offline - generated scores could be used : ( 1 ) as an offline metric for estimating relevance quality impact , ( 2 ) as a re - ranking feature covering head / torso queries , and ( 3 ) as a training objective for optimization .", "forward": true, "src_ids": "2022.ecnlp-1.6_3517"}
+{"input": "relevance quality impact is done by using Metric| context: ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term .", "entity": "relevance quality impact", "output": "offline metric", "neg_sample": ["relevance quality impact is done by using Metric", "ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term ."], "relation": "used for", "id": "2022.ecnlp-1.6", "year": 2022, "rel_sent": "In this work we leverage a high - precision cross - encoder BERT model for semantic similarity between customer query and products and survey its effectiveness for three ranking applications where offline - generated scores could be used : ( 1 ) as an offline metric for estimating relevance quality impact , ( 2 ) as a re - ranking feature covering head / torso queries , and ( 3 ) as a training objective for optimization .", "forward": false, "src_ids": "2022.ecnlp-1.6_3518"}
+{"input": "optimization is done by using OtherScientificTerm| context: ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term .", "entity": "optimization", "output": "training objective", "neg_sample": ["optimization is done by using OtherScientificTerm", "ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term ."], "relation": "used for", "id": "2022.ecnlp-1.6", "year": 2022, "rel_sent": "In this work we leverage a high - precision cross - encoder BERT model for semantic similarity between customer query and products and survey its effectiveness for three ranking applications where offline - generated scores could be used : ( 1 ) as an offline metric for estimating relevance quality impact , ( 2 ) as a re - ranking feature covering head / torso queries , and ( 3 ) as a training objective for optimization .", "forward": false, "src_ids": "2022.ecnlp-1.6_3519"}
+{"input": "training objective is used for Task| context: ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term .", "entity": "training objective", "output": "optimization", "neg_sample": ["training objective is used for Task", "ensuring relevance quality in product search is a critical task as it impacts the customer 's ability tofind intended products in the short - term as well as the general perception and trust of the e - commerce system in the long term ."], "relation": "used for", "id": "2022.ecnlp-1.6", "year": 2022, "rel_sent": "In this work we leverage a high - precision cross - encoder BERT model for semantic similarity between customer query and products and survey its effectiveness for three ranking applications where offline - generated scores could be used : ( 1 ) as an offline metric for estimating relevance quality impact , ( 2 ) as a re - ranking feature covering head / torso queries , and ( 3 ) as a training objective for optimization .", "forward": true, "src_ids": "2022.ecnlp-1.6_3520"}
+{"input": "nlp task is used for Task| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "nlp task", "output": "detecting political bias", "neg_sample": ["nlp task is used for Task", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": true, "src_ids": "2022.acl-srw.17_3521"}
+{"input": "detecting political bias is done by using Task| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "detecting political bias", "output": "nlp task", "neg_sample": ["detecting political bias is done by using Task", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": false, "src_ids": "2022.acl-srw.17_3522"}
+{"input": "pre - trained network is done by using Method| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "pre - trained network", "output": "transformer - based transfer learning method", "neg_sample": ["pre - trained network is done by using Method", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": false, "src_ids": "2022.acl-srw.17_3523"}
+{"input": "bias detection is done by using Method| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "bias detection", "output": "transformer - based transfer learning method", "neg_sample": ["bias detection is done by using Method", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": false, "src_ids": "2022.acl-srw.17_3524"}
+{"input": "transformer - based transfer learning method is used for Method| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "transformer - based transfer learning method", "output": "pre - trained network", "neg_sample": ["transformer - based transfer learning method is used for Method", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": true, "src_ids": "2022.acl-srw.17_3525"}
+{"input": "pre - trained network is used for Task| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "pre - trained network", "output": "bias detection", "neg_sample": ["pre - trained network is used for Task", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": true, "src_ids": "2022.acl-srw.17_3526"}
+{"input": "transformer - based transfer learning method is used for Task| context: political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party .", "entity": "transformer - based transfer learning method", "output": "bias detection", "neg_sample": ["transformer - based transfer learning method is used for Task", "political propaganda in recent times has been amplified by media news portals through biased reporting , creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party ."], "relation": "used for", "id": "2022.acl-srw.17", "year": 2022, "rel_sent": "This issue proposes a challenging NLP task of detecting political bias in news articles . We propose a transformer - based transfer learning method tofine - tune the pre - trained network on our data for this bias detection .", "forward": true, "src_ids": "2022.acl-srw.17_3527"}
+{"input": "natural language inference ( nli ) is done by using Method| context: moral values as commonsense norms shape our everyday individual and community behavior . the possibility to extract moral attitude rapidly from natural language is an appealing perspective that would enable a deeper understanding of social interaction dynamics and the individual cognitive and behavioral dimension .", "entity": "natural language inference ( nli )", "output": "zero - shot machine learning model", "neg_sample": ["natural language inference ( nli ) is done by using Method", "moral values as commonsense norms shape our everyday individual and community behavior .", "the possibility to extract moral attitude rapidly from natural language is an appealing perspective that would enable a deeper understanding of social interaction dynamics and the individual cognitive and behavioral dimension ."], "relation": "used for", "id": "2022.deelio-1.4", "year": 2022, "rel_sent": "We develop and compare two different approaches : ( i ) a frame - based symbolic value detector based on knowledge graphs and ( ii ) a zero - shot machine learning model fine - tuned on a task of Natural Language Inference ( NLI ) and a task of emotion detection .", "forward": false, "src_ids": "2022.deelio-1.4_3528"}
+{"input": "emotion detection is done by using Method| context: moral values as commonsense norms shape our everyday individual and community behavior . the possibility to extract moral attitude rapidly from natural language is an appealing perspective that would enable a deeper understanding of social interaction dynamics and the individual cognitive and behavioral dimension .", "entity": "emotion detection", "output": "zero - shot machine learning model", "neg_sample": ["emotion detection is done by using Method", "moral values as commonsense norms shape our everyday individual and community behavior .", "the possibility to extract moral attitude rapidly from natural language is an appealing perspective that would enable a deeper understanding of social interaction dynamics and the individual cognitive and behavioral dimension ."], "relation": "used for", "id": "2022.deelio-1.4", "year": 2022, "rel_sent": "We develop and compare two different approaches : ( i ) a frame - based symbolic value detector based on knowledge graphs and ( ii ) a zero - shot machine learning model fine - tuned on a task of Natural Language Inference ( NLI ) and a task of emotion detection .", "forward": false, "src_ids": "2022.deelio-1.4_3529"}
+{"input": "zero - shot machine learning model is used for Task| context: moral values as commonsense norms shape our everyday individual and community behavior . the possibility to extract moral attitude rapidly from natural language is an appealing perspective that would enable a deeper understanding of social interaction dynamics and the individual cognitive and behavioral dimension .", "entity": "zero - shot machine learning model", "output": "natural language inference ( nli )", "neg_sample": ["zero - shot machine learning model is used for Task", "moral values as commonsense norms shape our everyday individual and community behavior .", "the possibility to extract moral attitude rapidly from natural language is an appealing perspective that would enable a deeper understanding of social interaction dynamics and the individual cognitive and behavioral dimension ."], "relation": "used for", "id": "2022.deelio-1.4", "year": 2022, "rel_sent": "We develop and compare two different approaches : ( i ) a frame - based symbolic value detector based on knowledge graphs and ( ii ) a zero - shot machine learning model fine - tuned on a task of Natural Language Inference ( NLI ) and a task of emotion detection .", "forward": true, "src_ids": "2022.deelio-1.4_3530"}
+{"input": "equality is done by using Task| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "entity": "equality", "output": "homophobia / transphobia detection", "neg_sample": ["equality is done by using Task", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "NAYEL @LT - EDI - ACL2022 : Homophobia / Transphobia Detection for Equality , Diversity , and Inclusion using SVM.", "forward": false, "src_ids": "2022.ltedi-1.42_3531"}
+{"input": "diversity is done by using Task| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "entity": "diversity", "output": "homophobia / transphobia detection", "neg_sample": ["diversity is done by using Task", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "NAYEL @LT - EDI - ACL2022 : Homophobia / Transphobia Detection for Equality , Diversity , and Inclusion using SVM.", "forward": false, "src_ids": "2022.ltedi-1.42_3532"}
+{"input": "homophobia / transphobia detection is used for OtherScientificTerm| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "homophobia / transphobia detection", "output": "equality", "neg_sample": ["homophobia / transphobia detection is used for OtherScientificTerm", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "NAYEL @LT - EDI - ACL2022 : Homophobia / Transphobia Detection for Equality , Diversity , and Inclusion using SVM.", "forward": true, "src_ids": "2022.ltedi-1.42_3533"}
+{"input": "automatic detection of homophobia is done by using Method| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "automatic detection of homophobia", "output": "classification algorithms", "neg_sample": ["automatic detection of homophobia is done by using Method", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "A machine learning - based model has been designed and various classification algorithms have been implemented for automatic detection of homophobia in YouTube comments .", "forward": false, "src_ids": "2022.ltedi-1.42_3534"}
+{"input": "classification algorithms is used for Task| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "classification algorithms", "output": "automatic detection of homophobia", "neg_sample": ["classification algorithms is used for Task", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "A machine learning - based model has been designed and various classification algorithms have been implemented for automatic detection of homophobia in YouTube comments .", "forward": true, "src_ids": "2022.ltedi-1.42_3535"}
+{"input": "vectorization of comments is done by using OtherScientificTerm| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "vectorization of comments", "output": "tf / idf", "neg_sample": ["vectorization of comments is done by using OtherScientificTerm", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "TF / IDF has been used with a range of bigram model for vectorization of comments .", "forward": false, "src_ids": "2022.ltedi-1.42_3536"}
+{"input": "vectorization of comments is done by using Method| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "vectorization of comments", "output": "bigram model", "neg_sample": ["vectorization of comments is done by using Method", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "TF / IDF has been used with a range of bigram model for vectorization of comments .", "forward": false, "src_ids": "2022.ltedi-1.42_3537"}
+{"input": "bigram model is used for Task| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "bigram model", "output": "vectorization of comments", "neg_sample": ["bigram model is used for Task", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "TF / IDF has been used with a range of bigram model for vectorization of comments .", "forward": true, "src_ids": "2022.ltedi-1.42_3538"}
+{"input": "tf / idf is used for Task| context: analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users . one of the important tasks is homophobia / transphobia detection .", "entity": "tf / idf", "output": "vectorization of comments", "neg_sample": ["tf / idf is used for Task", "analysing the contents of social media platforms such as youtube , facebook and twitter gained interest due to the vast number of users .", "one of the important tasks is homophobia / transphobia detection ."], "relation": "used for", "id": "2022.ltedi-1.42", "year": 2022, "rel_sent": "TF / IDF has been used with a range of bigram model for vectorization of comments .", "forward": true, "src_ids": "2022.ltedi-1.42_3539"}
+{"input": "low resource multilingual translation is done by using Method| context: these alternate segmentations function like related languages in multilingual translation .", "entity": "low resource multilingual translation", "output": "auxiliary subword segmentations", "neg_sample": ["low resource multilingual translation is done by using Method", "these alternate segmentations function like related languages in multilingual translation ."], "relation": "used for", "id": "2022.eamt-1.16", "year": 2022, "rel_sent": "Auxiliary Subword Segmentations as Related Languages for Low Resource Multilingual Translation.", "forward": false, "src_ids": "2022.eamt-1.16_3540"}
+{"input": "auxiliary subword segmentations is used for Task| context: these alternate segmentations function like related languages in multilingual translation .", "entity": "auxiliary subword segmentations", "output": "low resource multilingual translation", "neg_sample": ["auxiliary subword segmentations is used for Task", "these alternate segmentations function like related languages in multilingual translation ."], "relation": "used for", "id": "2022.eamt-1.16", "year": 2022, "rel_sent": "Auxiliary Subword Segmentations as Related Languages for Low Resource Multilingual Translation.", "forward": true, "src_ids": "2022.eamt-1.16_3541"}
+{"input": "multi - modal sarcasm detection is done by using Method| context: with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection .", "entity": "multi - modal sarcasm detection", "output": "cross - modal graph convolutional network", "neg_sample": ["multi - modal sarcasm detection is done by using Method", "with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection ."], "relation": "used for", "id": "2022.acl-long.124", "year": 2022, "rel_sent": "Furthermore , we devise a cross - modal graph convolutional network to make sense of the incongruity relations between modalities for multi - modal sarcasm detection .", "forward": false, "src_ids": "2022.acl-long.124_3542"}
+{"input": "ironic relations is done by using Method| context: with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection .", "entity": "ironic relations", "output": "cross - modal graph", "neg_sample": ["ironic relations is done by using Method", "with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection ."], "relation": "used for", "id": "2022.acl-long.124", "year": 2022, "rel_sent": "In this paper , we investigate multi - modal sarcasm detection from a novel perspective by constructing a cross - modal graph for each instance to explicitly draw the ironic relations between textual and visual modalities .", "forward": false, "src_ids": "2022.acl-long.124_3543"}
+{"input": "multi - modal instance is done by using Method| context: with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection .", "entity": "multi - modal instance", "output": "cross - modal graph", "neg_sample": ["multi - modal instance is done by using Method", "with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection ."], "relation": "used for", "id": "2022.acl-long.124", "year": 2022, "rel_sent": "Then , the descriptions of the objects are served as a bridge to determine the importance of the association between the objects of image modality and the contextual words of text modality , so as to build a cross - modal graph for each multi - modal instance .", "forward": false, "src_ids": "2022.acl-long.124_3544"}
+{"input": "cross - modal graph is used for OtherScientificTerm| context: with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection .", "entity": "cross - modal graph", "output": "ironic relations", "neg_sample": ["cross - modal graph is used for OtherScientificTerm", "with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection ."], "relation": "used for", "id": "2022.acl-long.124", "year": 2022, "rel_sent": "In this paper , we investigate multi - modal sarcasm detection from a novel perspective by constructing a cross - modal graph for each instance to explicitly draw the ironic relations between textual and visual modalities .", "forward": true, "src_ids": "2022.acl-long.124_3545"}
+{"input": "cross - modal graph is used for OtherScientificTerm| context: with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection .", "entity": "cross - modal graph", "output": "multi - modal instance", "neg_sample": ["cross - modal graph is used for OtherScientificTerm", "with the increasing popularity of posting multimodal messages online , many recent studies have been carried out utilizing both textual and visual information for multi - modal sarcasm detection ."], "relation": "used for", "id": "2022.acl-long.124", "year": 2022, "rel_sent": "Then , the descriptions of the objects are served as a bridge to determine the importance of the association between the objects of image modality and the contextual words of text modality , so as to build a cross - modal graph for each multi - modal instance .", "forward": true, "src_ids": "2022.acl-long.124_3546"}
+{"input": "latency reduction is done by using Method| context: standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user . this could be slow when the program contains expensive function calls . we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking .", "entity": "latency reduction", "output": "online semantic parsing", "neg_sample": ["latency reduction is done by using Method", "standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user .", "this could be slow when the program contains expensive function calls .", "we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking ."], "relation": "used for", "id": "2022.acl-long.110", "year": 2022, "rel_sent": "Online Semantic Parsing for Latency Reduction in Task - Oriented Dialogue.", "forward": false, "src_ids": "2022.acl-long.110_3547"}
+{"input": "online semantic parsing is used for Task| context: standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user . this could be slow when the program contains expensive function calls . we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking .", "entity": "online semantic parsing", "output": "latency reduction", "neg_sample": ["online semantic parsing is used for Task", "standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user .", "this could be slow when the program contains expensive function calls .", "we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking ."], "relation": "used for", "id": "2022.acl-long.110", "year": 2022, "rel_sent": "Online Semantic Parsing for Latency Reduction in Task - Oriented Dialogue.", "forward": true, "src_ids": "2022.acl-long.110_3548"}
+{"input": "subprogram selection is done by using Method| context: standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user . this could be slow when the program contains expensive function calls . we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking .", "entity": "subprogram selection", "output": "thresholding heuristic", "neg_sample": ["subprogram selection is done by using Method", "standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user .", "this could be slow when the program contains expensive function calls .", "we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking ."], "relation": "used for", "id": "2022.acl-long.110", "year": 2022, "rel_sent": "We propose a general framework with first a learned prefix - to - program prediction module , and then a simple yet effective thresholding heuristic for subprogram selection for early execution .", "forward": false, "src_ids": "2022.acl-long.110_3549"}
+{"input": "early execution is done by using Task| context: standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user . this could be slow when the program contains expensive function calls . we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking .", "entity": "early execution", "output": "subprogram selection", "neg_sample": ["early execution is done by using Task", "standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user .", "this could be slow when the program contains expensive function calls .", "we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking ."], "relation": "used for", "id": "2022.acl-long.110", "year": 2022, "rel_sent": "We propose a general framework with first a learned prefix - to - program prediction module , and then a simple yet effective thresholding heuristic for subprogram selection for early execution .", "forward": false, "src_ids": "2022.acl-long.110_3550"}
+{"input": "thresholding heuristic is used for Task| context: standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user . this could be slow when the program contains expensive function calls . we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking .", "entity": "thresholding heuristic", "output": "subprogram selection", "neg_sample": ["thresholding heuristic is used for Task", "standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user .", "this could be slow when the program contains expensive function calls .", "we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking ."], "relation": "used for", "id": "2022.acl-long.110", "year": 2022, "rel_sent": "We propose a general framework with first a learned prefix - to - program prediction module , and then a simple yet effective thresholding heuristic for subprogram selection for early execution .", "forward": true, "src_ids": "2022.acl-long.110_3551"}
+{"input": "subprogram selection is used for OtherScientificTerm| context: standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user . this could be slow when the program contains expensive function calls . we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking .", "entity": "subprogram selection", "output": "early execution", "neg_sample": ["subprogram selection is used for OtherScientificTerm", "standard conversational semantic parsing maps a complete user utterance into an executable program , after which the program is executed to respond to the user .", "this could be slow when the program contains expensive function calls .", "we investigate the opportunity to reduce latency by predicting and executing function calls while the user is still speaking ."], "relation": "used for", "id": "2022.acl-long.110", "year": 2022, "rel_sent": "We propose a general framework with first a learned prefix - to - program prediction module , and then a simple yet effective thresholding heuristic for subprogram selection for early execution .", "forward": true, "src_ids": "2022.acl-long.110_3552"}
+{"input": "distantly supervised learning is used for Material| context: however , there is still a problem of low prediction performance due to the inclusion of mislabeled data .", "entity": "distantly supervised learning", "output": "pseudo - training data", "neg_sample": ["distantly supervised learning is used for Material", "however , there is still a problem of low prediction performance due to the inclusion of mislabeled data ."], "relation": "used for", "id": "2022.bionlp-1.16", "year": 2022, "rel_sent": "Distantly supervised learning has been proposed to generate a large amount of pseudo - training data at low cost .", "forward": true, "src_ids": "2022.bionlp-1.16_3553"}
+{"input": "pseudo - training data is done by using Method| context: however , there is still a problem of low prediction performance due to the inclusion of mislabeled data .", "entity": "pseudo - training data", "output": "distantly supervised learning", "neg_sample": ["pseudo - training data is done by using Method", "however , there is still a problem of low prediction performance due to the inclusion of mislabeled data ."], "relation": "used for", "id": "2022.bionlp-1.16", "year": 2022, "rel_sent": "Distantly supervised learning has been proposed to generate a large amount of pseudo - training data at low cost .", "forward": false, "src_ids": "2022.bionlp-1.16_3554"}
+{"input": "classifier learning is done by using OtherScientificTerm| context: languages around the world employ classifier systems as a method of semantic organization and categorization . these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time .", "entity": "classifier learning", "output": "homophony avoidance", "neg_sample": ["classifier learning is done by using OtherScientificTerm", "languages around the world employ classifier systems as a method of semantic organization and categorization .", "these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time ."], "relation": "used for", "id": "2022.lchange-1.2", "year": 2022, "rel_sent": "Our research consists of two parts : a contrastive corpus study of Cantonese and Mandarin child - directed speech to determine the role that ambiguity and homophony avoidance may play in classifier learning and change followed by a series of population - level learning simulations of an abstract classifier system .", "forward": false, "src_ids": "2022.lchange-1.2_3555"}
+{"input": "homophony avoidance is used for Task| context: languages around the world employ classifier systems as a method of semantic organization and categorization . these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time .", "entity": "homophony avoidance", "output": "classifier learning", "neg_sample": ["homophony avoidance is used for Task", "languages around the world employ classifier systems as a method of semantic organization and categorization .", "these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time ."], "relation": "used for", "id": "2022.lchange-1.2", "year": 2022, "rel_sent": "Our research consists of two parts : a contrastive corpus study of Cantonese and Mandarin child - directed speech to determine the role that ambiguity and homophony avoidance may play in classifier learning and change followed by a series of population - level learning simulations of an abstract classifier system .", "forward": true, "src_ids": "2022.lchange-1.2_3556"}
+{"input": "classifier death is done by using OtherScientificTerm| context: languages around the world employ classifier systems as a method of semantic organization and categorization . these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time .", "entity": "classifier death", "output": "adults", "neg_sample": ["classifier death is done by using OtherScientificTerm", "languages around the world employ classifier systems as a method of semantic organization and categorization .", "these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time ."], "relation": "used for", "id": "2022.lchange-1.2", "year": 2022, "rel_sent": "We find that acquisition without reference to ambiguity avoidance is sufficient to drive broad trends in classifier change and suggest an additional role for adults and discourse factors in classifier death .", "forward": false, "src_ids": "2022.lchange-1.2_3557"}
+{"input": "discourse factors is used for Task| context: languages around the world employ classifier systems as a method of semantic organization and categorization . these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time .", "entity": "discourse factors", "output": "classifier death", "neg_sample": ["discourse factors is used for Task", "languages around the world employ classifier systems as a method of semantic organization and categorization .", "these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time ."], "relation": "used for", "id": "2022.lchange-1.2", "year": 2022, "rel_sent": "We find that acquisition without reference to ambiguity avoidance is sufficient to drive broad trends in classifier change and suggest an additional role for adults and discourse factors in classifier death .", "forward": true, "src_ids": "2022.lchange-1.2_3558"}
+{"input": "adults is used for Task| context: languages around the world employ classifier systems as a method of semantic organization and categorization . these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time .", "entity": "adults", "output": "classifier death", "neg_sample": ["adults is used for Task", "languages around the world employ classifier systems as a method of semantic organization and categorization .", "these systems are rife with variability , violability , and ambiguity , and are prone to constant change over time ."], "relation": "used for", "id": "2022.lchange-1.2", "year": 2022, "rel_sent": "We find that acquisition without reference to ambiguity avoidance is sufficient to drive broad trends in classifier change and suggest an additional role for adults and discourse factors in classifier death .", "forward": true, "src_ids": "2022.lchange-1.2_3559"}
+{"input": "endangered language is done by using Task| context: these include ( i ) a screenreading facility for visually impaired people , ( ii ) an application to help develop phonological awareness and early literacy for dyslexic people ( iii ) a speech - enabled aac system for non - speaking people . each of these is at a different stage of development and poses unique challenges : these are dis - cussed along with the approaches adopted to address them .", "entity": "endangered language", "output": "assistive technology development", "neg_sample": ["endangered language is done by using Task", "these include ( i ) a screenreading facility for visually impaired people , ( ii ) an application to help develop phonological awareness and early literacy for dyslexic people ( iii ) a speech - enabled aac system for non - speaking people .", "each of these is at a different stage of development and poses unique challenges : these are dis - cussed along with the approaches adopted to address them ."], "relation": "used for", "id": "2022.slpat-1.10", "year": 2022, "rel_sent": "Challenges in assistive technology development for an endangered language : an Irish ( Gaelic ) perspective.", "forward": false, "src_ids": "2022.slpat-1.10_3560"}
+{"input": "assistive technology development is used for Material| context: these include ( i ) a screenreading facility for visually impaired people , ( ii ) an application to help develop phonological awareness and early literacy for dyslexic people ( iii ) a speech - enabled aac system for non - speaking people . each of these is at a different stage of development and poses unique challenges : these are dis - cussed along with the approaches adopted to address them .", "entity": "assistive technology development", "output": "endangered language", "neg_sample": ["assistive technology development is used for Material", "these include ( i ) a screenreading facility for visually impaired people , ( ii ) an application to help develop phonological awareness and early literacy for dyslexic people ( iii ) a speech - enabled aac system for non - speaking people .", "each of these is at a different stage of development and poses unique challenges : these are dis - cussed along with the approaches adopted to address them ."], "relation": "used for", "id": "2022.slpat-1.10", "year": 2022, "rel_sent": "Challenges in assistive technology development for an endangered language : an Irish ( Gaelic ) perspective.", "forward": true, "src_ids": "2022.slpat-1.10_3561"}
+{"input": "multiple tonal association is done by using Method| context: association of tones to prosodic trees was introduced in pierrehumbert and beckman ( 1988 ) . this included : ( i ) tonal association to higher - level prosodic nodes such as intonational phrases , and ( ii ) multiple association of a tone to a higher - level prosodic node in addi - tion to a tone bearing unit such as a syllable . since then , these concepts have been broadly assumed in intonational phonology without much comment , even though pierrehumbert and beckman ( 1988 ) 's stipulation that tones associated to higher - level prosodic nodes are peripherally realized does not fit all the empirical data .", "entity": "multiple tonal association", "output": "multi bottom - up tree transducers", "neg_sample": ["multiple tonal association is done by using Method", "association of tones to prosodic trees was introduced in pierrehumbert and beckman ( 1988 ) .", "this included : ( i ) tonal association to higher - level prosodic nodes such as intonational phrases , and ( ii ) multiple association of a tone to a higher - level prosodic node in addi - tion to a tone bearing unit such as a syllable .", "since then , these concepts have been broadly assumed in intonational phonology without much comment , even though pierrehumbert and beckman ( 1988 ) 's stipulation that tones associated to higher - level prosodic nodes are peripherally realized does not fit all the empirical data ."], "relation": "used for", "id": "2022.scil-1.14", "year": 2022, "rel_sent": "Additionally , multi bottom - up tree transducers provide a way to represent non - peripheral boundary tones and multiple tonal association , as well as multiple dependencies in prosodic structures in general , including prosodically - conditioned segmental allophony .", "forward": false, "src_ids": "2022.scil-1.14_3562"}
+{"input": "non - peripheral boundary tones is done by using Method| context: association of tones to prosodic trees was introduced in pierrehumbert and beckman ( 1988 ) . this included : ( i ) tonal association to higher - level prosodic nodes such as intonational phrases , and ( ii ) multiple association of a tone to a higher - level prosodic node in addi - tion to a tone bearing unit such as a syllable . since then , these concepts have been broadly assumed in intonational phonology without much comment , even though pierrehumbert and beckman ( 1988 ) 's stipulation that tones associated to higher - level prosodic nodes are peripherally realized does not fit all the empirical data .", "entity": "non - peripheral boundary tones", "output": "multi bottom - up tree transducers", "neg_sample": ["non - peripheral boundary tones is done by using Method", "association of tones to prosodic trees was introduced in pierrehumbert and beckman ( 1988 ) .", "this included : ( i ) tonal association to higher - level prosodic nodes such as intonational phrases , and ( ii ) multiple association of a tone to a higher - level prosodic node in addi - tion to a tone bearing unit such as a syllable .", "since then , these concepts have been broadly assumed in intonational phonology without much comment , even though pierrehumbert and beckman ( 1988 ) 's stipulation that tones associated to higher - level prosodic nodes are peripherally realized does not fit all the empirical data ."], "relation": "used for", "id": "2022.scil-1.14", "year": 2022, "rel_sent": "Additionally , multi bottom - up tree transducers provide a way to represent non - peripheral boundary tones and multiple tonal association , as well as multiple dependencies in prosodic structures in general , including prosodically - conditioned segmental allophony .", "forward": false, "src_ids": "2022.scil-1.14_3563"}
+{"input": "multi bottom - up tree transducers is used for OtherScientificTerm| context: association of tones to prosodic trees was introduced in pierrehumbert and beckman ( 1988 ) . this included : ( i ) tonal association to higher - level prosodic nodes such as intonational phrases , and ( ii ) multiple association of a tone to a higher - level prosodic node in addi - tion to a tone bearing unit such as a syllable . since then , these concepts have been broadly assumed in intonational phonology without much comment , even though pierrehumbert and beckman ( 1988 ) 's stipulation that tones associated to higher - level prosodic nodes are peripherally realized does not fit all the empirical data .", "entity": "multi bottom - up tree transducers", "output": "non - peripheral boundary tones", "neg_sample": ["multi bottom - up tree transducers is used for OtherScientificTerm", "association of tones to prosodic trees was introduced in pierrehumbert and beckman ( 1988 ) .", "this included : ( i ) tonal association to higher - level prosodic nodes such as intonational phrases , and ( ii ) multiple association of a tone to a higher - level prosodic node in addi - tion to a tone bearing unit such as a syllable .", "since then , these concepts have been broadly assumed in intonational phonology without much comment , even though pierrehumbert and beckman ( 1988 ) 's stipulation that tones associated to higher - level prosodic nodes are peripherally realized does not fit all the empirical data ."], "relation": "used for", "id": "2022.scil-1.14", "year": 2022, "rel_sent": "Additionally , multi bottom - up tree transducers provide a way to represent non - peripheral boundary tones and multiple tonal association , as well as multiple dependencies in prosodic structures in general , including prosodically - conditioned segmental allophony .", "forward": true, "src_ids": "2022.scil-1.14_3564"}
+{"input": "program generation is done by using Method| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "program generation", "output": "structure - aware semantic parsing", "neg_sample": ["program generation is done by using Method", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Through analyzing the connection between the program tree and the dependency tree , we define a unified concept , operation - oriented tree , to mine structure features , and introduce Structure - Aware Semantic Parsing to integrate structure features into program generation .", "forward": false, "src_ids": "2022.acl-long.525_3565"}
+{"input": "violation punishments is used for OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "entity": "violation punishments", "output": "spurious programs", "neg_sample": ["violation punishments is used for OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Moreover , we design a refined objective function with lexical features and violation punishments tofurther avoid spurious programs .", "forward": true, "src_ids": "2022.acl-long.525_3566"}
+{"input": "program generation is done by using OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "program generation", "output": "lexical features", "neg_sample": ["program generation is done by using OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "In this paper , we address the challenge by leveraging both lexical features and structure features for program generation .", "forward": false, "src_ids": "2022.acl-long.525_3567"}
+{"input": "operation - oriented tree is used for OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "operation - oriented tree", "output": "structure features", "neg_sample": ["operation - oriented tree is used for OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Through analyzing the connection between the program tree and the dependency tree , we define a unified concept , operation - oriented tree , to mine structure features , and introduce Structure - Aware Semantic Parsing to integrate structure features into program generation .", "forward": true, "src_ids": "2022.acl-long.525_3568"}
+{"input": "unified concept is used for OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "unified concept", "output": "structure features", "neg_sample": ["unified concept is used for OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Through analyzing the connection between the program tree and the dependency tree , we define a unified concept , operation - oriented tree , to mine structure features , and introduce Structure - Aware Semantic Parsing to integrate structure features into program generation .", "forward": true, "src_ids": "2022.acl-long.525_3569"}
+{"input": "structure features is used for Task| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "structure features", "output": "program generation", "neg_sample": ["structure features is used for Task", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "In this paper , we address the challenge by leveraging both lexical features and structure features for program generation .", "forward": true, "src_ids": "2022.acl-long.525_3570"}
+{"input": "lexical features is used for Task| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "lexical features", "output": "program generation", "neg_sample": ["lexical features is used for Task", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "In this paper , we address the challenge by leveraging both lexical features and structure features for program generation .", "forward": true, "src_ids": "2022.acl-long.525_3571"}
+{"input": "structure - aware semantic parsing is used for Task| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "structure - aware semantic parsing", "output": "program generation", "neg_sample": ["structure - aware semantic parsing is used for Task", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Through analyzing the connection between the program tree and the dependency tree , we define a unified concept , operation - oriented tree , to mine structure features , and introduce Structure - Aware Semantic Parsing to integrate structure features into program generation .", "forward": true, "src_ids": "2022.acl-long.525_3572"}
+{"input": "structure features is done by using Generic| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "structure features", "output": "unified concept", "neg_sample": ["structure features is done by using Generic", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Through analyzing the connection between the program tree and the dependency tree , we define a unified concept , operation - oriented tree , to mine structure features , and introduce Structure - Aware Semantic Parsing to integrate structure features into program generation .", "forward": false, "src_ids": "2022.acl-long.525_3573"}
+{"input": "structure features is done by using OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "structure features", "output": "operation - oriented tree", "neg_sample": ["structure features is done by using OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Through analyzing the connection between the program tree and the dependency tree , we define a unified concept , operation - oriented tree , to mine structure features , and introduce Structure - Aware Semantic Parsing to integrate structure features into program generation .", "forward": false, "src_ids": "2022.acl-long.525_3574"}
+{"input": "spurious programs is done by using OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "spurious programs", "output": "violation punishments", "neg_sample": ["spurious programs is done by using OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.525", "year": 2022, "rel_sent": "Moreover , we design a refined objective function with lexical features and violation punishments tofurther avoid spurious programs .", "forward": false, "src_ids": "2022.acl-long.525_3575"}
+{"input": "unsupervised pos tagging task is done by using Method| context: in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks . but , in the unsupervised pos tagging task , works utilizing plms are few and fail to achieve state - of - the - art ( sota ) performance . the recent sota performance is yielded by a guassian hmm variant proposed by he et al . ( 2018 ) . however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms .", "entity": "unsupervised pos tagging task", "output": "bridging pre - trained language models", "neg_sample": ["unsupervised pos tagging task is done by using Method", "in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks .", "but , in the unsupervised pos tagging task , works utilizing plms are few and fail to achieve state - of - the - art ( sota ) performance .", "the recent sota performance is yielded by a guassian hmm variant proposed by he et al .", "( 2018 ) .", "however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms ."], "relation": "used for", "id": "2022.findings-acl.259", "year": 2022, "rel_sent": "Bridging Pre - trained Language Models and Hand - crafted Features for Unsupervised POS Tagging.", "forward": false, "src_ids": "2022.findings-acl.259_3576"}
+{"input": "unsupervised pos tagging task is done by using OtherScientificTerm| context: in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks . but , in the unsupervised pos tagging task , works utilizing plms are few and fail to achieve state - of - the - art ( sota ) performance . the recent sota performance is yielded by a guassian hmm variant proposed by he et al . ( 2018 ) . however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms .", "entity": "unsupervised pos tagging task", "output": "hand - crafted features", "neg_sample": ["unsupervised pos tagging task is done by using OtherScientificTerm", "in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks .", "but , in the unsupervised pos tagging task , works utilizing plms are few and fail to achieve state - of - the - art ( sota ) performance .", "the recent sota performance is yielded by a guassian hmm variant proposed by he et al .", "( 2018 ) .", "however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms ."], "relation": "used for", "id": "2022.findings-acl.259", "year": 2022, "rel_sent": "Bridging Pre - trained Language Models and Hand - crafted Features for Unsupervised POS Tagging.", "forward": false, "src_ids": "2022.findings-acl.259_3577"}
+{"input": "hand - crafted features is used for Task| context: in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks . the recent sota performance is yielded by a guassian hmm variant proposed by he et al . ( 2018 ) . however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms .", "entity": "hand - crafted features", "output": "unsupervised pos tagging task", "neg_sample": ["hand - crafted features is used for Task", "in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks .", "the recent sota performance is yielded by a guassian hmm variant proposed by he et al .", "( 2018 ) .", "however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms ."], "relation": "used for", "id": "2022.findings-acl.259", "year": 2022, "rel_sent": "Bridging Pre - trained Language Models and Hand - crafted Features for Unsupervised POS Tagging.", "forward": true, "src_ids": "2022.findings-acl.259_3578"}
+{"input": "bridging pre - trained language models is used for Task| context: in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks . the recent sota performance is yielded by a guassian hmm variant proposed by he et al . ( 2018 ) . however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms .", "entity": "bridging pre - trained language models", "output": "unsupervised pos tagging task", "neg_sample": ["bridging pre - trained language models is used for Task", "in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks .", "the recent sota performance is yielded by a guassian hmm variant proposed by he et al .", "( 2018 ) .", "however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms ."], "relation": "used for", "id": "2022.findings-acl.259", "year": 2022, "rel_sent": "Bridging Pre - trained Language Models and Hand - crafted Features for Unsupervised POS Tagging.", "forward": true, "src_ids": "2022.findings-acl.259_3579"}
+{"input": "attention maps is done by using OtherScientificTerm| context: word alignment is essential for the downstream cross - lingual language understanding and generation tasks . recently , the performance of the neural word alignment models has exceeded that of statistical models . however , they heavily rely on sophisticated translation models .", "entity": "attention maps", "output": "agreement loss", "neg_sample": ["attention maps is done by using OtherScientificTerm", "word alignment is essential for the downstream cross - lingual language understanding and generation tasks .", "recently , the performance of the neural word alignment models has exceeded that of statistical models .", "however , they heavily rely on sophisticated translation models ."], "relation": "used for", "id": "2022.iwslt-1.8", "year": 2022, "rel_sent": "In this study , we propose a super lightweight unsupervised word alignment model named MirrorAlign , in which bidirectional symmetric attention trained with a contrastive learning objective is introduced , and an agreement loss is employed to bind the attention maps , such that the alignments follow mirror - like symmetry hypothesis .", "forward": false, "src_ids": "2022.iwslt-1.8_3580"}
+{"input": "agreement loss is used for OtherScientificTerm| context: word alignment is essential for the downstream cross - lingual language understanding and generation tasks . recently , the performance of the neural word alignment models has exceeded that of statistical models . however , they heavily rely on sophisticated translation models .", "entity": "agreement loss", "output": "attention maps", "neg_sample": ["agreement loss is used for OtherScientificTerm", "word alignment is essential for the downstream cross - lingual language understanding and generation tasks .", "recently , the performance of the neural word alignment models has exceeded that of statistical models .", "however , they heavily rely on sophisticated translation models ."], "relation": "used for", "id": "2022.iwslt-1.8", "year": 2022, "rel_sent": "In this study , we propose a super lightweight unsupervised word alignment model named MirrorAlign , in which bidirectional symmetric attention trained with a contrastive learning objective is introduced , and an agreement loss is employed to bind the attention maps , such that the alignments follow mirror - like symmetry hypothesis .", "forward": true, "src_ids": "2022.iwslt-1.8_3581"}
+{"input": "instance is used for Method| context: transformer architectures have achieved state- of - the - art results on a variety of natural language processing ( nlp ) tasks . however , their attention mechanism comes with a quadratic complexity in sequence lengths , making the computational overhead prohibitive , especially for long sequences . attention context can be seen as a random - access memory with each token taking a slot . under this perspective , the memory size grows linearly with the sequence length , and so does the overhead of reading from it .", "entity": "instance", "output": "attention with bounded - memory control", "neg_sample": ["instance is used for Method", "transformer architectures have achieved state- of - the - art results on a variety of natural language processing ( nlp ) tasks .", "however , their attention mechanism comes with a quadratic complexity in sequence lengths , making the computational overhead prohibitive , especially for long sequences .", "attention context can be seen as a random - access memory with each token taking a slot .", "under this perspective , the memory size grows linearly with the sequence length , and so does the overhead of reading from it ."], "relation": "used for", "id": "2022.acl-long.515", "year": 2022, "rel_sent": "Last , we present a new instance of ABC , which draws inspiration from existing ABC approaches , but replaces their heuristic memory - organizing functions with a learned , contextualized one .", "forward": true, "src_ids": "2022.acl-long.515_3582"}
+{"input": "attention with bounded - memory control is done by using Generic| context: transformer architectures have achieved state- of - the - art results on a variety of natural language processing ( nlp ) tasks . however , their attention mechanism comes with a quadratic complexity in sequence lengths , making the computational overhead prohibitive , especially for long sequences . attention context can be seen as a random - access memory with each token taking a slot . under this perspective , the memory size grows linearly with the sequence length , and so does the overhead of reading from it .", "entity": "attention with bounded - memory control", "output": "instance", "neg_sample": ["attention with bounded - memory control is done by using Generic", "transformer architectures have achieved state- of - the - art results on a variety of natural language processing ( nlp ) tasks .", "however , their attention mechanism comes with a quadratic complexity in sequence lengths , making the computational overhead prohibitive , especially for long sequences .", "attention context can be seen as a random - access memory with each token taking a slot .", "under this perspective , the memory size grows linearly with the sequence length , and so does the overhead of reading from it ."], "relation": "used for", "id": "2022.acl-long.515", "year": 2022, "rel_sent": "Last , we present a new instance of ABC , which draws inspiration from existing ABC approaches , but replaces their heuristic memory - organizing functions with a learned , contextualized one .", "forward": false, "src_ids": "2022.acl-long.515_3583"}
+{"input": "entity retrieval is done by using Method| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "entity retrieval", "output": "tabi", "neg_sample": ["entity retrieval is done by using Method", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "In this work , we introduce TABi , a method to jointly train bi - encoders on knowledge graph types and unstructured text for entity retrieval for open - domain tasks .", "forward": false, "src_ids": "2022.findings-acl.169_3584"}
+{"input": "retrieval of rare entities is done by using Method| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "retrieval of rare entities", "output": "tabi", "neg_sample": ["retrieval of rare entities is done by using Method", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi improves retrieval of rare entities on the Ambiguous Entity Retrieval ( AmbER ) sets , while maintaining strong overall retrieval performance on open - domain tasks in the KILT benchmark compared to state - of - the - art retrievers .", "forward": false, "src_ids": "2022.findings-acl.169_3585"}
+{"input": "incomplete type systems is done by using Method| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "incomplete type systems", "output": "tabi", "neg_sample": ["incomplete type systems is done by using Method", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi is also robust to incomplete type systems , improving rare entity retrieval over baselines with only 5 % type coverage of the training dataset .", "forward": false, "src_ids": "2022.findings-acl.169_3586"}
+{"input": "retrieval of rare entities is done by using Method| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "retrieval of rare entities", "output": "tabi", "neg_sample": ["retrieval of rare entities is done by using Method", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi is also robust to incomplete type systems , improving rare entity retrieval over baselines with only 5 % type coverage of the training dataset .", "forward": false, "src_ids": "2022.findings-acl.169_3587"}
+{"input": "open - domain tasks is done by using Task| context: however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "open - domain tasks", "output": "entity retrieval", "neg_sample": ["open - domain tasks is done by using Task", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "In this work , we introduce TABi , a method to jointly train bi - encoders on knowledge graph types and unstructured text for entity retrieval for open - domain tasks .", "forward": false, "src_ids": "2022.findings-acl.169_3588"}
+{"input": "tabi is used for Task| context: however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "tabi", "output": "entity retrieval", "neg_sample": ["tabi is used for Task", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "In this work , we introduce TABi , a method to jointly train bi - encoders on knowledge graph types and unstructured text for entity retrieval for open - domain tasks .", "forward": true, "src_ids": "2022.findings-acl.169_3589"}
+{"input": "bi - encoders is used for Task| context: however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "bi - encoders", "output": "entity retrieval", "neg_sample": ["bi - encoders is used for Task", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "In this work , we introduce TABi , a method to jointly train bi - encoders on knowledge graph types and unstructured text for entity retrieval for open - domain tasks .", "forward": true, "src_ids": "2022.findings-acl.169_3590"}
+{"input": "entity retrieval is used for Task| context: however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "entity retrieval", "output": "open - domain tasks", "neg_sample": ["entity retrieval is used for Task", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "In this work , we introduce TABi , a method to jointly train bi - encoders on knowledge graph types and unstructured text for entity retrieval for open - domain tasks .", "forward": true, "src_ids": "2022.findings-acl.169_3591"}
+{"input": "popularity biases is done by using OtherScientificTerm| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "popularity biases", "output": "knowledge graph types", "neg_sample": ["popularity biases is done by using OtherScientificTerm", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "Incorporating knowledge graph types during training could help overcome popularity biases , but there are several challenges : ( 1 ) existing type - based retrieval methods require mention boundaries as input , but open - domain tasks run on unstructured text , ( 2 ) type - based methods should not compromise overall performance , and ( 3 ) type - based methods should be robust to noisy and missing types .", "forward": false, "src_ids": "2022.findings-acl.169_3592"}
+{"input": "knowledge graph types is used for OtherScientificTerm| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "knowledge graph types", "output": "popularity biases", "neg_sample": ["knowledge graph types is used for OtherScientificTerm", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "Incorporating knowledge graph types during training could help overcome popularity biases , but there are several challenges : ( 1 ) existing type - based retrieval methods require mention boundaries as input , but open - domain tasks run on unstructured text , ( 2 ) type - based methods should not compromise overall performance , and ( 3 ) type - based methods should be robust to noisy and missing types .", "forward": true, "src_ids": "2022.findings-acl.169_3593"}
+{"input": "noisy and missing types is done by using Method| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "noisy and missing types", "output": "type - based methods", "neg_sample": ["noisy and missing types is done by using Method", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "Incorporating knowledge graph types during training could help overcome popularity biases , but there are several challenges : ( 1 ) existing type - based retrieval methods require mention boundaries as input , but open - domain tasks run on unstructured text , ( 2 ) type - based methods should not compromise overall performance , and ( 3 ) type - based methods should be robust to noisy and missing types .", "forward": false, "src_ids": "2022.findings-acl.169_3594"}
+{"input": "type - based methods is used for OtherScientificTerm| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "type - based methods", "output": "noisy and missing types", "neg_sample": ["type - based methods is used for OtherScientificTerm", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "Incorporating knowledge graph types during training could help overcome popularity biases , but there are several challenges : ( 1 ) existing type - based retrieval methods require mention boundaries as input , but open - domain tasks run on unstructured text , ( 2 ) type - based methods should not compromise overall performance , and ( 3 ) type - based methods should be robust to noisy and missing types .", "forward": true, "src_ids": "2022.findings-acl.169_3595"}
+{"input": "entities is done by using OtherScientificTerm| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "entities", "output": "type - enforced contrastive loss", "neg_sample": ["entities is done by using OtherScientificTerm", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi leverages a type - enforced contrastive loss to encourage entities and queries of similar types to be close in the embedding space .", "forward": false, "src_ids": "2022.findings-acl.169_3596"}
+{"input": "type - enforced contrastive loss is used for OtherScientificTerm| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "entity": "type - enforced contrastive loss", "output": "entities", "neg_sample": ["type - enforced contrastive loss is used for OtherScientificTerm", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi leverages a type - enforced contrastive loss to encourage entities and queries of similar types to be close in the embedding space .", "forward": true, "src_ids": "2022.findings-acl.169_3597"}
+{"input": "tabi is used for Task| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "tabi", "output": "retrieval of rare entities", "neg_sample": ["tabi is used for Task", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi improves retrieval of rare entities on the Ambiguous Entity Retrieval ( AmbER ) sets , while maintaining strong overall retrieval performance on open - domain tasks in the KILT benchmark compared to state - of - the - art retrievers .", "forward": true, "src_ids": "2022.findings-acl.169_3598"}
+{"input": "tabi is used for Task| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "tabi", "output": "retrieval of rare entities", "neg_sample": ["tabi is used for Task", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi is also robust to incomplete type systems , improving rare entity retrieval over baselines with only 5 % type coverage of the training dataset .", "forward": true, "src_ids": "2022.findings-acl.169_3599"}
+{"input": "tabi is used for Method| context: entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking . however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities .", "entity": "tabi", "output": "incomplete type systems", "neg_sample": ["tabi is used for Method", "entity retrieval - retrieving information about entity mentions in a query - is a key step in open - domain tasks , such as question answering or fact checking .", "however , state - of - the - art entity retrievers struggle to retrieve rare entities for ambiguous mentions due to biases towards popular entities ."], "relation": "used for", "id": "2022.findings-acl.169", "year": 2022, "rel_sent": "TABi is also robust to incomplete type systems , improving rare entity retrieval over baselines with only 5 % type coverage of the training dataset .", "forward": true, "src_ids": "2022.findings-acl.169_3600"}
+{"input": "abuse detection is done by using Task| context: with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community .", "entity": "abuse detection", "output": "transliteration", "neg_sample": ["abuse detection is done by using Task", "with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community ."], "relation": "used for", "id": "2022.dravidianlangtech-1.5", "year": 2022, "rel_sent": "DE - ABUSE@TamilNLP - ACL 2022 : Transliteration as Data Augmentation for Abuse Detection in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.5_3601"}
+{"input": "abuse detection is done by using Method| context: with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community .", "entity": "abuse detection", "output": "data augmentation", "neg_sample": ["abuse detection is done by using Method", "with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community ."], "relation": "used for", "id": "2022.dravidianlangtech-1.5", "year": 2022, "rel_sent": "DE - ABUSE@TamilNLP - ACL 2022 : Transliteration as Data Augmentation for Abuse Detection in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.5_3602"}
+{"input": "data augmentation is used for Task| context: with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community .", "entity": "data augmentation", "output": "abuse detection", "neg_sample": ["data augmentation is used for Task", "with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community ."], "relation": "used for", "id": "2022.dravidianlangtech-1.5", "year": 2022, "rel_sent": "DE - ABUSE@TamilNLP - ACL 2022 : Transliteration as Data Augmentation for Abuse Detection in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.5_3603"}
+{"input": "transliteration is used for Task| context: with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community .", "entity": "transliteration", "output": "abuse detection", "neg_sample": ["transliteration is used for Task", "with the rise of social media and internet , thereis a necessity to provide an inclusive space andprevent the abusive topics against any gender , race or community ."], "relation": "used for", "id": "2022.dravidianlangtech-1.5", "year": 2022, "rel_sent": "DE - ABUSE@TamilNLP - ACL 2022 : Transliteration as Data Augmentation for Abuse Detection in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.5_3604"}
+{"input": "generating sentiment tuples is done by using Method| context: recent generative methods such as seq2seq models have achieved good performance by formulating the output as a sequence of sentiment tuples . however , the orders between the sentiment tuples do not naturally exist and the generation of the current tuple should not condition on the previous ones .", "entity": "generating sentiment tuples", "output": "seq2path", "neg_sample": ["generating sentiment tuples is done by using Method", "recent generative methods such as seq2seq models have achieved good performance by formulating the output as a sequence of sentiment tuples .", "however , the orders between the sentiment tuples do not naturally exist and the generation of the current tuple should not condition on the previous ones ."], "relation": "used for", "id": "2022.findings-acl.174", "year": 2022, "rel_sent": "Seq2Path : Generating Sentiment Tuples as Paths of a Tree.", "forward": false, "src_ids": "2022.findings-acl.174_3605"}
+{"input": "sentiment tuples is done by using Method| context: recent generative methods such as seq2seq models have achieved good performance by formulating the output as a sequence of sentiment tuples . however , the orders between the sentiment tuples do not naturally exist and the generation of the current tuple should not condition on the previous ones .", "entity": "sentiment tuples", "output": "seq2path", "neg_sample": ["sentiment tuples is done by using Method", "recent generative methods such as seq2seq models have achieved good performance by formulating the output as a sequence of sentiment tuples .", "however , the orders between the sentiment tuples do not naturally exist and the generation of the current tuple should not condition on the previous ones ."], "relation": "used for", "id": "2022.findings-acl.174", "year": 2022, "rel_sent": "In this paper , we propose Seq2Path to generate sentiment tuples as paths of a tree .", "forward": false, "src_ids": "2022.findings-acl.174_3606"}
+{"input": "seq2path is used for Task| context: recent generative methods such as seq2seq models have achieved good performance by formulating the output as a sequence of sentiment tuples . however , the orders between the sentiment tuples do not naturally exist and the generation of the current tuple should not condition on the previous ones .", "entity": "seq2path", "output": "generating sentiment tuples", "neg_sample": ["seq2path is used for Task", "recent generative methods such as seq2seq models have achieved good performance by formulating the output as a sequence of sentiment tuples .", "however , the orders between the sentiment tuples do not naturally exist and the generation of the current tuple should not condition on the previous ones ."], "relation": "used for", "id": "2022.findings-acl.174", "year": 2022, "rel_sent": "Seq2Path : Generating Sentiment Tuples as Paths of a Tree.", "forward": true, "src_ids": "2022.findings-acl.174_3607"}
+{"input": "text classification is done by using Method| context: in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce . in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance .", "entity": "text classification", "output": "cluster & tune", "neg_sample": ["text classification is done by using Method", "in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce .", "in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance ."], "relation": "used for", "id": "2022.acl-long.526", "year": 2022, "rel_sent": "Cluster & Tune : Boost Cold Start Performance in Text Classification.", "forward": false, "src_ids": "2022.acl-long.526_3608"}
+{"input": "cluster & tune is used for Task| context: in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance .", "entity": "cluster & tune", "output": "text classification", "neg_sample": ["cluster & tune is used for Task", "in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance ."], "relation": "used for", "id": "2022.acl-long.526", "year": 2022, "rel_sent": "Cluster & Tune : Boost Cold Start Performance in Text Classification.", "forward": true, "src_ids": "2022.acl-long.526_3609"}
+{"input": "topical classification tasks is done by using Task| context: in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce . in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance .", "entity": "topical classification tasks", "output": "classification phase", "neg_sample": ["topical classification tasks is done by using Task", "in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce .", "in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance ."], "relation": "used for", "id": "2022.acl-long.526", "year": 2022, "rel_sent": "As such an intermediate task , we perform clustering and train the pre - trained model on predicting the cluster labels . We test this hypothesis on various data sets , and show that this additional classification phase can significantly improve performance , mainly for topical classification tasks , when the number of labeled instances available for fine - tuning is only a couple of dozen to a few hundred .", "forward": false, "src_ids": "2022.acl-long.526_3610"}
+{"input": "classification phase is used for Task| context: in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce . in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance .", "entity": "classification phase", "output": "topical classification tasks", "neg_sample": ["classification phase is used for Task", "in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce .", "in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance ."], "relation": "used for", "id": "2022.acl-long.526", "year": 2022, "rel_sent": "As such an intermediate task , we perform clustering and train the pre - trained model on predicting the cluster labels . We test this hypothesis on various data sets , and show that this additional classification phase can significantly improve performance , mainly for topical classification tasks , when the number of labeled instances available for fine - tuning is only a couple of dozen to a few hundred .", "forward": true, "src_ids": "2022.acl-long.526_3611"}
+{"input": "post - editing machine translation is done by using Task| context: as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology .", "entity": "post - editing machine translation", "output": "error annotation", "neg_sample": ["post - editing machine translation is done by using Task", "as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology ."], "relation": "used for", "id": "2022.eamt-1.28", "year": 2022, "rel_sent": "Error Annotation in Post - Editing Machine Translation : Investigating the Impact of Text - to - Speech Technology.", "forward": false, "src_ids": "2022.eamt-1.28_3612"}
+{"input": "creative text is done by using Task| context: as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology .", "entity": "creative text", "output": "error annotation", "neg_sample": ["creative text is done by using Task", "as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology ."], "relation": "used for", "id": "2022.eamt-1.28", "year": 2022, "rel_sent": "Our study was conducted with university students and included both PEMT and error annotation of a creative text with and without T2S.", "forward": false, "src_ids": "2022.eamt-1.28_3613"}
+{"input": "error annotation is used for Task| context: as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology .", "entity": "error annotation", "output": "post - editing machine translation", "neg_sample": ["error annotation is used for Task", "as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology ."], "relation": "used for", "id": "2022.eamt-1.28", "year": 2022, "rel_sent": "Error Annotation in Post - Editing Machine Translation : Investigating the Impact of Text - to - Speech Technology.", "forward": true, "src_ids": "2022.eamt-1.28_3614"}
+{"input": "post - editors is done by using Method| context: as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology .", "entity": "post - editors", "output": "text - to - speech ( t2s )", "neg_sample": ["post - editors is done by using Method", "as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology ."], "relation": "used for", "id": "2022.eamt-1.28", "year": 2022, "rel_sent": "We present text - to - speech ( T2S ) as a potential attention - raising technology for post - editors .", "forward": false, "src_ids": "2022.eamt-1.28_3615"}
+{"input": "text - to - speech ( t2s ) is used for OtherScientificTerm| context: as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology .", "entity": "text - to - speech ( t2s )", "output": "post - editors", "neg_sample": ["text - to - speech ( t2s ) is used for OtherScientificTerm", "as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology ."], "relation": "used for", "id": "2022.eamt-1.28", "year": 2022, "rel_sent": "We present text - to - speech ( T2S ) as a potential attention - raising technology for post - editors .", "forward": true, "src_ids": "2022.eamt-1.28_3616"}
+{"input": "error annotation is used for Material| context: as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology .", "entity": "error annotation", "output": "creative text", "neg_sample": ["error annotation is used for Material", "as post - editing of machine translation ( pemt ) is becoming one of the most dominant services offered by the language services industry ( lsi ) , efforts are being made to support provision of this service with additional technology ."], "relation": "used for", "id": "2022.eamt-1.28", "year": 2022, "rel_sent": "Our study was conducted with university students and included both PEMT and error annotation of a creative text with and without T2S.", "forward": true, "src_ids": "2022.eamt-1.28_3617"}
+{"input": "compilable code generation is done by using Generic| context: automatically generating compilable programs with ( or without ) natural language descriptions has always been a touchstone problem for computational linguistics and automated software engineering . existing deep - learning approaches model code generation as text generation , either constrained by grammar structures in decoder , or driven by pre - trained language models on large - scale code corpus ( e.g. , codegpt , plbart , and codet5 ) . however , few of them account for compilability of the generated programs .", "entity": "compilable code generation", "output": "three - stage pipeline", "neg_sample": ["compilable code generation is done by using Generic", "automatically generating compilable programs with ( or without ) natural language descriptions has always been a touchstone problem for computational linguistics and automated software engineering .", "existing deep - learning approaches model code generation as text generation , either constrained by grammar structures in decoder , or driven by pre - trained language models on large - scale code corpus ( e.g.", ", codegpt , plbart , and codet5 ) .", "however , few of them account for compilability of the generated programs ."], "relation": "used for", "id": "2022.findings-acl.2", "year": 2022, "rel_sent": "To improve compilability of the generated programs , this paper proposes COMPCODER , a three - stage pipeline utilizing compiler feedback for compilable code generation , including language model fine - tuning , compilability reinforcement , and compilability discrimination .", "forward": false, "src_ids": "2022.findings-acl.2_3618"}
+{"input": "three - stage pipeline is used for Task| context: automatically generating compilable programs with ( or without ) natural language descriptions has always been a touchstone problem for computational linguistics and automated software engineering . existing deep - learning approaches model code generation as text generation , either constrained by grammar structures in decoder , or driven by pre - trained language models on large - scale code corpus ( e.g. , codegpt , plbart , and codet5 ) . however , few of them account for compilability of the generated programs .", "entity": "three - stage pipeline", "output": "compilable code generation", "neg_sample": ["three - stage pipeline is used for Task", "automatically generating compilable programs with ( or without ) natural language descriptions has always been a touchstone problem for computational linguistics and automated software engineering .", "existing deep - learning approaches model code generation as text generation , either constrained by grammar structures in decoder , or driven by pre - trained language models on large - scale code corpus ( e.g.", ", codegpt , plbart , and codet5 ) .", "however , few of them account for compilability of the generated programs ."], "relation": "used for", "id": "2022.findings-acl.2", "year": 2022, "rel_sent": "To improve compilability of the generated programs , this paper proposes COMPCODER , a three - stage pipeline utilizing compiler feedback for compilable code generation , including language model fine - tuning , compilability reinforcement , and compilability discrimination .", "forward": true, "src_ids": "2022.findings-acl.2_3619"}
+{"input": "fine - grained entity typing is done by using Method| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "fine - grained entity typing", "output": "cross - lingual contrastive learning framework", "neg_sample": ["fine - grained entity typing is done by using Method", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "Cross - Lingual Contrastive Learning for Fine - Grained Entity Typing for Low - Resource Languages.", "forward": false, "src_ids": "2022.acl-long.159_3620"}
+{"input": "fget models is done by using Method| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "fget models", "output": "cross - lingual contrastive learning framework", "neg_sample": ["fget models is done by using Method", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "In this paper , we propose a cross - lingual contrastive learning framework to learn FGET models for low - resource languages .", "forward": false, "src_ids": "2022.acl-long.159_3621"}
+{"input": "cross - lingual contrastive learning framework is used for Task| context: especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "cross - lingual contrastive learning framework", "output": "fine - grained entity typing", "neg_sample": ["cross - lingual contrastive learning framework is used for Task", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "Cross - Lingual Contrastive Learning for Fine - Grained Entity Typing for Low - Resource Languages.", "forward": true, "src_ids": "2022.acl-long.159_3622"}
+{"input": "cross - lingual contrastive learning framework is used for Method| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "cross - lingual contrastive learning framework", "output": "fget models", "neg_sample": ["cross - lingual contrastive learning framework is used for Method", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "In this paper , we propose a cross - lingual contrastive learning framework to learn FGET models for low - resource languages .", "forward": true, "src_ids": "2022.acl-long.159_3623"}
+{"input": "cross - lingual distantly - supervised data is done by using Method| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "cross - lingual distantly - supervised data", "output": "entity - pair - oriented heuristic rules", "neg_sample": ["cross - lingual distantly - supervised data is done by using Method", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "Furthermore , we introduce entity - pair - oriented heuristic rules as well as machine translation to obtain cross - lingual distantly - supervised data , and apply cross - lingual contrastive learning on the distantly - supervised data to enhance the backbone PLMs .", "forward": false, "src_ids": "2022.acl-long.159_3624"}
+{"input": "cross - lingual distantly - supervised data is done by using Task| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "cross - lingual distantly - supervised data", "output": "machine translation", "neg_sample": ["cross - lingual distantly - supervised data is done by using Task", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "Furthermore , we introduce entity - pair - oriented heuristic rules as well as machine translation to obtain cross - lingual distantly - supervised data , and apply cross - lingual contrastive learning on the distantly - supervised data to enhance the backbone PLMs .", "forward": false, "src_ids": "2022.acl-long.159_3625"}
+{"input": "entity - pair - oriented heuristic rules is used for Material| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "entity - pair - oriented heuristic rules", "output": "cross - lingual distantly - supervised data", "neg_sample": ["entity - pair - oriented heuristic rules is used for Material", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "Furthermore , we introduce entity - pair - oriented heuristic rules as well as machine translation to obtain cross - lingual distantly - supervised data , and apply cross - lingual contrastive learning on the distantly - supervised data to enhance the backbone PLMs .", "forward": true, "src_ids": "2022.acl-long.159_3626"}
+{"input": "machine translation is used for Material| context: for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "machine translation", "output": "cross - lingual distantly - supervised data", "neg_sample": ["machine translation is used for Material", "for fget , a key challenge is the low - resource problem - the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.159", "year": 2022, "rel_sent": "Furthermore , we introduce entity - pair - oriented heuristic rules as well as machine translation to obtain cross - lingual distantly - supervised data , and apply cross - lingual contrastive learning on the distantly - supervised data to enhance the backbone PLMs .", "forward": true, "src_ids": "2022.acl-long.159_3627"}
+{"input": "mesh terms is done by using OtherScientificTerm| context: currently , medical subject headings ( mesh ) are manually assigned to every biomedical article published and subsequently recorded in the pubmed database tofacilitate retrieving relevant information . with the rapid growth of the pubmed database , large - scale biomedical document indexing becomes increasingly important . mesh indexing is a challenging task for machine learning , as it needs to assign multiple labels to each article from an extremely large hierachically organized collection .", "entity": "mesh terms", "output": "document features", "neg_sample": ["mesh terms is done by using OtherScientificTerm", "currently , medical subject headings ( mesh ) are manually assigned to every biomedical article published and subsequently recorded in the pubmed database tofacilitate retrieving relevant information .", "with the rapid growth of the pubmed database , large - scale biomedical document indexing becomes increasingly important .", "mesh indexing is a challenging task for machine learning , as it needs to assign multiple labels to each article from an extremely large hierachically organized collection ."], "relation": "used for", "id": "2022.acl-long.210", "year": 2022, "rel_sent": "To address this challenge , we propose KenMeSH , an end - to - end model that combines new text features and a dynamic knowledge - enhanced mask attention that integrates document features with MeSH label hierarchy and journal correlation features to index MeSH terms .", "forward": false, "src_ids": "2022.acl-long.210_3628"}
+{"input": "journal correlation features is used for OtherScientificTerm| context: currently , medical subject headings ( mesh ) are manually assigned to every biomedical article published and subsequently recorded in the pubmed database tofacilitate retrieving relevant information . with the rapid growth of the pubmed database , large - scale biomedical document indexing becomes increasingly important . mesh indexing is a challenging task for machine learning , as it needs to assign multiple labels to each article from an extremely large hierachically organized collection .", "entity": "journal correlation features", "output": "mesh terms", "neg_sample": ["journal correlation features is used for OtherScientificTerm", "currently , medical subject headings ( mesh ) are manually assigned to every biomedical article published and subsequently recorded in the pubmed database tofacilitate retrieving relevant information .", "with the rapid growth of the pubmed database , large - scale biomedical document indexing becomes increasingly important .", "mesh indexing is a challenging task for machine learning , as it needs to assign multiple labels to each article from an extremely large hierachically organized collection ."], "relation": "used for", "id": "2022.acl-long.210", "year": 2022, "rel_sent": "To address this challenge , we propose KenMeSH , an end - to - end model that combines new text features and a dynamic knowledge - enhanced mask attention that integrates document features with MeSH label hierarchy and journal correlation features to index MeSH terms .", "forward": true, "src_ids": "2022.acl-long.210_3629"}
+{"input": "document features is used for OtherScientificTerm| context: currently , medical subject headings ( mesh ) are manually assigned to every biomedical article published and subsequently recorded in the pubmed database tofacilitate retrieving relevant information . with the rapid growth of the pubmed database , large - scale biomedical document indexing becomes increasingly important . mesh indexing is a challenging task for machine learning , as it needs to assign multiple labels to each article from an extremely large hierachically organized collection .", "entity": "document features", "output": "mesh terms", "neg_sample": ["document features is used for OtherScientificTerm", "currently , medical subject headings ( mesh ) are manually assigned to every biomedical article published and subsequently recorded in the pubmed database tofacilitate retrieving relevant information .", "with the rapid growth of the pubmed database , large - scale biomedical document indexing becomes increasingly important .", "mesh indexing is a challenging task for machine learning , as it needs to assign multiple labels to each article from an extremely large hierachically organized collection ."], "relation": "used for", "id": "2022.acl-long.210", "year": 2022, "rel_sent": "To address this challenge , we propose KenMeSH , an end - to - end model that combines new text features and a dynamic knowledge - enhanced mask attention that integrates document features with MeSH label hierarchy and journal correlation features to index MeSH terms .", "forward": true, "src_ids": "2022.acl-long.210_3630"}
+{"input": "hope - speech detection is done by using Generic| context: hope is an inherent part of human life and essential for improving the quality of life . hope increases happiness and reduces stress and feelings of helplessness . hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes .", "entity": "hope - speech detection", "output": "deep - learning approach", "neg_sample": ["hope - speech detection is done by using Generic", "hope is an inherent part of human life and essential for improving the quality of life .", "hope increases happiness and reduces stress and feelings of helplessness .", "hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes ."], "relation": "used for", "id": "2022.ltedi-1.28", "year": 2022, "rel_sent": "In this paper , we address a deep - learning approach with a combination of linguistic and psycho - linguistic features for hope - speech detection .", "forward": false, "src_ids": "2022.ltedi-1.28_3631"}
+{"input": "hope - speech detection is done by using OtherScientificTerm| context: hope is an inherent part of human life and essential for improving the quality of life . hope increases happiness and reduces stress and feelings of helplessness . hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes .", "entity": "hope - speech detection", "output": "linguistic and psycho - linguistic features", "neg_sample": ["hope - speech detection is done by using OtherScientificTerm", "hope is an inherent part of human life and essential for improving the quality of life .", "hope increases happiness and reduces stress and feelings of helplessness .", "hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes ."], "relation": "used for", "id": "2022.ltedi-1.28", "year": 2022, "rel_sent": "In this paper , we address a deep - learning approach with a combination of linguistic and psycho - linguistic features for hope - speech detection .", "forward": false, "src_ids": "2022.ltedi-1.28_3632"}
+{"input": "linguistic and psycho - linguistic features is used for Task| context: hope is an inherent part of human life and essential for improving the quality of life . hope increases happiness and reduces stress and feelings of helplessness . hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes .", "entity": "linguistic and psycho - linguistic features", "output": "hope - speech detection", "neg_sample": ["linguistic and psycho - linguistic features is used for Task", "hope is an inherent part of human life and essential for improving the quality of life .", "hope increases happiness and reduces stress and feelings of helplessness .", "hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes ."], "relation": "used for", "id": "2022.ltedi-1.28", "year": 2022, "rel_sent": "In this paper , we address a deep - learning approach with a combination of linguistic and psycho - linguistic features for hope - speech detection .", "forward": true, "src_ids": "2022.ltedi-1.28_3633"}
+{"input": "deep - learning approach is used for Task| context: hope is an inherent part of human life and essential for improving the quality of life . hope increases happiness and reduces stress and feelings of helplessness . hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes .", "entity": "deep - learning approach", "output": "hope - speech detection", "neg_sample": ["deep - learning approach is used for Task", "hope is an inherent part of human life and essential for improving the quality of life .", "hope increases happiness and reduces stress and feelings of helplessness .", "hope speech is the desired outcome for better and can be studied using text from various online sources where people express their desires and outcomes ."], "relation": "used for", "id": "2022.ltedi-1.28", "year": 2022, "rel_sent": "In this paper , we address a deep - learning approach with a combination of linguistic and psycho - linguistic features for hope - speech detection .", "forward": true, "src_ids": "2022.ltedi-1.28_3634"}
+{"input": "training strategy is used for Task| context: most state - of - the - art matching models , e.g. , bert , directly perform text comparison by processing each word uniformly . however , a query sentence generally comprises content that calls for different levels of matching granularity . specifically , keywords represent factual information such as action , entity , and event that should be strictly matched , while intents convey abstract concepts and ideas that can be paraphrased into various expressions .", "entity": "training strategy", "output": "text semantic matching", "neg_sample": ["training strategy is used for Task", "most state - of - the - art matching models , e.g.", ", bert , directly perform text comparison by processing each word uniformly .", "however , a query sentence generally comprises content that calls for different levels of matching granularity .", "specifically , keywords represent factual information such as action , entity , and event that should be strictly matched , while intents convey abstract concepts and ideas that can be paraphrased into various expressions ."], "relation": "used for", "id": "2022.findings-acl.287", "year": 2022, "rel_sent": "In this work , we propose a simple yet effective training strategy for text semantic matching in a divide - and - conquer manner by disentangling keywords from intents .", "forward": true, "src_ids": "2022.findings-acl.287_3635"}
+{"input": "text semantic matching is done by using Method| context: text semantic matching is a fundamental task that has been widely used in various scenarios , such as community question answering , information retrieval , and recommendation . most state - of - the - art matching models , e.g. , bert , directly perform text comparison by processing each word uniformly . however , a query sentence generally comprises content that calls for different levels of matching granularity . specifically , keywords represent factual information such as action , entity , and event that should be strictly matched , while intents convey abstract concepts and ideas that can be paraphrased into various expressions .", "entity": "text semantic matching", "output": "training strategy", "neg_sample": ["text semantic matching is done by using Method", "text semantic matching is a fundamental task that has been widely used in various scenarios , such as community question answering , information retrieval , and recommendation .", "most state - of - the - art matching models , e.g.", ", bert , directly perform text comparison by processing each word uniformly .", "however , a query sentence generally comprises content that calls for different levels of matching granularity .", "specifically , keywords represent factual information such as action , entity , and event that should be strictly matched , while intents convey abstract concepts and ideas that can be paraphrased into various expressions ."], "relation": "used for", "id": "2022.findings-acl.287", "year": 2022, "rel_sent": "In this work , we propose a simple yet effective training strategy for text semantic matching in a divide - and - conquer manner by disentangling keywords from intents .", "forward": false, "src_ids": "2022.findings-acl.287_3636"}
+{"input": "human - like perception biases is done by using Method| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "human - like perception biases", "output": "self - supervised speech models", "neg_sample": ["human - like perception biases is done by using Method", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "Do self - supervised speech models develop human - like perception biases ?.", "forward": false, "src_ids": "2022.acl-long.523_3637"}
+{"input": "self - supervised speech models is used for OtherScientificTerm| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "self - supervised speech models", "output": "human - like perception biases", "neg_sample": ["self - supervised speech models is used for OtherScientificTerm", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "Do self - supervised speech models develop human - like perception biases ?.", "forward": true, "src_ids": "2022.acl-long.523_3638"}
+{"input": "fine - grained perceptual phenomena is done by using Method| context: increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "entity": "fine - grained perceptual phenomena", "output": "self - supervised models", "neg_sample": ["fine - grained perceptual phenomena is done by using Method", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages ."], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "A comparison against the predictions of supervised phone recognisers suggests that all three self - supervised models capture relatively fine - grained perceptual phenomena , while supervised models are better at capturing coarser , phone - level effects , and effects of listeners ' native language , on perception .", "forward": false, "src_ids": "2022.acl-long.523_3639"}
+{"input": "phone - level effects is done by using Method| context: increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "entity": "phone - level effects", "output": "self - supervised models", "neg_sample": ["phone - level effects is done by using Method", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages ."], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "A comparison against the predictions of supervised phone recognisers suggests that all three self - supervised models capture relatively fine - grained perceptual phenomena , while supervised models are better at capturing coarser , phone - level effects , and effects of listeners ' native language , on perception .", "forward": false, "src_ids": "2022.acl-long.523_3640"}
+{"input": "universal speech perception space is done by using Method| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "universal speech perception space", "output": "wav2vec", "neg_sample": ["universal speech perception space is done by using Method", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "We show that the CPC model shows a small native language effect , but that wav2vec and HuBERT seem to develop a universal speech perception space which is not language specific .", "forward": false, "src_ids": "2022.acl-long.523_3641"}
+{"input": "wav2vec is used for OtherScientificTerm| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "wav2vec", "output": "universal speech perception space", "neg_sample": ["wav2vec is used for OtherScientificTerm", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "We show that the CPC model shows a small native language effect , but that wav2vec and HuBERT seem to develop a universal speech perception space which is not language specific .", "forward": true, "src_ids": "2022.acl-long.523_3642"}
+{"input": "hubert is used for OtherScientificTerm| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "hubert", "output": "universal speech perception space", "neg_sample": ["hubert is used for OtherScientificTerm", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "We show that the CPC model shows a small native language effect , but that wav2vec and HuBERT seem to develop a universal speech perception space which is not language specific .", "forward": true, "src_ids": "2022.acl-long.523_3643"}
+{"input": "self - supervised models is used for OtherScientificTerm| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "self - supervised models", "output": "fine - grained perceptual phenomena", "neg_sample": ["self - supervised models is used for OtherScientificTerm", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "A comparison against the predictions of supervised phone recognisers suggests that all three self - supervised models capture relatively fine - grained perceptual phenomena , while supervised models are better at capturing coarser , phone - level effects , and effects of listeners ' native language , on perception .", "forward": true, "src_ids": "2022.acl-long.523_3644"}
+{"input": "perception is done by using Method| context: increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "entity": "perception", "output": "supervised models", "neg_sample": ["perception is done by using Method", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages ."], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "A comparison against the predictions of supervised phone recognisers suggests that all three self - supervised models capture relatively fine - grained perceptual phenomena , while supervised models are better at capturing coarser , phone - level effects , and effects of listeners ' native language , on perception .", "forward": false, "src_ids": "2022.acl-long.523_3645"}
+{"input": "supervised models is used for OtherScientificTerm| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages . does the same thing happen in self - supervised models ?", "entity": "supervised models", "output": "phone - level effects", "neg_sample": ["supervised models is used for OtherScientificTerm", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "but what kind of representational spaces do these models construct?human perception specializes to the sounds of listeners ' native languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "A comparison against the predictions of supervised phone recognisers suggests that all three self - supervised models capture relatively fine - grained perceptual phenomena , while supervised models are better at capturing coarser , phone - level effects , and effects of listeners ' native language , on perception .", "forward": true, "src_ids": "2022.acl-long.523_3646"}
+{"input": "supervised models is used for OtherScientificTerm| context: self - supervised models for speech processing form representational spaces without using any external labels . increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages . does the same thing happen in self - supervised models ?", "entity": "supervised models", "output": "perception", "neg_sample": ["supervised models is used for OtherScientificTerm", "self - supervised models for speech processing form representational spaces without using any external labels .", "increasingly , they appear to be a feasible way of at least partially eliminating costly manual annotations , a problem of particular concern for low - resource languages .", "does the same thing happen in self - supervised models ?"], "relation": "used for", "id": "2022.acl-long.523", "year": 2022, "rel_sent": "A comparison against the predictions of supervised phone recognisers suggests that all three self - supervised models capture relatively fine - grained perceptual phenomena , while supervised models are better at capturing coarser , phone - level effects , and effects of listeners ' native language , on perception .", "forward": true, "src_ids": "2022.acl-long.523_3647"}
+{"input": "program understanding is done by using Method| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "program understanding", "output": "neural network architecture", "neg_sample": ["program understanding is done by using Method", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "A Neural Network Architecture for Program Understanding Inspired by Human Behaviors.", "forward": false, "src_ids": "2022.acl-long.353_3648"}
+{"input": "neural network architecture is used for Task| context: despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "neural network architecture", "output": "program understanding", "neg_sample": ["neural network architecture is used for Task", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "A Neural Network Architecture for Program Understanding Inspired by Human Behaviors.", "forward": true, "src_ids": "2022.acl-long.353_3649"}
+{"input": "ast of codes is done by using Method| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "ast of codes", "output": "partitioning - based graph neural network model pgnn", "neg_sample": ["ast of codes is done by using Method", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "On the one hand , inspired by the ' divide - and - conquer ' reading behaviors of humans , we present a partitioning - based graph neural network model PGNN on the upgraded AST of codes .", "forward": false, "src_ids": "2022.acl-long.353_3650"}
+{"input": "partitioning - based graph neural network model pgnn is used for Material| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "partitioning - based graph neural network model pgnn", "output": "ast of codes", "neg_sample": ["partitioning - based graph neural network model pgnn is used for Material", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "On the one hand , inspired by the ' divide - and - conquer ' reading behaviors of humans , we present a partitioning - based graph neural network model PGNN on the upgraded AST of codes .", "forward": true, "src_ids": "2022.acl-long.353_3651"}
+{"input": "information extraction is done by using Method| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "information extraction", "output": "pre - training techniques", "neg_sample": ["information extraction is done by using Method", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "On the other hand , to characterize human behaviors of resorting to other resources to help code comprehension , we transform raw codes with external knowledge and apply pre - training techniques for information extraction .", "forward": false, "src_ids": "2022.acl-long.353_3652"}
+{"input": "pre - training techniques is used for Task| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "pre - training techniques", "output": "information extraction", "neg_sample": ["pre - training techniques is used for Task", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "On the other hand , to characterize human behaviors of resorting to other resources to help code comprehension , we transform raw codes with external knowledge and apply pre - training techniques for information extraction .", "forward": true, "src_ids": "2022.acl-long.353_3653"}
+{"input": "code embeddings is done by using OtherScientificTerm| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "code embeddings", "output": "embeddings", "neg_sample": ["code embeddings is done by using OtherScientificTerm", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "Finally , we combine the two embeddings generated from the two components to output code embeddings .", "forward": false, "src_ids": "2022.acl-long.353_3654"}
+{"input": "embeddings is used for Method| context: program understanding is a fundamental task in program language processing . despite the success , existing works fail to take human behaviors as reference in understanding programs .", "entity": "embeddings", "output": "code embeddings", "neg_sample": ["embeddings is used for Method", "program understanding is a fundamental task in program language processing .", "despite the success , existing works fail to take human behaviors as reference in understanding programs ."], "relation": "used for", "id": "2022.acl-long.353", "year": 2022, "rel_sent": "Finally , we combine the two embeddings generated from the two components to output code embeddings .", "forward": true, "src_ids": "2022.acl-long.353_3655"}
+{"input": "data augmentation is used for Task| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "data augmentation", "output": "intent classification", "neg_sample": ["data augmentation is used for Task", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "Data Augmentation for Intent Classification with Off - the - shelf Large Language Models.", "forward": true, "src_ids": "2022.nlp4convai-1.5_3656"}
+{"input": "labelled training data is used for Task| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "labelled training data", "output": "intent classification", "neg_sample": ["labelled training data is used for Task", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "In this work , we propose a prompting - based approach to generate labelled training data for intent classification with off - the - shelf language models ( LMs ) such as GPT-3 .", "forward": true, "src_ids": "2022.nlp4convai-1.5_3657"}
+{"input": "prompting - based approach is used for Task| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "prompting - based approach", "output": "intent classification", "neg_sample": ["prompting - based approach is used for Task", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "In this work , we propose a prompting - based approach to generate labelled training data for intent classification with off - the - shelf language models ( LMs ) such as GPT-3 .", "forward": true, "src_ids": "2022.nlp4convai-1.5_3658"}
+{"input": "intent classification is done by using Method| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "intent classification", "output": "prompting - based approach", "neg_sample": ["intent classification is done by using Method", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "In this work , we propose a prompting - based approach to generate labelled training data for intent classification with off - the - shelf language models ( LMs ) such as GPT-3 .", "forward": false, "src_ids": "2022.nlp4convai-1.5_3659"}
+{"input": "labelled training data is done by using Method| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "labelled training data", "output": "prompting - based approach", "neg_sample": ["labelled training data is done by using Method", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "In this work , we propose a prompting - based approach to generate labelled training data for intent classification with off - the - shelf language models ( LMs ) such as GPT-3 .", "forward": false, "src_ids": "2022.nlp4convai-1.5_3660"}
+{"input": "intent classification is done by using Material| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "intent classification", "output": "labelled training data", "neg_sample": ["intent classification is done by using Material", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "In this work , we propose a prompting - based approach to generate labelled training data for intent classification with off - the - shelf language models ( LMs ) such as GPT-3 .", "forward": false, "src_ids": "2022.nlp4convai-1.5_3661"}
+{"input": "prompting - based approach is used for Material| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "prompting - based approach", "output": "labelled training data", "neg_sample": ["prompting - based approach is used for Material", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "In this work , we propose a prompting - based approach to generate labelled training data for intent classification with off - the - shelf language models ( LMs ) such as GPT-3 .", "forward": true, "src_ids": "2022.nlp4convai-1.5_3662"}
+{"input": "data generation is done by using Method| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "data generation", "output": "task - specific lm - fine - tuning", "neg_sample": ["data generation is done by using Method", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "An advantage of this method is that no task - specific LM - fine - tuning for data generation is required ; hence the method requires no hyper parameter tuning and is applicable even when the available training data is very scarce .", "forward": false, "src_ids": "2022.nlp4convai-1.5_3663"}
+{"input": "task - specific lm - fine - tuning is used for Task| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "task - specific lm - fine - tuning", "output": "data generation", "neg_sample": ["task - specific lm - fine - tuning is used for Task", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "An advantage of this method is that no task - specific LM - fine - tuning for data generation is required ; hence the method requires no hyper parameter tuning and is applicable even when the available training data is very scarce .", "forward": true, "src_ids": "2022.nlp4convai-1.5_3664"}
+{"input": "intent classifiers is done by using Material| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "intent classifiers", "output": "gpt - generated data", "neg_sample": ["intent classifiers is done by using Material", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "We find that GPT - generated data significantly boosts the performance of intent classifiers when intents in consideration are sufficiently distinct from each other .", "forward": false, "src_ids": "2022.nlp4convai-1.5_3665"}
+{"input": "gpt - generated data is used for Method| context: data augmentation is a widely employed technique to alleviate the problem of data scarcity .", "entity": "gpt - generated data", "output": "intent classifiers", "neg_sample": ["gpt - generated data is used for Method", "data augmentation is a widely employed technique to alleviate the problem of data scarcity ."], "relation": "used for", "id": "2022.nlp4convai-1.5", "year": 2022, "rel_sent": "We find that GPT - generated data significantly boosts the performance of intent classifiers when intents in consideration are sufficiently distinct from each other .", "forward": true, "src_ids": "2022.nlp4convai-1.5_3666"}
+{"input": "cause - effect relation classification is done by using Method| context: previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models . however , these methods behave differently across domains and downstream tasks .", "entity": "cause - effect relation classification", "output": "knowledge - augmented language models", "neg_sample": ["cause - effect relation classification is done by using Method", "previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models .", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "Knowledge - Augmented Language Models for Cause - Effect Relation Classification.", "forward": false, "src_ids": "2022.csrr-1.6_3667"}
+{"input": "knowledge - augmented language models is used for Task| context: previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models . however , these methods behave differently across domains and downstream tasks .", "entity": "knowledge - augmented language models", "output": "cause - effect relation classification", "neg_sample": ["knowledge - augmented language models is used for Task", "previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models .", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "Knowledge - Augmented Language Models for Cause - Effect Relation Classification.", "forward": true, "src_ids": "2022.csrr-1.6_3668"}
+{"input": "pretrained language models is used for Task| context: previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models . however , these methods behave differently across domains and downstream tasks .", "entity": "pretrained language models", "output": "cause - effect relation classification", "neg_sample": ["pretrained language models is used for Task", "previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models .", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "In this work , we investigate the augmentation of pretrained language models with knowledge graph data in the cause - effect relation classification and commonsense causal reasoning tasks .", "forward": true, "src_ids": "2022.csrr-1.6_3669"}
+{"input": "cause - effect relation classification is done by using Method| context: however , these methods behave differently across domains and downstream tasks .", "entity": "cause - effect relation classification", "output": "pretrained language models", "neg_sample": ["cause - effect relation classification is done by using Method", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "In this work , we investigate the augmentation of pretrained language models with knowledge graph data in the cause - effect relation classification and commonsense causal reasoning tasks .", "forward": false, "src_ids": "2022.csrr-1.6_3670"}
+{"input": "commonsense causal reasoning questions is done by using Method| context: however , these methods behave differently across domains and downstream tasks .", "entity": "commonsense causal reasoning questions", "output": "pretrained language models", "neg_sample": ["commonsense causal reasoning questions is done by using Method", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "In this work , we investigate the augmentation of pretrained language models with knowledge graph data in the cause - effect relation classification and commonsense causal reasoning tasks .", "forward": false, "src_ids": "2022.csrr-1.6_3671"}
+{"input": "fine - tuning is done by using Material| context: previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models . however , these methods behave differently across domains and downstream tasks .", "entity": "fine - tuning", "output": "quality - enhanced data", "neg_sample": ["fine - tuning is done by using Material", "previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models .", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "Our results show that a continually pretrained language model augmented with commonsense reasoning knowledge outperforms our baselines on two commonsense causal reasoning benchmarks , COPA and BCOPA - CE , and a Temporal and Causal Reasoning ( TCR ) dataset , without additional improvement in model architecture or using quality - enhanced data for fine - tuning .", "forward": false, "src_ids": "2022.csrr-1.6_3672"}
+{"input": "quality - enhanced data is used for Method| context: previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models . however , these methods behave differently across domains and downstream tasks .", "entity": "quality - enhanced data", "output": "fine - tuning", "neg_sample": ["quality - enhanced data is used for Method", "previous studies have shown the efficacy of knowledge augmentation methods in pretrained language models .", "however , these methods behave differently across domains and downstream tasks ."], "relation": "used for", "id": "2022.csrr-1.6", "year": 2022, "rel_sent": "Our results show that a continually pretrained language model augmented with commonsense reasoning knowledge outperforms our baselines on two commonsense causal reasoning benchmarks , COPA and BCOPA - CE , and a Temporal and Causal Reasoning ( TCR ) dataset , without additional improvement in model architecture or using quality - enhanced data for fine - tuning .", "forward": true, "src_ids": "2022.csrr-1.6_3673"}
+{"input": "two - tier bert architecture is used for OtherScientificTerm| context: pre - trained language models such as bert have been successful at tackling many natural language processing tasks . however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g. , byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages .", "entity": "two - tier bert architecture", "output": "morphological compositionality", "neg_sample": ["two - tier bert architecture is used for OtherScientificTerm", "pre - trained language models such as bert have been successful at tackling many natural language processing tasks .", "however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g.", ", byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages ."], "relation": "used for", "id": "2022.acl-long.367", "year": 2022, "rel_sent": "We address these challenges by proposing a simple yet effective two - tier BERT architecture that leverages a morphological analyzer and explicitly represents morphological compositionality . Despite the success of BERT , most of its evaluations have been conducted on high - resource languages , obscuring its applicability on low - resource languages .", "forward": true, "src_ids": "2022.acl-long.367_3674"}
+{"input": "morphological compositionality is done by using Method| context: pre - trained language models such as bert have been successful at tackling many natural language processing tasks . however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g. , byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages . even given a morphological analyzer , naive sequencing of morphemes into a standard bert architecture is inefficient at capturing morphological compositionality and expressing word - relative syntactic regularities .", "entity": "morphological compositionality", "output": "two - tier bert architecture", "neg_sample": ["morphological compositionality is done by using Method", "pre - trained language models such as bert have been successful at tackling many natural language processing tasks .", "however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g.", ", byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages .", "even given a morphological analyzer , naive sequencing of morphemes into a standard bert architecture is inefficient at capturing morphological compositionality and expressing word - relative syntactic regularities ."], "relation": "used for", "id": "2022.acl-long.367", "year": 2022, "rel_sent": "We address these challenges by proposing a simple yet effective two - tier BERT architecture that leverages a morphological analyzer and explicitly represents morphological compositionality . Despite the success of BERT , most of its evaluations have been conducted on high - resource languages , obscuring its applicability on low - resource languages .", "forward": false, "src_ids": "2022.acl-long.367_3675"}
+{"input": "weak supervision is done by using Method| context: a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias .", "entity": "weak supervision", "output": "generative modeling", "neg_sample": ["weak supervision is done by using Method", "a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias ."], "relation": "used for", "id": "2022.repl4nlp-1.27", "year": 2022, "rel_sent": "In this work , we explore a novel direction of generative modeling for weak supervision ' : ' Instead of modeling the output of the annotation process ( the labeling function matches ) , we generatively model the input - side data distributions ( the feature space ) covered by labeling functions .", "forward": false, "src_ids": "2022.repl4nlp-1.27_3676"}
+{"input": "density is used for OtherScientificTerm| context: a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias . methods for overcoming these problems have either relied on discriminative models , trained with cost functions specific to weak supervision , and more recently , generative models , trying to model the output of the automatic annotation process .", "entity": "density", "output": "labeling functions", "neg_sample": ["density is used for OtherScientificTerm", "a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias .", "methods for overcoming these problems have either relied on discriminative models , trained with cost functions specific to weak supervision , and more recently , generative models , trying to model the output of the automatic annotation process ."], "relation": "used for", "id": "2022.repl4nlp-1.27", "year": 2022, "rel_sent": "Specifically , we estimate a density for each weak labeling source , or labeling function , by using normalizing flows .", "forward": true, "src_ids": "2022.repl4nlp-1.27_3677"}
+{"input": "weak labeling source is done by using OtherScientificTerm| context: a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias . methods for overcoming these problems have either relied on discriminative models , trained with cost functions specific to weak supervision , and more recently , generative models , trying to model the output of the automatic annotation process .", "entity": "weak labeling source", "output": "density", "neg_sample": ["weak labeling source is done by using OtherScientificTerm", "a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias .", "methods for overcoming these problems have either relied on discriminative models , trained with cost functions specific to weak supervision , and more recently , generative models , trying to model the output of the automatic annotation process ."], "relation": "used for", "id": "2022.repl4nlp-1.27", "year": 2022, "rel_sent": "Specifically , we estimate a density for each weak labeling source , or labeling function , by using normalizing flows .", "forward": false, "src_ids": "2022.repl4nlp-1.27_3678"}
+{"input": "labeling functions is done by using OtherScientificTerm| context: a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias . methods for overcoming these problems have either relied on discriminative models , trained with cost functions specific to weak supervision , and more recently , generative models , trying to model the output of the automatic annotation process .", "entity": "labeling functions", "output": "density", "neg_sample": ["labeling functions is done by using OtherScientificTerm", "a popular approach to decrease the need for costly manual annotation of large data sets is weak supervision , which introduces problems of noisy labels , coverage and bias .", "methods for overcoming these problems have either relied on discriminative models , trained with cost functions specific to weak supervision , and more recently , generative models , trying to model the output of the automatic annotation process ."], "relation": "used for", "id": "2022.repl4nlp-1.27", "year": 2022, "rel_sent": "Specifically , we estimate a density for each weak labeling source , or labeling function , by using normalizing flows .", "forward": false, "src_ids": "2022.repl4nlp-1.27_3679"}
+{"input": "empathy and emotion detection is done by using Method| context: our system , iucl , participated in the wassa 2022 shared task on empathy detection and emotion classification .", "entity": "empathy and emotion detection", "output": "text - only approach", "neg_sample": ["empathy and emotion detection is done by using Method", "our system , iucl , participated in the wassa 2022 shared task on empathy detection and emotion classification ."], "relation": "used for", "id": "2022.wassa-1.21", "year": 2022, "rel_sent": "IUCL at WASSA 2022 Shared Task : A Text - only Approach to Empathy and Emotion Detection.", "forward": false, "src_ids": "2022.wassa-1.21_3680"}
+{"input": "text - only approach is used for Task| context: our system , iucl , participated in the wassa 2022 shared task on empathy detection and emotion classification .", "entity": "text - only approach", "output": "empathy and emotion detection", "neg_sample": ["text - only approach is used for Task", "our system , iucl , participated in the wassa 2022 shared task on empathy detection and emotion classification ."], "relation": "used for", "id": "2022.wassa-1.21", "year": 2022, "rel_sent": "IUCL at WASSA 2022 Shared Task : A Text - only Approach to Empathy and Emotion Detection.", "forward": true, "src_ids": "2022.wassa-1.21_3681"}
+{"input": "dialogue summarization is done by using OtherScientificTerm| context: dialogue summarization helps users capture salient information from various types of dialogues has received much attention recently . however , current works mainly focus on english dialogue summarization , leaving other languages less well explored .", "entity": "dialogue summarization", "output": "multi - lingual settings", "neg_sample": ["dialogue summarization is done by using OtherScientificTerm", "dialogue summarization helps users capture salient information from various types of dialogues has received much attention recently .", "however , current works mainly focus on english dialogue summarization , leaving other languages less well explored ."], "relation": "used for", "id": "2022.dialdoc-1.1", "year": 2022, "rel_sent": "Given the proposed MSAMum , we systematically set up five multi - lingual settings for this task , including a novel mix - lingual dialogue summarization setting .", "forward": false, "src_ids": "2022.dialdoc-1.1_3682"}
+{"input": "emotional support conversation is done by using Method| context: applying existing methods to emotional support conversation - which provides valuable assistance to people who are in need - has two major limitations : ( a ) they generally employ a conversation - level emotion label , which is too coarse - grained to capture user 's instant mental state ; ( b ) most of them focus on expressing empathy in the response(s ) rather than gradually reducing user 's distress .", "entity": "emotional support conversation", "output": "comet", "neg_sample": ["emotional support conversation is done by using Method", "applying existing methods to emotional support conversation - which provides valuable assistance to people who are in need - has two major limitations : ( a ) they generally employ a conversation - level emotion label , which is too coarse - grained to capture user 's instant mental state ; ( b ) most of them focus on expressing empathy in the response(s ) rather than gradually reducing user 's distress ."], "relation": "used for", "id": "2022.acl-long.25", "year": 2022, "rel_sent": "MISC : A Mixed Strategy - Aware Model integrating COMET for Emotional Support Conversation.", "forward": false, "src_ids": "2022.acl-long.25_3683"}
+{"input": "inference is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "inference", "output": "multi - level mutual promotion mechanism", "neg_sample": ["inference is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "MPII : Multi - Level Mutual Promotion for Inference and Interpretation.", "forward": false, "src_ids": "2022.acl-long.488_3684"}
+{"input": "interpretation is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "interpretation", "output": "multi - level mutual promotion mechanism", "neg_sample": ["interpretation is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "MPII : Multi - Level Mutual Promotion for Inference and Interpretation.", "forward": false, "src_ids": "2022.acl-long.488_3685"}
+{"input": "sentence - level interpretation ( mpii ) is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "sentence - level interpretation ( mpii )", "output": "multi - level mutual promotion mechanism", "neg_sample": ["sentence - level interpretation ( mpii ) is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "In this paper , we propose a multi - level Mutual Promotion mechanism for self - evolved Inference and sentence - level Interpretation ( MPII ) .", "forward": false, "src_ids": "2022.acl-long.488_3686"}
+{"input": "self - evolved inference is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "self - evolved inference", "output": "multi - level mutual promotion mechanism", "neg_sample": ["self - evolved inference is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "In this paper , we propose a multi - level Mutual Promotion mechanism for self - evolved Inference and sentence - level Interpretation ( MPII ) .", "forward": false, "src_ids": "2022.acl-long.488_3687"}
+{"input": "multi - level mutual promotion mechanism is used for Task| context: however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "entity": "multi - level mutual promotion mechanism", "output": "inference", "neg_sample": ["multi - level mutual promotion mechanism is used for Task", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "MPII : Multi - Level Mutual Promotion for Inference and Interpretation.", "forward": true, "src_ids": "2022.acl-long.488_3688"}
+{"input": "step - wise integration mechanism is used for Task| context: however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "entity": "step - wise integration mechanism", "output": "inference", "neg_sample": ["step - wise integration mechanism is used for Task", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "Specifically , from the model - level , we propose a Step - wise Integration Mechanism to jointly perform and deeply integrate inference and interpretation in an autoregressive manner .", "forward": true, "src_ids": "2022.acl-long.488_3689"}
+{"input": "multi - level mutual promotion mechanism is used for Task| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "multi - level mutual promotion mechanism", "output": "self - evolved inference", "neg_sample": ["multi - level mutual promotion mechanism is used for Task", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "In this paper , we propose a multi - level Mutual Promotion mechanism for self - evolved Inference and sentence - level Interpretation ( MPII ) .", "forward": true, "src_ids": "2022.acl-long.488_3690"}
+{"input": "inference is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "inference", "output": "step - wise integration mechanism", "neg_sample": ["inference is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "Specifically , from the model - level , we propose a Step - wise Integration Mechanism to jointly perform and deeply integrate inference and interpretation in an autoregressive manner .", "forward": false, "src_ids": "2022.acl-long.488_3691"}
+{"input": "interpretation is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "interpretation", "output": "step - wise integration mechanism", "neg_sample": ["interpretation is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "Specifically , from the model - level , we propose a Step - wise Integration Mechanism to jointly perform and deeply integrate inference and interpretation in an autoregressive manner .", "forward": false, "src_ids": "2022.acl-long.488_3692"}
+{"input": "interpretation is done by using Method| context: in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction . however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e. either inference promotion with interpretation or vice versa .", "entity": "interpretation", "output": "adversarial fidelity regularization", "neg_sample": ["interpretation is done by using Method", "in order to better understand the rationale behind model behavior , recent works have exploited providing interpretation to support the inference prediction .", "however , existing methods tend to provide human - unfriendly interpretation , and are prone to sub - optimal performance due to one - side promotion , i.e.", "either inference promotion with interpretation or vice versa ."], "relation": "used for", "id": "2022.acl-long.488", "year": 2022, "rel_sent": "From the optimization - level , we propose an Adversarial Fidelity Regularization to improve the fidelity between inference and interpretation with the Adversarial Mutual Information training strategy .", "forward": false, "src_ids": "2022.acl-long.488_3693"}
+{"input": "low - resource stereotype detection is done by using Method| context: as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus . existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms . blodgett et . al . ( 2021 ) show that there are significant reliability issues with the existing benchmark datasets . annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text .", "entity": "low - resource stereotype detection", "output": "reinforcement guided multi - task learning framework", "neg_sample": ["low - resource stereotype detection is done by using Method", "as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus .", "existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms .", "blodgett et .", "al .", "( 2021 ) show that there are significant reliability issues with the existing benchmark datasets .", "annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text ."], "relation": "used for", "id": "2022.acl-long.462", "year": 2022, "rel_sent": "Reinforcement Guided Multi - Task Learning Framework for Low - Resource Stereotype Detection.", "forward": false, "src_ids": "2022.acl-long.462_3694"}
+{"input": "reinforcement guided multi - task learning framework is used for Task| context: as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus . existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms . blodgett et . al . ( 2021 ) show that there are significant reliability issues with the existing benchmark datasets . annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text .", "entity": "reinforcement guided multi - task learning framework", "output": "low - resource stereotype detection", "neg_sample": ["reinforcement guided multi - task learning framework is used for Task", "as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus .", "existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms .", "blodgett et .", "al .", "( 2021 ) show that there are significant reliability issues with the existing benchmark datasets .", "annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text ."], "relation": "used for", "id": "2022.acl-long.462", "year": 2022, "rel_sent": "Reinforcement Guided Multi - Task Learning Framework for Low - Resource Stereotype Detection.", "forward": true, "src_ids": "2022.acl-long.462_3695"}
+{"input": "multi - task learning model is done by using Method| context: as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus . existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms . blodgett et . al . ( 2021 ) show that there are significant reliability issues with the existing benchmark datasets . annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text .", "entity": "multi - task learning model", "output": "reinforcement - learning agent", "neg_sample": ["multi - task learning model is done by using Method", "as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus .", "existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms .", "blodgett et .", "al .", "( 2021 ) show that there are significant reliability issues with the existing benchmark datasets .", "annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text ."], "relation": "used for", "id": "2022.acl-long.462", "year": 2022, "rel_sent": "We then propose a reinforcement - learning agent that guides the multi - task learning model by learning to identify the training examples from the neighboring tasks that help the target task the most .", "forward": false, "src_ids": "2022.acl-long.462_3696"}
+{"input": "reinforcement - learning agent is used for Method| context: as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus . existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms . blodgett et . al . ( 2021 ) show that there are significant reliability issues with the existing benchmark datasets . annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text .", "entity": "reinforcement - learning agent", "output": "multi - task learning model", "neg_sample": ["reinforcement - learning agent is used for Method", "as large pre - trained language models ( plms ) trained on large amounts of data in an unsupervised manner become more ubiquitous , identifying various types of bias in the text has come into sharp focus .", "existing ' stereotype detection ' datasets mainly adopt a diagnostic approach toward large plms .", "blodgett et .", "al .", "( 2021 ) show that there are significant reliability issues with the existing benchmark datasets .", "annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text ."], "relation": "used for", "id": "2022.acl-long.462", "year": 2022, "rel_sent": "We then propose a reinforcement - learning agent that guides the multi - task learning model by learning to identify the training examples from the neighboring tasks that help the target task the most .", "forward": true, "src_ids": "2022.acl-long.462_3697"}
+{"input": "coarse summary is done by using Method| context: text summarization helps readers capture salient information from documents , news , interviews , and meetings . however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks .", "entity": "coarse summary", "output": "summ^n", "neg_sample": ["coarse summary is done by using Method", "text summarization helps readers capture salient information from documents , news , interviews , and meetings .", "however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks ."], "relation": "used for", "id": "2022.acl-long.112", "year": 2022, "rel_sent": "Summ^N first splits the data samples and generates a coarse summary in multiple stages and then produces the final fine - grained summary based on it .", "forward": false, "src_ids": "2022.acl-long.112_3698"}
+{"input": "long input dialogues is done by using Method| context: text summarization helps readers capture salient information from documents , news , interviews , and meetings . however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks .", "entity": "long input dialogues", "output": "multi - stage summarization framework", "neg_sample": ["long input dialogues is done by using Method", "text summarization helps readers capture salient information from documents , news , interviews , and meetings .", "however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks ."], "relation": "used for", "id": "2022.acl-long.112", "year": 2022, "rel_sent": "Summ^N : A Multi - Stage Summarization Framework for Long Input Dialogues and Documents.", "forward": false, "src_ids": "2022.acl-long.112_3699"}
+{"input": "multi - stage summarization framework is used for Material| context: text summarization helps readers capture salient information from documents , news , interviews , and meetings . however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks .", "entity": "multi - stage summarization framework", "output": "long input dialogues", "neg_sample": ["multi - stage summarization framework is used for Material", "text summarization helps readers capture salient information from documents , news , interviews , and meetings .", "however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks ."], "relation": "used for", "id": "2022.acl-long.112", "year": 2022, "rel_sent": "Summ^N : A Multi - Stage Summarization Framework for Long Input Dialogues and Documents.", "forward": true, "src_ids": "2022.acl-long.112_3700"}
+{"input": "summ^n is used for OtherScientificTerm| context: text summarization helps readers capture salient information from documents , news , interviews , and meetings . however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks .", "entity": "summ^n", "output": "coarse summary", "neg_sample": ["summ^n is used for OtherScientificTerm", "text summarization helps readers capture salient information from documents , news , interviews , and meetings .", "however , most state - of - the - art pretrained language models ( lm ) are unable to efficiently process long text for many summarization tasks ."], "relation": "used for", "id": "2022.acl-long.112", "year": 2022, "rel_sent": "Summ^N first splits the data samples and generates a coarse summary in multiple stages and then produces the final fine - grained summary based on it .", "forward": true, "src_ids": "2022.acl-long.112_3701"}
+{"input": "neural machine translation is done by using Method| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner . although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context .", "entity": "neural machine translation", "output": "confidence based bidirectional global context aware training framework", "neg_sample": ["neural machine translation is done by using Method", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "Confidence Based Bidirectional Global Context Aware Training Framework for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.206_3702"}
+{"input": "confidence based bidirectional global context aware training framework is used for Task| context: although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context .", "entity": "confidence based bidirectional global context aware training framework", "output": "neural machine translation", "neg_sample": ["confidence based bidirectional global context aware training framework is used for Task", "although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "Confidence Based Bidirectional Global Context Aware Training Framework for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.206_3703"}
+{"input": "encoder parameters is used for Method| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "entity": "encoder parameters", "output": "nmt models", "neg_sample": ["encoder parameters is used for Method", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "At the first stage , by sharing encoder parameters , the NMT model is additionally supervised by the signal from the CMLM decoder that contains bidirectional global contexts .", "forward": true, "src_ids": "2022.acl-long.206_3704"}
+{"input": "bidirectional global contexts is used for Method| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "entity": "bidirectional global contexts", "output": "nmt models", "neg_sample": ["bidirectional global contexts is used for Method", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "Moreover , at the second stage , using the CMLM as teacher , we further pertinently incorporate bidirectional global context to the NMT model on its unconfidently - predicted target words via knowledge distillation .", "forward": true, "src_ids": "2022.acl-long.206_3705"}
+{"input": "cbbgca training framework is used for Method| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "entity": "cbbgca training framework", "output": "nmt models", "neg_sample": ["cbbgca training framework is used for Method", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "Experimental results show that our proposed CBBGCA training framework significantly improves the NMT model by +1.02 , +1.30 and +0.57 BLEU scores on three large - scale translation datasets , namely WMT'14 English - to - German , WMT'19 Chinese - to - English and WMT'14 English - to - French , respectively .", "forward": true, "src_ids": "2022.acl-long.206_3706"}
+{"input": "nmt models is done by using OtherScientificTerm| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "entity": "nmt models", "output": "bidirectional global contexts", "neg_sample": ["nmt models is done by using OtherScientificTerm", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "Moreover , at the second stage , using the CMLM as teacher , we further pertinently incorporate bidirectional global context to the NMT model on its unconfidently - predicted target words via knowledge distillation .", "forward": false, "src_ids": "2022.acl-long.206_3707"}
+{"input": "nmt models is done by using OtherScientificTerm| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner . although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context .", "entity": "nmt models", "output": "encoder parameters", "neg_sample": ["nmt models is done by using OtherScientificTerm", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "At the first stage , by sharing encoder parameters , the NMT model is additionally supervised by the signal from the CMLM decoder that contains bidirectional global contexts .", "forward": false, "src_ids": "2022.acl-long.206_3708"}
+{"input": "nmt models is done by using Method| context: most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner . although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context .", "entity": "nmt models", "output": "cbbgca training framework", "neg_sample": ["nmt models is done by using Method", "most dominant neural machine translation ( nmt ) models are restricted to make predictions only according to the local context of preceding words in a left - to - right manner .", "although many previous studies try to incorporate global information into nmt models , there still exist limitations on how to effectively exploit bidirectional global context ."], "relation": "used for", "id": "2022.acl-long.206", "year": 2022, "rel_sent": "Experimental results show that our proposed CBBGCA training framework significantly improves the NMT model by +1.02 , +1.30 and +0.57 BLEU scores on three large - scale translation datasets , namely WMT'14 English - to - German , WMT'19 Chinese - to - English and WMT'14 English - to - French , respectively .", "forward": false, "src_ids": "2022.acl-long.206_3709"}
+{"input": "interpretability is used for OtherScientificTerm| context: grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning . however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored . a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections .", "entity": "interpretability", "output": "language learners", "neg_sample": ["interpretability is used for OtherScientificTerm", "grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning .", "however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored .", "a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections ."], "relation": "used for", "id": "2022.acl-long.496", "year": 2022, "rel_sent": "Interpretability for Language Learners Using Example - Based Grammatical Error Correction.", "forward": true, "src_ids": "2022.acl-long.496_3710"}
+{"input": "eb - gec is used for OtherScientificTerm| context: grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning . however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored . a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections .", "entity": "eb - gec", "output": "language learners", "neg_sample": ["eb - gec is used for OtherScientificTerm", "grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning .", "however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored .", "a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections ."], "relation": "used for", "id": "2022.acl-long.496", "year": 2022, "rel_sent": "In addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing . Therefore , we hypothesize that incorporating an example - based method into GEC can improve interpretability as well as support language learners . In this study , we introduce an Example - Based GEC ( EB - GEC ) that presents examples to language learners as a basis for a correction result . The examples consist of pairs of correct and incorrect sentences similar to a given input and its predicted correction . Experiments demonstrate that the examples presented by EB - GEC help language learners decide to accept or refuse suggestions from the GEC output . Furthermore , the experiments also show that retrieved examples improve the accuracy of corrections .", "forward": true, "src_ids": "2022.acl-long.496_3711"}
+{"input": "language learners is done by using Method| context: grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning . however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored . a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections .", "entity": "language learners", "output": "eb - gec", "neg_sample": ["language learners is done by using Method", "grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning .", "however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored .", "a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections ."], "relation": "used for", "id": "2022.acl-long.496", "year": 2022, "rel_sent": "In addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing . Therefore , we hypothesize that incorporating an example - based method into GEC can improve interpretability as well as support language learners . In this study , we introduce an Example - Based GEC ( EB - GEC ) that presents examples to language learners as a basis for a correction result . The examples consist of pairs of correct and incorrect sentences similar to a given input and its predicted correction . Experiments demonstrate that the examples presented by EB - GEC help language learners decide to accept or refuse suggestions from the GEC output . Furthermore , the experiments also show that retrieved examples improve the accuracy of corrections .", "forward": false, "src_ids": "2022.acl-long.496_3712"}
+{"input": "bbai task is done by using Method| context: the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks . though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities .", "entity": "bbai task", "output": "mars encoder", "neg_sample": ["bbai task is done by using Method", "the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks .", "though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities ."], "relation": "used for", "id": "2022.findings-acl.257", "year": 2022, "rel_sent": "Specifically , using the MARS encoder we achieve the highest accuracy on our BBAI task , outperforming strong baselines .", "forward": false, "src_ids": "2022.findings-acl.257_3713"}
+{"input": "encoder model is used for Task| context: the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks . though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities .", "entity": "encoder model", "output": "question response pairing", "neg_sample": ["encoder model is used for Task", "the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks .", "though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities ."], "relation": "used for", "id": "2022.findings-acl.257", "year": 2022, "rel_sent": "Additionally , we introduce MARS : Multi - Agent Response Selection , a new encoder model for question response pairing that jointly encodes user question and agent response pairs .", "forward": true, "src_ids": "2022.findings-acl.257_3714"}
+{"input": "mars encoder is used for Task| context: the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks . though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities .", "entity": "mars encoder", "output": "bbai task", "neg_sample": ["mars encoder is used for Task", "the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks .", "though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities ."], "relation": "used for", "id": "2022.findings-acl.257", "year": 2022, "rel_sent": "Specifically , using the MARS encoder we achieve the highest accuracy on our BBAI task , outperforming strong baselines .", "forward": true, "src_ids": "2022.findings-acl.257_3715"}
+{"input": "question response pairing is done by using Method| context: the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks . though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities .", "entity": "question response pairing", "output": "encoder model", "neg_sample": ["question response pairing is done by using Method", "the increasing volume of commercially available conversational agents ( cas ) on the market has resulted in users being burdened with learning and adopting multiple agents to accomplish their tasks .", "though prior work has explored supporting a multitude of domains within the design of a single agent , the interaction experience suffers due to the large action space of desired capabilities ."], "relation": "used for", "id": "2022.findings-acl.257", "year": 2022, "rel_sent": "Additionally , we introduce MARS : Multi - Agent Response Selection , a new encoder model for question response pairing that jointly encodes user question and agent response pairs .", "forward": false, "src_ids": "2022.findings-acl.257_3716"}
+{"input": "multi task learning is done by using Method| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "multi task learning", "output": "q - learning scheduler", "neg_sample": ["multi task learning is done by using Method", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "Q - Learning Scheduler for Multi Task Learning Through the use of Histogram of Task Uncertainty.", "forward": false, "src_ids": "2022.repl4nlp-1.2_3717"}
+{"input": "q - learning scheduler is used for Method| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "q - learning scheduler", "output": "multi task learning", "neg_sample": ["q - learning scheduler is used for Method", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "Q - Learning Scheduler for Multi Task Learning Through the use of Histogram of Task Uncertainty.", "forward": true, "src_ids": "2022.repl4nlp-1.2_3718"}
+{"input": "shared features is done by using Method| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "shared features", "output": "deep q - learning scheduler ( qls )", "neg_sample": ["shared features is done by using Method", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "We proposed a deep Q - Learning Scheduler ( QLS ) that monitors the state of the tasks and the shared features using a novel histogram of task uncertainty , and through trial - and - error , learns an optimal policy for task scheduling .", "forward": false, "src_ids": "2022.repl4nlp-1.2_3719"}
+{"input": "deep q - learning scheduler ( qls ) is used for OtherScientificTerm| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "deep q - learning scheduler ( qls )", "output": "shared features", "neg_sample": ["deep q - learning scheduler ( qls ) is used for OtherScientificTerm", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "We proposed a deep Q - Learning Scheduler ( QLS ) that monitors the state of the tasks and the shared features using a novel histogram of task uncertainty , and through trial - and - error , learns an optimal policy for task scheduling .", "forward": true, "src_ids": "2022.repl4nlp-1.2_3720"}
+{"input": "optimal policy is done by using Method| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "optimal policy", "output": "trial - and - error", "neg_sample": ["optimal policy is done by using Method", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "We proposed a deep Q - Learning Scheduler ( QLS ) that monitors the state of the tasks and the shared features using a novel histogram of task uncertainty , and through trial - and - error , learns an optimal policy for task scheduling .", "forward": false, "src_ids": "2022.repl4nlp-1.2_3721"}
+{"input": "task scheduling is done by using OtherScientificTerm| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "task scheduling", "output": "optimal policy", "neg_sample": ["task scheduling is done by using OtherScientificTerm", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "We proposed a deep Q - Learning Scheduler ( QLS ) that monitors the state of the tasks and the shared features using a novel histogram of task uncertainty , and through trial - and - error , learns an optimal policy for task scheduling .", "forward": false, "src_ids": "2022.repl4nlp-1.2_3722"}
+{"input": "trial - and - error is used for OtherScientificTerm| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "trial - and - error", "output": "optimal policy", "neg_sample": ["trial - and - error is used for OtherScientificTerm", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "We proposed a deep Q - Learning Scheduler ( QLS ) that monitors the state of the tasks and the shared features using a novel histogram of task uncertainty , and through trial - and - error , learns an optimal policy for task scheduling .", "forward": true, "src_ids": "2022.repl4nlp-1.2_3723"}
+{"input": "optimal policy is used for Task| context: simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward . it could lead to inferior performance or generalization compared to the corresponding single - task networks . an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning .", "entity": "optimal policy", "output": "task scheduling", "neg_sample": ["optimal policy is used for Task", "simultaneous training of a multi - task learning network on different domains or tasks is not always straightforward .", "it could lead to inferior performance or generalization compared to the corresponding single - task networks .", "an effective training scheduling method is deemed necessary to maximize the benefits of multi - task learning ."], "relation": "used for", "id": "2022.repl4nlp-1.2", "year": 2022, "rel_sent": "We proposed a deep Q - Learning Scheduler ( QLS ) that monitors the state of the tasks and the shared features using a novel histogram of task uncertainty , and through trial - and - error , learns an optimal policy for task scheduling .", "forward": true, "src_ids": "2022.repl4nlp-1.2_3724"}
+{"input": "active learning is done by using Method| context: active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings . as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date . this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings .", "entity": "active learning", "output": "uncertainty - based query strategies", "neg_sample": ["active learning is done by using Method", "active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings .", "as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date .", "this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings ."], "relation": "used for", "id": "2022.findings-acl.172", "year": 2022, "rel_sent": "Revisiting Uncertainty - based Query Strategies for Active Learning with Transformers.", "forward": false, "src_ids": "2022.findings-acl.172_3725"}
+{"input": "fine - tuning transformers is done by using Method| context: active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings . as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date . this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings .", "entity": "fine - tuning transformers", "output": "uncertainty - based query strategies", "neg_sample": ["fine - tuning transformers is done by using Method", "active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings .", "as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date .", "this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings ."], "relation": "used for", "id": "2022.findings-acl.172", "year": 2022, "rel_sent": "For this reason , we revisit uncertainty - based query strategies , which had been largely outperformed before , but are particularly suited in the context of fine - tuning transformers .", "forward": false, "src_ids": "2022.findings-acl.172_3726"}
+{"input": "text classification is done by using Method| context: this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings .", "entity": "text classification", "output": "active learning", "neg_sample": ["text classification is done by using Method", "this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings ."], "relation": "used for", "id": "2022.findings-acl.172", "year": 2022, "rel_sent": "For active learning with transformers , several other uncertainty - based approaches outperform the well - known prediction entropy query strategy , thereby challenging its status as most popular uncertainty baseline in active learning for text classification .", "forward": false, "src_ids": "2022.findings-acl.172_3727"}
+{"input": "uncertainty - based query strategies is used for Method| context: this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings .", "entity": "uncertainty - based query strategies", "output": "active learning", "neg_sample": ["uncertainty - based query strategies is used for Method", "this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings ."], "relation": "used for", "id": "2022.findings-acl.172", "year": 2022, "rel_sent": "Revisiting Uncertainty - based Query Strategies for Active Learning with Transformers.", "forward": true, "src_ids": "2022.findings-acl.172_3728"}
+{"input": "uncertainty - based query strategies is used for Method| context: active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings . as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date . this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings .", "entity": "uncertainty - based query strategies", "output": "fine - tuning transformers", "neg_sample": ["uncertainty - based query strategies is used for Method", "active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings .", "as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date .", "this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings ."], "relation": "used for", "id": "2022.findings-acl.172", "year": 2022, "rel_sent": "For this reason , we revisit uncertainty - based query strategies , which had been largely outperformed before , but are particularly suited in the context of fine - tuning transformers .", "forward": true, "src_ids": "2022.findings-acl.172_3729"}
+{"input": "active learning is used for Task| context: active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings . as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date . this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings .", "entity": "active learning", "output": "text classification", "neg_sample": ["active learning is used for Task", "active learning is the iterative construction of a classification model through targeted labeling , enabling significant labeling cost savings .", "as most research on active learning has been carried out before transformer - based language models ( ' transformers ' ) became popular , despite its practical importance , comparably few papers have investigated how transformers can be combined with active learning to date .", "this can be attributed to the fact that using state - of - the - art query strategies for transformers induces a prohibitive runtime overhead , which effectively nullifies , or even outweighs the desired cost savings ."], "relation": "used for", "id": "2022.findings-acl.172", "year": 2022, "rel_sent": "For active learning with transformers , several other uncertainty - based approaches outperform the well - known prediction entropy query strategy , thereby challenging its status as most popular uncertainty baseline in active learning for text classification .", "forward": true, "src_ids": "2022.findings-acl.172_3730"}
+{"input": "end - to - end text augmentation is done by using Method| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "end - to - end text augmentation", "output": "multi - level optimization framework", "neg_sample": ["end - to - end text augmentation is done by using Method", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "A Multi - Level Optimization Framework for End - to - End Text Augmentation.", "forward": false, "src_ids": "2022.tacl-1.20_3731"}
+{"input": "multi - level optimization framework is used for Task| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "multi - level optimization framework", "output": "end - to - end text augmentation", "neg_sample": ["multi - level optimization framework is used for Task", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "A Multi - Level Optimization Framework for End - to - End Text Augmentation.", "forward": true, "src_ids": "2022.tacl-1.20_3732"}
+{"input": "three - level optimization framework is used for Task| context: as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "three - level optimization framework", "output": "text augmentation", "neg_sample": ["three - level optimization framework is used for Task", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "To address this problem , we propose a three - level optimization framework to perform text augmentation and the downstream task end - to- end .", "forward": true, "src_ids": "2022.tacl-1.20_3733"}
+{"input": "three - level optimization framework is used for Task| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "three - level optimization framework", "output": "downstream tasks", "neg_sample": ["three - level optimization framework is used for Task", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "To address this problem , we propose a three - level optimization framework to perform text augmentation and the downstream task end - to- end .", "forward": true, "src_ids": "2022.tacl-1.20_3734"}
+{"input": "augmentation model is used for Task| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "augmentation model", "output": "downstream tasks", "neg_sample": ["augmentation model is used for Task", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "The augmentation model is trained in a way tailored to the downstream task .", "forward": true, "src_ids": "2022.tacl-1.20_3735"}
+{"input": "text augmentation is done by using Method| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "text augmentation", "output": "three - level optimization framework", "neg_sample": ["text augmentation is done by using Method", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "To address this problem , we propose a three - level optimization framework to perform text augmentation and the downstream task end - to- end .", "forward": false, "src_ids": "2022.tacl-1.20_3736"}
+{"input": "downstream tasks is done by using Method| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "downstream tasks", "output": "three - level optimization framework", "neg_sample": ["downstream tasks is done by using Method", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "To address this problem , we propose a three - level optimization framework to perform text augmentation and the downstream task end - to- end .", "forward": false, "src_ids": "2022.tacl-1.20_3737"}
+{"input": "downstream tasks is done by using Method| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "downstream tasks", "output": "augmentation model", "neg_sample": ["downstream tasks is done by using Method", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "The augmentation model is trained in a way tailored to the downstream task .", "forward": false, "src_ids": "2022.tacl-1.20_3738"}
+{"input": "data augmentation is done by using Method| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "data augmentation", "output": "text summarization model", "neg_sample": ["data augmentation is done by using Method", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "A text summarization model is trained to perform data augmentation at the first stage .", "forward": false, "src_ids": "2022.tacl-1.20_3739"}
+{"input": "text summarization model is used for Method| context: text augmentation is an effective technique in alleviating overfitting in nlp tasks . in existing methods , text augmentation and downstream tasks are mostly performed separately . as a result , the augmented texts may not be optimal to train the downstream model .", "entity": "text summarization model", "output": "data augmentation", "neg_sample": ["text summarization model is used for Method", "text augmentation is an effective technique in alleviating overfitting in nlp tasks .", "in existing methods , text augmentation and downstream tasks are mostly performed separately .", "as a result , the augmented texts may not be optimal to train the downstream model ."], "relation": "used for", "id": "2022.tacl-1.20", "year": 2022, "rel_sent": "A text summarization model is trained to perform data augmentation at the first stage .", "forward": true, "src_ids": "2022.tacl-1.20_3740"}
+{"input": "language communication is done by using Method| context: human communication is a collaborative process . speakers , on top of conveying their own intent , adjust the content and language expressions by taking the listeners into account , including their knowledge background , personalities , and physical capabilities .", "entity": "language communication", "output": "ai agents", "neg_sample": ["language communication is done by using Method", "human communication is a collaborative process .", "speakers , on top of conveying their own intent , adjust the content and language expressions by taking the listeners into account , including their knowledge background , personalities , and physical capabilities ."], "relation": "used for", "id": "2022.acl-long.202", "year": 2022, "rel_sent": "Towards building AI agents with similar abilities in language communication , we propose a novel rational reasoning framework , Pragmatic Rational Speaker ( PRS ) , where the speaker attempts to learn the speaker - listener disparity and adjust the speech accordingly , by adding a light - weighted disparity adjustment layer into working memory on top of speaker 's long - term memory system .", "forward": false, "src_ids": "2022.acl-long.202_3741"}
+{"input": "ai agents is used for Task| context: human communication is a collaborative process . speakers , on top of conveying their own intent , adjust the content and language expressions by taking the listeners into account , including their knowledge background , personalities , and physical capabilities .", "entity": "ai agents", "output": "language communication", "neg_sample": ["ai agents is used for Task", "human communication is a collaborative process .", "speakers , on top of conveying their own intent , adjust the content and language expressions by taking the listeners into account , including their knowledge background , personalities , and physical capabilities ."], "relation": "used for", "id": "2022.acl-long.202", "year": 2022, "rel_sent": "Towards building AI agents with similar abilities in language communication , we propose a novel rational reasoning framework , Pragmatic Rational Speaker ( PRS ) , where the speaker attempts to learn the speaker - listener disparity and adjust the speech accordingly , by adding a light - weighted disparity adjustment layer into working memory on top of speaker 's long - term memory system .", "forward": true, "src_ids": "2022.acl-long.202_3742"}
+{"input": "contrastive learning is done by using OtherScientificTerm| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "entity": "contrastive learning", "output": "negative samples", "neg_sample": ["contrastive learning is done by using OtherScientificTerm", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "Exploring the Impact of Negative Samples of Contrastive Learning : A Case Study of Sentence Embedding.", "forward": false, "src_ids": "2022.findings-acl.248_3743"}
+{"input": "negative samples is used for Method| context: this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . this is typically achieved by maintaining a queue of negative samples during training . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "entity": "negative samples", "output": "contrastive learning", "neg_sample": ["negative samples is used for Method", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "this is typically achieved by maintaining a queue of negative samples during training .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "Exploring the Impact of Negative Samples of Contrastive Learning : A Case Study of Sentence Embedding.", "forward": true, "src_ids": "2022.findings-acl.248_3744"}
+{"input": "momentum contrastive learning model is used for Task| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . this is typically achieved by maintaining a queue of negative samples during training . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear . the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration .", "entity": "momentum contrastive learning model", "output": "sentence embedding", "neg_sample": ["momentum contrastive learning model is used for Task", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "this is typically achieved by maintaining a queue of negative samples during training .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "This paper presents a momentum contrastive learning model with negative sample queue for sentence embedding , namely MoCoSE .", "forward": true, "src_ids": "2022.findings-acl.248_3745"}
+{"input": "historical information is used for OtherScientificTerm| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear . the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration .", "entity": "historical information", "output": "queue of negative samples", "neg_sample": ["historical information is used for OtherScientificTerm", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "Our experiments find that the best results are obtained when the maximum traceable distance is at a certain range , demonstrating that there is an optimal range of historical information for a negative sample queue .", "forward": true, "src_ids": "2022.findings-acl.248_3746"}
+{"input": "sentence embedding is done by using Method| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . this is typically achieved by maintaining a queue of negative samples during training . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear . the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration .", "entity": "sentence embedding", "output": "momentum contrastive learning model", "neg_sample": ["sentence embedding is done by using Method", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "this is typically achieved by maintaining a queue of negative samples during training .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "This paper presents a momentum contrastive learning model with negative sample queue for sentence embedding , namely MoCoSE .", "forward": false, "src_ids": "2022.findings-acl.248_3747"}
+{"input": "text contrastive learning is done by using Metric| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . this is typically achieved by maintaining a queue of negative samples during training . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear . the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration .", "entity": "text contrastive learning", "output": "maximum traceable distance metric", "neg_sample": ["text contrastive learning is done by using Metric", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "this is typically achieved by maintaining a queue of negative samples during training .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "We define a maximum traceable distance metric , through which we learn to what extent the text contrastive learning benefits from the historical information of negative samples .", "forward": false, "src_ids": "2022.findings-acl.248_3748"}
+{"input": "maximum traceable distance metric is used for Method| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . this is typically achieved by maintaining a queue of negative samples during training . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear . the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration .", "entity": "maximum traceable distance metric", "output": "text contrastive learning", "neg_sample": ["maximum traceable distance metric is used for Method", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "this is typically achieved by maintaining a queue of negative samples during training .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "We define a maximum traceable distance metric , through which we learn to what extent the text contrastive learning benefits from the historical information of negative samples .", "forward": true, "src_ids": "2022.findings-acl.248_3749"}
+{"input": "queue of negative samples is done by using OtherScientificTerm| context: contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data . this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples . this is typically achieved by maintaining a queue of negative samples during training . prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear . the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration .", "entity": "queue of negative samples", "output": "historical information", "neg_sample": ["queue of negative samples is done by using OtherScientificTerm", "contrastive learning is emerging as a powerful technique for extracting knowledge from unlabeled data .", "this technique requires a balanced mixture of two ingredients : positive ( similar ) and negative ( dissimilar ) samples .", "this is typically achieved by maintaining a queue of negative samples during training .", "prior works in the area typically uses a fixed - length negative sample queue , but how the negative sample size affects the model performance remains unclear .", "the opaque impact of the number of negative samples on performance when employing contrastive learning aroused our in - depth exploration ."], "relation": "used for", "id": "2022.findings-acl.248", "year": 2022, "rel_sent": "Our experiments find that the best results are obtained when the maximum traceable distance is at a certain range , demonstrating that there is an optimal range of historical information for a negative sample queue .", "forward": false, "src_ids": "2022.findings-acl.248_3750"}
+{"input": "output translation is done by using Method| context: simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications . simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed . however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario . indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed .", "entity": "output translation", "output": "re - segmentation approach", "neg_sample": ["output translation is done by using Method", "simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications .", "simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed .", "however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario .", "indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed ."], "relation": "used for", "id": "2022.acl-long.480", "year": 2022, "rel_sent": "This work proposes a stream - level adaptation of the current latency measures based on a re - segmentation approach applied to the output translation , that is successfully evaluated on streaming conditions for a reference IWSLT task", "forward": false, "src_ids": "2022.acl-long.480_3751"}
+{"input": "re - segmentation approach is used for Task| context: simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications . simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed . however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario . indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed .", "entity": "re - segmentation approach", "output": "output translation", "neg_sample": ["re - segmentation approach is used for Task", "simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications .", "simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed .", "however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario .", "indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed ."], "relation": "used for", "id": "2022.acl-long.480", "year": 2022, "rel_sent": "This work proposes a stream - level adaptation of the current latency measures based on a re - segmentation approach applied to the output translation , that is successfully evaluated on streaming conditions for a reference IWSLT task", "forward": true, "src_ids": "2022.acl-long.480_3752"}
+{"input": "iwslt task is done by using OtherScientificTerm| context: simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications . simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed . however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario . indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed .", "entity": "iwslt task", "output": "streaming conditions", "neg_sample": ["iwslt task is done by using OtherScientificTerm", "simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications .", "simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed .", "however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario .", "indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed ."], "relation": "used for", "id": "2022.acl-long.480", "year": 2022, "rel_sent": "This work proposes a stream - level adaptation of the current latency measures based on a re - segmentation approach applied to the output translation , that is successfully evaluated on streaming conditions for a reference IWSLT task", "forward": false, "src_ids": "2022.acl-long.480_3753"}
+{"input": "streaming conditions is used for Task| context: simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications . simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed . however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario . indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed .", "entity": "streaming conditions", "output": "iwslt task", "neg_sample": ["streaming conditions is used for Task", "simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications .", "simultaneous translation systems need tofind a trade - off between translation quality and response time , and with this purpose multiple latency measures have been proposed .", "however , latency evaluations for simultaneous translation are estimated at the sentence level , not taking into account the sequential nature of a streaming scenario .", "indeed , these sentence - level latency measures are not well suited for continuous stream translation , resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed ."], "relation": "used for", "id": "2022.acl-long.480", "year": 2022, "rel_sent": "This work proposes a stream - level adaptation of the current latency measures based on a re - segmentation approach applied to the output translation , that is successfully evaluated on streaming conditions for a reference IWSLT task", "forward": true, "src_ids": "2022.acl-long.480_3754"}
+{"input": "ai - guided misinformation detection and mitigation is done by using Task| context: even to a simple and short news headline , readers react in a multitude of ways : cognitively ( e.g. inferring the writer 's intent ) , emotionally ( e.g. feeling distrust ) , and behaviorally ( e.g. sharing the news with their friends ) .", "entity": "ai - guided misinformation detection and mitigation", "output": "pragmatic inferences", "neg_sample": ["ai - guided misinformation detection and mitigation is done by using Task", "even to a simple and short news headline , readers react in a multitude of ways : cognitively ( e.g.", "inferring the writer 's intent ) , emotionally ( e.g.", "feeling distrust ) , and behaviorally ( e.g.", "sharing the news with their friends ) ."], "relation": "used for", "id": "2022.acl-long.222", "year": 2022, "rel_sent": "Our work demonstrates the feasibility and importance of pragmatic inferences on news headlines to help enhance AI - guided misinformation detection and mitigation .", "forward": false, "src_ids": "2022.acl-long.222_3755"}
+{"input": "pragmatic inferences is used for Task| context: even to a simple and short news headline , readers react in a multitude of ways : cognitively ( e.g. inferring the writer 's intent ) , emotionally ( e.g. feeling distrust ) , and behaviorally ( e.g. sharing the news with their friends ) .", "entity": "pragmatic inferences", "output": "ai - guided misinformation detection and mitigation", "neg_sample": ["pragmatic inferences is used for Task", "even to a simple and short news headline , readers react in a multitude of ways : cognitively ( e.g.", "inferring the writer 's intent ) , emotionally ( e.g.", "feeling distrust ) , and behaviorally ( e.g.", "sharing the news with their friends ) ."], "relation": "used for", "id": "2022.acl-long.222", "year": 2022, "rel_sent": "Our work demonstrates the feasibility and importance of pragmatic inferences on news headlines to help enhance AI - guided misinformation detection and mitigation .", "forward": true, "src_ids": "2022.acl-long.222_3756"}
+{"input": "classification is done by using Method| context: however , these models are still quite behind the sota kgc models in terms of performance . in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting .", "entity": "classification", "output": "pre - trained language models", "neg_sample": ["classification is done by using Method", "however , these models are still quite behind the sota kgc models in terms of performance .", "in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting ."], "relation": "used for", "id": "2022.findings-acl.282", "year": 2022, "rel_sent": "The basic idea is to convert each triple and its support information into natural prompt sentences , which is further fed into PLMs for classification .", "forward": false, "src_ids": "2022.findings-acl.282_3757"}
+{"input": "natural prompt sentences is used for Method| context: however , these models are still quite behind the sota kgc models in terms of performance . in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting .", "entity": "natural prompt sentences", "output": "pre - trained language models", "neg_sample": ["natural prompt sentences is used for Method", "however , these models are still quite behind the sota kgc models in terms of performance .", "in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting ."], "relation": "used for", "id": "2022.findings-acl.282", "year": 2022, "rel_sent": "The basic idea is to convert each triple and its support information into natural prompt sentences , which is further fed into PLMs for classification .", "forward": true, "src_ids": "2022.findings-acl.282_3758"}
+{"input": "pre - trained language models is done by using OtherScientificTerm| context: in recent years , pre - trained language models ( plms ) have been shown to capture factual knowledge from massive texts , which encourages the proposal of plm - based knowledge graph completion ( kgc ) models . however , these models are still quite behind the sota kgc models in terms of performance . in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting . the evaluation setting under the closed - world assumption ( cwa ) may underestimate the plm - based kgc models since they introduce more external knowledge ; ( 2 ) inappropriate utilization of plms . most plm - based kgc models simply splice the labels of entities and relations as inputs , leading to incoherent sentences that do not take full advantage of the implicit knowledge in plms .", "entity": "pre - trained language models", "output": "natural prompt sentences", "neg_sample": ["pre - trained language models is done by using OtherScientificTerm", "in recent years , pre - trained language models ( plms ) have been shown to capture factual knowledge from massive texts , which encourages the proposal of plm - based knowledge graph completion ( kgc ) models .", "however , these models are still quite behind the sota kgc models in terms of performance .", "in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting .", "the evaluation setting under the closed - world assumption ( cwa ) may underestimate the plm - based kgc models since they introduce more external knowledge ; ( 2 ) inappropriate utilization of plms .", "most plm - based kgc models simply splice the labels of entities and relations as inputs , leading to incoherent sentences that do not take full advantage of the implicit knowledge in plms ."], "relation": "used for", "id": "2022.findings-acl.282", "year": 2022, "rel_sent": "The basic idea is to convert each triple and its support information into natural prompt sentences , which is further fed into PLMs for classification .", "forward": false, "src_ids": "2022.findings-acl.282_3759"}
+{"input": "pre - trained language models is used for Task| context: in recent years , pre - trained language models ( plms ) have been shown to capture factual knowledge from massive texts , which encourages the proposal of plm - based knowledge graph completion ( kgc ) models . however , these models are still quite behind the sota kgc models in terms of performance . in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting . the evaluation setting under the closed - world assumption ( cwa ) may underestimate the plm - based kgc models since they introduce more external knowledge ; ( 2 ) inappropriate utilization of plms . most plm - based kgc models simply splice the labels of entities and relations as inputs , leading to incoherent sentences that do not take full advantage of the implicit knowledge in plms .", "entity": "pre - trained language models", "output": "classification", "neg_sample": ["pre - trained language models is used for Task", "in recent years , pre - trained language models ( plms ) have been shown to capture factual knowledge from massive texts , which encourages the proposal of plm - based knowledge graph completion ( kgc ) models .", "however , these models are still quite behind the sota kgc models in terms of performance .", "in this work , we find two main reasons for the weak performance : ( 1 ) inaccurate evaluation setting .", "the evaluation setting under the closed - world assumption ( cwa ) may underestimate the plm - based kgc models since they introduce more external knowledge ; ( 2 ) inappropriate utilization of plms .", "most plm - based kgc models simply splice the labels of entities and relations as inputs , leading to incoherent sentences that do not take full advantage of the implicit knowledge in plms ."], "relation": "used for", "id": "2022.findings-acl.282", "year": 2022, "rel_sent": "The basic idea is to convert each triple and its support information into natural prompt sentences , which is further fed into PLMs for classification .", "forward": true, "src_ids": "2022.findings-acl.282_3760"}
+{"input": "goal - oriented document - grounded dialogue is done by using Method| context: existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation . however , such pipeline methods would unavoidably suffer from the error propagation issue .", "entity": "goal - oriented document - grounded dialogue", "output": "unified generative framework", "neg_sample": ["goal - oriented document - grounded dialogue is done by using Method", "existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation .", "however , such pipeline methods would unavoidably suffer from the error propagation issue ."], "relation": "used for", "id": "2022.acl-short.66", "year": 2022, "rel_sent": "UniGDD : A Unified Generative Framework for Goal - Oriented Document - Grounded Dialogue.", "forward": false, "src_ids": "2022.acl-short.66_3761"}
+{"input": "unified generative framework is used for Task| context: existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation . however , such pipeline methods would unavoidably suffer from the error propagation issue .", "entity": "unified generative framework", "output": "goal - oriented document - grounded dialogue", "neg_sample": ["unified generative framework is used for Task", "existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation .", "however , such pipeline methods would unavoidably suffer from the error propagation issue ."], "relation": "used for", "id": "2022.acl-short.66", "year": 2022, "rel_sent": "UniGDD : A Unified Generative Framework for Goal - Oriented Document - Grounded Dialogue.", "forward": true, "src_ids": "2022.acl-short.66_3762"}
+{"input": "irrelevant document information is done by using Method| context: existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation . however , such pipeline methods would unavoidably suffer from the error propagation issue .", "entity": "irrelevant document information", "output": "linear temperature scheduling", "neg_sample": ["irrelevant document information is done by using Method", "existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation .", "however , such pipeline methods would unavoidably suffer from the error propagation issue ."], "relation": "used for", "id": "2022.acl-short.66", "year": 2022, "rel_sent": "We further develop a prompt - connected multi - task learning strategy to model the characteristics and connections of different tasks and introduce linear temperature scheduling to reduce the negative effect of irrelevant document information .", "forward": false, "src_ids": "2022.acl-short.66_3763"}
+{"input": "linear temperature scheduling is used for OtherScientificTerm| context: existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation . however , such pipeline methods would unavoidably suffer from the error propagation issue .", "entity": "linear temperature scheduling", "output": "irrelevant document information", "neg_sample": ["linear temperature scheduling is used for OtherScientificTerm", "existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation .", "however , such pipeline methods would unavoidably suffer from the error propagation issue ."], "relation": "used for", "id": "2022.acl-short.66", "year": 2022, "rel_sent": "We further develop a prompt - connected multi - task learning strategy to model the characteristics and connections of different tasks and introduce linear temperature scheduling to reduce the negative effect of irrelevant document information .", "forward": true, "src_ids": "2022.acl-short.66_3764"}
+{"input": "dependency parsing is used for Method| context: compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years . to test compositional generalization in semantic parsing , keysers et al . this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases . dependency parsing , however , lacks a compositional generalization benchmark .", "entity": "dependency parsing", "output": "compositional freebase queries ( cfq )", "neg_sample": ["dependency parsing is used for Method", "compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years .", "to test compositional generalization in semantic parsing , keysers et al .", "this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases .", "dependency parsing , however , lacks a compositional generalization benchmark ."], "relation": "used for", "id": "2022.acl-long.448", "year": 2022, "rel_sent": "In this work , we introduce a gold - standard set of dependency parses for CFQ , and use this to analyze the behaviour of a state - of - the art dependency parser ( Qi et al . , 2020 ) on the CFQ dataset .", "forward": true, "src_ids": "2022.acl-long.448_3765"}
+{"input": "compositional freebase queries ( cfq ) is done by using Task| context: compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years . to test compositional generalization in semantic parsing , keysers et al . ( 2020 ) introduced compositional freebase queries ( cfq ) . this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases .", "entity": "compositional freebase queries ( cfq )", "output": "dependency parsing", "neg_sample": ["compositional freebase queries ( cfq ) is done by using Task", "compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years .", "to test compositional generalization in semantic parsing , keysers et al .", "( 2020 ) introduced compositional freebase queries ( cfq ) .", "this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases ."], "relation": "used for", "id": "2022.acl-long.448", "year": 2022, "rel_sent": "In this work , we introduce a gold - standard set of dependency parses for CFQ , and use this to analyze the behaviour of a state - of - the art dependency parser ( Qi et al . , 2020 ) on the CFQ dataset .", "forward": false, "src_ids": "2022.acl-long.448_3766"}
+{"input": "syntactic structures is used for Task| context: compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years . to test compositional generalization in semantic parsing , keysers et al . ( 2020 ) introduced compositional freebase queries ( cfq ) . this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases .", "entity": "syntactic structures", "output": "dependency parsing", "neg_sample": ["syntactic structures is used for Task", "compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years .", "to test compositional generalization in semantic parsing , keysers et al .", "( 2020 ) introduced compositional freebase queries ( cfq ) .", "this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases ."], "relation": "used for", "id": "2022.acl-long.448", "year": 2022, "rel_sent": "We explore a number of hypotheses for what causes the non - uniform degradation in dependency parsing performance , and identify a number of syntactic structures that drive the dependency parser 's lower performance on the most challenging splits .", "forward": true, "src_ids": "2022.acl-long.448_3767"}
+{"input": "dependency parsing is done by using OtherScientificTerm| context: compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years . to test compositional generalization in semantic parsing , keysers et al . ( 2020 ) introduced compositional freebase queries ( cfq ) . this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases . dependency parsing , however , lacks a compositional generalization benchmark .", "entity": "dependency parsing", "output": "syntactic structures", "neg_sample": ["dependency parsing is done by using OtherScientificTerm", "compositionality- the ability to combine familiar units like words into novel phrases and sentences- has been the focus of intense interest in artificial intelligence in recent years .", "to test compositional generalization in semantic parsing , keysers et al .", "( 2020 ) introduced compositional freebase queries ( cfq ) .", "this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases .", "dependency parsing , however , lacks a compositional generalization benchmark ."], "relation": "used for", "id": "2022.acl-long.448", "year": 2022, "rel_sent": "We explore a number of hypotheses for what causes the non - uniform degradation in dependency parsing performance , and identify a number of syntactic structures that drive the dependency parser 's lower performance on the most challenging splits .", "forward": false, "src_ids": "2022.acl-long.448_3768"}
+{"input": "language models is done by using Material| context: models pre - trained with a language modeling objective possess ample world knowledge and language skills , but are known to struggle in tasks that require reasoning .", "entity": "language models", "output": "semi - structured tables", "neg_sample": ["language models is done by using Material", "models pre - trained with a language modeling objective possess ample world knowledge and language skills , but are known to struggle in tasks that require reasoning ."], "relation": "used for", "id": "2022.acl-long.416", "year": 2022, "rel_sent": "Turning Tables : Generating Examples from Semi - structured Tables for Endowing Language Models with Reasoning Skills.", "forward": false, "src_ids": "2022.acl-long.416_3769"}
+{"input": "semi - structured tables is used for Method| context: models pre - trained with a language modeling objective possess ample world knowledge and language skills , but are known to struggle in tasks that require reasoning .", "entity": "semi - structured tables", "output": "language models", "neg_sample": ["semi - structured tables is used for Method", "models pre - trained with a language modeling objective possess ample world knowledge and language skills , but are known to struggle in tasks that require reasoning ."], "relation": "used for", "id": "2022.acl-long.416", "year": 2022, "rel_sent": "Turning Tables : Generating Examples from Semi - structured Tables for Endowing Language Models with Reasoning Skills.", "forward": true, "src_ids": "2022.acl-long.416_3770"}
+{"input": "interactive theorem prover is used for Method| context: a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al . , 2011 ; chomsky , 2013 ) .", "entity": "interactive theorem prover", "output": "minimalist grammar ( mg )", "neg_sample": ["interactive theorem prover is used for Method", "a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al .", ", 2011 ; chomsky , 2013 ) ."], "relation": "used for", "id": "2022.scil-1.19", "year": 2022, "rel_sent": "This study addresses this question by introducing a novel procedure , implemented as a working computer program , that uses an interactive theorem prover to incrementally infer a Minimalist Grammar ( MG ) ( Stabler , 1996 ) .", "forward": true, "src_ids": "2022.scil-1.19_3771"}
+{"input": "thematic roles is done by using OtherScientificTerm| context: a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al . , 2011 ; chomsky , 2013 ) .", "entity": "thematic roles", "output": "logical form ( lf )", "neg_sample": ["thematic roles is done by using OtherScientificTerm", "a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al .", ", 2011 ; chomsky , 2013 ) ."], "relation": "used for", "id": "2022.scil-1.19", "year": 2022, "rel_sent": "each entry is a Phonological Form ( PF ) , encoding a sentence , paired with a Logical Form ( LF ) , which encodes thematic roles for each predicate as well as agreement relations ; the input sequence , which corresponds to the PLD that a child is exposed to , is organized into a sequence of batches that the procedure consumes incrementally .", "forward": false, "src_ids": "2022.scil-1.19_3772"}
+{"input": "logical form ( lf ) is used for OtherScientificTerm| context: a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al . , 2011 ; chomsky , 2013 ) .", "entity": "logical form ( lf )", "output": "thematic roles", "neg_sample": ["logical form ( lf ) is used for OtherScientificTerm", "a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al .", ", 2011 ; chomsky , 2013 ) ."], "relation": "used for", "id": "2022.scil-1.19", "year": 2022, "rel_sent": "each entry is a Phonological Form ( PF ) , encoding a sentence , paired with a Logical Form ( LF ) , which encodes thematic roles for each predicate as well as agreement relations ; the input sequence , which corresponds to the PLD that a child is exposed to , is organized into a sequence of batches that the procedure consumes incrementally .", "forward": true, "src_ids": "2022.scil-1.19_3773"}
+{"input": "minimalist grammar ( mg ) is done by using Method| context: a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al . , 2011 ; chomsky , 2013 ) .", "entity": "minimalist grammar ( mg )", "output": "interactive theorem prover", "neg_sample": ["minimalist grammar ( mg ) is done by using Method", "a central question in cognitive linguistics is how children everywhere can readily acquire knowledge of language ( kol ) from ( restrictive ) primary linguistic data ( pld ) ( chomsky , 1986 ; berwick et al .", ", 2011 ; chomsky , 2013 ) ."], "relation": "used for", "id": "2022.scil-1.19", "year": 2022, "rel_sent": "This study addresses this question by introducing a novel procedure , implemented as a working computer program , that uses an interactive theorem prover to incrementally infer a Minimalist Grammar ( MG ) ( Stabler , 1996 ) .", "forward": false, "src_ids": "2022.scil-1.19_3774"}
+{"input": "hate speech detection is done by using Method| context: behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data . while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development .", "entity": "hate speech detection", "output": "behaviour - aware learning", "neg_sample": ["hate speech detection is done by using Method", "behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data .", "while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development ."], "relation": "used for", "id": "2022.nlppower-1.8", "year": 2022, "rel_sent": "Checking HateCheck : a cross - functional analysis of behaviour - aware learning for hate speech detection.", "forward": false, "src_ids": "2022.nlppower-1.8_3775"}
+{"input": "behaviour - aware learning is used for Task| context: behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data . while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development .", "entity": "behaviour - aware learning", "output": "hate speech detection", "neg_sample": ["behaviour - aware learning is used for Task", "behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data .", "while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development ."], "relation": "used for", "id": "2022.nlppower-1.8", "year": 2022, "rel_sent": "Checking HateCheck : a cross - functional analysis of behaviour - aware learning for hate speech detection.", "forward": true, "src_ids": "2022.nlppower-1.8_3776"}
+{"input": "hate speech detection systems is done by using OtherScientificTerm| context: behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data . while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development .", "entity": "hate speech detection systems", "output": "functional tests", "neg_sample": ["hate speech detection systems is done by using OtherScientificTerm", "behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data .", "while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development ."], "relation": "used for", "id": "2022.nlppower-1.8", "year": 2022, "rel_sent": "With this in mind , we explore behaviour - aware learning by examining several fine - tuning schemes using HateCheck , a suite of functional tests for hate speech detection systems .", "forward": false, "src_ids": "2022.nlppower-1.8_3777"}
+{"input": "functional tests is used for Method| context: behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data . while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development .", "entity": "functional tests", "output": "hate speech detection systems", "neg_sample": ["functional tests is used for Method", "behavioural testing - verifying system capabilities by validating human - designed input - output pairs - is an alternative evaluation method of natural language processing systems proposed to address the shortcomings of the standard approach : computing metrics on held - out data .", "while behavioural tests capture human prior knowledge and insights , there has been little exploration on how to leverage them for model training and development ."], "relation": "used for", "id": "2022.nlppower-1.8", "year": 2022, "rel_sent": "With this in mind , we explore behaviour - aware learning by examining several fine - tuning schemes using HateCheck , a suite of functional tests for hate speech detection systems .", "forward": true, "src_ids": "2022.nlppower-1.8_3778"}
+{"input": "wizard - of - oz dialogues is done by using Method| context: previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set . approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations .", "entity": "wizard - of - oz dialogues", "output": "few - shot semantic parser", "neg_sample": ["wizard - of - oz dialogues is done by using Method", "previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set .", "approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations ."], "relation": "used for", "id": "2022.findings-acl.317", "year": 2022, "rel_sent": "A Few - Shot Semantic Parser for Wizard - of - Oz Dialogues with the Precise ThingTalk Representation.", "forward": false, "src_ids": "2022.findings-acl.317_3779"}
+{"input": "few - shot semantic parser is used for Material| context: previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set . approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations .", "entity": "few - shot semantic parser", "output": "wizard - of - oz dialogues", "neg_sample": ["few - shot semantic parser is used for Material", "previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set .", "approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations ."], "relation": "used for", "id": "2022.findings-acl.317", "year": 2022, "rel_sent": "A Few - Shot Semantic Parser for Wizard - of - Oz Dialogues with the Precise ThingTalk Representation.", "forward": true, "src_ids": "2022.findings-acl.317_3780"}
+{"input": "precise dialogue states is done by using Method| context: previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set . approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations .", "entity": "precise dialogue states", "output": "sample - efficient methodology", "neg_sample": ["precise dialogue states is done by using Method", "previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set .", "approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations ."], "relation": "used for", "id": "2022.findings-acl.317", "year": 2022, "rel_sent": "Furthermore , previously proposed dialogue state representations are ambiguous and lack the precision necessary for building an effective agent . This paper proposes a new dialogue representation and a sample - efficient methodology that can predict precise dialogue states in WOZ conversations .", "forward": false, "src_ids": "2022.findings-acl.317_3781"}
+{"input": "sample - efficient methodology is used for OtherScientificTerm| context: previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set . approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations .", "entity": "sample - efficient methodology", "output": "precise dialogue states", "neg_sample": ["sample - efficient methodology is used for OtherScientificTerm", "previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set .", "approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations ."], "relation": "used for", "id": "2022.findings-acl.317", "year": 2022, "rel_sent": "Furthermore , previously proposed dialogue state representations are ambiguous and lack the precision necessary for building an effective agent . This paper proposes a new dialogue representation and a sample - efficient methodology that can predict precise dialogue states in WOZ conversations .", "forward": true, "src_ids": "2022.findings-acl.317_3782"}
+{"input": "log - bilinear model is used for Task| context: we investigate how to use pretrained static word embeddings to deliver improved estimates of bilexical co - occurrence probabilities : conditional probabilities of one word given a single other word in a specific relationship . such probabilities play important roles in psycholinguistics , corpus linguistics , and usage - based cognitive modeling of language more generally .", "entity": "log - bilinear model", "output": "generalization", "neg_sample": ["log - bilinear model is used for Task", "we investigate how to use pretrained static word embeddings to deliver improved estimates of bilexical co - occurrence probabilities : conditional probabilities of one word given a single other word in a specific relationship .", "such probabilities play important roles in psycholinguistics , corpus linguistics , and usage - based cognitive modeling of language more generally ."], "relation": "used for", "id": "2022.cmcl-1.6", "year": 2022, "rel_sent": "We propose a log - bilinear model taking pretrained vector representations of the two words as input , enabling generalization based on the distributional information contained in both vectors .", "forward": true, "src_ids": "2022.cmcl-1.6_3783"}
+{"input": "generalization is done by using Method| context: we investigate how to use pretrained static word embeddings to deliver improved estimates of bilexical co - occurrence probabilities : conditional probabilities of one word given a single other word in a specific relationship . such probabilities play important roles in psycholinguistics , corpus linguistics , and usage - based cognitive modeling of language more generally .", "entity": "generalization", "output": "log - bilinear model", "neg_sample": ["generalization is done by using Method", "we investigate how to use pretrained static word embeddings to deliver improved estimates of bilexical co - occurrence probabilities : conditional probabilities of one word given a single other word in a specific relationship .", "such probabilities play important roles in psycholinguistics , corpus linguistics , and usage - based cognitive modeling of language more generally ."], "relation": "used for", "id": "2022.cmcl-1.6", "year": 2022, "rel_sent": "We propose a log - bilinear model taking pretrained vector representations of the two words as input , enabling generalization based on the distributional information contained in both vectors .", "forward": false, "src_ids": "2022.cmcl-1.6_3784"}
+{"input": "entropy estimators is used for Task| context: shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language . however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution . while entropy estimation is a well - studied problem in other fields , there is not yet a comprehensive exploration of the efficacy of entropy estimators for use with linguistic data .", "entity": "entropy estimators", "output": "linguistic studies", "neg_sample": ["entropy estimators is used for Task", "shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language .", "however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution .", "while entropy estimation is a well - studied problem in other fields , there is not yet a comprehensive exploration of the efficacy of entropy estimators for use with linguistic data ."], "relation": "used for", "id": "2022.acl-short.20", "year": 2022, "rel_sent": "We end this paper with a concrete recommendation for the entropy estimators that should be used in future linguistic studies .", "forward": true, "src_ids": "2022.acl-short.20_3785"}
+{"input": "linguistic distributions is done by using Method| context: shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language . however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution .", "entity": "linguistic distributions", "output": "entropy estimators", "neg_sample": ["linguistic distributions is done by using Method", "shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language .", "however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution ."], "relation": "used for", "id": "2022.acl-short.20", "year": 2022, "rel_sent": "In this work , we fill this void , studying the empirical effectiveness of different entropy estimators for linguistic distributions .", "forward": false, "src_ids": "2022.acl-short.20_3786"}
+{"input": "linguistic studies is done by using Method| context: shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language . however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution .", "entity": "linguistic studies", "output": "entropy estimators", "neg_sample": ["linguistic studies is done by using Method", "shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language .", "however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution ."], "relation": "used for", "id": "2022.acl-short.20", "year": 2022, "rel_sent": "We end this paper with a concrete recommendation for the entropy estimators that should be used in future linguistic studies .", "forward": false, "src_ids": "2022.acl-short.20_3787"}
+{"input": "entropy estimators is used for OtherScientificTerm| context: shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language . however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution . while entropy estimation is a well - studied problem in other fields , there is not yet a comprehensive exploration of the efficacy of entropy estimators for use with linguistic data .", "entity": "entropy estimators", "output": "linguistic distributions", "neg_sample": ["entropy estimators is used for OtherScientificTerm", "shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language .", "however , entropymust typically be estimated from observed data because researchers do not have access to the underlying probability distribution .", "while entropy estimation is a well - studied problem in other fields , there is not yet a comprehensive exploration of the efficacy of entropy estimators for use with linguistic data ."], "relation": "used for", "id": "2022.acl-short.20", "year": 2022, "rel_sent": "In this work , we fill this void , studying the empirical effectiveness of different entropy estimators for linguistic distributions .", "forward": true, "src_ids": "2022.acl-short.20_3788"}
+{"input": "logic reasoning is done by using Method| context: multi - hop reading comprehension requires an ability to reason across multiple documents . on the one hand , deep learning approaches only implicitly encode query - related information into distributed embeddings which fail to uncover the discrete relational reasoning process to infer the correct answer . on the other hand , logic - based approaches provide interpretable rules to infer the target answer , but mostly work on structured data where entities and relations are well - defined .", "entity": "logic reasoning", "output": "deep - learning based inductive logic reasoning method", "neg_sample": ["logic reasoning is done by using Method", "multi - hop reading comprehension requires an ability to reason across multiple documents .", "on the one hand , deep learning approaches only implicitly encode query - related information into distributed embeddings which fail to uncover the discrete relational reasoning process to infer the correct answer .", "on the other hand , logic - based approaches provide interpretable rules to infer the target answer , but mostly work on structured data where entities and relations are well - defined ."], "relation": "used for", "id": "2022.acl-long.343", "year": 2022, "rel_sent": "In this paper , we propose a deep - learning based inductive logic reasoning method that firstly extracts query - related ( candidate - related ) information , and then conducts logic reasoning among the filtered information by inducing feasible rules that entail the target relation .", "forward": false, "src_ids": "2022.acl-long.343_3789"}
+{"input": "deep - learning based inductive logic reasoning method is used for Method| context: multi - hop reading comprehension requires an ability to reason across multiple documents . on the one hand , deep learning approaches only implicitly encode query - related information into distributed embeddings which fail to uncover the discrete relational reasoning process to infer the correct answer . on the other hand , logic - based approaches provide interpretable rules to infer the target answer , but mostly work on structured data where entities and relations are well - defined .", "entity": "deep - learning based inductive logic reasoning method", "output": "logic reasoning", "neg_sample": ["deep - learning based inductive logic reasoning method is used for Method", "multi - hop reading comprehension requires an ability to reason across multiple documents .", "on the one hand , deep learning approaches only implicitly encode query - related information into distributed embeddings which fail to uncover the discrete relational reasoning process to infer the correct answer .", "on the other hand , logic - based approaches provide interpretable rules to infer the target answer , but mostly work on structured data where entities and relations are well - defined ."], "relation": "used for", "id": "2022.acl-long.343", "year": 2022, "rel_sent": "In this paper , we propose a deep - learning based inductive logic reasoning method that firstly extracts query - related ( candidate - related ) information , and then conducts logic reasoning among the filtered information by inducing feasible rules that entail the target relation .", "forward": true, "src_ids": "2022.acl-long.343_3790"}
+{"input": "force dynamics theory is used for OtherScientificTerm| context: humans are able to perceive , understand and reason about causal events . developing models with similar physical and causal understanding capabilities is a long - standing goal of artificial intelligence .", "entity": "force dynamics theory", "output": "causal question category", "neg_sample": ["force dynamics theory is used for OtherScientificTerm", "humans are able to perceive , understand and reason about causal events .", "developing models with similar physical and causal understanding capabilities is a long - standing goal of artificial intelligence ."], "relation": "used for", "id": "2022.findings-acl.205", "year": 2022, "rel_sent": "Additionally , inspired by the Force Dynamics Theory in cognitive linguistics , we introduce a new causal question category that involves understanding the causal interactions between objects through notions like cause , enable , and prevent .", "forward": true, "src_ids": "2022.findings-acl.205_3791"}
+{"input": "causal question category is done by using Method| context: humans are able to perceive , understand and reason about causal events . developing models with similar physical and causal understanding capabilities is a long - standing goal of artificial intelligence .", "entity": "causal question category", "output": "force dynamics theory", "neg_sample": ["causal question category is done by using Method", "humans are able to perceive , understand and reason about causal events .", "developing models with similar physical and causal understanding capabilities is a long - standing goal of artificial intelligence ."], "relation": "used for", "id": "2022.findings-acl.205", "year": 2022, "rel_sent": "Additionally , inspired by the Force Dynamics Theory in cognitive linguistics , we introduce a new causal question category that involves understanding the causal interactions between objects through notions like cause , enable , and prevent .", "forward": false, "src_ids": "2022.findings-acl.205_3792"}
+{"input": "semantics is done by using Method| context: humans create internal mental models of their observations that greatly aid in their ability to understand and reason about a large variety of problems .", "entity": "semantics", "output": "latent world model", "neg_sample": ["semantics is done by using Method", "humans create internal mental models of their observations that greatly aid in their ability to understand and reason about a large variety of problems ."], "relation": "used for", "id": "2022.tacl-1.19", "year": 2022, "rel_sent": "We derive and implement an inference algorithm that reads sentences by parsing and abducing updates to its latent world model that capture the semantics of those sentences , and evaluate it on two out - of - domain question - answering datasets : ( 1 ) ProofWriter and ( 2 ) a new dataset we call FictionalGeoQA , designed to be more representative of real language but still simple enough tofocus on evaluating reasoning ability , while being robust against heuristics .", "forward": false, "src_ids": "2022.tacl-1.19_3793"}
+{"input": "latent world model is used for OtherScientificTerm| context: humans create internal mental models of their observations that greatly aid in their ability to understand and reason about a large variety of problems .", "entity": "latent world model", "output": "semantics", "neg_sample": ["latent world model is used for OtherScientificTerm", "humans create internal mental models of their observations that greatly aid in their ability to understand and reason about a large variety of problems ."], "relation": "used for", "id": "2022.tacl-1.19", "year": 2022, "rel_sent": "We derive and implement an inference algorithm that reads sentences by parsing and abducing updates to its latent world model that capture the semantics of those sentences , and evaluate it on two out - of - domain question - answering datasets : ( 1 ) ProofWriter and ( 2 ) a new dataset we call FictionalGeoQA , designed to be more representative of real language but still simple enough tofocus on evaluating reasoning ability , while being robust against heuristics .", "forward": true, "src_ids": "2022.tacl-1.19_3794"}
+{"input": "less- represented languages is done by using OtherScientificTerm| context: embeddia project developed a range of resources and methods for less - resourced eu languages , focusing on applications for media industry , including keyword extraction , comment moderation and article generation .", "entity": "less- represented languages", "output": "cross - lingual embeddings", "neg_sample": ["less- represented languages is done by using OtherScientificTerm", "embeddia project developed a range of resources and methods for less - resourced eu languages , focusing on applications for media industry , including keyword extraction , comment moderation and article generation ."], "relation": "used for", "id": "2022.eamt-1.36", "year": 2022, "rel_sent": "EMBEDDIA project : Cross - Lingual Embeddings for Less- Represented Languages in European News Media.", "forward": false, "src_ids": "2022.eamt-1.36_3795"}
+{"input": "cross - lingual embeddings is used for OtherScientificTerm| context: embeddia project developed a range of resources and methods for less - resourced eu languages , focusing on applications for media industry , including keyword extraction , comment moderation and article generation .", "entity": "cross - lingual embeddings", "output": "less- represented languages", "neg_sample": ["cross - lingual embeddings is used for OtherScientificTerm", "embeddia project developed a range of resources and methods for less - resourced eu languages , focusing on applications for media industry , including keyword extraction , comment moderation and article generation ."], "relation": "used for", "id": "2022.eamt-1.36", "year": 2022, "rel_sent": "EMBEDDIA project : Cross - Lingual Embeddings for Less- Represented Languages in European News Media.", "forward": true, "src_ids": "2022.eamt-1.36_3796"}
+{"input": "continual ner is done by using Method| context: traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types . however , in many real - world scenarios , new entity types are incrementally involved . to investigate this problem , continual learning is introduced for ner .", "entity": "continual ner", "output": "learn - and - review ( l&r )", "neg_sample": ["continual ner is done by using Method", "traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types .", "however , in many real - world scenarios , new entity types are incrementally involved .", "to investigate this problem , continual learning is introduced for ner ."], "relation": "used for", "id": "2022.findings-acl.179", "year": 2022, "rel_sent": "However , the existing method depends on the relevance between tasks and is prone to inter - type confusion . In this paper , we propose a novel two - stage framework Learn - and - Review ( L&R ) for continual NER under the type - incremental setting to alleviate the above issues . Specifically , for the learning stage , we distill the old knowledge from teacher to a student on the current dataset .", "forward": false, "src_ids": "2022.findings-acl.179_3797"}
+{"input": "learn - and - review ( l&r ) is used for Task| context: traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types . however , in many real - world scenarios , new entity types are incrementally involved . to investigate this problem , continual learning is introduced for ner .", "entity": "learn - and - review ( l&r )", "output": "continual ner", "neg_sample": ["learn - and - review ( l&r ) is used for Task", "traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types .", "however , in many real - world scenarios , new entity types are incrementally involved .", "to investigate this problem , continual learning is introduced for ner ."], "relation": "used for", "id": "2022.findings-acl.179", "year": 2022, "rel_sent": "However , the existing method depends on the relevance between tasks and is prone to inter - type confusion . In this paper , we propose a novel two - stage framework Learn - and - Review ( L&R ) for continual NER under the type - incremental setting to alleviate the above issues . Specifically , for the learning stage , we distill the old knowledge from teacher to a student on the current dataset .", "forward": true, "src_ids": "2022.findings-acl.179_3798"}
+{"input": "distillation is done by using Material| context: traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types . however , in many real - world scenarios , new entity types are incrementally involved . to investigate this problem , continual learning is introduced for ner .", "entity": "distillation", "output": "synthetic samples", "neg_sample": ["distillation is done by using Material", "traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types .", "however , in many real - world scenarios , new entity types are incrementally involved .", "to investigate this problem , continual learning is introduced for ner ."], "relation": "used for", "id": "2022.findings-acl.179", "year": 2022, "rel_sent": "This stage has the following advantages : ( 1 ) The synthetic samples mitigate the gap between the old and new task and thus enhance the further distillation ; ( 2 ) Different types of entities are jointly seen during training which alleviates the inter - type confusion .", "forward": false, "src_ids": "2022.findings-acl.179_3799"}
+{"input": "synthetic samples is used for Method| context: traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types . however , in many real - world scenarios , new entity types are incrementally involved . to investigate this problem , continual learning is introduced for ner .", "entity": "synthetic samples", "output": "distillation", "neg_sample": ["synthetic samples is used for Method", "traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types .", "however , in many real - world scenarios , new entity types are incrementally involved .", "to investigate this problem , continual learning is introduced for ner ."], "relation": "used for", "id": "2022.findings-acl.179", "year": 2022, "rel_sent": "This stage has the following advantages : ( 1 ) The synthetic samples mitigate the gap between the old and new task and thus enhance the further distillation ; ( 2 ) Different types of entities are jointly seen during training which alleviates the inter - type confusion .", "forward": true, "src_ids": "2022.findings-acl.179_3800"}
+{"input": "icd-9 coding is done by using Method| context: medical document coding is the process of assigning labels from a structured label space ( ontology - e.g. , icd-9 ) to medical documents . this process is laborious , costly , and error - prone . in recent years , efforts have been made to automate this process with neural models . the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios .", "entity": "icd-9 coding", "output": "ontology - guided data augmentation and synthesis", "neg_sample": ["icd-9 coding is done by using Method", "medical document coding is the process of assigning labels from a structured label space ( ontology - e.g.", ", icd-9 ) to medical documents .", "this process is laborious , costly , and error - prone .", "in recent years , efforts have been made to automate this process with neural models .", "the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios ."], "relation": "used for", "id": "2022.bionlp-1.39", "year": 2022, "rel_sent": "Horses to Zebras : Ontology - Guided Data Augmentation and Synthesis for ICD-9 Coding.", "forward": false, "src_ids": "2022.bionlp-1.39_3801"}
+{"input": "ontology - guided data augmentation and synthesis is used for OtherScientificTerm| context: medical document coding is the process of assigning labels from a structured label space ( ontology - e.g. , icd-9 ) to medical documents . this process is laborious , costly , and error - prone . in recent years , efforts have been made to automate this process with neural models . the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios .", "entity": "ontology - guided data augmentation and synthesis", "output": "icd-9 coding", "neg_sample": ["ontology - guided data augmentation and synthesis is used for OtherScientificTerm", "medical document coding is the process of assigning labels from a structured label space ( ontology - e.g.", ", icd-9 ) to medical documents .", "this process is laborious , costly , and error - prone .", "in recent years , efforts have been made to automate this process with neural models .", "the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios ."], "relation": "used for", "id": "2022.bionlp-1.39", "year": 2022, "rel_sent": "Horses to Zebras : Ontology - Guided Data Augmentation and Synthesis for ICD-9 Coding.", "forward": true, "src_ids": "2022.bionlp-1.39_3802"}
+{"input": "analysis technique is used for Method| context: medical document coding is the process of assigning labels from a structured label space ( ontology - e.g. , icd-9 ) to medical documents . this process is laborious , costly , and error - prone . in recent years , efforts have been made to automate this process with neural models . the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios .", "entity": "analysis technique", "output": "data augmentation and synthesis techniques", "neg_sample": ["analysis technique is used for Method", "medical document coding is the process of assigning labels from a structured label space ( ontology - e.g.", ", icd-9 ) to medical documents .", "this process is laborious , costly , and error - prone .", "in recent years , efforts have been made to automate this process with neural models .", "the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios ."], "relation": "used for", "id": "2022.bionlp-1.39", "year": 2022, "rel_sent": "This analysis technique points to the positive impact of data augmentation and synthesis , but also highlights more general issues of confusion within families of codes , and underprediction .", "forward": true, "src_ids": "2022.bionlp-1.39_3803"}
+{"input": "data augmentation and synthesis techniques is done by using Method| context: medical document coding is the process of assigning labels from a structured label space ( ontology - e.g. , icd-9 ) to medical documents . this process is laborious , costly , and error - prone . in recent years , efforts have been made to automate this process with neural models . the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios .", "entity": "data augmentation and synthesis techniques", "output": "analysis technique", "neg_sample": ["data augmentation and synthesis techniques is done by using Method", "medical document coding is the process of assigning labels from a structured label space ( ontology - e.g.", ", icd-9 ) to medical documents .", "this process is laborious , costly , and error - prone .", "in recent years , efforts have been made to automate this process with neural models .", "the label spaces are large ( in the order of thousands of labels ) and follow a big - head long - tail label distribution , giving rise tofew - shot and zero - shot scenarios ."], "relation": "used for", "id": "2022.bionlp-1.39", "year": 2022, "rel_sent": "This analysis technique points to the positive impact of data augmentation and synthesis , but also highlights more general issues of confusion within families of codes , and underprediction .", "forward": false, "src_ids": "2022.bionlp-1.39_3804"}
+{"input": "unsupervised keyphrase extraction is done by using Method| context: keyphrase extraction ( kpe ) automatically extracts phrases in a document that provide a concise summary of the core content , which benefits downstream information retrieval and nlp tasks . previous state - of - the - art methods select candidate keyphrases based on the similarity between learned representations of the candidates and the document . they suffer performance degradation on long documents due to discrepancy between sequence lengths which causes mismatch between representations of keyphrase candidates and the document .", "entity": "unsupervised keyphrase extraction", "output": "masked document embedding rank approach", "neg_sample": ["unsupervised keyphrase extraction is done by using Method", "keyphrase extraction ( kpe ) automatically extracts phrases in a document that provide a concise summary of the core content , which benefits downstream information retrieval and nlp tasks .", "previous state - of - the - art methods select candidate keyphrases based on the similarity between learned representations of the candidates and the document .", "they suffer performance degradation on long documents due to discrepancy between sequence lengths which causes mismatch between representations of keyphrase candidates and the document ."], "relation": "used for", "id": "2022.findings-acl.34", "year": 2022, "rel_sent": "MDERank : A Masked Document Embedding Rank Approach for Unsupervised Keyphrase Extraction.", "forward": false, "src_ids": "2022.findings-acl.34_3805"}
+{"input": "masked document embedding rank approach is used for Task| context: keyphrase extraction ( kpe ) automatically extracts phrases in a document that provide a concise summary of the core content , which benefits downstream information retrieval and nlp tasks . previous state - of - the - art methods select candidate keyphrases based on the similarity between learned representations of the candidates and the document . they suffer performance degradation on long documents due to discrepancy between sequence lengths which causes mismatch between representations of keyphrase candidates and the document .", "entity": "masked document embedding rank approach", "output": "unsupervised keyphrase extraction", "neg_sample": ["masked document embedding rank approach is used for Task", "keyphrase extraction ( kpe ) automatically extracts phrases in a document that provide a concise summary of the core content , which benefits downstream information retrieval and nlp tasks .", "previous state - of - the - art methods select candidate keyphrases based on the similarity between learned representations of the candidates and the document .", "they suffer performance degradation on long documents due to discrepancy between sequence lengths which causes mismatch between representations of keyphrase candidates and the document ."], "relation": "used for", "id": "2022.findings-acl.34", "year": 2022, "rel_sent": "MDERank : A Masked Document Embedding Rank Approach for Unsupervised Keyphrase Extraction.", "forward": true, "src_ids": "2022.findings-acl.34_3806"}
+{"input": "australian aboriginal community is done by using Task| context: most low resource language technology development is premised on the need to collect data for training statistical models . when we follow the typical process of recording and transcribing text for small indigenous languages , we hit up against the so - called ' transcription bottleneck . ' therefore it is worth exploring new ways of engaging with speakers which generate data while avoiding the transcription bottleneck .", "entity": "australian aboriginal community", "output": "data capture", "neg_sample": ["australian aboriginal community is done by using Task", "most low resource language technology development is premised on the need to collect data for training statistical models .", "when we follow the typical process of recording and transcribing text for small indigenous languages , we hit up against the so - called ' transcription bottleneck . '", "therefore it is worth exploring new ways of engaging with speakers which generate data while avoiding the transcription bottleneck ."], "relation": "used for", "id": "2022.acl-long.342", "year": 2022, "rel_sent": "Learning From Failure : Data Capture in an Australian Aboriginal Community.", "forward": false, "src_ids": "2022.acl-long.342_3807"}
+{"input": "data capture is used for Material| context: most low resource language technology development is premised on the need to collect data for training statistical models . when we follow the typical process of recording and transcribing text for small indigenous languages , we hit up against the so - called ' transcription bottleneck . ' therefore it is worth exploring new ways of engaging with speakers which generate data while avoiding the transcription bottleneck .", "entity": "data capture", "output": "australian aboriginal community", "neg_sample": ["data capture is used for Material", "most low resource language technology development is premised on the need to collect data for training statistical models .", "when we follow the typical process of recording and transcribing text for small indigenous languages , we hit up against the so - called ' transcription bottleneck . '", "therefore it is worth exploring new ways of engaging with speakers which generate data while avoiding the transcription bottleneck ."], "relation": "used for", "id": "2022.acl-long.342", "year": 2022, "rel_sent": "Learning From Failure : Data Capture in an Australian Aboriginal Community.", "forward": true, "src_ids": "2022.acl-long.342_3808"}
+{"input": "fitting of token - level samples is done by using OtherScientificTerm| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "fitting of token - level samples", "output": "prediction difference", "neg_sample": ["fitting of token - level samples is done by using OtherScientificTerm", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "In this paper , we utilize prediction difference for ground - truth tokens to analyze the fitting of token - level samples and find that under - fitting is almost as common as over - fitting .", "forward": false, "src_ids": "2022.acl-long.528_3809"}
+{"input": "ground - truth tokens is done by using OtherScientificTerm| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "ground - truth tokens", "output": "prediction difference", "neg_sample": ["ground - truth tokens is done by using OtherScientificTerm", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "In this paper , we utilize prediction difference for ground - truth tokens to analyze the fitting of token - level samples and find that under - fitting is almost as common as over - fitting .", "forward": false, "src_ids": "2022.acl-long.528_3810"}
+{"input": "pd - r is used for OtherScientificTerm| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "pd - r", "output": "prediction difference", "neg_sample": ["pd - r is used for OtherScientificTerm", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "For all token - level samples , PD - R minimizes the prediction difference between the original pass and the input - perturbed pass , making the model less sensitive to small input changes , thus more robust to both perturbations and under - fitted training data .", "forward": true, "src_ids": "2022.acl-long.528_3811"}
+{"input": "prediction difference is used for OtherScientificTerm| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "prediction difference", "output": "ground - truth tokens", "neg_sample": ["prediction difference is used for OtherScientificTerm", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "In this paper , we utilize prediction difference for ground - truth tokens to analyze the fitting of token - level samples and find that under - fitting is almost as common as over - fitting .", "forward": true, "src_ids": "2022.acl-long.528_3812"}
+{"input": "prediction difference is used for Task| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "prediction difference", "output": "fitting of token - level samples", "neg_sample": ["prediction difference is used for Task", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "In this paper , we utilize prediction difference for ground - truth tokens to analyze the fitting of token - level samples and find that under - fitting is almost as common as over - fitting .", "forward": true, "src_ids": "2022.acl-long.528_3813"}
+{"input": "prediction difference regularization ( pd - r ) is used for OtherScientificTerm| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "entity": "prediction difference regularization ( pd - r )", "output": "under - fitting", "neg_sample": ["prediction difference regularization ( pd - r ) is used for OtherScientificTerm", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "We introduce prediction difference regularization ( PD - R ) , a simple and effective method that can reduce over - fitting and under - fitting at the same time .", "forward": true, "src_ids": "2022.acl-long.528_3814"}
+{"input": "prediction difference regularization ( pd - r ) is used for OtherScientificTerm| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "prediction difference regularization ( pd - r )", "output": "over - fitting", "neg_sample": ["prediction difference regularization ( pd - r ) is used for OtherScientificTerm", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "We introduce prediction difference regularization ( PD - R ) , a simple and effective method that can reduce over - fitting and under - fitting at the same time .", "forward": true, "src_ids": "2022.acl-long.528_3815"}
+{"input": "under - fitting is done by using Method| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "under - fitting", "output": "prediction difference regularization ( pd - r )", "neg_sample": ["under - fitting is done by using Method", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "We introduce prediction difference regularization ( PD - R ) , a simple and effective method that can reduce over - fitting and under - fitting at the same time .", "forward": false, "src_ids": "2022.acl-long.528_3816"}
+{"input": "over - fitting is done by using Method| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "over - fitting", "output": "prediction difference regularization ( pd - r )", "neg_sample": ["over - fitting is done by using Method", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "We introduce prediction difference regularization ( PD - R ) , a simple and effective method that can reduce over - fitting and under - fitting at the same time .", "forward": false, "src_ids": "2022.acl-long.528_3817"}
+{"input": "prediction difference is done by using Method| context: regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years . despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data .", "entity": "prediction difference", "output": "pd - r", "neg_sample": ["prediction difference is done by using Method", "regularization methods applying input perturbation have drawn considerable attention and have been frequently explored for nmt tasks in recent years .", "despite their simplicity and effectiveness , we argue that these methods are limited by the under - fitting of training data ."], "relation": "used for", "id": "2022.acl-long.528", "year": 2022, "rel_sent": "For all token - level samples , PD - R minimizes the prediction difference between the original pass and the input - perturbed pass , making the model less sensitive to small input changes , thus more robust to both perturbations and under - fitted training data .", "forward": false, "src_ids": "2022.acl-long.528_3818"}
+{"input": "question answering is done by using Method| context: recent advances in nlp and information retrieval have given rise to a diverse set of question answering tasks that are of different formats ( e.g. , extractive , abstractive ) , require different model architectures ( e.g. , generative , discriminative ) , and setups ( e.g. , with or without retrieval ) . despite having a large number of powerful , specialized qa pipelines ( which we refer to as skills ) that consider a single domain , model or setup , there exists noframework where users can easily explore and compare such pipelines and can extend them according to their needs .", "entity": "question answering", "output": "ukp - square", "neg_sample": ["question answering is done by using Method", "recent advances in nlp and information retrieval have given rise to a diverse set of question answering tasks that are of different formats ( e.g.", ", extractive , abstractive ) , require different model architectures ( e.g.", ", generative , discriminative ) , and setups ( e.g.", ", with or without retrieval ) .", "despite having a large number of powerful , specialized qa pipelines ( which we refer to as skills ) that consider a single domain , model or setup , there exists noframework where users can easily explore and compare such pipelines and can extend them according to their needs ."], "relation": "used for", "id": "2022.acl-demo.2", "year": 2022, "rel_sent": "UKP - SQUARE : An Online Platform for Question Answering Research.", "forward": false, "src_ids": "2022.acl-demo.2_3819"}
+{"input": "ukp - square is used for Task| context: despite having a large number of powerful , specialized qa pipelines ( which we refer to as skills ) that consider a single domain , model or setup , there exists noframework where users can easily explore and compare such pipelines and can extend them according to their needs .", "entity": "ukp - square", "output": "question answering", "neg_sample": ["ukp - square is used for Task", "despite having a large number of powerful , specialized qa pipelines ( which we refer to as skills ) that consider a single domain , model or setup , there exists noframework where users can easily explore and compare such pipelines and can extend them according to their needs ."], "relation": "used for", "id": "2022.acl-demo.2", "year": 2022, "rel_sent": "UKP - SQUARE : An Online Platform for Question Answering Research.", "forward": true, "src_ids": "2022.acl-demo.2_3820"}
+{"input": "hierarchical inductive bias is done by using Method| context: relations between words are governed by hierarchical structure rather than linear ordering . sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions .", "entity": "hierarchical inductive bias", "output": "pre - training", "neg_sample": ["hierarchical inductive bias is done by using Method", "relations between words are governed by hierarchical structure rather than linear ordering .", "sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions ."], "relation": "used for", "id": "2022.findings-acl.106", "year": 2022, "rel_sent": "Coloring the Blank Slate : Pre - training Imparts a Hierarchical Inductive Bias to Sequence - to - sequence Models.", "forward": false, "src_ids": "2022.findings-acl.106_3821"}
+{"input": "pre - training is used for OtherScientificTerm| context: relations between words are governed by hierarchical structure rather than linear ordering . sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions . however , syntactic evaluations of seq2seq models have only observed models that were not pre - trained on natural language data before being trained to perform syntactic transformations , in spite of the fact that pre - training has been found to induce hierarchical linguistic generalizations in language models ; in other words , the syntactic capabilities of seq2seq models may have been greatly understated .", "entity": "pre - training", "output": "hierarchical inductive bias", "neg_sample": ["pre - training is used for OtherScientificTerm", "relations between words are governed by hierarchical structure rather than linear ordering .", "sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions .", "however , syntactic evaluations of seq2seq models have only observed models that were not pre - trained on natural language data before being trained to perform syntactic transformations , in spite of the fact that pre - training has been found to induce hierarchical linguistic generalizations in language models ; in other words , the syntactic capabilities of seq2seq models may have been greatly understated ."], "relation": "used for", "id": "2022.findings-acl.106", "year": 2022, "rel_sent": "Coloring the Blank Slate : Pre - training Imparts a Hierarchical Inductive Bias to Sequence - to - sequence Models.", "forward": true, "src_ids": "2022.findings-acl.106_3822"}
+{"input": "seq2seq models is used for OtherScientificTerm| context: relations between words are governed by hierarchical structure rather than linear ordering .", "entity": "seq2seq models", "output": "syntactic transformations", "neg_sample": ["seq2seq models is used for OtherScientificTerm", "relations between words are governed by hierarchical structure rather than linear ordering ."], "relation": "used for", "id": "2022.findings-acl.106", "year": 2022, "rel_sent": "We find that pre - trained seq2seq models generalize hierarchically when performing syntactic transformations , whereas models trained from scratch on syntactic transformations do not .", "forward": true, "src_ids": "2022.findings-acl.106_3823"}
+{"input": "syntactic transformations is done by using Method| context: relations between words are governed by hierarchical structure rather than linear ordering . sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions .", "entity": "syntactic transformations", "output": "seq2seq models", "neg_sample": ["syntactic transformations is done by using Method", "relations between words are governed by hierarchical structure rather than linear ordering .", "sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions ."], "relation": "used for", "id": "2022.findings-acl.106", "year": 2022, "rel_sent": "We find that pre - trained seq2seq models generalize hierarchically when performing syntactic transformations , whereas models trained from scratch on syntactic transformations do not .", "forward": false, "src_ids": "2022.findings-acl.106_3824"}
+{"input": "syntactic generalization is done by using Method| context: relations between words are governed by hierarchical structure rather than linear ordering . sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions .", "entity": "syntactic generalization", "output": "seq2seq models", "neg_sample": ["syntactic generalization is done by using Method", "relations between words are governed by hierarchical structure rather than linear ordering .", "sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions ."], "relation": "used for", "id": "2022.findings-acl.106", "year": 2022, "rel_sent": "This result presents evidence for the learnability of hierarchical syntactic information from non - annotated natural language text while also demonstrating that seq2seq models are capable of syntactic generalization , though only after exposure to much more language data than human learners receive .", "forward": false, "src_ids": "2022.findings-acl.106_3825"}
+{"input": "seq2seq models is used for Task| context: relations between words are governed by hierarchical structure rather than linear ordering . sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions . however , syntactic evaluations of seq2seq models have only observed models that were not pre - trained on natural language data before being trained to perform syntactic transformations , in spite of the fact that pre - training has been found to induce hierarchical linguistic generalizations in language models ; in other words , the syntactic capabilities of seq2seq models may have been greatly understated .", "entity": "seq2seq models", "output": "syntactic generalization", "neg_sample": ["seq2seq models is used for Task", "relations between words are governed by hierarchical structure rather than linear ordering .", "sequence - to - sequence ( seq2seq ) models , despite their success in downstream nlp applications , often fail to generalize in a hierarchy - sensitive manner when performing syntactic transformations - for example , transforming declarative sentences into questions .", "however , syntactic evaluations of seq2seq models have only observed models that were not pre - trained on natural language data before being trained to perform syntactic transformations , in spite of the fact that pre - training has been found to induce hierarchical linguistic generalizations in language models ; in other words , the syntactic capabilities of seq2seq models may have been greatly understated ."], "relation": "used for", "id": "2022.findings-acl.106", "year": 2022, "rel_sent": "This result presents evidence for the learnability of hierarchical syntactic information from non - annotated natural language text while also demonstrating that seq2seq models are capable of syntactic generalization , though only after exposure to much more language data than human learners receive .", "forward": true, "src_ids": "2022.findings-acl.106_3826"}
+{"input": "generalization is done by using Method| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "generalization", "output": "expert guided adversarial augmentation", "neg_sample": ["generalization is done by using Method", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "Leveraging Expert Guided Adversarial Augmentation For Improving Generalization in Named Entity Recognition.", "forward": false, "src_ids": "2022.findings-acl.154_3827"}
+{"input": "named entity recognition is done by using Method| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "named entity recognition", "output": "expert guided adversarial augmentation", "neg_sample": ["named entity recognition is done by using Method", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "Leveraging Expert Guided Adversarial Augmentation For Improving Generalization in Named Entity Recognition.", "forward": false, "src_ids": "2022.findings-acl.154_3828"}
+{"input": "expert guided adversarial augmentation is used for Task| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "entity": "expert guided adversarial augmentation", "output": "generalization", "neg_sample": ["expert guided adversarial augmentation is used for Task", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "Leveraging Expert Guided Adversarial Augmentation For Improving Generalization in Named Entity Recognition.", "forward": true, "src_ids": "2022.findings-acl.154_3829"}
+{"input": "expert guided adversarial augmentation is used for Task| context: one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "expert guided adversarial augmentation", "output": "named entity recognition", "neg_sample": ["expert guided adversarial augmentation is used for Task", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "Leveraging Expert Guided Adversarial Augmentation For Improving Generalization in Named Entity Recognition.", "forward": true, "src_ids": "2022.findings-acl.154_3830"}
+{"input": "entity tokens is done by using Method| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "entity tokens", "output": "expert - guided heuristics", "neg_sample": ["entity tokens is done by using Method", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "To this end , we propose leveraging expert - guided heuristics to change the entity tokens and their surrounding contexts thereby altering their entity types as adversarial attacks .", "forward": false, "src_ids": "2022.findings-acl.154_3831"}
+{"input": "conll 2003 test set is done by using Method| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "conll 2003 test set", "output": "expert - guided heuristics", "neg_sample": ["conll 2003 test set is done by using Method", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "Using expert - guided heuristics , we augmented the CoNLL 2003 test set and manually annotated it to construct a high - quality challenging set .", "forward": false, "src_ids": "2022.findings-acl.154_3832"}
+{"input": "expert - guided heuristics is used for OtherScientificTerm| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "expert - guided heuristics", "output": "entity tokens", "neg_sample": ["expert - guided heuristics is used for OtherScientificTerm", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "To this end , we propose leveraging expert - guided heuristics to change the entity tokens and their surrounding contexts thereby altering their entity types as adversarial attacks .", "forward": true, "src_ids": "2022.findings-acl.154_3833"}
+{"input": "expert - guided heuristics is used for Material| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "expert - guided heuristics", "output": "conll 2003 test set", "neg_sample": ["expert - guided heuristics is used for Material", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "Using expert - guided heuristics , we augmented the CoNLL 2003 test set and manually annotated it to construct a high - quality challenging set .", "forward": true, "src_ids": "2022.findings-acl.154_3834"}
+{"input": "regularization is done by using Method| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "regularization", "output": "mixup", "neg_sample": ["regularization is done by using Method", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "By training on adversarial augmented training examples and using mixup for regularization , we were able to significantly improve the performance on the challenging set as well as improve out - of - domain generalization which we evaluated by using OntoNotes data .", "forward": false, "src_ids": "2022.findings-acl.154_3835"}
+{"input": "mixup is used for OtherScientificTerm| context: named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution . one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered .", "entity": "mixup", "output": "regularization", "neg_sample": ["mixup is used for OtherScientificTerm", "named entity recognition ( ner ) systems often demonstrate great performance on in - distribution data , but perform poorly on examples drawn from a shifted distribution .", "one way to evaluate the generalization ability of ner models is to use adversarial examples , on which the specific variations associated with named entities are rarely considered ."], "relation": "used for", "id": "2022.findings-acl.154", "year": 2022, "rel_sent": "By training on adversarial augmented training examples and using mixup for regularization , we were able to significantly improve the performance on the challenging set as well as improve out - of - domain generalization which we evaluated by using OntoNotes data .", "forward": true, "src_ids": "2022.findings-acl.154_3836"}
+{"input": "systematic generalization is done by using Method| context: semantic parsing is the task of producing structured meaning representations for natural language sentences . recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "systematic generalization", "output": "lagr", "neg_sample": ["systematic generalization is done by using Method", "semantic parsing is the task of producing structured meaning representations for natural language sentences .", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "LAGr : Label Aligned Graphs for Better Systematic Generalization in Semantic Parsing.", "forward": false, "src_ids": "2022.acl-long.233_3837"}
+{"input": "label aligned graphs is used for Task| context: semantic parsing is the task of producing structured meaning representations for natural language sentences . recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "label aligned graphs", "output": "systematic generalization", "neg_sample": ["label aligned graphs is used for Task", "semantic parsing is the task of producing structured meaning representations for natural language sentences .", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "LAGr : Label Aligned Graphs for Better Systematic Generalization in Semantic Parsing.", "forward": true, "src_ids": "2022.acl-long.233_3838"}
+{"input": "lagr is used for Task| context: semantic parsing is the task of producing structured meaning representations for natural language sentences . recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "lagr", "output": "systematic generalization", "neg_sample": ["lagr is used for Task", "semantic parsing is the task of producing structured meaning representations for natural language sentences .", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "LAGr : Label Aligned Graphs for Better Systematic Generalization in Semantic Parsing.", "forward": true, "src_ids": "2022.acl-long.233_3839"}
+{"input": "lagr ( label aligned graphs ) is used for Task| context: recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "lagr ( label aligned graphs )", "output": "semantic parsing", "neg_sample": ["lagr ( label aligned graphs ) is used for Task", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "To this end we propose LAGr ( Label Aligned Graphs ) , a general framework to produce semantic parses by independently predicting node and edge labels for a complete multi - layer input - aligned graph .", "forward": true, "src_ids": "2022.acl-long.233_3840"}
+{"input": "semantic parsing is done by using Method| context: semantic parsing is the task of producing structured meaning representations for natural language sentences . recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "semantic parsing", "output": "lagr ( label aligned graphs )", "neg_sample": ["semantic parsing is done by using Method", "semantic parsing is the task of producing structured meaning representations for natural language sentences .", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "To this end we propose LAGr ( Label Aligned Graphs ) , a general framework to produce semantic parses by independently predicting node and edge labels for a complete multi - layer input - aligned graph .", "forward": false, "src_ids": "2022.acl-long.233_3841"}
+{"input": "alignments is done by using Method| context: semantic parsing is the task of producing structured meaning representations for natural language sentences . recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "alignments", "output": "weakly - supervised lagr", "neg_sample": ["alignments is done by using Method", "semantic parsing is the task of producing structured meaning representations for natural language sentences .", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "The strongly - supervised LAGr algorithm requires aligned graphs as inputs , whereas weakly - supervised LAGr infers alignments for originally unaligned target graphs using approximate maximum - a - posteriori inference .", "forward": false, "src_ids": "2022.acl-long.233_3842"}
+{"input": "weakly - supervised lagr is used for OtherScientificTerm| context: semantic parsing is the task of producing structured meaning representations for natural language sentences . recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e. to handle examples that require recombining known knowledge in novel settings .", "entity": "weakly - supervised lagr", "output": "alignments", "neg_sample": ["weakly - supervised lagr is used for OtherScientificTerm", "semantic parsing is the task of producing structured meaning representations for natural language sentences .", "recent research has pointed out that the commonly - used sequence - to - sequence ( seq2seq ) semantic parsers struggle to generalize systematically , i.e.", "to handle examples that require recombining known knowledge in novel settings ."], "relation": "used for", "id": "2022.acl-long.233", "year": 2022, "rel_sent": "The strongly - supervised LAGr algorithm requires aligned graphs as inputs , whereas weakly - supervised LAGr infers alignments for originally unaligned target graphs using approximate maximum - a - posteriori inference .", "forward": true, "src_ids": "2022.acl-long.233_3843"}
+{"input": "coshc is used for Task| context: however , previous methods focus on retrieval accuracy , but lacked attention to the efficiency of the retrieval process .", "entity": "coshc", "output": "code search", "neg_sample": ["coshc is used for Task", "however , previous methods focus on retrieval accuracy , but lacked attention to the efficiency of the retrieval process ."], "relation": "used for", "id": "2022.acl-long.181", "year": 2022, "rel_sent": "We propose a novel method CoSHC to accelerate code search with deep hashing and code classification , aiming to perform efficient code search without sacrificing too much accuracy .", "forward": true, "src_ids": "2022.acl-long.181_3844"}
+{"input": "code search is done by using Method| context: code search is to search reusable code snippets from source code corpus based on natural languages queries . deep learning - based methods on code search have shown promising results . however , previous methods focus on retrieval accuracy , but lacked attention to the efficiency of the retrieval process .", "entity": "code search", "output": "coshc", "neg_sample": ["code search is done by using Method", "code search is to search reusable code snippets from source code corpus based on natural languages queries .", "deep learning - based methods on code search have shown promising results .", "however , previous methods focus on retrieval accuracy , but lacked attention to the efficiency of the retrieval process ."], "relation": "used for", "id": "2022.acl-long.181", "year": 2022, "rel_sent": "We propose a novel method CoSHC to accelerate code search with deep hashing and code classification , aiming to perform efficient code search without sacrificing too much accuracy .", "forward": false, "src_ids": "2022.acl-long.181_3845"}
+{"input": "latent space is done by using OtherScientificTerm| context: automatic transfer of text between domains has become popular in recent times . however , it does not explicitly maintain other attributes between the source and translated text : e.g. , text length and descriptiveness . maintaining constraints in transfer has several downstream applications , including data augmentation and debiasing .", "entity": "latent space", "output": "classification loss", "neg_sample": ["latent space is done by using OtherScientificTerm", "automatic transfer of text between domains has become popular in recent times .", "however , it does not explicitly maintain other attributes between the source and translated text : e.g.", ", text length and descriptiveness .", "maintaining constraints in transfer has several downstream applications , including data augmentation and debiasing ."], "relation": "used for", "id": "2022.acl-long.32", "year": 2022, "rel_sent": "The first is a contrastive loss and the second is a classification loss - aiming to regularize the latent space further and bring similar sentences closer together .", "forward": false, "src_ids": "2022.acl-long.32_3846"}
+{"input": "classification loss is used for OtherScientificTerm| context: automatic transfer of text between domains has become popular in recent times . however , it does not explicitly maintain other attributes between the source and translated text : e.g. , text length and descriptiveness . maintaining constraints in transfer has several downstream applications , including data augmentation and debiasing .", "entity": "classification loss", "output": "latent space", "neg_sample": ["classification loss is used for OtherScientificTerm", "automatic transfer of text between domains has become popular in recent times .", "however , it does not explicitly maintain other attributes between the source and translated text : e.g.", ", text length and descriptiveness .", "maintaining constraints in transfer has several downstream applications , including data augmentation and debiasing ."], "relation": "used for", "id": "2022.acl-long.32", "year": 2022, "rel_sent": "The first is a contrastive loss and the second is a classification loss - aiming to regularize the latent space further and bring similar sentences closer together .", "forward": true, "src_ids": "2022.acl-long.32_3847"}
+{"input": "sentiment of multimodal movie reviews is done by using OtherScientificTerm| context: with the proliferation of internet usage , a massive growth of consumer - generated content on social media has been witnessed in recent years that provide people 's opinions on diverse issues . through social media , users can convey their emotions and thoughts in distinctive forms such as text , image , audio , video , and emoji , which leads to the advancement of the multimodality of the content users on social networking sites .", "entity": "sentiment of multimodal movie reviews", "output": "textual features", "neg_sample": ["sentiment of multimodal movie reviews is done by using OtherScientificTerm", "with the proliferation of internet usage , a massive growth of consumer - generated content on social media has been witnessed in recent years that provide people 's opinions on diverse issues .", "through social media , users can convey their emotions and thoughts in distinctive forms such as text , image , audio , video , and emoji , which leads to the advancement of the multimodality of the content users on social networking sites ."], "relation": "used for", "id": "2022.dravidianlangtech-1.30", "year": 2022, "rel_sent": "CUET - NLP@DravidianLangTech - ACL2022 : Exploiting Textual Features to Classify Sentiment of Multimodal Movie Reviews.", "forward": false, "src_ids": "2022.dravidianlangtech-1.30_3848"}
+{"input": "textual features is used for Task| context: with the proliferation of internet usage , a massive growth of consumer - generated content on social media has been witnessed in recent years that provide people 's opinions on diverse issues . through social media , users can convey their emotions and thoughts in distinctive forms such as text , image , audio , video , and emoji , which leads to the advancement of the multimodality of the content users on social networking sites .", "entity": "textual features", "output": "sentiment of multimodal movie reviews", "neg_sample": ["textual features is used for Task", "with the proliferation of internet usage , a massive growth of consumer - generated content on social media has been witnessed in recent years that provide people 's opinions on diverse issues .", "through social media , users can convey their emotions and thoughts in distinctive forms such as text , image , audio , video , and emoji , which leads to the advancement of the multimodality of the content users on social networking sites ."], "relation": "used for", "id": "2022.dravidianlangtech-1.30", "year": 2022, "rel_sent": "CUET - NLP@DravidianLangTech - ACL2022 : Exploiting Textual Features to Classify Sentiment of Multimodal Movie Reviews.", "forward": true, "src_ids": "2022.dravidianlangtech-1.30_3849"}
+{"input": "multilingual crossover encoder - decoder ( mxencdec ) is used for OtherScientificTerm| context: multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs . the dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages ; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct sub - spaces .", "entity": "multilingual crossover encoder - decoder ( mxencdec )", "output": "instance level", "neg_sample": ["multilingual crossover encoder - decoder ( mxencdec ) is used for OtherScientificTerm", "multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs .", "the dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages ; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct sub - spaces ."], "relation": "used for", "id": "2022.acl-long.282", "year": 2022, "rel_sent": "In this paper , we introduce multilingual crossover encoder - decoder ( mXEncDec ) tofuse language pairs at an instance level .", "forward": true, "src_ids": "2022.acl-long.282_3850"}
+{"input": "language pairs is done by using Method| context: multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs . the dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages ; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct sub - spaces .", "entity": "language pairs", "output": "multilingual crossover encoder - decoder ( mxencdec )", "neg_sample": ["language pairs is done by using Method", "multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs .", "the dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages ; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct sub - spaces ."], "relation": "used for", "id": "2022.acl-long.282", "year": 2022, "rel_sent": "In this paper , we introduce multilingual crossover encoder - decoder ( mXEncDec ) tofuse language pairs at an instance level .", "forward": false, "src_ids": "2022.acl-long.282_3851"}
+{"input": "instance level is done by using Method| context: multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs . the dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages ; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct sub - spaces .", "entity": "instance level", "output": "multilingual crossover encoder - decoder ( mxencdec )", "neg_sample": ["instance level is done by using Method", "multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs .", "the dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages ; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct sub - spaces ."], "relation": "used for", "id": "2022.acl-long.282", "year": 2022, "rel_sent": "In this paper , we introduce multilingual crossover encoder - decoder ( mXEncDec ) tofuse language pairs at an instance level .", "forward": false, "src_ids": "2022.acl-long.282_3852"}
+{"input": "multimodality is used for Task| context: sarcasm is important to sentiment analysis on social media . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "multimodality", "output": "sarcasm target identification ( sti )", "neg_sample": ["multimodality is used for Task", "sarcasm is important to sentiment analysis on social media .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "In this paper , we introduce multimodality to STI and present Multimodal Sarcasm Target Identification ( MSTI ) task .", "forward": true, "src_ids": "2022.acl-long.562_3853"}
+{"input": "sarcasm target identification ( sti ) is done by using OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "sarcasm target identification ( sti )", "output": "multimodality", "neg_sample": ["sarcasm target identification ( sti ) is done by using OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "In this paper , we introduce multimodality to STI and present Multimodal Sarcasm Target Identification ( MSTI ) task .", "forward": false, "src_ids": "2022.acl-long.562_3854"}
+{"input": "textual target labeling is done by using Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "textual target labeling", "output": "multi - scale cross - modality model", "neg_sample": ["textual target labeling is done by using Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We propose a novel multi - scale cross - modality model that can simultaneously perform textual target labeling and visual target detection .", "forward": false, "src_ids": "2022.acl-long.562_3855"}
+{"input": "visual target detection is done by using Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "visual target detection", "output": "multi - scale cross - modality model", "neg_sample": ["visual target detection is done by using Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We propose a novel multi - scale cross - modality model that can simultaneously perform textual target labeling and visual target detection .", "forward": false, "src_ids": "2022.acl-long.562_3856"}
+{"input": "multi - scale cross - modality model is used for Task| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "multi - scale cross - modality model", "output": "textual target labeling", "neg_sample": ["multi - scale cross - modality model is used for Task", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We propose a novel multi - scale cross - modality model that can simultaneously perform textual target labeling and visual target detection .", "forward": true, "src_ids": "2022.acl-long.562_3857"}
+{"input": "spatial information is done by using OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "spatial information", "output": "multi - scale visual features", "neg_sample": ["spatial information is done by using OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "In the model , we extract multi - scale visual features to enrich spatial information for different sized visual sarcasm targets .", "forward": false, "src_ids": "2022.acl-long.562_3858"}
+{"input": "sized visual sarcasm targets is done by using OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "sized visual sarcasm targets", "output": "multi - scale visual features", "neg_sample": ["sized visual sarcasm targets is done by using OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "In the model , we extract multi - scale visual features to enrich spatial information for different sized visual sarcasm targets .", "forward": false, "src_ids": "2022.acl-long.562_3859"}
+{"input": "convolution networks is used for OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "convolution networks", "output": "multi - scale visual features", "neg_sample": ["convolution networks is used for OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": true, "src_ids": "2022.acl-long.562_3860"}
+{"input": "multi - scale visual features is used for OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "multi - scale visual features", "output": "spatial information", "neg_sample": ["multi - scale visual features is used for OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "In the model , we extract multi - scale visual features to enrich spatial information for different sized visual sarcasm targets .", "forward": true, "src_ids": "2022.acl-long.562_3861"}
+{"input": "spatial information is used for OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "spatial information", "output": "sized visual sarcasm targets", "neg_sample": ["spatial information is used for OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "In the model , we extract multi - scale visual features to enrich spatial information for different sized visual sarcasm targets .", "forward": true, "src_ids": "2022.acl-long.562_3862"}
+{"input": "cross - modal attention learning is done by using Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "cross - modal attention learning", "output": "convolution networks", "neg_sample": ["cross - modal attention learning is done by using Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": false, "src_ids": "2022.acl-long.562_3863"}
+{"input": "multi - scale visual features is done by using Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "multi - scale visual features", "output": "convolution networks", "neg_sample": ["multi - scale visual features is done by using Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": false, "src_ids": "2022.acl-long.562_3864"}
+{"input": "cross - modal attention learning is done by using OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "cross - modal attention learning", "output": "textual features", "neg_sample": ["cross - modal attention learning is done by using OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": false, "src_ids": "2022.acl-long.562_3865"}
+{"input": "convolution networks is used for Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "convolution networks", "output": "cross - modal attention learning", "neg_sample": ["convolution networks is used for Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": true, "src_ids": "2022.acl-long.562_3866"}
+{"input": "textual features is used for Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "textual features", "output": "cross - modal attention learning", "neg_sample": ["textual features is used for Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": true, "src_ids": "2022.acl-long.562_3867"}
+{"input": "multi - scale visual information is done by using Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "multi - scale visual information", "output": "transposed convolution networks", "neg_sample": ["multi - scale visual information is done by using Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": false, "src_ids": "2022.acl-long.562_3868"}
+{"input": "transposed convolution networks is used for OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "transposed convolution networks", "output": "multi - scale visual information", "neg_sample": ["transposed convolution networks is used for OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "We design a set of convolution networks to unify multi - scale visual features with textual features for cross - modal attention learning , and correspondingly a set of transposed convolution networks to restore multi - scale visual information .", "forward": true, "src_ids": "2022.acl-long.562_3869"}
+{"input": "tsti is done by using OtherScientificTerm| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "tsti", "output": "visual clues", "neg_sample": ["tsti is done by using OtherScientificTerm", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "The results show that visual clues can improve the performance of TSTI by a large margin , and VSTI achieves good accuracy .", "forward": false, "src_ids": "2022.acl-long.562_3870"}
+{"input": "visual clues is used for Method| context: sarcasm is important to sentiment analysis on social media . sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth . however , text lacking context or missing sarcasm target makes target identification very difficult .", "entity": "visual clues", "output": "tsti", "neg_sample": ["visual clues is used for Method", "sarcasm is important to sentiment analysis on social media .", "sarcasm target identification ( sti ) deserves further study to understand sarcasm in depth .", "however , text lacking context or missing sarcasm target makes target identification very difficult ."], "relation": "used for", "id": "2022.acl-long.562", "year": 2022, "rel_sent": "The results show that visual clues can improve the performance of TSTI by a large margin , and VSTI achieves good accuracy .", "forward": true, "src_ids": "2022.acl-long.562_3871"}
+{"input": "document - level relation extraction ( docre ) is done by using Method| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "document - level relation extraction ( docre )", "output": "eider", "neg_sample": ["document - level relation extraction ( docre ) is done by using Method", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "Eider : Empowering Document - level Relation Extraction with Efficient Evidence Extraction and Inference - stage Fusion.", "forward": false, "src_ids": "2022.findings-acl.23_3872"}
+{"input": "document - level relation extraction ( docre ) is done by using Method| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "document - level relation extraction ( docre )", "output": "eider", "neg_sample": ["document - level relation extraction ( docre ) is done by using Method", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "In this paper , we propose an evidence - enhanced framework , Eider , that empowers DocRE by efficiently extracting evidence and effectively fusing the extracted evidence in inference .", "forward": false, "src_ids": "2022.findings-acl.23_3873"}
+{"input": "semantic relations is done by using Task| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "semantic relations", "output": "document - level relation extraction ( docre )", "neg_sample": ["semantic relations is done by using Task", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "Document - level relation extraction ( DocRE ) aims to extract semantic relations among entity pairs in a document .", "forward": false, "src_ids": "2022.findings-acl.23_3874"}
+{"input": "eider is used for Task| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "eider", "output": "document - level relation extraction ( docre )", "neg_sample": ["eider is used for Task", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "Eider : Empowering Document - level Relation Extraction with Efficient Evidence Extraction and Inference - stage Fusion.", "forward": true, "src_ids": "2022.findings-acl.23_3875"}
+{"input": "evidence - enhanced framework is used for Task| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "evidence - enhanced framework", "output": "document - level relation extraction ( docre )", "neg_sample": ["evidence - enhanced framework is used for Task", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "In this paper , we propose an evidence - enhanced framework , Eider , that empowers DocRE by efficiently extracting evidence and effectively fusing the extracted evidence in inference .", "forward": true, "src_ids": "2022.findings-acl.23_3876"}
+{"input": "eider is used for Task| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "eider", "output": "document - level relation extraction ( docre )", "neg_sample": ["eider is used for Task", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "In this paper , we propose an evidence - enhanced framework , Eider , that empowers DocRE by efficiently extracting evidence and effectively fusing the extracted evidence in inference .", "forward": true, "src_ids": "2022.findings-acl.23_3877"}
+{"input": "document - level relation extraction ( docre ) is used for OtherScientificTerm| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "document - level relation extraction ( docre )", "output": "semantic relations", "neg_sample": ["document - level relation extraction ( docre ) is used for OtherScientificTerm", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "Document - level relation extraction ( DocRE ) aims to extract semantic relations among entity pairs in a document .", "forward": true, "src_ids": "2022.findings-acl.23_3878"}
+{"input": "re is done by using Method| context: typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair .", "entity": "re", "output": "evidence model", "neg_sample": ["re is done by using Method", "typical docre methods blindly take the full document as input , while a subset of the sentences in the document , noted as the evidence , are often sufficient for humans to predict the relation of an entity pair ."], "relation": "used for", "id": "2022.findings-acl.23", "year": 2022, "rel_sent": "Empirically , even training the evidence model on silver labels constructed by our heuristic rules can lead to better RE performance .", "forward": false, "src_ids": "2022.findings-acl.23_3879"}
+{"input": "dialogue skills is done by using Method| context: pre - trained models have achieved excellent performance on the dialogue task . however , for the continual increase of online chit - chat scenarios , directly fine - tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre - trained models and knowledge interference among diverse dialogue tasks .", "entity": "dialogue skills", "output": "hierarchical inductive transfer framework", "neg_sample": ["dialogue skills is done by using Method", "pre - trained models have achieved excellent performance on the dialogue task .", "however , for the continual increase of online chit - chat scenarios , directly fine - tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre - trained models and knowledge interference among diverse dialogue tasks ."], "relation": "used for", "id": "2022.findings-acl.57", "year": 2022, "rel_sent": "In this work , we propose a hierarchical inductive transfer framework to learn and deploy the dialogue skills continually and efficiently .", "forward": false, "src_ids": "2022.findings-acl.57_3880"}
+{"input": "dialogue skills is done by using Method| context: pre - trained models have achieved excellent performance on the dialogue task .", "entity": "dialogue skills", "output": "dialogue system", "neg_sample": ["dialogue skills is done by using Method", "pre - trained models have achieved excellent performance on the dialogue task ."], "relation": "used for", "id": "2022.findings-acl.57", "year": 2022, "rel_sent": "As the only trainable module , it is beneficial for the dialogue system on the embedded devices to acquire new dialogue skills with negligible additional parameters .", "forward": false, "src_ids": "2022.findings-acl.57_3881"}
+{"input": "hierarchical inductive transfer framework is used for OtherScientificTerm| context: pre - trained models have achieved excellent performance on the dialogue task . however , for the continual increase of online chit - chat scenarios , directly fine - tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre - trained models and knowledge interference among diverse dialogue tasks .", "entity": "hierarchical inductive transfer framework", "output": "dialogue skills", "neg_sample": ["hierarchical inductive transfer framework is used for OtherScientificTerm", "pre - trained models have achieved excellent performance on the dialogue task .", "however , for the continual increase of online chit - chat scenarios , directly fine - tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre - trained models and knowledge interference among diverse dialogue tasks ."], "relation": "used for", "id": "2022.findings-acl.57", "year": 2022, "rel_sent": "In this work , we propose a hierarchical inductive transfer framework to learn and deploy the dialogue skills continually and efficiently .", "forward": true, "src_ids": "2022.findings-acl.57_3882"}
+{"input": "dialogue system is used for OtherScientificTerm| context: pre - trained models have achieved excellent performance on the dialogue task . however , for the continual increase of online chit - chat scenarios , directly fine - tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre - trained models and knowledge interference among diverse dialogue tasks .", "entity": "dialogue system", "output": "dialogue skills", "neg_sample": ["dialogue system is used for OtherScientificTerm", "pre - trained models have achieved excellent performance on the dialogue task .", "however , for the continual increase of online chit - chat scenarios , directly fine - tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre - trained models and knowledge interference among diverse dialogue tasks ."], "relation": "used for", "id": "2022.findings-acl.57", "year": 2022, "rel_sent": "As the only trainable module , it is beneficial for the dialogue system on the embedded devices to acquire new dialogue skills with negligible additional parameters .", "forward": true, "src_ids": "2022.findings-acl.57_3883"}
+{"input": "probing datasets is used for Method| context: as large and powerful neural language models are developed , researchers have been increasingly interested in developing diagnostic tools to probe them . there are many papers with conclusions of the form ' observation x is found in model y ' , using their own datasets with varying sizes . larger probing datasets bring more reliability , but are also expensive to collect . there is yet to be a quantitative method for estimating reasonable probing dataset sizes .", "entity": "probing datasets", "output": "neural nlp models", "neg_sample": ["probing datasets is used for Method", "as large and powerful neural language models are developed , researchers have been increasingly interested in developing diagnostic tools to probe them .", "there are many papers with conclusions of the form ' observation x is found in model y ' , using their own datasets with varying sizes .", "larger probing datasets bring more reliability , but are also expensive to collect .", "there is yet to be a quantitative method for estimating reasonable probing dataset sizes ."], "relation": "used for", "id": "2022.findings-acl.326", "year": 2022, "rel_sent": "Our framework helps to systematically construct probing datasets to diagnose neural NLP models .", "forward": true, "src_ids": "2022.findings-acl.326_3884"}
+{"input": "neural nlp models is done by using Material| context: as large and powerful neural language models are developed , researchers have been increasingly interested in developing diagnostic tools to probe them . there are many papers with conclusions of the form ' observation x is found in model y ' , using their own datasets with varying sizes . there is yet to be a quantitative method for estimating reasonable probing dataset sizes .", "entity": "neural nlp models", "output": "probing datasets", "neg_sample": ["neural nlp models is done by using Material", "as large and powerful neural language models are developed , researchers have been increasingly interested in developing diagnostic tools to probe them .", "there are many papers with conclusions of the form ' observation x is found in model y ' , using their own datasets with varying sizes .", "there is yet to be a quantitative method for estimating reasonable probing dataset sizes ."], "relation": "used for", "id": "2022.findings-acl.326", "year": 2022, "rel_sent": "Our framework helps to systematically construct probing datasets to diagnose neural NLP models .", "forward": false, "src_ids": "2022.findings-acl.326_3885"}
+{"input": "robust text classification is done by using Method| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "robust text classification", "output": "sibylvariant transformations", "neg_sample": ["robust text classification is done by using Method", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "Sibylvariant Transformations for Robust Text Classification.", "forward": false, "src_ids": "2022.findings-acl.140_3886"}
+{"input": "sibylvariant transformations is used for Task| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "sibylvariant transformations", "output": "robust text classification", "neg_sample": ["sibylvariant transformations is used for Task", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "Sibylvariant Transformations for Robust Text Classification.", "forward": true, "src_ids": "2022.findings-acl.140_3887"}
+{"input": "transforms is done by using OtherScientificTerm| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "transforms", "output": "sibylvariance ( sib )", "neg_sample": ["transforms is done by using OtherScientificTerm", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "In this work , we propose the notion of sibylvariance ( SIB ) to describe the broader set of transforms that relax the label - preserving constraint , knowably vary the expected class , and lead to significantly more diverse input distributions .", "forward": false, "src_ids": "2022.findings-acl.140_3888"}
+{"input": "sibylvariance ( sib ) is used for Generic| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "sibylvariance ( sib )", "output": "transforms", "neg_sample": ["sibylvariance ( sib ) is used for Generic", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "In this work , we propose the notion of sibylvariance ( SIB ) to describe the broader set of transforms that relax the label - preserving constraint , knowably vary the expected class , and lead to significantly more diverse input distributions .", "forward": true, "src_ids": "2022.findings-acl.140_3889"}
+{"input": "nlp is done by using OtherScientificTerm| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "nlp", "output": "sibylvariance", "neg_sample": ["nlp is done by using OtherScientificTerm", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "To explore the role of sibylvariance within NLP , we implemented 41 text transformations , including several novel techniques like Concept2Sentence and SentMix .", "forward": false, "src_ids": "2022.findings-acl.140_3890"}
+{"input": "adaptive training is done by using OtherScientificTerm| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "adaptive training", "output": "sibylvariance", "neg_sample": ["adaptive training is done by using OtherScientificTerm", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "Sibylvariance also enables a unique form of adaptive training that generates new input mixtures for the most confused class pairs , challenging the learner to differentiate with greater nuance .", "forward": false, "src_ids": "2022.findings-acl.140_3891"}
+{"input": "sibylvariance is used for Method| context: the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label .", "entity": "sibylvariance", "output": "adaptive training", "neg_sample": ["sibylvariance is used for Method", "the vast majority of text transformation techniques in nlp are inherently limited in their ability to expand input space coverage due to an implicit constraint to preserve the original class label ."], "relation": "used for", "id": "2022.findings-acl.140", "year": 2022, "rel_sent": "Sibylvariance also enables a unique form of adaptive training that generates new input mixtures for the most confused class pairs , challenging the learner to differentiate with greater nuance .", "forward": true, "src_ids": "2022.findings-acl.140_3892"}
+{"input": "qa ( bbq ) is done by using Generic| context: it is well documented that nlp models learn social biases , but little work has been done on how these biases manifest in model outputs for applied tasks like question answering ( qa ) .", "entity": "qa ( bbq )", "output": "bias benchmark", "neg_sample": ["qa ( bbq ) is done by using Generic", "it is well documented that nlp models learn social biases , but little work has been done on how these biases manifest in model outputs for applied tasks like question answering ( qa ) ."], "relation": "used for", "id": "2022.findings-acl.165", "year": 2022, "rel_sent": "We introduce the Bias Benchmark for QA ( BBQ ) , a dataset of question - sets constructed by the authors that highlight attested social biases against people belonging to protected classes along nine social dimensions relevant for U.S. English - speaking contexts .", "forward": false, "src_ids": "2022.findings-acl.165_3893"}
+{"input": "bias benchmark is used for Task| context: it is well documented that nlp models learn social biases , but little work has been done on how these biases manifest in model outputs for applied tasks like question answering ( qa ) .", "entity": "bias benchmark", "output": "qa ( bbq )", "neg_sample": ["bias benchmark is used for Task", "it is well documented that nlp models learn social biases , but little work has been done on how these biases manifest in model outputs for applied tasks like question answering ( qa ) ."], "relation": "used for", "id": "2022.findings-acl.165", "year": 2022, "rel_sent": "We introduce the Bias Benchmark for QA ( BBQ ) , a dataset of question - sets constructed by the authors that highlight attested social biases against people belonging to protected classes along nine social dimensions relevant for U.S. English - speaking contexts .", "forward": true, "src_ids": "2022.findings-acl.165_3894"}
+{"input": "naturalistic social dialogs is done by using Task| context: current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors .", "entity": "naturalistic social dialogs", "output": "word sense disambiguation", "neg_sample": ["naturalistic social dialogs is done by using Task", "current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors ."], "relation": "used for", "id": "2022.humeval-1.10", "year": 2022, "rel_sent": "Towards Human Evaluation of Mutual Understanding in Human - Computer Spontaneous Conversation : An Empirical Study of Word Sense Disambiguation for Naturalistic Social Dialogs in American English.", "forward": false, "src_ids": "2022.humeval-1.10_3895"}
+{"input": "word sense disambiguation is used for Material| context: current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors .", "entity": "word sense disambiguation", "output": "naturalistic social dialogs", "neg_sample": ["word sense disambiguation is used for Material", "current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors ."], "relation": "used for", "id": "2022.humeval-1.10", "year": 2022, "rel_sent": "Towards Human Evaluation of Mutual Understanding in Human - Computer Spontaneous Conversation : An Empirical Study of Word Sense Disambiguation for Naturalistic Social Dialogs in American English.", "forward": true, "src_ids": "2022.humeval-1.10_3896"}
+{"input": "human evaluation framework is done by using Task| context: current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors .", "entity": "human evaluation framework", "output": "word sense disambiguation ( wsd )", "neg_sample": ["human evaluation framework is done by using Task", "current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors ."], "relation": "used for", "id": "2022.humeval-1.10", "year": 2022, "rel_sent": "This work proposes Word Sense Disambiguation ( WSD ) as an essential component of a valid and reliable human evaluation framework , whose long - term goal is to radically improve the usability of dialog systems in real - life human - computer collaboration .", "forward": false, "src_ids": "2022.humeval-1.10_3897"}
+{"input": "word sense disambiguation ( wsd ) is used for Method| context: current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors .", "entity": "word sense disambiguation ( wsd )", "output": "human evaluation framework", "neg_sample": ["word sense disambiguation ( wsd ) is used for Method", "current evaluation practices for social dialog systems , dedicated to human - computer spontaneous conversation , exclusively focus on the quality of system - generated surface text , but not human - verifiable aspects of mutual understanding between the systems and their interlocutors ."], "relation": "used for", "id": "2022.humeval-1.10", "year": 2022, "rel_sent": "This work proposes Word Sense Disambiguation ( WSD ) as an essential component of a valid and reliable human evaluation framework , whose long - term goal is to radically improve the usability of dialog systems in real - life human - computer collaboration .", "forward": true, "src_ids": "2022.humeval-1.10_3898"}
+{"input": "zero - shot relation extraction ( zsre ) is done by using Method| context: while most previous studies have focused on fully supervised relation extraction and achieved considerably high performance , less effort has been made towards zsre .", "entity": "zero - shot relation extraction ( zsre )", "output": "discriminative learning", "neg_sample": ["zero - shot relation extraction ( zsre ) is done by using Method", "while most previous studies have focused on fully supervised relation extraction and achieved considerably high performance , less effort has been made towards zsre ."], "relation": "used for", "id": "2022.spanlp-1.1", "year": 2022, "rel_sent": "Improving Discriminative Learning for Zero - Shot Relation Extraction.", "forward": false, "src_ids": "2022.spanlp-1.1_3899"}
+{"input": "sentences and semantic relations is done by using Method| context: while most previous studies have focused on fully supervised relation extraction and achieved considerably high performance , less effort has been made towards zsre .", "entity": "sentences and semantic relations", "output": "discriminative embedding learning", "neg_sample": ["sentences and semantic relations is done by using Method", "while most previous studies have focused on fully supervised relation extraction and achieved considerably high performance , less effort has been made towards zsre ."], "relation": "used for", "id": "2022.spanlp-1.1", "year": 2022, "rel_sent": "This study proposes a new model incorporating discriminative embedding learning for both sentences and semantic relations .", "forward": false, "src_ids": "2022.spanlp-1.1_3900"}
+{"input": "discriminative embedding learning is used for OtherScientificTerm| context: while most previous studies have focused on fully supervised relation extraction and achieved considerably high performance , less effort has been made towards zsre .", "entity": "discriminative embedding learning", "output": "sentences and semantic relations", "neg_sample": ["discriminative embedding learning is used for OtherScientificTerm", "while most previous studies have focused on fully supervised relation extraction and achieved considerably high performance , less effort has been made towards zsre ."], "relation": "used for", "id": "2022.spanlp-1.1", "year": 2022, "rel_sent": "This study proposes a new model incorporating discriminative embedding learning for both sentences and semantic relations .", "forward": true, "src_ids": "2022.spanlp-1.1_3901"}
+{"input": "sequence - to - sequence models is done by using Task| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "sequence - to - sequence models", "output": "decoding", "neg_sample": ["sequence - to - sequence models is done by using Task", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "Uncertainty Determines the Adequacy of the Mode and the Tractability of Decoding in Sequence - to - Sequence Models.", "forward": false, "src_ids": "2022.acl-long.591_3902"}
+{"input": "decoding is used for Method| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "decoding", "output": "sequence - to - sequence models", "neg_sample": ["decoding is used for Method", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "Uncertainty Determines the Adequacy of the Mode and the Tractability of Decoding in Sequence - to - Sequence Models.", "forward": true, "src_ids": "2022.acl-long.591_3903"}
+{"input": "search is done by using OtherScientificTerm| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "search", "output": "intrinsic uncertainty", "neg_sample": ["search is done by using OtherScientificTerm", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "At both the sentence- and the task - level , intrinsic uncertainty has major implications for various aspects of search such as the inductive biases in beam search and the complexity of exact search .", "forward": false, "src_ids": "2022.acl-long.591_3904"}
+{"input": "model uncertainty is done by using OtherScientificTerm| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "model uncertainty", "output": "intrinsic uncertainty", "neg_sample": ["model uncertainty is done by using OtherScientificTerm", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "Furthermore , we propose a novel exact n - best search algorithm for neural sequence models , and show that intrinsic uncertainty affects model uncertainty as the model tends to overly spread out the probability mass for uncertain tasks and sentences .", "forward": false, "src_ids": "2022.acl-long.591_3905"}
+{"input": "exact n - best search algorithm is used for Method| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "exact n - best search algorithm", "output": "neural sequence models", "neg_sample": ["exact n - best search algorithm is used for Method", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "Furthermore , we propose a novel exact n - best search algorithm for neural sequence models , and show that intrinsic uncertainty affects model uncertainty as the model tends to overly spread out the probability mass for uncertain tasks and sentences .", "forward": true, "src_ids": "2022.acl-long.591_3906"}
+{"input": "intrinsic uncertainty is used for Task| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "intrinsic uncertainty", "output": "search", "neg_sample": ["intrinsic uncertainty is used for Task", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "At both the sentence- and the task - level , intrinsic uncertainty has major implications for various aspects of search such as the inductive biases in beam search and the complexity of exact search .", "forward": true, "src_ids": "2022.acl-long.591_3907"}
+{"input": "neural sequence models is done by using Method| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "neural sequence models", "output": "exact n - best search algorithm", "neg_sample": ["neural sequence models is done by using Method", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "Furthermore , we propose a novel exact n - best search algorithm for neural sequence models , and show that intrinsic uncertainty affects model uncertainty as the model tends to overly spread out the probability mass for uncertain tasks and sentences .", "forward": false, "src_ids": "2022.acl-long.591_3908"}
+{"input": "intrinsic uncertainty is used for OtherScientificTerm| context: in many natural language processing ( nlp ) tasks the same input ( e.g. source sentence ) can have multiple possible outputs ( e.g. translations ) .", "entity": "intrinsic uncertainty", "output": "model uncertainty", "neg_sample": ["intrinsic uncertainty is used for OtherScientificTerm", "in many natural language processing ( nlp ) tasks the same input ( e.g.", "source sentence ) can have multiple possible outputs ( e.g.", "translations ) ."], "relation": "used for", "id": "2022.acl-long.591", "year": 2022, "rel_sent": "Furthermore , we propose a novel exact n - best search algorithm for neural sequence models , and show that intrinsic uncertainty affects model uncertainty as the model tends to overly spread out the probability mass for uncertain tasks and sentences .", "forward": true, "src_ids": "2022.acl-long.591_3909"}
+{"input": "word - level recognition is done by using Material| context: ai technologies for natural languages have made tremendous progress recently .", "entity": "word - level recognition", "output": "sign languages", "neg_sample": ["word - level recognition is done by using Material", "ai technologies for natural languages have made tremendous progress recently ."], "relation": "used for", "id": "2022.acl-long.150", "year": 2022, "rel_sent": "We introduce OpenHands , a library where we take four key ideas from the NLP community for low - resource languages and apply them to sign languages for word - level recognition .", "forward": false, "src_ids": "2022.acl-long.150_3910"}
+{"input": "sign languages is used for Task| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "sign languages", "output": "word - level recognition", "neg_sample": ["sign languages is used for Task", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.150", "year": 2022, "rel_sent": "We introduce OpenHands , a library where we take four key ideas from the NLP community for low - resource languages and apply them to sign languages for word - level recognition .", "forward": true, "src_ids": "2022.acl-long.150_3911"}
+{"input": "inference is done by using OtherScientificTerm| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "inference", "output": "pose", "neg_sample": ["inference is done by using OtherScientificTerm", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.150", "year": 2022, "rel_sent": "First , we propose using pose extracted through pretrained models as the standard modality of data in this work to reduce training time and enable efficient inference , and we release standardized pose datasets for different existing sign language datasets .", "forward": false, "src_ids": "2022.acl-long.150_3912"}
+{"input": "pose is used for Task| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "pose", "output": "inference", "neg_sample": ["pose is used for Task", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.150", "year": 2022, "rel_sent": "First , we propose using pose extracted through pretrained models as the standard modality of data in this work to reduce training time and enable efficient inference , and we release standardized pose datasets for different existing sign language datasets .", "forward": true, "src_ids": "2022.acl-long.150_3913"}
+{"input": "sign language datasets is done by using Material| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "sign language datasets", "output": "standardized pose datasets", "neg_sample": ["sign language datasets is done by using Material", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.150", "year": 2022, "rel_sent": "First , we propose using pose extracted through pretrained models as the standard modality of data in this work to reduce training time and enable efficient inference , and we release standardized pose datasets for different existing sign language datasets .", "forward": false, "src_ids": "2022.acl-long.150_3914"}
+{"input": "standardized pose datasets is used for Material| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "standardized pose datasets", "output": "sign language datasets", "neg_sample": ["standardized pose datasets is used for Material", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.150", "year": 2022, "rel_sent": "First , we propose using pose extracted through pretrained models as the standard modality of data in this work to reduce training time and enable efficient inference , and we release standardized pose datasets for different existing sign language datasets .", "forward": true, "src_ids": "2022.acl-long.150_3915"}
+{"input": "webpage information extraction ( wie ) is done by using Method| context: webpage information extraction ( wie ) is an important step to create knowledge bases . for this , classical wie methods leverage the document object model ( dom ) tree of a website . however , use of the dom tree poses significant challenges as context and appearance are encoded in an abstract manner .", "entity": "webpage information extraction ( wie )", "output": "context - aware visual attention", "neg_sample": ["webpage information extraction ( wie ) is done by using Method", "webpage information extraction ( wie ) is an important step to create knowledge bases .", "for this , classical wie methods leverage the document object model ( dom ) tree of a website .", "however , use of the dom tree poses significant challenges as context and appearance are encoded in an abstract manner ."], "relation": "used for", "id": "2022.ecnlp-1.11", "year": 2022, "rel_sent": "CoVA : Context - aware Visual Attention for Webpage Information Extraction.", "forward": false, "src_ids": "2022.ecnlp-1.11_3916"}
+{"input": "context - aware visual attention is used for Task| context: for this , classical wie methods leverage the document object model ( dom ) tree of a website . however , use of the dom tree poses significant challenges as context and appearance are encoded in an abstract manner .", "entity": "context - aware visual attention", "output": "webpage information extraction ( wie )", "neg_sample": ["context - aware visual attention is used for Task", "for this , classical wie methods leverage the document object model ( dom ) tree of a website .", "however , use of the dom tree poses significant challenges as context and appearance are encoded in an abstract manner ."], "relation": "used for", "id": "2022.ecnlp-1.11", "year": 2022, "rel_sent": "CoVA : Context - aware Visual Attention for Webpage Information Extraction.", "forward": true, "src_ids": "2022.ecnlp-1.11_3917"}
+{"input": "swiss german tv content is done by using Material| context: in this study we compare two approaches ( neural machine translation and edit - based ) and the use of synthetic data for the task of translating normalised swiss german asr output into correct written standard german for subtitles , with a special focus on syntactic differences .", "entity": "swiss german tv content", "output": "german subtitles", "neg_sample": ["swiss german tv content is done by using Material", "in this study we compare two approaches ( neural machine translation and edit - based ) and the use of synthetic data for the task of translating normalised swiss german asr output into correct written standard german for subtitles , with a special focus on syntactic differences ."], "relation": "used for", "id": "2022.slpat-1.5", "year": 2022, "rel_sent": "Producing Standard German Subtitles for Swiss German TV Content.", "forward": false, "src_ids": "2022.slpat-1.5_3918"}
+{"input": "german subtitles is used for Material| context: in this study we compare two approaches ( neural machine translation and edit - based ) and the use of synthetic data for the task of translating normalised swiss german asr output into correct written standard german for subtitles , with a special focus on syntactic differences .", "entity": "german subtitles", "output": "swiss german tv content", "neg_sample": ["german subtitles is used for Material", "in this study we compare two approaches ( neural machine translation and edit - based ) and the use of synthetic data for the task of translating normalised swiss german asr output into correct written standard german for subtitles , with a special focus on syntactic differences ."], "relation": "used for", "id": "2022.slpat-1.5", "year": 2022, "rel_sent": "Producing Standard German Subtitles for Swiss German TV Content.", "forward": true, "src_ids": "2022.slpat-1.5_3919"}
+{"input": "byte sequences is done by using Method| context: most widely used pre - trained language models operate on sequences of tokens corresponding to word or subword units . by comparison , token - free models that operate directly on raw text ( bytes or characters ) have many benefits : they can process text in any language out of the box , they are more robust to noise , and they minimize technical debt by removing complex and error - prone text preprocessing pipelines . because byte or character sequences are longer than token sequences , past work on token - free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text .", "entity": "byte sequences", "output": "transformer architecture", "neg_sample": ["byte sequences is done by using Method", "most widely used pre - trained language models operate on sequences of tokens corresponding to word or subword units .", "by comparison , token - free models that operate directly on raw text ( bytes or characters ) have many benefits : they can process text in any language out of the box , they are more robust to noise , and they minimize technical debt by removing complex and error - prone text preprocessing pipelines .", "because byte or character sequences are longer than token sequences , past work on token - free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text ."], "relation": "used for", "id": "2022.tacl-1.17", "year": 2022, "rel_sent": "In this paper , we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences .", "forward": false, "src_ids": "2022.tacl-1.17_3920"}
+{"input": "transformer architecture is used for Material| context: most widely used pre - trained language models operate on sequences of tokens corresponding to word or subword units . by comparison , token - free models that operate directly on raw text ( bytes or characters ) have many benefits : they can process text in any language out of the box , they are more robust to noise , and they minimize technical debt by removing complex and error - prone text preprocessing pipelines . because byte or character sequences are longer than token sequences , past work on token - free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text .", "entity": "transformer architecture", "output": "byte sequences", "neg_sample": ["transformer architecture is used for Material", "most widely used pre - trained language models operate on sequences of tokens corresponding to word or subword units .", "by comparison , token - free models that operate directly on raw text ( bytes or characters ) have many benefits : they can process text in any language out of the box , they are more robust to noise , and they minimize technical debt by removing complex and error - prone text preprocessing pipelines .", "because byte or character sequences are longer than token sequences , past work on token - free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text ."], "relation": "used for", "id": "2022.tacl-1.17", "year": 2022, "rel_sent": "In this paper , we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences .", "forward": true, "src_ids": "2022.tacl-1.17_3921"}
+{"input": "natural language processing of user - generated content is done by using Material| context: modern irish is a minority language lacking sufficient computational resources for the task of accurate automatic syntactic parsing of user - generated content such as tweets . although language technology for the irish language has been developing in recent years , these tools tend to perform poorly on user - generated content . as with other languages , the linguistic style observed in irish tweets differs , in terms of orthography , lexicon , and syntax , from that of standard texts more commonly used for the development of language models and parsers .", "entity": "natural language processing of user - generated content", "output": "universal dependencies treebank of irish tweets", "neg_sample": ["natural language processing of user - generated content is done by using Material", "modern irish is a minority language lacking sufficient computational resources for the task of accurate automatic syntactic parsing of user - generated content such as tweets .", "although language technology for the irish language has been developing in recent years , these tools tend to perform poorly on user - generated content .", "as with other languages , the linguistic style observed in irish tweets differs , in terms of orthography , lexicon , and syntax , from that of standard texts more commonly used for the development of language models and parsers ."], "relation": "used for", "id": "2022.acl-long.473", "year": 2022, "rel_sent": "We release the first Universal Dependencies treebank of Irish tweets , facilitating natural language processing of user - generated content in Irish .", "forward": false, "src_ids": "2022.acl-long.473_3922"}
+{"input": "universal dependencies treebank of irish tweets is used for Task| context: modern irish is a minority language lacking sufficient computational resources for the task of accurate automatic syntactic parsing of user - generated content such as tweets . although language technology for the irish language has been developing in recent years , these tools tend to perform poorly on user - generated content . as with other languages , the linguistic style observed in irish tweets differs , in terms of orthography , lexicon , and syntax , from that of standard texts more commonly used for the development of language models and parsers .", "entity": "universal dependencies treebank of irish tweets", "output": "natural language processing of user - generated content", "neg_sample": ["universal dependencies treebank of irish tweets is used for Task", "modern irish is a minority language lacking sufficient computational resources for the task of accurate automatic syntactic parsing of user - generated content such as tweets .", "although language technology for the irish language has been developing in recent years , these tools tend to perform poorly on user - generated content .", "as with other languages , the linguistic style observed in irish tweets differs , in terms of orthography , lexicon , and syntax , from that of standard texts more commonly used for the development of language models and parsers ."], "relation": "used for", "id": "2022.acl-long.473", "year": 2022, "rel_sent": "We release the first Universal Dependencies treebank of Irish tweets , facilitating natural language processing of user - generated content in Irish .", "forward": true, "src_ids": "2022.acl-long.473_3923"}
+{"input": "neural machine translation is done by using Task| context: the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time . however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) . a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature . nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus . however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains .", "entity": "neural machine translation", "output": "domain adaptation problems", "neg_sample": ["neural machine translation is done by using Task", "the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time .", "however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) .", "a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature .", "nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus .", "however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains ."], "relation": "used for", "id": "2022.eamt-1.3", "year": 2022, "rel_sent": "Domain Adaptation for Neural Machine Translation.", "forward": false, "src_ids": "2022.eamt-1.3_3924"}
+{"input": "neural machine translation is done by using Task| context: the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time . however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) . a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature . nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus . however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains .", "entity": "neural machine translation", "output": "domain adaptation problems", "neg_sample": ["neural machine translation is done by using Task", "the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time .", "however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) .", "a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature .", "nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus .", "however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains ."], "relation": "used for", "id": "2022.eamt-1.3", "year": 2022, "rel_sent": "This thesis focuses instead on more robust approaches to domain adaptation for NMT .", "forward": false, "src_ids": "2022.eamt-1.3_3925"}
+{"input": "domain adaptation problems is used for Task| context: however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) . a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature . nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus . however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains .", "entity": "domain adaptation problems", "output": "neural machine translation", "neg_sample": ["domain adaptation problems is used for Task", "however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) .", "a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature .", "nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus .", "however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains ."], "relation": "used for", "id": "2022.eamt-1.3", "year": 2022, "rel_sent": "Domain Adaptation for Neural Machine Translation.", "forward": true, "src_ids": "2022.eamt-1.3_3926"}
+{"input": "domain adaptation problems is used for Task| context: however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) . a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature . nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus . however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains .", "entity": "domain adaptation problems", "output": "neural machine translation", "neg_sample": ["domain adaptation problems is used for Task", "however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) .", "a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature .", "nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus .", "however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains ."], "relation": "used for", "id": "2022.eamt-1.3", "year": 2022, "rel_sent": "This thesis focuses instead on more robust approaches to domain adaptation for NMT .", "forward": true, "src_ids": "2022.eamt-1.3_3927"}
+{"input": "machine translation of gendered terms is done by using Method| context: the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time . however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) . a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature . nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus . however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains .", "entity": "machine translation of gendered terms", "output": "language representations", "neg_sample": ["machine translation of gendered terms is done by using Method", "the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time .", "however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) .", "a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature .", "nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus .", "however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains ."], "relation": "used for", "id": "2022.eamt-1.3", "year": 2022, "rel_sent": "adapting machine translation to a biomedical domain can also be used when making use of language representations beyond the surface - level , or when encouraging better machine translation of gendered terms .", "forward": false, "src_ids": "2022.eamt-1.3_3928"}
+{"input": "language representations is used for Task| context: the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time . however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) . a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature . nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus . however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains .", "entity": "language representations", "output": "machine translation of gendered terms", "neg_sample": ["language representations is used for Task", "the development of deep learning techniques has allowed neural machine translation ( nmt ) models to become extremely powerful , given sufficient training data and training time .", "however , such translation models struggle when translating text of a new or unfamiliar domain ( koehn and knowles , 2017 ) .", "a domain may be a well - defined topic , text of a specific provenance , text of unknown provenance with an identifiable vocabulary distribution , or language with some other stylometric feature .", "nmt models can achieve good translation performance on domain - specific data via simple tuning on a representative training corpus .", "however , such data - centric approaches have negative sideeffects , including over - fitting and brittleness on narrow - distribution samples and catastrophic forgetting of previously seen domains ."], "relation": "used for", "id": "2022.eamt-1.3", "year": 2022, "rel_sent": "adapting machine translation to a biomedical domain can also be used when making use of language representations beyond the surface - level , or when encouraging better machine translation of gendered terms .", "forward": true, "src_ids": "2022.eamt-1.3_3929"}
+{"input": "multiple - session dialogue datasets is used for Task| context: most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations . the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information .", "entity": "multiple - session dialogue datasets", "output": "model training", "neg_sample": ["multiple - session dialogue datasets is used for Task", "most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations .", "the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information ."], "relation": "used for", "id": "2022.findings-acl.207", "year": 2022, "rel_sent": "This LTM mechanism enables our system to accurately extract and continuously update long - term persona memory without requiring multiple - session dialogue datasets for model training .", "forward": true, "src_ids": "2022.findings-acl.207_3930"}
+{"input": "ltm mechanism is used for OtherScientificTerm| context: most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations . the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information .", "entity": "ltm mechanism", "output": "long - term persona memory", "neg_sample": ["ltm mechanism is used for OtherScientificTerm", "most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations .", "the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information ."], "relation": "used for", "id": "2022.findings-acl.207", "year": 2022, "rel_sent": "This LTM mechanism enables our system to accurately extract and continuously update long - term persona memory without requiring multiple - session dialogue datasets for model training .", "forward": true, "src_ids": "2022.findings-acl.207_3931"}
+{"input": "long - term persona memory is done by using Method| context: most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations . the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information .", "entity": "long - term persona memory", "output": "ltm mechanism", "neg_sample": ["long - term persona memory is done by using Method", "most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations .", "the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information ."], "relation": "used for", "id": "2022.findings-acl.207", "year": 2022, "rel_sent": "This LTM mechanism enables our system to accurately extract and continuously update long - term persona memory without requiring multiple - session dialogue datasets for model training .", "forward": false, "src_ids": "2022.findings-acl.207_3932"}
+{"input": "model training is done by using Material| context: most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations . the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information .", "entity": "model training", "output": "multiple - session dialogue datasets", "neg_sample": ["model training is done by using Material", "most of the open - domain dialogue models tend to perform poorly in the setting of long - term human - bot conversations .", "the possible reason is that they lack the capability of understanding and memorizing long - term dialogue history information ."], "relation": "used for", "id": "2022.findings-acl.207", "year": 2022, "rel_sent": "This LTM mechanism enables our system to accurately extract and continuously update long - term persona memory without requiring multiple - session dialogue datasets for model training .", "forward": false, "src_ids": "2022.findings-acl.207_3933"}
+{"input": "neuroscientific ideas is used for Task| context: information integration from different modalities is an active area of research . human beings and , in general , biological neural systems are quite adept at using a multitude of signals from different sensory perceptive fields to interact with the environment and each other . recent work in deep fusion models via neural networks has led to substantial improvements over unimodal approaches in areas like speech recognition , emotion recognition and analysis , captioning and image description .", "entity": "neuroscientific ideas", "output": "multisensory integration and processing", "neg_sample": ["neuroscientific ideas is used for Task", "information integration from different modalities is an active area of research .", "human beings and , in general , biological neural systems are quite adept at using a multitude of signals from different sensory perceptive fields to interact with the environment and each other .", "recent work in deep fusion models via neural networks has led to substantial improvements over unimodal approaches in areas like speech recognition , emotion recognition and analysis , captioning and image description ."], "relation": "used for", "id": "2022.acl-long.83", "year": 2022, "rel_sent": "However , such research has mostly focused on architectural changes allowing for fusion of different modalities while keeping the model complexity manageable . Inspired by neuroscientific ideas about multisensory integration and processing , we investigate the effect of introducing neural dependencies in the loss functions .", "forward": true, "src_ids": "2022.acl-long.83_3934"}
+{"input": "multisensory integration and processing is done by using OtherScientificTerm| context: information integration from different modalities is an active area of research . human beings and , in general , biological neural systems are quite adept at using a multitude of signals from different sensory perceptive fields to interact with the environment and each other . recent work in deep fusion models via neural networks has led to substantial improvements over unimodal approaches in areas like speech recognition , emotion recognition and analysis , captioning and image description .", "entity": "multisensory integration and processing", "output": "neuroscientific ideas", "neg_sample": ["multisensory integration and processing is done by using OtherScientificTerm", "information integration from different modalities is an active area of research .", "human beings and , in general , biological neural systems are quite adept at using a multitude of signals from different sensory perceptive fields to interact with the environment and each other .", "recent work in deep fusion models via neural networks has led to substantial improvements over unimodal approaches in areas like speech recognition , emotion recognition and analysis , captioning and image description ."], "relation": "used for", "id": "2022.acl-long.83", "year": 2022, "rel_sent": "However , such research has mostly focused on architectural changes allowing for fusion of different modalities while keeping the model complexity manageable . Inspired by neuroscientific ideas about multisensory integration and processing , we investigate the effect of introducing neural dependencies in the loss functions .", "forward": false, "src_ids": "2022.acl-long.83_3935"}
+{"input": "ptlm is used for Material| context: pretrained language models ( ptlms ) are typically learned over a large , static corpus and further fine - tuned for various downstream tasks . however , when deployed in the real world , a ptlm - based model must deal with data distributions that deviates from what the ptlm was initially trained on .", "entity": "ptlm", "output": "emerging data", "neg_sample": ["ptlm is used for Material", "pretrained language models ( ptlms ) are typically learned over a large , static corpus and further fine - tuned for various downstream tasks .", "however , when deployed in the real world , a ptlm - based model must deal with data distributions that deviates from what the ptlm was initially trained on ."], "relation": "used for", "id": "2022.bigscience-1.1", "year": 2022, "rel_sent": "In this paper , we study a lifelong language model pretraining challenge where a PTLM is continually updated so as to adapt to emerging data .", "forward": true, "src_ids": "2022.bigscience-1.1_3936"}
+{"input": "probing biomedical knowledge is done by using Method| context: knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) . despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored .", "entity": "probing biomedical knowledge", "output": "contrastive recipe", "neg_sample": ["probing biomedical knowledge is done by using Method", "knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) .", "despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored ."], "relation": "used for", "id": "2022.acl-long.329", "year": 2022, "rel_sent": "Rewire - then - Probe : A Contrastive Recipe for Probing Biomedical Knowledge of Pre - trained Language Models.", "forward": false, "src_ids": "2022.acl-long.329_3937"}
+{"input": "contrastive recipe is used for Task| context: knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) . despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored .", "entity": "contrastive recipe", "output": "probing biomedical knowledge", "neg_sample": ["contrastive recipe is used for Task", "knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) .", "despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored ."], "relation": "used for", "id": "2022.acl-long.329", "year": 2022, "rel_sent": "Rewire - then - Probe : A Contrastive Recipe for Probing Biomedical Knowledge of Pre - trained Language Models.", "forward": true, "src_ids": "2022.acl-long.329_3938"}
+{"input": "probing techniques is done by using Method| context: knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) . despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored .", "entity": "probing techniques", "output": "medlama", "neg_sample": ["probing techniques is done by using Method", "knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) .", "despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored ."], "relation": "used for", "id": "2022.acl-long.329", "year": 2022, "rel_sent": "We hope MedLAMA and Contrastive - Probe facilitate further developments of more suited probing techniques for this domain .", "forward": false, "src_ids": "2022.acl-long.329_3939"}
+{"input": "pre - trained language models is used for Task| context: knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) . despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored .", "entity": "pre - trained language models", "output": "probing tasks", "neg_sample": ["pre - trained language models is used for Task", "knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) .", "despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored ."], "relation": "used for", "id": "2022.acl-long.329", "year": 2022, "rel_sent": "While highlighting various sources of domain - specific challenges that amount to this underwhelming performance , we illustrate that the underlying PLMs have a higher potential for probing tasks .", "forward": true, "src_ids": "2022.acl-long.329_3940"}
+{"input": "contrastive - probe is used for Method| context: knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) . despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored .", "entity": "contrastive - probe", "output": "probing techniques", "neg_sample": ["contrastive - probe is used for Method", "knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) .", "despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored ."], "relation": "used for", "id": "2022.acl-long.329", "year": 2022, "rel_sent": "We hope MedLAMA and Contrastive - Probe facilitate further developments of more suited probing techniques for this domain .", "forward": true, "src_ids": "2022.acl-long.329_3941"}
+{"input": "medlama is used for Method| context: knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) . despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored .", "entity": "medlama", "output": "probing techniques", "neg_sample": ["medlama is used for Method", "knowledge probing is crucial for understanding the knowledge transfer mechanism behind the pre - trained language models ( plms ) .", "despite the growing progress of probing knowledge for plms in the general domain , specialised areas such as the biomedical domain are vastly under - explored ."], "relation": "used for", "id": "2022.acl-long.329", "year": 2022, "rel_sent": "We hope MedLAMA and Contrastive - Probe facilitate further developments of more suited probing techniques for this domain .", "forward": true, "src_ids": "2022.acl-long.329_3942"}
+{"input": "neural network based detector is used for Material| context: in this work , we focus on the problem of distinguishing a human written news article from a news article that is created by manipulating entities in a human written news article ( e.g. , replacing entities with factually incorrect entities ) . such manipulated articles can mislead the reader by posing as a human written news article .", "entity": "neural network based detector", "output": "manipulated news articles", "neg_sample": ["neural network based detector is used for Material", "in this work , we focus on the problem of distinguishing a human written news article from a news article that is created by manipulating entities in a human written news article ( e.g.", ", replacing entities with factually incorrect entities ) .", "such manipulated articles can mislead the reader by posing as a human written news article ."], "relation": "used for", "id": "2022.acl-short.10", "year": 2022, "rel_sent": "We propose a neural network based detector that detects manipulated news articles by reasoning about the facts mentioned in the article .", "forward": true, "src_ids": "2022.acl-short.10_3943"}
+{"input": "manipulated news articles is done by using Method| context: in this work , we focus on the problem of distinguishing a human written news article from a news article that is created by manipulating entities in a human written news article ( e.g. , replacing entities with factually incorrect entities ) . such manipulated articles can mislead the reader by posing as a human written news article .", "entity": "manipulated news articles", "output": "neural network based detector", "neg_sample": ["manipulated news articles is done by using Method", "in this work , we focus on the problem of distinguishing a human written news article from a news article that is created by manipulating entities in a human written news article ( e.g.", ", replacing entities with factually incorrect entities ) .", "such manipulated articles can mislead the reader by posing as a human written news article ."], "relation": "used for", "id": "2022.acl-short.10", "year": 2022, "rel_sent": "We propose a neural network based detector that detects manipulated news articles by reasoning about the facts mentioned in the article .", "forward": false, "src_ids": "2022.acl-short.10_3944"}
+{"input": "automatic dialog evaluation is done by using OtherScientificTerm| context: accurate automatic evaluation metrics for open - domain dialogs are in high demand . existing model - based metrics for system response evaluation are trained on human annotated data , which is cumbersome to collect .", "entity": "automatic dialog evaluation", "output": "user sentiment", "neg_sample": ["automatic dialog evaluation is done by using OtherScientificTerm", "accurate automatic evaluation metrics for open - domain dialogs are in high demand .", "existing model - based metrics for system response evaluation are trained on human annotated data , which is cumbersome to collect ."], "relation": "used for", "id": "2022.findings-acl.331", "year": 2022, "rel_sent": "What is wrong with you ? : Leveraging User Sentiment for Automatic Dialog Evaluation.", "forward": false, "src_ids": "2022.findings-acl.331_3945"}
+{"input": "user sentiment is used for Task| context: accurate automatic evaluation metrics for open - domain dialogs are in high demand . existing model - based metrics for system response evaluation are trained on human annotated data , which is cumbersome to collect .", "entity": "user sentiment", "output": "automatic dialog evaluation", "neg_sample": ["user sentiment is used for Task", "accurate automatic evaluation metrics for open - domain dialogs are in high demand .", "existing model - based metrics for system response evaluation are trained on human annotated data , which is cumbersome to collect ."], "relation": "used for", "id": "2022.findings-acl.331", "year": 2022, "rel_sent": "What is wrong with you ? : Leveraging User Sentiment for Automatic Dialog Evaluation.", "forward": true, "src_ids": "2022.findings-acl.331_3946"}
+{"input": "xlm - r is done by using Method| context: static and contextual multilingual embeddings have complementary strengths . static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages .", "entity": "xlm - r", "output": "continued pre - training approach", "neg_sample": ["xlm - r is done by using Method", "static and contextual multilingual embeddings have complementary strengths .", "static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages ."], "relation": "used for", "id": "2022.findings-acl.182", "year": 2022, "rel_sent": "Then we apply a novel continued pre - training approach to XLM - R , leveraging the high quality alignment of our static embeddings to better align the representation space of XLM - R. We show positive results for multiple complex semantic tasks .", "forward": false, "src_ids": "2022.findings-acl.182_3947"}
+{"input": "multilingual representations is done by using Method| context: static and contextual multilingual embeddings have complementary strengths . static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages .", "entity": "multilingual representations", "output": "static and contextual models", "neg_sample": ["multilingual representations is done by using Method", "static and contextual multilingual embeddings have complementary strengths .", "static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages ."], "relation": "used for", "id": "2022.findings-acl.182", "year": 2022, "rel_sent": "We combine the strengths of static and contextual models to improve multilingual representations .", "forward": false, "src_ids": "2022.findings-acl.182_3948"}
+{"input": "static and contextual models is used for Method| context: static and contextual multilingual embeddings have complementary strengths . static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages .", "entity": "static and contextual models", "output": "multilingual representations", "neg_sample": ["static and contextual models is used for Method", "static and contextual multilingual embeddings have complementary strengths .", "static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages ."], "relation": "used for", "id": "2022.findings-acl.182", "year": 2022, "rel_sent": "We combine the strengths of static and contextual models to improve multilingual representations .", "forward": true, "src_ids": "2022.findings-acl.182_3949"}
+{"input": "continued pre - training approach is used for Method| context: static and contextual multilingual embeddings have complementary strengths . static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages .", "entity": "continued pre - training approach", "output": "xlm - r", "neg_sample": ["continued pre - training approach is used for Method", "static and contextual multilingual embeddings have complementary strengths .", "static embeddings , while less expressive than contextual language models , can be more straightforwardly aligned across multiple languages ."], "relation": "used for", "id": "2022.findings-acl.182", "year": 2022, "rel_sent": "Then we apply a novel continued pre - training approach to XLM - R , leveraging the high quality alignment of our static embeddings to better align the representation space of XLM - R. We show positive results for multiple complex semantic tasks .", "forward": true, "src_ids": "2022.findings-acl.182_3950"}
+{"input": "chinese word segmentation ( cws ) is done by using Method| context: recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) . however , these methods rely heavily on such additional information mentioned above and focus less on the model itself .", "entity": "chinese word segmentation ( cws )", "output": "weighted self distillation", "neg_sample": ["chinese word segmentation ( cws ) is done by using Method", "recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) .", "however , these methods rely heavily on such additional information mentioned above and focus less on the model itself ."], "relation": "used for", "id": "2022.findings-acl.139", "year": 2022, "rel_sent": "Weighted self Distillation for Chinese word segmentation.", "forward": false, "src_ids": "2022.findings-acl.139_3951"}
+{"input": "chinese word segmentation ( weidc ) is done by using Method| context: recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) . however , these methods rely heavily on such additional information mentioned above and focus less on the model itself .", "entity": "chinese word segmentation ( weidc )", "output": "weighted self distillation", "neg_sample": ["chinese word segmentation ( weidc ) is done by using Method", "recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) .", "however , these methods rely heavily on such additional information mentioned above and focus less on the model itself ."], "relation": "used for", "id": "2022.findings-acl.139", "year": 2022, "rel_sent": "We thus propose a novel neural framework , named Weighted self Distillation for Chinese word segmentation ( WeiDC ) .", "forward": false, "src_ids": "2022.findings-acl.139_3952"}
+{"input": "weighted self distillation is used for Task| context: however , these methods rely heavily on such additional information mentioned above and focus less on the model itself .", "entity": "weighted self distillation", "output": "chinese word segmentation ( cws )", "neg_sample": ["weighted self distillation is used for Task", "however , these methods rely heavily on such additional information mentioned above and focus less on the model itself ."], "relation": "used for", "id": "2022.findings-acl.139", "year": 2022, "rel_sent": "Weighted self Distillation for Chinese word segmentation.", "forward": true, "src_ids": "2022.findings-acl.139_3953"}
+{"input": "weighted self distillation is used for Task| context: recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) . however , these methods rely heavily on such additional information mentioned above and focus less on the model itself .", "entity": "weighted self distillation", "output": "chinese word segmentation ( weidc )", "neg_sample": ["weighted self distillation is used for Task", "recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) .", "however , these methods rely heavily on such additional information mentioned above and focus less on the model itself ."], "relation": "used for", "id": "2022.findings-acl.139", "year": 2022, "rel_sent": "We thus propose a novel neural framework , named Weighted self Distillation for Chinese word segmentation ( WeiDC ) .", "forward": true, "src_ids": "2022.findings-acl.139_3954"}
+{"input": "contextual knowledge is done by using OtherScientificTerm| context: recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) . however , these methods rely heavily on such additional information mentioned above and focus less on the model itself .", "entity": "contextual knowledge", "output": "character features", "neg_sample": ["contextual knowledge is done by using OtherScientificTerm", "recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) .", "however , these methods rely heavily on such additional information mentioned above and focus less on the model itself ."], "relation": "used for", "id": "2022.findings-acl.139", "year": 2022, "rel_sent": "Experiment results show that WeiDC can make use of character features to learn contextual knowledge and successfully achieve state - of - the - art or competitive performance in terms of strictly closed test settings on SIGHAN Bakeoff benchmark datasets .", "forward": false, "src_ids": "2022.findings-acl.139_3955"}
+{"input": "character features is used for OtherScientificTerm| context: recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) . however , these methods rely heavily on such additional information mentioned above and focus less on the model itself .", "entity": "character features", "output": "contextual knowledge", "neg_sample": ["character features is used for OtherScientificTerm", "recent researches show that multi - criteria resources and n - gram features are beneficial to chinese word segmentation ( cws ) .", "however , these methods rely heavily on such additional information mentioned above and focus less on the model itself ."], "relation": "used for", "id": "2022.findings-acl.139", "year": 2022, "rel_sent": "Experiment results show that WeiDC can make use of character features to learn contextual knowledge and successfully achieve state - of - the - art or competitive performance in terms of strictly closed test settings on SIGHAN Bakeoff benchmark datasets .", "forward": true, "src_ids": "2022.findings-acl.139_3956"}
+{"input": "abstractive summarization is done by using Task| context: previous length - controllable summarization models mostly control lengths at the decoding stage , whereas the encoding or the selection of information from the source document is not sensitive to the designed length . they also tend to generate summaries as long as those in the training data .", "entity": "abstractive summarization", "output": "length control", "neg_sample": ["abstractive summarization is done by using Task", "previous length - controllable summarization models mostly control lengths at the decoding stage , whereas the encoding or the selection of information from the source document is not sensitive to the designed length .", "they also tend to generate summaries as long as those in the training data ."], "relation": "used for", "id": "2022.acl-long.474", "year": 2022, "rel_sent": "Length Control in Abstractive Summarization by Pretraining Information Selection.", "forward": false, "src_ids": "2022.acl-long.474_3957"}
+{"input": "length control is used for Task| context: previous length - controllable summarization models mostly control lengths at the decoding stage , whereas the encoding or the selection of information from the source document is not sensitive to the designed length . they also tend to generate summaries as long as those in the training data .", "entity": "length control", "output": "abstractive summarization", "neg_sample": ["length control is used for Task", "previous length - controllable summarization models mostly control lengths at the decoding stage , whereas the encoding or the selection of information from the source document is not sensitive to the designed length .", "they also tend to generate summaries as long as those in the training data ."], "relation": "used for", "id": "2022.acl-long.474", "year": 2022, "rel_sent": "Length Control in Abstractive Summarization by Pretraining Information Selection.", "forward": true, "src_ids": "2022.acl-long.474_3958"}
+{"input": "neural machine translation is done by using Method| context: restricted machine translation incorporates human prior knowledge into translation . it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios . existing work typically imposes constraints on beam search decoding . although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality .", "entity": "neural machine translation", "output": "training framework", "neg_sample": ["neural machine translation is done by using Method", "restricted machine translation incorporates human prior knowledge into translation .", "it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios .", "existing work typically imposes constraints on beam search decoding .", "although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality ."], "relation": "used for", "id": "2022.acl-srw.18", "year": 2022, "rel_sent": "Restricted or Not : A General Training Framework for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-srw.18_3959"}
+{"input": "training framework is used for Task| context: restricted machine translation incorporates human prior knowledge into translation . it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios . existing work typically imposes constraints on beam search decoding . although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality .", "entity": "training framework", "output": "neural machine translation", "neg_sample": ["training framework is used for Task", "restricted machine translation incorporates human prior knowledge into translation .", "it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios .", "existing work typically imposes constraints on beam search decoding .", "although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality ."], "relation": "used for", "id": "2022.acl-srw.18", "year": 2022, "rel_sent": "Restricted or Not : A General Training Framework for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-srw.18_3960"}
+{"input": "unrestricted and restricted translation is done by using Method| context: restricted machine translation incorporates human prior knowledge into translation . it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios . existing work typically imposes constraints on beam search decoding . although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality .", "entity": "unrestricted and restricted translation", "output": "general training framework", "neg_sample": ["unrestricted and restricted translation is done by using Method", "restricted machine translation incorporates human prior knowledge into translation .", "it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios .", "existing work typically imposes constraints on beam search decoding .", "although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality ."], "relation": "used for", "id": "2022.acl-srw.18", "year": 2022, "rel_sent": "In this paper , we propose a general training framework that allows a model to simultaneously support both unrestricted and restricted translation by adopting an additional auxiliary training process without constraining the decoding process .", "forward": false, "src_ids": "2022.acl-srw.18_3961"}
+{"input": "general training framework is used for Task| context: restricted machine translation incorporates human prior knowledge into translation . it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios . existing work typically imposes constraints on beam search decoding . although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality .", "entity": "general training framework", "output": "unrestricted and restricted translation", "neg_sample": ["general training framework is used for Task", "restricted machine translation incorporates human prior knowledge into translation .", "it restricts the flexibility of the translation to satisfy the demands of translation in specific scenarios .", "existing work typically imposes constraints on beam search decoding .", "although this can satisfy the requirements overall , it usually requires a larger beam size and far longer decoding time than unrestricted translation , which limits the concurrent processing ability of the translation model in deployment , and thus its practicality ."], "relation": "used for", "id": "2022.acl-srw.18", "year": 2022, "rel_sent": "In this paper , we propose a general training framework that allows a model to simultaneously support both unrestricted and restricted translation by adopting an additional auxiliary training process without constraining the decoding process .", "forward": true, "src_ids": "2022.acl-srw.18_3962"}
+{"input": "symbolic reasoning is done by using Method| context: numerical reasoning over hybrid data containing both textual and tabular content ( e.g. , financial reports ) has recently attracted much attention in the nlp community . however , existing question answering ( qa ) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi - step numerical reasoning across multiple hierarchical tables . multihiertt is built from a wealth of financial reports and has the following unique characteristics : 1 ) each document contain multiple tables and longer unstructured texts ; 2 ) most of tables contained are hierarchical ; 3 ) the reasoning process required for each question is more complex and challenging than existing benchmarks ; and 4 ) fine - grained annotations of reasoning processes and supporting facts are provided to reveal complex numerical reasoning .", "entity": "symbolic reasoning", "output": "reasoning module", "neg_sample": ["symbolic reasoning is done by using Method", "numerical reasoning over hybrid data containing both textual and tabular content ( e.g.", ", financial reports ) has recently attracted much attention in the nlp community .", "however , existing question answering ( qa ) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi - step numerical reasoning across multiple hierarchical tables .", "multihiertt is built from a wealth of financial reports and has the following unique characteristics : 1 ) each document contain multiple tables and longer unstructured texts ; 2 ) most of tables contained are hierarchical ; 3 ) the reasoning process required for each question is more complex and challenging than existing benchmarks ; and 4 ) fine - grained annotations of reasoning processes and supporting facts are provided to reveal complex numerical reasoning ."], "relation": "used for", "id": "2022.acl-long.454", "year": 2022, "rel_sent": "We further introduce a novel QA model termed MT2Net , which first applies facts retrieving to extract relevant supporting facts from both tables and text and then uses a reasoning module to perform symbolic reasoning over retrieved facts .", "forward": false, "src_ids": "2022.acl-long.454_3963"}
+{"input": "reasoning module is used for Method| context: numerical reasoning over hybrid data containing both textual and tabular content ( e.g. , financial reports ) has recently attracted much attention in the nlp community . however , existing question answering ( qa ) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi - step numerical reasoning across multiple hierarchical tables . multihiertt is built from a wealth of financial reports and has the following unique characteristics : 1 ) each document contain multiple tables and longer unstructured texts ; 2 ) most of tables contained are hierarchical ; 3 ) the reasoning process required for each question is more complex and challenging than existing benchmarks ; and 4 ) fine - grained annotations of reasoning processes and supporting facts are provided to reveal complex numerical reasoning .", "entity": "reasoning module", "output": "symbolic reasoning", "neg_sample": ["reasoning module is used for Method", "numerical reasoning over hybrid data containing both textual and tabular content ( e.g.", ", financial reports ) has recently attracted much attention in the nlp community .", "however , existing question answering ( qa ) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi - step numerical reasoning across multiple hierarchical tables .", "multihiertt is built from a wealth of financial reports and has the following unique characteristics : 1 ) each document contain multiple tables and longer unstructured texts ; 2 ) most of tables contained are hierarchical ; 3 ) the reasoning process required for each question is more complex and challenging than existing benchmarks ; and 4 ) fine - grained annotations of reasoning processes and supporting facts are provided to reveal complex numerical reasoning ."], "relation": "used for", "id": "2022.acl-long.454", "year": 2022, "rel_sent": "We further introduce a novel QA model termed MT2Net , which first applies facts retrieving to extract relevant supporting facts from both tables and text and then uses a reasoning module to perform symbolic reasoning over retrieved facts .", "forward": true, "src_ids": "2022.acl-long.454_3964"}
+{"input": "simultaneous machine translation is done by using Method| context: simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available . it is difficult due to limited context and word order difference between languages . existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality . however , the long - distance reordering would make the simulmt models learn translation mistakenly . specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read . this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon .", "entity": "simultaneous machine translation", "output": "anticipation - free training", "neg_sample": ["simultaneous machine translation is done by using Method", "simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available .", "it is difficult due to limited context and word order difference between languages .", "existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality .", "however , the long - distance reordering would make the simulmt models learn translation mistakenly .", "specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read .", "this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon ."], "relation": "used for", "id": "2022.iwslt-1.5", "year": 2022, "rel_sent": "Anticipation - Free Training for Simultaneous Machine Translation.", "forward": false, "src_ids": "2022.iwslt-1.5_3965"}
+{"input": "anticipation - free training is used for Task| context: it is difficult due to limited context and word order difference between languages . existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality . however , the long - distance reordering would make the simulmt models learn translation mistakenly . specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read . this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon .", "entity": "anticipation - free training", "output": "simultaneous machine translation", "neg_sample": ["anticipation - free training is used for Task", "it is difficult due to limited context and word order difference between languages .", "existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality .", "however , the long - distance reordering would make the simulmt models learn translation mistakenly .", "specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read .", "this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon ."], "relation": "used for", "id": "2022.iwslt-1.5", "year": 2022, "rel_sent": "Anticipation - Free Training for Simultaneous Machine Translation.", "forward": true, "src_ids": "2022.iwslt-1.5_3966"}
+{"input": "hidden states is done by using OtherScientificTerm| context: simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available . it is difficult due to limited context and word order difference between languages . existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality . however , the long - distance reordering would make the simulmt models learn translation mistakenly . specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read . this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon .", "entity": "hidden states", "output": "auxiliary sorting network", "neg_sample": ["hidden states is done by using OtherScientificTerm", "simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available .", "it is difficult due to limited context and word order difference between languages .", "existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality .", "however , the long - distance reordering would make the simulmt models learn translation mistakenly .", "specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read .", "this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon ."], "relation": "used for", "id": "2022.iwslt-1.5", "year": 2022, "rel_sent": "The ASN rearranges the hidden states to match the order in the target language , so that the SimulMT model could learn to translate more reasonably .", "forward": false, "src_ids": "2022.iwslt-1.5_3967"}
+{"input": "streaming is done by using OtherScientificTerm| context: simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available . it is difficult due to limited context and word order difference between languages . existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality . however , the long - distance reordering would make the simulmt models learn translation mistakenly . specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read . this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon .", "entity": "streaming", "output": "auxiliary sorting network", "neg_sample": ["streaming is done by using OtherScientificTerm", "simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available .", "it is difficult due to limited context and word order difference between languages .", "existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality .", "however , the long - distance reordering would make the simulmt models learn translation mistakenly .", "specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read .", "this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon ."], "relation": "used for", "id": "2022.iwslt-1.5", "year": 2022, "rel_sent": "During inference , ASN is removed to achieve streaming .", "forward": false, "src_ids": "2022.iwslt-1.5_3968"}
+{"input": "auxiliary sorting network is used for OtherScientificTerm| context: simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available . it is difficult due to limited context and word order difference between languages . existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality . however , the long - distance reordering would make the simulmt models learn translation mistakenly . specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read . this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon .", "entity": "auxiliary sorting network", "output": "hidden states", "neg_sample": ["auxiliary sorting network is used for OtherScientificTerm", "simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available .", "it is difficult due to limited context and word order difference between languages .", "existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality .", "however , the long - distance reordering would make the simulmt models learn translation mistakenly .", "specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read .", "this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon ."], "relation": "used for", "id": "2022.iwslt-1.5", "year": 2022, "rel_sent": "The ASN rearranges the hidden states to match the order in the target language , so that the SimulMT model could learn to translate more reasonably .", "forward": true, "src_ids": "2022.iwslt-1.5_3969"}
+{"input": "auxiliary sorting network is used for Task| context: simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available . it is difficult due to limited context and word order difference between languages . existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality . however , the long - distance reordering would make the simulmt models learn translation mistakenly . specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read . this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon .", "entity": "auxiliary sorting network", "output": "streaming", "neg_sample": ["auxiliary sorting network is used for Task", "simultaneous machine translation ( simulmt ) speeds up the translation process by starting to translate before the source sentence is completely available .", "it is difficult due to limited context and word order difference between languages .", "existing methods increase latency or introduce adaptive read - write policies for simulmt models to handle local reordering and improve translation quality .", "however , the long - distance reordering would make the simulmt models learn translation mistakenly .", "specifically , the model may be forced to predict target tokens when the corresponding source tokens have not been read .", "this leads to aggressive anticipation during inference , resulting in the hallucination phenomenon ."], "relation": "used for", "id": "2022.iwslt-1.5", "year": 2022, "rel_sent": "During inference , ASN is removed to achieve streaming .", "forward": true, "src_ids": "2022.iwslt-1.5_3970"}
+{"input": "multilingual style transfer is done by using Method| context: detoxification is a task of generating text in polite style while preserving meaning and fluency of the original toxic text . existing detoxification methods are monolingual i.e. designed to work in one exact language .", "entity": "multilingual style transfer", "output": "multilingual models", "neg_sample": ["multilingual style transfer is done by using Method", "detoxification is a task of generating text in polite style while preserving meaning and fluency of the original toxic text .", "existing detoxification methods are monolingual i.e.", "designed to work in one exact language ."], "relation": "used for", "id": "2022.acl-srw.26", "year": 2022, "rel_sent": "Experiments show that multilingual models are capable of performing multilingual style transfer .", "forward": false, "src_ids": "2022.acl-srw.26_3971"}
+{"input": "multilingual models is used for Task| context: detoxification is a task of generating text in polite style while preserving meaning and fluency of the original toxic text . existing detoxification methods are monolingual i.e. designed to work in one exact language .", "entity": "multilingual models", "output": "multilingual style transfer", "neg_sample": ["multilingual models is used for Task", "detoxification is a task of generating text in polite style while preserving meaning and fluency of the original toxic text .", "existing detoxification methods are monolingual i.e.", "designed to work in one exact language ."], "relation": "used for", "id": "2022.acl-srw.26", "year": 2022, "rel_sent": "Experiments show that multilingual models are capable of performing multilingual style transfer .", "forward": true, "src_ids": "2022.acl-srw.26_3972"}
+{"input": "large language models is used for Task| context: designed to work in one exact language .", "entity": "large language models", "output": "detoxification", "neg_sample": ["large language models is used for Task", "designed to work in one exact language ."], "relation": "used for", "id": "2022.acl-srw.26", "year": 2022, "rel_sent": "Unlike previous works we aim to make large language models able to perform detoxification without direct fine - tuning in a given language .", "forward": true, "src_ids": "2022.acl-srw.26_3973"}
+{"input": "detoxification is done by using Method| context: detoxification is a task of generating text in polite style while preserving meaning and fluency of the original toxic text . existing detoxification methods are monolingual i.e. designed to work in one exact language .", "entity": "detoxification", "output": "large language models", "neg_sample": ["detoxification is done by using Method", "detoxification is a task of generating text in polite style while preserving meaning and fluency of the original toxic text .", "existing detoxification methods are monolingual i.e.", "designed to work in one exact language ."], "relation": "used for", "id": "2022.acl-srw.26", "year": 2022, "rel_sent": "Unlike previous works we aim to make large language models able to perform detoxification without direct fine - tuning in a given language .", "forward": false, "src_ids": "2022.acl-srw.26_3974"}
+{"input": "multi - level product issue identification problem is done by using Method| context: in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews . the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services .", "entity": "multi - level product issue identification problem", "output": "seq2seq language generation", "neg_sample": ["multi - level product issue identification problem is done by using Method", "in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews .", "the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services ."], "relation": "used for", "id": "2022.ecnlp-1.3", "year": 2022, "rel_sent": "Leveraging Seq2seq Language Generation for Multi - level Product Issue Identification.", "forward": false, "src_ids": "2022.ecnlp-1.3_3975"}
+{"input": "seq2seq language generation is used for Task| context: in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews . the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services .", "entity": "seq2seq language generation", "output": "multi - level product issue identification problem", "neg_sample": ["seq2seq language generation is used for Task", "in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews .", "the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services ."], "relation": "used for", "id": "2022.ecnlp-1.3", "year": 2022, "rel_sent": "Leveraging Seq2seq Language Generation for Multi - level Product Issue Identification.", "forward": true, "src_ids": "2022.ecnlp-1.3_3976"}
+{"input": "product issues is done by using Method| context: in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews . the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services .", "entity": "product issues", "output": "machine learning approach", "neg_sample": ["product issues is done by using Method", "in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews .", "the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services ."], "relation": "used for", "id": "2022.ecnlp-1.3", "year": 2022, "rel_sent": "To harness such information to better serve customers , in this paper , we created a machine learning approach to automatically identify product issues and uncover root causes from the customer feedback text .", "forward": false, "src_ids": "2022.ecnlp-1.3_3977"}
+{"input": "machine learning approach is used for OtherScientificTerm| context: in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews . the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services .", "entity": "machine learning approach", "output": "product issues", "neg_sample": ["machine learning approach is used for OtherScientificTerm", "in a leading e - commerce business , we receive hundreds of millions of customer feedback from different text communication channels such as product reviews .", "the feedback can contain rich information regarding customers ' dissatisfaction in the quality of goods and services ."], "relation": "used for", "id": "2022.ecnlp-1.3", "year": 2022, "rel_sent": "To harness such information to better serve customers , in this paper , we created a machine learning approach to automatically identify product issues and uncover root causes from the customer feedback text .", "forward": true, "src_ids": "2022.ecnlp-1.3_3978"}
+{"input": "answerable queries is done by using OtherScientificTerm| context: existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e. , answerable . however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e. , unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image . a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding .", "entity": "answerable queries", "output": "ground - truth regions", "neg_sample": ["answerable queries is done by using OtherScientificTerm", "existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e.", ", answerable .", "however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e.", ", unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image .", "a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding ."], "relation": "used for", "id": "2022.acl-srw.22", "year": 2022, "rel_sent": "The model is then trained to ground to ground - truth regions for answerable queries and pseudo regions for unanswerable queries .", "forward": false, "src_ids": "2022.acl-srw.22_3979"}
+{"input": "ground - truth regions is used for OtherScientificTerm| context: existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e. , answerable . however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e. , unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image . a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding .", "entity": "ground - truth regions", "output": "answerable queries", "neg_sample": ["ground - truth regions is used for OtherScientificTerm", "existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e.", ", answerable .", "however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e.", ", unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image .", "a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding ."], "relation": "used for", "id": "2022.acl-srw.22", "year": 2022, "rel_sent": "The model is then trained to ground to ground - truth regions for answerable queries and pseudo regions for unanswerable queries .", "forward": true, "src_ids": "2022.acl-srw.22_3980"}
+{"input": "unanswerable queries is done by using OtherScientificTerm| context: existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e. , answerable . however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e. , unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image . a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding .", "entity": "unanswerable queries", "output": "pseudo regions", "neg_sample": ["unanswerable queries is done by using OtherScientificTerm", "existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e.", ", answerable .", "however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e.", ", unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image .", "a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding ."], "relation": "used for", "id": "2022.acl-srw.22", "year": 2022, "rel_sent": "The model is then trained to ground to ground - truth regions for answerable queries and pseudo regions for unanswerable queries .", "forward": false, "src_ids": "2022.acl-srw.22_3981"}
+{"input": "pseudo regions is used for Material| context: existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e. , answerable . however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e. , unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image . a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding .", "entity": "pseudo regions", "output": "unanswerable queries", "neg_sample": ["pseudo regions is used for Material", "existing visual grounding datasets are artificially made , where every query regarding an entity must be able to be grounded to a corresponding image region , i.e.", ", answerable .", "however , in real - world multimedia data such as news articles and social media , many entities in the text can not be grounded to the image , i.e.", ", unanswerable , due to the fact that the text is unnecessarily directly describing the accompanying image .", "a robust visual grounding model should be able toflexibly deal with both answerable and unanswerable visual grounding ."], "relation": "used for", "id": "2022.acl-srw.22", "year": 2022, "rel_sent": "The model is then trained to ground to ground - truth regions for answerable queries and pseudo regions for unanswerable queries .", "forward": true, "src_ids": "2022.acl-srw.22_3982"}
+{"input": "few - shot ner is done by using Method| context: few - shot ner needs to effectively capture information from limited instances and transfer useful knowledge from external resources .", "entity": "few - shot ner", "output": "self - describing mechanism", "neg_sample": ["few - shot ner is done by using Method", "few - shot ner needs to effectively capture information from limited instances and transfer useful knowledge from external resources ."], "relation": "used for", "id": "2022.acl-long.392", "year": 2022, "rel_sent": "In this paper , we propose a self - describing mechanism for few - shot NER , which can effectively leverage illustrative instances and precisely transfer knowledge from external resources by describing both entity types and mentions using a universal concept set .", "forward": false, "src_ids": "2022.acl-long.392_3983"}
+{"input": "representation techniques is used for Method| context: we address the problem of learning fixed - length vector representations of characters in novels . recent advances in word embeddings have proven successful in learning entity representations from short texts , but fall short on longer documents because they do not capture full book - level information .", "entity": "representation techniques", "output": "character representations", "neg_sample": ["representation techniques is used for Method", "we address the problem of learning fixed - length vector representations of characters in novels .", "recent advances in word embeddings have proven successful in learning entity representations from short texts , but fall short on longer documents because they do not capture full book - level information ."], "relation": "used for", "id": "2022.findings-acl.81", "year": 2022, "rel_sent": "We show that our representation techniques combined with text - based embeddings lead to the best character representations , outperforming text - based embeddings in four tasks .", "forward": true, "src_ids": "2022.findings-acl.81_3984"}
+{"input": "text - based embeddings is used for Method| context: we address the problem of learning fixed - length vector representations of characters in novels . recent advances in word embeddings have proven successful in learning entity representations from short texts , but fall short on longer documents because they do not capture full book - level information .", "entity": "text - based embeddings", "output": "character representations", "neg_sample": ["text - based embeddings is used for Method", "we address the problem of learning fixed - length vector representations of characters in novels .", "recent advances in word embeddings have proven successful in learning entity representations from short texts , but fall short on longer documents because they do not capture full book - level information ."], "relation": "used for", "id": "2022.findings-acl.81", "year": 2022, "rel_sent": "We show that our representation techniques combined with text - based embeddings lead to the best character representations , outperforming text - based embeddings in four tasks .", "forward": true, "src_ids": "2022.findings-acl.81_3985"}
+{"input": "character representations is done by using Method| context: we address the problem of learning fixed - length vector representations of characters in novels . recent advances in word embeddings have proven successful in learning entity representations from short texts , but fall short on longer documents because they do not capture full book - level information .", "entity": "character representations", "output": "text - based embeddings", "neg_sample": ["character representations is done by using Method", "we address the problem of learning fixed - length vector representations of characters in novels .", "recent advances in word embeddings have proven successful in learning entity representations from short texts , but fall short on longer documents because they do not capture full book - level information ."], "relation": "used for", "id": "2022.findings-acl.81", "year": 2022, "rel_sent": "We show that our representation techniques combined with text - based embeddings lead to the best character representations , outperforming text - based embeddings in four tasks .", "forward": false, "src_ids": "2022.findings-acl.81_3986"}
+{"input": "table pretraining is done by using OtherScientificTerm| context: tables store rich numerical data , but numerical reasoning over tables is still a challenge . in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables .", "entity": "table pretraining", "output": "formulas", "neg_sample": ["table pretraining is done by using OtherScientificTerm", "tables store rich numerical data , but numerical reasoning over tables is still a challenge .", "in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables ."], "relation": "used for", "id": "2022.acl-long.82", "year": 2022, "rel_sent": "FORTAP outperforms state - of - the - art methods by large margins on three representative datasets of formula prediction , question answering , and cell type classification , showing the great potential of leveraging formulas for table pretraining .", "forward": false, "src_ids": "2022.acl-long.82_3987"}
+{"input": "table pretraining is done by using OtherScientificTerm| context: tables store rich numerical data , but numerical reasoning over tables is still a challenge .", "entity": "table pretraining", "output": "spreadsheet formulas", "neg_sample": ["table pretraining is done by using OtherScientificTerm", "tables store rich numerical data , but numerical reasoning over tables is still a challenge ."], "relation": "used for", "id": "2022.acl-long.82", "year": 2022, "rel_sent": "Considering large amounts of spreadsheets available on the web , we propose FORTAP , the first exploration to leverage spreadsheet formulas for table pretraining .", "forward": false, "src_ids": "2022.acl-long.82_3988"}
+{"input": "spreadsheet formulas is used for Task| context: tables store rich numerical data , but numerical reasoning over tables is still a challenge . in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables .", "entity": "spreadsheet formulas", "output": "table pretraining", "neg_sample": ["spreadsheet formulas is used for Task", "tables store rich numerical data , but numerical reasoning over tables is still a challenge .", "in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables ."], "relation": "used for", "id": "2022.acl-long.82", "year": 2022, "rel_sent": "Considering large amounts of spreadsheets available on the web , we propose FORTAP , the first exploration to leverage spreadsheet formulas for table pretraining .", "forward": true, "src_ids": "2022.acl-long.82_3989"}
+{"input": "formulas is used for Task| context: tables store rich numerical data , but numerical reasoning over tables is still a challenge . in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables .", "entity": "formulas", "output": "table pretraining", "neg_sample": ["formulas is used for Task", "tables store rich numerical data , but numerical reasoning over tables is still a challenge .", "in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables ."], "relation": "used for", "id": "2022.acl-long.82", "year": 2022, "rel_sent": "FORTAP outperforms state - of - the - art methods by large margins on three representative datasets of formula prediction , question answering , and cell type classification , showing the great potential of leveraging formulas for table pretraining .", "forward": true, "src_ids": "2022.acl-long.82_3990"}
+{"input": "spreadsheet table pretraining is done by using Method| context: tables store rich numerical data , but numerical reasoning over tables is still a challenge . in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables .", "entity": "spreadsheet table pretraining", "output": "transformer - based method", "neg_sample": ["spreadsheet table pretraining is done by using Method", "tables store rich numerical data , but numerical reasoning over tables is still a challenge .", "in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables ."], "relation": "used for", "id": "2022.acl-long.82", "year": 2022, "rel_sent": "While our proposed objectives are generic for encoders , to better capture spreadsheet table layouts and structures , FORTAP is built upon TUTA , the first transformer - based method for spreadsheet table pretraining with tree attention .", "forward": false, "src_ids": "2022.acl-long.82_3991"}
+{"input": "transformer - based method is used for Task| context: tables store rich numerical data , but numerical reasoning over tables is still a challenge . in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables .", "entity": "transformer - based method", "output": "spreadsheet table pretraining", "neg_sample": ["transformer - based method is used for Task", "tables store rich numerical data , but numerical reasoning over tables is still a challenge .", "in this paper , we find that the spreadsheet formula , a commonly used language to perform computations on numerical values in spreadsheets , is a valuable supervision for numerical reasoning in tables ."], "relation": "used for", "id": "2022.acl-long.82", "year": 2022, "rel_sent": "While our proposed objectives are generic for encoders , to better capture spreadsheet table layouts and structures , FORTAP is built upon TUTA , the first transformer - based method for spreadsheet table pretraining with tree attention .", "forward": true, "src_ids": "2022.acl-long.82_3992"}
+{"input": "oie evaluation is done by using Generic| context: intrinsic evaluations of oie systems are carried out either manually - with human evaluators judging the correctness of extractions - or automatically , on standardized benchmarks . the latter , while much more cost - effective , is less reliable , primarily because of the incompleteness of the existing oie benchmarks : the ground truth extractions do not include all acceptable variants of the same fact , leading to unreliable assessment of the models ' performance . moreover , the existing oie benchmarks are available for english only .", "entity": "oie evaluation", "output": "benchmark variants", "neg_sample": ["oie evaluation is done by using Generic", "intrinsic evaluations of oie systems are carried out either manually - with human evaluators judging the correctness of extractions - or automatically , on standardized benchmarks .", "the latter , while much more cost - effective , is less reliable , primarily because of the incompleteness of the existing oie benchmarks : the ground truth extractions do not include all acceptable variants of the same fact , leading to unreliable assessment of the models ' performance .", "moreover , the existing oie benchmarks are available for english only ."], "relation": "used for", "id": "2022.acl-long.307", "year": 2022, "rel_sent": "Moreover , having in mind common downstream applications for OIE , we make BenchIE multi - faceted ; i.e. , we create benchmark variants that focus on different facets of OIE evaluation , e.g. , compactness or minimality of extractions .", "forward": false, "src_ids": "2022.acl-long.307_3993"}
+{"input": "benchmark variants is used for Task| context: intrinsic evaluations of oie systems are carried out either manually - with human evaluators judging the correctness of extractions - or automatically , on standardized benchmarks . the latter , while much more cost - effective , is less reliable , primarily because of the incompleteness of the existing oie benchmarks : the ground truth extractions do not include all acceptable variants of the same fact , leading to unreliable assessment of the models ' performance . moreover , the existing oie benchmarks are available for english only .", "entity": "benchmark variants", "output": "oie evaluation", "neg_sample": ["benchmark variants is used for Task", "intrinsic evaluations of oie systems are carried out either manually - with human evaluators judging the correctness of extractions - or automatically , on standardized benchmarks .", "the latter , while much more cost - effective , is less reliable , primarily because of the incompleteness of the existing oie benchmarks : the ground truth extractions do not include all acceptable variants of the same fact , leading to unreliable assessment of the models ' performance .", "moreover , the existing oie benchmarks are available for english only ."], "relation": "used for", "id": "2022.acl-long.307", "year": 2022, "rel_sent": "Moreover , having in mind common downstream applications for OIE , we make BenchIE multi - faceted ; i.e. , we create benchmark variants that focus on different facets of OIE evaluation , e.g. , compactness or minimality of extractions .", "forward": true, "src_ids": "2022.acl-long.307_3994"}
+{"input": "human and ai expertise is used for Material| context: natural language processing ( nlp ) systems are often used for adversarial tasks such as detecting spam , abuse , hate speech , and fake news . properly evaluating such systems requires dynamic evaluation that searches for weaknesses in the model , rather than a static test set . prior work has evaluated such models on both manually and automatically generated examples , but both approaches have limitations : manually constructed examples are time - consuming to create and are limited by the imagination and intuition of the creators , while automatically constructed examples are often ungrammatical or labeled inconsistently .", "entity": "human and ai expertise", "output": "adversarial examples", "neg_sample": ["human and ai expertise is used for Material", "natural language processing ( nlp ) systems are often used for adversarial tasks such as detecting spam , abuse , hate speech , and fake news .", "properly evaluating such systems requires dynamic evaluation that searches for weaknesses in the model , rather than a static test set .", "prior work has evaluated such models on both manually and automatically generated examples , but both approaches have limitations : manually constructed examples are time - consuming to create and are limited by the imagination and intuition of the creators , while automatically constructed examples are often ungrammatical or labeled inconsistently ."], "relation": "used for", "id": "2022.nlppower-1.2", "year": 2022, "rel_sent": "We propose to combine human and AI expertise in generating adversarial examples , benefiting from humans ' expertise in language and automated attacks ' ability to probe the target system more quickly and thoroughly .", "forward": true, "src_ids": "2022.nlppower-1.2_3995"}
+{"input": "adversarial examples is done by using OtherScientificTerm| context: natural language processing ( nlp ) systems are often used for adversarial tasks such as detecting spam , abuse , hate speech , and fake news . properly evaluating such systems requires dynamic evaluation that searches for weaknesses in the model , rather than a static test set . prior work has evaluated such models on both manually and automatically generated examples , but both approaches have limitations : manually constructed examples are time - consuming to create and are limited by the imagination and intuition of the creators , while automatically constructed examples are often ungrammatical or labeled inconsistently .", "entity": "adversarial examples", "output": "human and ai expertise", "neg_sample": ["adversarial examples is done by using OtherScientificTerm", "natural language processing ( nlp ) systems are often used for adversarial tasks such as detecting spam , abuse , hate speech , and fake news .", "properly evaluating such systems requires dynamic evaluation that searches for weaknesses in the model , rather than a static test set .", "prior work has evaluated such models on both manually and automatically generated examples , but both approaches have limitations : manually constructed examples are time - consuming to create and are limited by the imagination and intuition of the creators , while automatically constructed examples are often ungrammatical or labeled inconsistently ."], "relation": "used for", "id": "2022.nlppower-1.2", "year": 2022, "rel_sent": "We propose to combine human and AI expertise in generating adversarial examples , benefiting from humans ' expertise in language and automated attacks ' ability to probe the target system more quickly and thoroughly .", "forward": false, "src_ids": "2022.nlppower-1.2_3996"}
+{"input": "hierarchical approach is used for Task| context: large pretrained models enable transfer learning to low - resource domains for language generation tasks . however , previous end - to - end approaches do not account for the fact that some generation sub - tasks , specifically aggregation and lexicalisation , can benefit from transfer learning in different extents .", "entity": "hierarchical approach", "output": "few - shot and zero - shot generation", "neg_sample": ["hierarchical approach is used for Task", "large pretrained models enable transfer learning to low - resource domains for language generation tasks .", "however , previous end - to - end approaches do not account for the fact that some generation sub - tasks , specifically aggregation and lexicalisation , can benefit from transfer learning in different extents ."], "relation": "used for", "id": "2022.findings-acl.170", "year": 2022, "rel_sent": "To exploit these varying potentials for transfer learning , we propose a new hierarchical approach for few - shot and zero - shot generation .", "forward": true, "src_ids": "2022.findings-acl.170_3997"}
+{"input": "few - shot and zero - shot generation is done by using Method| context: large pretrained models enable transfer learning to low - resource domains for language generation tasks . however , previous end - to - end approaches do not account for the fact that some generation sub - tasks , specifically aggregation and lexicalisation , can benefit from transfer learning in different extents .", "entity": "few - shot and zero - shot generation", "output": "hierarchical approach", "neg_sample": ["few - shot and zero - shot generation is done by using Method", "large pretrained models enable transfer learning to low - resource domains for language generation tasks .", "however , previous end - to - end approaches do not account for the fact that some generation sub - tasks , specifically aggregation and lexicalisation , can benefit from transfer learning in different extents ."], "relation": "used for", "id": "2022.findings-acl.170", "year": 2022, "rel_sent": "To exploit these varying potentials for transfer learning , we propose a new hierarchical approach for few - shot and zero - shot generation .", "forward": false, "src_ids": "2022.findings-acl.170_3998"}
+{"input": "low - resource named entity recognition is done by using Method| context: pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data . however , their performance in low - resource scenarios , where such data is not available , remains an open question .", "entity": "low - resource named entity recognition", "output": "pre - trained encoders", "neg_sample": ["low - resource named entity recognition is done by using Method", "pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data .", "however , their performance in low - resource scenarios , where such data is not available , remains an open question ."], "relation": "used for", "id": "2022.repl4nlp-1.6", "year": 2022, "rel_sent": "A Comparative Study of Pre - trained Encoders for Low - Resource Named Entity Recognition.", "forward": false, "src_ids": "2022.repl4nlp-1.6_3999"}
+{"input": "pre - trained encoders is used for Task| context: pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data . however , their performance in low - resource scenarios , where such data is not available , remains an open question .", "entity": "pre - trained encoders", "output": "low - resource named entity recognition", "neg_sample": ["pre - trained encoders is used for Task", "pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data .", "however , their performance in low - resource scenarios , where such data is not available , remains an open question ."], "relation": "used for", "id": "2022.repl4nlp-1.6", "year": 2022, "rel_sent": "A Comparative Study of Pre - trained Encoders for Low - Resource Named Entity Recognition.", "forward": true, "src_ids": "2022.repl4nlp-1.6_4000"}
+{"input": "pre - trained representations is used for Task| context: pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data . however , their performance in low - resource scenarios , where such data is not available , remains an open question .", "entity": "pre - trained representations", "output": "low - resource ner", "neg_sample": ["pre - trained representations is used for Task", "pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data .", "however , their performance in low - resource scenarios , where such data is not available , remains an open question ."], "relation": "used for", "id": "2022.repl4nlp-1.6", "year": 2022, "rel_sent": "We introduce an encoder evaluation framework , and use it to systematically compare the performance of state - of - the - art pre - trained representations on the task of low - resource NER .", "forward": true, "src_ids": "2022.repl4nlp-1.6_4001"}
+{"input": "low - resource ner is done by using Method| context: pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data . however , their performance in low - resource scenarios , where such data is not available , remains an open question .", "entity": "low - resource ner", "output": "pre - trained representations", "neg_sample": ["low - resource ner is done by using Method", "pre - trained language models ( plm ) are effective components of few - shot named entity recognition ( ner ) approaches when augmented with continued pre - training on task - specific out - of - domain data or fine - tuning on in - domain data .", "however , their performance in low - resource scenarios , where such data is not available , remains an open question ."], "relation": "used for", "id": "2022.repl4nlp-1.6", "year": 2022, "rel_sent": "We introduce an encoder evaluation framework , and use it to systematically compare the performance of state - of - the - art pre - trained representations on the task of low - resource NER .", "forward": false, "src_ids": "2022.repl4nlp-1.6_4002"}
+{"input": "equality is done by using Task| context: social media platforms have been provoking masses of people . the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities . hope significantly impacts the well - being of people , as suggested by health professionals .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "social media platforms have been provoking masses of people .", "the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities .", "hope significantly impacts the well - being of people , as suggested by health professionals ."], "relation": "used for", "id": "2022.ltedi-1.54", "year": 2022, "rel_sent": "IDIAP Submission@LT - EDI - ACL2022 : Hope Speech Detection for Equality , Diversity and Inclusion.", "forward": false, "src_ids": "2022.ltedi-1.54_4003"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: social media platforms have been provoking masses of people . the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities . hope significantly impacts the well - being of people , as suggested by health professionals .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "social media platforms have been provoking masses of people .", "the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities .", "hope significantly impacts the well - being of people , as suggested by health professionals ."], "relation": "used for", "id": "2022.ltedi-1.54", "year": 2022, "rel_sent": "IDIAP Submission@LT - EDI - ACL2022 : Hope Speech Detection for Equality , Diversity and Inclusion.", "forward": true, "src_ids": "2022.ltedi-1.54_4004"}
+{"input": "english language is done by using Method| context: social media platforms have been provoking masses of people . the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities . hope significantly impacts the well - being of people , as suggested by health professionals .", "entity": "english language", "output": "roberta", "neg_sample": ["english language is done by using Method", "social media platforms have been provoking masses of people .", "the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities .", "hope significantly impacts the well - being of people , as suggested by health professionals ."], "relation": "used for", "id": "2022.ltedi-1.54", "year": 2022, "rel_sent": "It reflects the belief to achieve an objective , discovers a new path , or become motivated toformulate pathways . In this paper we classify given a social media post , hope speech or not hope speech , using ensembled voting of BERT , ERNIE 2.0 and RoBERTa for English language with 0.54 macrof1 - score ( 2^{st } rank ) .", "forward": false, "src_ids": "2022.ltedi-1.54_4005"}
+{"input": "roberta is used for Material| context: social media platforms have been provoking masses of people . the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities . hope significantly impacts the well - being of people , as suggested by health professionals .", "entity": "roberta", "output": "english language", "neg_sample": ["roberta is used for Material", "social media platforms have been provoking masses of people .", "the individual comments affect a prevalent way of thinking by moving away from preoccupation with discrimination , loneliness , or influence in building confidence , support , and good qualities .", "hope significantly impacts the well - being of people , as suggested by health professionals ."], "relation": "used for", "id": "2022.ltedi-1.54", "year": 2022, "rel_sent": "It reflects the belief to achieve an objective , discovers a new path , or become motivated toformulate pathways . In this paper we classify given a social media post , hope speech or not hope speech , using ensembled voting of BERT , ERNIE 2.0 and RoBERTa for English language with 0.54 macrof1 - score ( 2^{st } rank ) .", "forward": true, "src_ids": "2022.ltedi-1.54_4006"}
+{"input": "data creation is done by using OtherScientificTerm| context: knowledge of difficulty level of questions helps a teacher in several ways , such as estimating students ' potential quickly by asking carefully selected questions and improving quality of examination by modifying trivial and hard questions . can we extract such benefits of instance difficulty in natural language processing ?", "entity": "data creation", "output": "dataset characteristics", "neg_sample": ["data creation is done by using OtherScientificTerm", "knowledge of difficulty level of questions helps a teacher in several ways , such as estimating students ' potential quickly by asking carefully selected questions and improving quality of examination by modifying trivial and hard questions .", "can we extract such benefits of instance difficulty in natural language processing ?"], "relation": "used for", "id": "2022.acl-long.240", "year": 2022, "rel_sent": "To this end , we conduct Instance - Level Difficulty Analysis of Evaluation data ( ILDAE ) in a large - scale setup of 23 datasets and demonstrate its five novel applications : 1 ) conducting efficient - yet - accurate evaluations with fewer instances saving computational cost and time , 2 ) improving quality of existing evaluation datasets by repairing erroneous and trivial instances , 3 ) selecting the best model based on application requirements , 4 ) analyzing dataset characteristics for guiding future data creation , 5 ) estimating Out - of - Domain performance reliably .", "forward": false, "src_ids": "2022.acl-long.240_4007"}
+{"input": "dataset characteristics is used for Task| context: knowledge of difficulty level of questions helps a teacher in several ways , such as estimating students ' potential quickly by asking carefully selected questions and improving quality of examination by modifying trivial and hard questions . can we extract such benefits of instance difficulty in natural language processing ?", "entity": "dataset characteristics", "output": "data creation", "neg_sample": ["dataset characteristics is used for Task", "knowledge of difficulty level of questions helps a teacher in several ways , such as estimating students ' potential quickly by asking carefully selected questions and improving quality of examination by modifying trivial and hard questions .", "can we extract such benefits of instance difficulty in natural language processing ?"], "relation": "used for", "id": "2022.acl-long.240", "year": 2022, "rel_sent": "To this end , we conduct Instance - Level Difficulty Analysis of Evaluation data ( ILDAE ) in a large - scale setup of 23 datasets and demonstrate its five novel applications : 1 ) conducting efficient - yet - accurate evaluations with fewer instances saving computational cost and time , 2 ) improving quality of existing evaluation datasets by repairing erroneous and trivial instances , 3 ) selecting the best model based on application requirements , 4 ) analyzing dataset characteristics for guiding future data creation , 5 ) estimating Out - of - Domain performance reliably .", "forward": true, "src_ids": "2022.acl-long.240_4008"}
+{"input": "dependency parsing is done by using Method| context: dependency parsing is a method for doing surface - level syntactic analysis on natural language texts . the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "dependency parsing", "output": "bert - based sequence labelling approach", "neg_sample": ["dependency parsing is done by using Method", "dependency parsing is a method for doing surface - level syntactic analysis on natural language texts .", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "BERT - Based Sequence Labelling Approach for Dependency Parsing in Tamil.", "forward": false, "src_ids": "2022.dravidianlangtech-1.1_4009"}
+{"input": "bert - based sequence labelling approach is used for Task| context: the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "bert - based sequence labelling approach", "output": "dependency parsing", "neg_sample": ["bert - based sequence labelling approach is used for Task", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "BERT - Based Sequence Labelling Approach for Dependency Parsing in Tamil.", "forward": true, "src_ids": "2022.dravidianlangtech-1.1_4010"}
+{"input": "malt parser is done by using Method| context: dependency parsing is a method for doing surface - level syntactic analysis on natural language texts . the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "malt parser", "output": "bert models", "neg_sample": ["malt parser is done by using Method", "dependency parsing is a method for doing surface - level syntactic analysis on natural language texts .", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "This paper focuses on a novel approach that uses word - to - word dependency tagging using BERT models to improve the malt parser performance .", "forward": false, "src_ids": "2022.dravidianlangtech-1.1_4011"}
+{"input": "bert models is used for Method| context: dependency parsing is a method for doing surface - level syntactic analysis on natural language texts . the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "bert models", "output": "malt parser", "neg_sample": ["bert models is used for Method", "dependency parsing is a method for doing surface - level syntactic analysis on natural language texts .", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "This paper focuses on a novel approach that uses word - to - word dependency tagging using BERT models to improve the malt parser performance .", "forward": true, "src_ids": "2022.dravidianlangtech-1.1_4012"}
+{"input": "data imbalance is done by using Method| context: dependency parsing is a method for doing surface - level syntactic analysis on natural language texts . the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "data imbalance", "output": "oversampling algorithms", "neg_sample": ["data imbalance is done by using Method", "dependency parsing is a method for doing surface - level syntactic analysis on natural language texts .", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "Oversampling algorithms such as SMOTE ( Chawla et al . , 2002 ) and ADASYN ( He et al . , 2008 ) are used to tackle data imbalance and consequently improve parsing results .", "forward": false, "src_ids": "2022.dravidianlangtech-1.1_4013"}
+{"input": "parsing is done by using Method| context: dependency parsing is a method for doing surface - level syntactic analysis on natural language texts . the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "parsing", "output": "oversampling algorithms", "neg_sample": ["parsing is done by using Method", "dependency parsing is a method for doing surface - level syntactic analysis on natural language texts .", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "Oversampling algorithms such as SMOTE ( Chawla et al . , 2002 ) and ADASYN ( He et al . , 2008 ) are used to tackle data imbalance and consequently improve parsing results .", "forward": false, "src_ids": "2022.dravidianlangtech-1.1_4014"}
+{"input": "oversampling algorithms is used for OtherScientificTerm| context: dependency parsing is a method for doing surface - level syntactic analysis on natural language texts . the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "oversampling algorithms", "output": "data imbalance", "neg_sample": ["oversampling algorithms is used for OtherScientificTerm", "dependency parsing is a method for doing surface - level syntactic analysis on natural language texts .", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "Oversampling algorithms such as SMOTE ( Chawla et al . , 2002 ) and ADASYN ( He et al . , 2008 ) are used to tackle data imbalance and consequently improve parsing results .", "forward": true, "src_ids": "2022.dravidianlangtech-1.1_4015"}
+{"input": "oversampling algorithms is used for Task| context: the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics .", "entity": "oversampling algorithms", "output": "parsing", "neg_sample": ["oversampling algorithms is used for Task", "the scarcity of any viable tools for doing these tasks in dravidian languages has introduced a new line of research into these topics ."], "relation": "used for", "id": "2022.dravidianlangtech-1.1", "year": 2022, "rel_sent": "Oversampling algorithms such as SMOTE ( Chawla et al . , 2002 ) and ADASYN ( He et al . , 2008 ) are used to tackle data imbalance and consequently improve parsing results .", "forward": true, "src_ids": "2022.dravidianlangtech-1.1_4016"}
+{"input": "e - commerce product copywriting generation is done by using Method| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "e - commerce product copywriting generation", "output": "interactive latent knowledge selection", "neg_sample": ["e - commerce product copywriting generation is done by using Method", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Interactive Latent Knowledge Selection for E - Commerce Product Copywriting Generation.", "forward": false, "src_ids": "2022.ecnlp-1.2_4017"}
+{"input": "interactive latent knowledge selection is used for Task| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "interactive latent knowledge selection", "output": "e - commerce product copywriting generation", "neg_sample": ["interactive latent knowledge selection is used for Task", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Interactive Latent Knowledge Selection for E - Commerce Product Copywriting Generation.", "forward": true, "src_ids": "2022.ecnlp-1.2_4018"}
+{"input": "auxiliary descriptions is used for Task| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "auxiliary descriptions", "output": "automatically generating product copywriting", "neg_sample": ["auxiliary descriptions is used for Task", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Automatically generating product copywriting has attracted noticeable interests from both academic and industrial communities , where existing solutions merely make use of a product 's title and attribute information to generate its corresponding description . However , in addition to the product title and attributes , we observe that there are various auxiliary descriptions created by the shoppers or marketers in the e - commerce platforms ( namely human knowledge ) , which contains valuable information for product copywriting generation , yet always accompanying lots of noises . In this work , we propose a novel solution to automatically generating product copywriting that involves all the title , attributes and denoised auxiliary knowledge . To be specific , we design an end - to - end generation framework equipped with two variational autoencoders that works interactively to select informative human knowledge and generate diverse copywriting .", "forward": true, "src_ids": "2022.ecnlp-1.2_4019"}
+{"input": "automatically generating product copywriting is done by using OtherScientificTerm| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "automatically generating product copywriting", "output": "auxiliary descriptions", "neg_sample": ["automatically generating product copywriting is done by using OtherScientificTerm", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Automatically generating product copywriting has attracted noticeable interests from both academic and industrial communities , where existing solutions merely make use of a product 's title and attribute information to generate its corresponding description . However , in addition to the product title and attributes , we observe that there are various auxiliary descriptions created by the shoppers or marketers in the e - commerce platforms ( namely human knowledge ) , which contains valuable information for product copywriting generation , yet always accompanying lots of noises . In this work , we propose a novel solution to automatically generating product copywriting that involves all the title , attributes and denoised auxiliary knowledge . To be specific , we design an end - to - end generation framework equipped with two variational autoencoders that works interactively to select informative human knowledge and generate diverse copywriting .", "forward": false, "src_ids": "2022.ecnlp-1.2_4020"}
+{"input": "diverse copywriting is done by using Method| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "diverse copywriting", "output": "end - to - end generation framework", "neg_sample": ["diverse copywriting is done by using Method", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Automatically generating product copywriting has attracted noticeable interests from both academic and industrial communities , where existing solutions merely make use of a product 's title and attribute information to generate its corresponding description . However , in addition to the product title and attributes , we observe that there are various auxiliary descriptions created by the shoppers or marketers in the e - commerce platforms ( namely human knowledge ) , which contains valuable information for product copywriting generation , yet always accompanying lots of noises . In this work , we propose a novel solution to automatically generating product copywriting that involves all the title , attributes and denoised auxiliary knowledge . To be specific , we design an end - to - end generation framework equipped with two variational autoencoders that works interactively to select informative human knowledge and generate diverse copywriting .", "forward": false, "src_ids": "2022.ecnlp-1.2_4021"}
+{"input": "informative human knowledge is done by using Method| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "informative human knowledge", "output": "variational autoencoders", "neg_sample": ["informative human knowledge is done by using Method", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Automatically generating product copywriting has attracted noticeable interests from both academic and industrial communities , where existing solutions merely make use of a product 's title and attribute information to generate its corresponding description . However , in addition to the product title and attributes , we observe that there are various auxiliary descriptions created by the shoppers or marketers in the e - commerce platforms ( namely human knowledge ) , which contains valuable information for product copywriting generation , yet always accompanying lots of noises . In this work , we propose a novel solution to automatically generating product copywriting that involves all the title , attributes and denoised auxiliary knowledge . To be specific , we design an end - to - end generation framework equipped with two variational autoencoders that works interactively to select informative human knowledge and generate diverse copywriting .", "forward": false, "src_ids": "2022.ecnlp-1.2_4022"}
+{"input": "variational autoencoders is used for OtherScientificTerm| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "variational autoencoders", "output": "informative human knowledge", "neg_sample": ["variational autoencoders is used for OtherScientificTerm", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Automatically generating product copywriting has attracted noticeable interests from both academic and industrial communities , where existing solutions merely make use of a product 's title and attribute information to generate its corresponding description . However , in addition to the product title and attributes , we observe that there are various auxiliary descriptions created by the shoppers or marketers in the e - commerce platforms ( namely human knowledge ) , which contains valuable information for product copywriting generation , yet always accompanying lots of noises . In this work , we propose a novel solution to automatically generating product copywriting that involves all the title , attributes and denoised auxiliary knowledge . To be specific , we design an end - to - end generation framework equipped with two variational autoencoders that works interactively to select informative human knowledge and generate diverse copywriting .", "forward": true, "src_ids": "2022.ecnlp-1.2_4023"}
+{"input": "end - to - end generation framework is used for OtherScientificTerm| context: as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms . the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase .", "entity": "end - to - end generation framework", "output": "diverse copywriting", "neg_sample": ["end - to - end generation framework is used for OtherScientificTerm", "as the multi - modal e - commerce is thriving , high - quality advertising product copywriting has gain more attentions , which plays a crucial role in the e - commerce recommender , advertising and even search platforms .", "the advertising product copywriting is able to enhance the user experience by highlighting the product 's characteristics with textual descriptions and thus to improve the likelihood of user click and purchase ."], "relation": "used for", "id": "2022.ecnlp-1.2", "year": 2022, "rel_sent": "Automatically generating product copywriting has attracted noticeable interests from both academic and industrial communities , where existing solutions merely make use of a product 's title and attribute information to generate its corresponding description . However , in addition to the product title and attributes , we observe that there are various auxiliary descriptions created by the shoppers or marketers in the e - commerce platforms ( namely human knowledge ) , which contains valuable information for product copywriting generation , yet always accompanying lots of noises . In this work , we propose a novel solution to automatically generating product copywriting that involves all the title , attributes and denoised auxiliary knowledge . To be specific , we design an end - to - end generation framework equipped with two variational autoencoders that works interactively to select informative human knowledge and generate diverse copywriting .", "forward": true, "src_ids": "2022.ecnlp-1.2_4024"}
+{"input": "revitalising endangered languages is done by using Method| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "revitalising endangered languages", "output": "cld^2 language documentation", "neg_sample": ["revitalising endangered languages is done by using Method", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 Language Documentation Meets Natural Language Processing for Revitalising Endangered Languages.", "forward": false, "src_ids": "2022.computel-1.4_4025"}
+{"input": "revitalising endangered languages is done by using Task| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "entity": "revitalising endangered languages", "output": "natural language processing", "neg_sample": ["revitalising endangered languages is done by using Task", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 Language Documentation Meets Natural Language Processing for Revitalising Endangered Languages.", "forward": false, "src_ids": "2022.computel-1.4_4026"}
+{"input": "natural language processing is used for Task| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "natural language processing", "output": "revitalising endangered languages", "neg_sample": ["natural language processing is used for Task", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 Language Documentation Meets Natural Language Processing for Revitalising Endangered Languages.", "forward": true, "src_ids": "2022.computel-1.4_4027"}
+{"input": "cld^2 language documentation is used for Task| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "cld^2 language documentation", "output": "revitalising endangered languages", "neg_sample": ["cld^2 language documentation is used for Task", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 Language Documentation Meets Natural Language Processing for Revitalising Endangered Languages.", "forward": true, "src_ids": "2022.computel-1.4_4028"}
+{"input": "language documentation projects is done by using Material| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "language documentation projects", "output": "nlp - friendly annotated data", "neg_sample": ["language documentation projects is done by using Material", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 calls for ( 1 ) the inclusion of NLP - friendly annotated data as a deliverable of future language documentation projects ; and ( 2 ) the exploitation of language documentation databases by the NLP community to promote the computerization of endangered languages , as one way to contribute to their revitalization .", "forward": false, "src_ids": "2022.computel-1.4_4029"}
+{"input": "nlp - friendly annotated data is used for Task| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "nlp - friendly annotated data", "output": "language documentation projects", "neg_sample": ["nlp - friendly annotated data is used for Task", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 calls for ( 1 ) the inclusion of NLP - friendly annotated data as a deliverable of future language documentation projects ; and ( 2 ) the exploitation of language documentation databases by the NLP community to promote the computerization of endangered languages , as one way to contribute to their revitalization .", "forward": true, "src_ids": "2022.computel-1.4_4030"}
+{"input": "computerization of endangered languages is done by using Material| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "computerization of endangered languages", "output": "language documentation databases", "neg_sample": ["computerization of endangered languages is done by using Material", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 calls for ( 1 ) the inclusion of NLP - friendly annotated data as a deliverable of future language documentation projects ; and ( 2 ) the exploitation of language documentation databases by the NLP community to promote the computerization of endangered languages , as one way to contribute to their revitalization .", "forward": false, "src_ids": "2022.computel-1.4_4031"}
+{"input": "language documentation databases is used for Task| context: language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories . natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages .", "entity": "language documentation databases", "output": "computerization of endangered languages", "neg_sample": ["language documentation databases is used for Task", "language revitalisation should not be understood as a direct outcome of language documentation , which is mainly focused on the creation of language repositories .", "natural language processing ( nlp ) offers the potential to complement and exploit these repositories through the development of language technologies that may contribute to improving the vitality status of endangered languages ."], "relation": "used for", "id": "2022.computel-1.4", "year": 2022, "rel_sent": "CLD^2 calls for ( 1 ) the inclusion of NLP - friendly annotated data as a deliverable of future language documentation projects ; and ( 2 ) the exploitation of language documentation databases by the NLP community to promote the computerization of endangered languages , as one way to contribute to their revitalization .", "forward": true, "src_ids": "2022.computel-1.4_4032"}
+{"input": "multi - domain task - oriented semantic parsing is done by using Method| context: while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering .", "entity": "multi - domain task - oriented semantic parsing", "output": "retrieval - based modeling ideas", "neg_sample": ["multi - domain task - oriented semantic parsing is done by using Method", "while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering ."], "relation": "used for", "id": "2022.nlp4convai-1.15", "year": 2022, "rel_sent": "In this work , we apply retrieval - based modeling ideas to the challenging complex task of multi - domain task - oriented semantic parsing for conversational assistants .", "forward": false, "src_ids": "2022.nlp4convai-1.15_4033"}
+{"input": "conversational assistants is done by using Task| context: while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering .", "entity": "conversational assistants", "output": "multi - domain task - oriented semantic parsing", "neg_sample": ["conversational assistants is done by using Task", "while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering ."], "relation": "used for", "id": "2022.nlp4convai-1.15", "year": 2022, "rel_sent": "In this work , we apply retrieval - based modeling ideas to the challenging complex task of multi - domain task - oriented semantic parsing for conversational assistants .", "forward": false, "src_ids": "2022.nlp4convai-1.15_4034"}
+{"input": "retrieval - based modeling ideas is used for Task| context: while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering .", "entity": "retrieval - based modeling ideas", "output": "multi - domain task - oriented semantic parsing", "neg_sample": ["retrieval - based modeling ideas is used for Task", "while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering ."], "relation": "used for", "id": "2022.nlp4convai-1.15", "year": 2022, "rel_sent": "In this work , we apply retrieval - based modeling ideas to the challenging complex task of multi - domain task - oriented semantic parsing for conversational assistants .", "forward": true, "src_ids": "2022.nlp4convai-1.15_4035"}
+{"input": "multi - domain task - oriented semantic parsing is used for Method| context: while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering .", "entity": "multi - domain task - oriented semantic parsing", "output": "conversational assistants", "neg_sample": ["multi - domain task - oriented semantic parsing is used for Method", "while large pre - trained language models accumulate a lot of knowledge in their parameters , it has been demonstrated that augmenting it with non - parametric retrieval - based memory has a number of benefits ranging from improved accuracy to data efficiency for knowledge - focused tasks such as question answering ."], "relation": "used for", "id": "2022.nlp4convai-1.15", "year": 2022, "rel_sent": "In this work , we apply retrieval - based modeling ideas to the challenging complex task of multi - domain task - oriented semantic parsing for conversational assistants .", "forward": true, "src_ids": "2022.nlp4convai-1.15_4036"}
+{"input": "clinical knowledge graphs is done by using Task| context: clinical knowledge graphs lack meaningful diagnostic relations ( e.g. comorbidities , sign / symptoms ) , limiting their ability to represent real - world diagnostic processes . previous methods in biomedical relation extraction have focused on concept relations , such as gene - disease and disease - drug , and largely ignored clinical processes .", "entity": "clinical knowledge graphs", "output": "extraction of diagnostic reasoning relations", "neg_sample": ["clinical knowledge graphs is done by using Task", "clinical knowledge graphs lack meaningful diagnostic relations ( e.g.", "comorbidities , sign / symptoms ) , limiting their ability to represent real - world diagnostic processes .", "previous methods in biomedical relation extraction have focused on concept relations , such as gene - disease and disease - drug , and largely ignored clinical processes ."], "relation": "used for", "id": "2022.acl-srw.33", "year": 2022, "rel_sent": "Extraction of Diagnostic Reasoning Relations for Clinical Knowledge Graphs.", "forward": false, "src_ids": "2022.acl-srw.33_4037"}
+{"input": "extraction of diagnostic reasoning relations is used for OtherScientificTerm| context: comorbidities , sign / symptoms ) , limiting their ability to represent real - world diagnostic processes . previous methods in biomedical relation extraction have focused on concept relations , such as gene - disease and disease - drug , and largely ignored clinical processes .", "entity": "extraction of diagnostic reasoning relations", "output": "clinical knowledge graphs", "neg_sample": ["extraction of diagnostic reasoning relations is used for OtherScientificTerm", "comorbidities , sign / symptoms ) , limiting their ability to represent real - world diagnostic processes .", "previous methods in biomedical relation extraction have focused on concept relations , such as gene - disease and disease - drug , and largely ignored clinical processes ."], "relation": "used for", "id": "2022.acl-srw.33", "year": 2022, "rel_sent": "Extraction of Diagnostic Reasoning Relations for Clinical Knowledge Graphs.", "forward": true, "src_ids": "2022.acl-srw.33_4038"}
+{"input": "model compression is done by using Method| context: pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications .", "entity": "model compression", "output": "textpruner", "neg_sample": ["model compression is done by using Method", "pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications ."], "relation": "used for", "id": "2022.acl-demo.4", "year": 2022, "rel_sent": "In this paper , we introduce TextPruner , an open - source model pruning toolkit designed for pre - trained language models , targeting fast and easy model compression .", "forward": false, "src_ids": "2022.acl-demo.4_4039"}
+{"input": "pre - trained language models is done by using Method| context: pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications .", "entity": "pre - trained language models", "output": "model pruning toolkit", "neg_sample": ["pre - trained language models is done by using Method", "pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications ."], "relation": "used for", "id": "2022.acl-demo.4", "year": 2022, "rel_sent": "TextPruner : A Model Pruning Toolkit for Pre - Trained Language Models.", "forward": false, "src_ids": "2022.acl-demo.4_4040"}
+{"input": "pre - trained language models is done by using Method| context: pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications .", "entity": "pre - trained language models", "output": "open - source model pruning toolkit", "neg_sample": ["pre - trained language models is done by using Method", "pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications ."], "relation": "used for", "id": "2022.acl-demo.4", "year": 2022, "rel_sent": "In this paper , we introduce TextPruner , an open - source model pruning toolkit designed for pre - trained language models , targeting fast and easy model compression .", "forward": false, "src_ids": "2022.acl-demo.4_4041"}
+{"input": "textpruner is used for Task| context: pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications .", "entity": "textpruner", "output": "model compression", "neg_sample": ["textpruner is used for Task", "pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications ."], "relation": "used for", "id": "2022.acl-demo.4", "year": 2022, "rel_sent": "In this paper , we introduce TextPruner , an open - source model pruning toolkit designed for pre - trained language models , targeting fast and easy model compression .", "forward": true, "src_ids": "2022.acl-demo.4_4042"}
+{"input": "open - source model pruning toolkit is used for Task| context: pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications .", "entity": "open - source model pruning toolkit", "output": "model compression", "neg_sample": ["open - source model pruning toolkit is used for Task", "pre - trained language models have been prevailed in natural language processing and become the backbones of many nlp tasks , but the demands for computational resources have limited their applications ."], "relation": "used for", "id": "2022.acl-demo.4", "year": 2022, "rel_sent": "In this paper , we introduce TextPruner , an open - source model pruning toolkit designed for pre - trained language models , targeting fast and easy model compression .", "forward": true, "src_ids": "2022.acl-demo.4_4043"}
+{"input": "gptk 's language is done by using Task| context: what kinds of instructional prompts are easier tofollow for language models ( lms ) ?", "entity": "gptk 's language", "output": "reframing instructional prompts", "neg_sample": ["gptk 's language is done by using Task", "what kinds of instructional prompts are easier tofollow for language models ( lms ) ?"], "relation": "used for", "id": "2022.findings-acl.50", "year": 2022, "rel_sent": "Reframing Instructional Prompts to GPTk 's Language.", "forward": false, "src_ids": "2022.findings-acl.50_4044"}
+{"input": "reframing instructional prompts is used for Material| context: what kinds of instructional prompts are easier tofollow for language models ( lms ) ?", "entity": "reframing instructional prompts", "output": "gptk 's language", "neg_sample": ["reframing instructional prompts is used for Material", "what kinds of instructional prompts are easier tofollow for language models ( lms ) ?"], "relation": "used for", "id": "2022.findings-acl.50", "year": 2022, "rel_sent": "Reframing Instructional Prompts to GPTk 's Language.", "forward": true, "src_ids": "2022.findings-acl.50_4045"}
+{"input": "manual reformulation of prompts is done by using Method| context: what kinds of instructional prompts are easier tofollow for language models ( lms ) ?", "entity": "manual reformulation of prompts", "output": "reframing techniques", "neg_sample": ["manual reformulation of prompts is done by using Method", "what kinds of instructional prompts are easier tofollow for language models ( lms ) ?"], "relation": "used for", "id": "2022.findings-acl.50", "year": 2022, "rel_sent": "Specifically , we study several classes of reframing techniques for manual reformulation of prompts into more effective ones .", "forward": false, "src_ids": "2022.findings-acl.50_4046"}
+{"input": "reframing techniques is used for Task| context: what kinds of instructional prompts are easier tofollow for language models ( lms ) ?", "entity": "reframing techniques", "output": "manual reformulation of prompts", "neg_sample": ["reframing techniques is used for Task", "what kinds of instructional prompts are easier tofollow for language models ( lms ) ?"], "relation": "used for", "id": "2022.findings-acl.50", "year": 2022, "rel_sent": "Specifically , we study several classes of reframing techniques for manual reformulation of prompts into more effective ones .", "forward": true, "src_ids": "2022.findings-acl.50_4047"}
+{"input": "prompting algorithms is done by using Method| context: what kinds of instructional prompts are easier tofollow for language models ( lms ) ?", "entity": "prompting algorithms", "output": "empirically - driven techniques", "neg_sample": ["prompting algorithms is done by using Method", "what kinds of instructional prompts are easier tofollow for language models ( lms ) ?"], "relation": "used for", "id": "2022.findings-acl.50", "year": 2022, "rel_sent": "We hope these empirically - driven techniques will pave the way towards more effective future prompting algorithms .", "forward": false, "src_ids": "2022.findings-acl.50_4048"}
+{"input": "empirically - driven techniques is used for Method| context: what kinds of instructional prompts are easier tofollow for language models ( lms ) ?", "entity": "empirically - driven techniques", "output": "prompting algorithms", "neg_sample": ["empirically - driven techniques is used for Method", "what kinds of instructional prompts are easier tofollow for language models ( lms ) ?"], "relation": "used for", "id": "2022.findings-acl.50", "year": 2022, "rel_sent": "We hope these empirically - driven techniques will pave the way towards more effective future prompting algorithms .", "forward": true, "src_ids": "2022.findings-acl.50_4049"}
+{"input": "cross - corpora hate speech detection is done by using Method| context: hate speech classifiers exhibit substantial performance degradation when evaluated on datasets different from the source . this is due to learning spurious correlations between words that are not necessarily relevant to hateful language , and hate speech labels from the training corpus . previous work has attempted to mitigate this problem by regularizing specific terms from pre - defined static dictionaries . while this has been demonstrated to improve the generalizability of classifiers , the coverage of such methods is limited and the dictionaries require regular manual updates from human experts .", "entity": "cross - corpora hate speech detection", "output": "dynamically refined regularization", "neg_sample": ["cross - corpora hate speech detection is done by using Method", "hate speech classifiers exhibit substantial performance degradation when evaluated on datasets different from the source .", "this is due to learning spurious correlations between words that are not necessarily relevant to hateful language , and hate speech labels from the training corpus .", "previous work has attempted to mitigate this problem by regularizing specific terms from pre - defined static dictionaries .", "while this has been demonstrated to improve the generalizability of classifiers , the coverage of such methods is limited and the dictionaries require regular manual updates from human experts ."], "relation": "used for", "id": "2022.findings-acl.32", "year": 2022, "rel_sent": "Dynamically Refined Regularization for Improving Cross - corpora Hate Speech Detection.", "forward": false, "src_ids": "2022.findings-acl.32_4050"}
+{"input": "dynamically refined regularization is used for Task| context: hate speech classifiers exhibit substantial performance degradation when evaluated on datasets different from the source . this is due to learning spurious correlations between words that are not necessarily relevant to hateful language , and hate speech labels from the training corpus . previous work has attempted to mitigate this problem by regularizing specific terms from pre - defined static dictionaries . while this has been demonstrated to improve the generalizability of classifiers , the coverage of such methods is limited and the dictionaries require regular manual updates from human experts .", "entity": "dynamically refined regularization", "output": "cross - corpora hate speech detection", "neg_sample": ["dynamically refined regularization is used for Task", "hate speech classifiers exhibit substantial performance degradation when evaluated on datasets different from the source .", "this is due to learning spurious correlations between words that are not necessarily relevant to hateful language , and hate speech labels from the training corpus .", "previous work has attempted to mitigate this problem by regularizing specific terms from pre - defined static dictionaries .", "while this has been demonstrated to improve the generalizability of classifiers , the coverage of such methods is limited and the dictionaries require regular manual updates from human experts ."], "relation": "used for", "id": "2022.findings-acl.32", "year": 2022, "rel_sent": "Dynamically Refined Regularization for Improving Cross - corpora Hate Speech Detection.", "forward": true, "src_ids": "2022.findings-acl.32_4051"}
+{"input": "meta - review domain is done by using Material| context: when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited . a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences . a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge .", "entity": "meta - review domain", "output": "mred", "neg_sample": ["meta - review domain is done by using Material", "when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited .", "a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences .", "a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge ."], "relation": "used for", "id": "2022.findings-acl.198", "year": 2022, "rel_sent": "Meanwhile , MReD also allows us to have a better understanding of the meta - review domain .", "forward": false, "src_ids": "2022.findings-acl.198_4052"}
+{"input": "structure - controllable text generation is done by using Material| context: when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited . a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences . a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge .", "entity": "structure - controllable text generation", "output": "meta - review dataset", "neg_sample": ["structure - controllable text generation is done by using Material", "when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited .", "a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences .", "a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge ."], "relation": "used for", "id": "2022.findings-acl.198", "year": 2022, "rel_sent": "MReD : A Meta - Review Dataset for Structure - Controllable Text Generation.", "forward": false, "src_ids": "2022.findings-acl.198_4053"}
+{"input": "meta - review dataset is used for Task| context: when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited . a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences . a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge .", "entity": "meta - review dataset", "output": "structure - controllable text generation", "neg_sample": ["meta - review dataset is used for Task", "when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited .", "a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences .", "a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge ."], "relation": "used for", "id": "2022.findings-acl.198", "year": 2022, "rel_sent": "MReD : A Meta - Review Dataset for Structure - Controllable Text Generation.", "forward": true, "src_ids": "2022.findings-acl.198_4054"}
+{"input": "mred is used for Material| context: when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited . a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences . a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge .", "entity": "mred", "output": "meta - review domain", "neg_sample": ["mred is used for Material", "when directly using existing text generation datasets for controllable generation , we are facing the problem of not having the domain knowledge and thus the aspects that could be controlled are limited .", "a typical example is when using cnn / daily mail dataset for controllable text summarization , there is no guided information on the emphasis of summary sentences .", "a more useful text generator should leverage both the input text and the control signal to guide the generation , which can only be built with deep understanding of the domain knowledge ."], "relation": "used for", "id": "2022.findings-acl.198", "year": 2022, "rel_sent": "Meanwhile , MReD also allows us to have a better understanding of the meta - review domain .", "forward": true, "src_ids": "2022.findings-acl.198_4055"}
+{"input": "translation process is done by using Method| context: prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks .", "entity": "translation process", "output": "msp", "neg_sample": ["translation process is done by using Method", "prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks ."], "relation": "used for", "id": "2022.acl-long.424", "year": 2022, "rel_sent": "To better mitigate the discrepancy between pre - training and translation , MSP divides the translation process via pre - trained language models into three separate stages : the encoding stage , the re - encoding stage , and the decoding stage .", "forward": false, "src_ids": "2022.acl-long.424_4056"}
+{"input": "pre - trained language models is done by using Method| context: prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks .", "entity": "pre - trained language models", "output": "multi - stage prompting", "neg_sample": ["pre - trained language models is done by using Method", "prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks ."], "relation": "used for", "id": "2022.acl-long.424", "year": 2022, "rel_sent": "MSP : Multi - Stage Prompting for Making Pre - trained Language Models Better Translators.", "forward": false, "src_ids": "2022.acl-long.424_4057"}
+{"input": "pre - trained language models is used for Task| context: prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks .", "entity": "pre - trained language models", "output": "translation tasks", "neg_sample": ["pre - trained language models is used for Task", "prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks ."], "relation": "used for", "id": "2022.acl-long.424", "year": 2022, "rel_sent": "We present Multi - Stage Prompting , a simple and automatic approach for leveraging pre - trained language models to translation tasks .", "forward": true, "src_ids": "2022.acl-long.424_4058"}
+{"input": "pre - trained language models is used for Task| context: prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks .", "entity": "pre - trained language models", "output": "translation tasks", "neg_sample": ["pre - trained language models is used for Task", "prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks ."], "relation": "used for", "id": "2022.acl-long.424", "year": 2022, "rel_sent": "During each stage , we independently apply different continuous prompts for allowing pre - trained language models better shift to translation tasks .", "forward": true, "src_ids": "2022.acl-long.424_4059"}
+{"input": "msp is used for Task| context: prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks .", "entity": "msp", "output": "translation process", "neg_sample": ["msp is used for Task", "prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks ."], "relation": "used for", "id": "2022.acl-long.424", "year": 2022, "rel_sent": "To better mitigate the discrepancy between pre - training and translation , MSP divides the translation process via pre - trained language models into three separate stages : the encoding stage , the re - encoding stage , and the decoding stage .", "forward": true, "src_ids": "2022.acl-long.424_4060"}
+{"input": "pre - trained language models is done by using OtherScientificTerm| context: prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks .", "entity": "pre - trained language models", "output": "continuous prompts", "neg_sample": ["pre - trained language models is done by using OtherScientificTerm", "prompting has recently been shown as a promising approach for applying pre - trained language models to perform downstream tasks ."], "relation": "used for", "id": "2022.acl-long.424", "year": 2022, "rel_sent": "During each stage , we independently apply different continuous prompts for allowing pre - trained language models better shift to translation tasks .", "forward": false, "src_ids": "2022.acl-long.424_4061"}
+{"input": "te reo maori is done by using Method| context: henceforth , part - of - speech will be referred to as pos throughout this paper and te reo maori will be referred to as maori , while universal dependencies will be referred to as ud . prior to the development of this tagger , there was no pos tagger for maori from aotearoa . pos taggers tag words according to their syntactic or grammatical category . however , many traditional syntactic categories , and by consequence pos labels , do not ' work for ' maori . by this we mean that , for some of the traditional categories , the definition of , or guidelines for , an existing category is not suitable for maori . they do not have an existing category for certain word classes of maori . they do not reflect a maori worldview of the maori language .", "entity": "te reo maori", "output": "part - of - speech tagger", "neg_sample": ["te reo maori is done by using Method", "henceforth , part - of - speech will be referred to as pos throughout this paper and te reo maori will be referred to as maori , while universal dependencies will be referred to as ud .", "prior to the development of this tagger , there was no pos tagger for maori from aotearoa .", "pos taggers tag words according to their syntactic or grammatical category .", "however , many traditional syntactic categories , and by consequence pos labels , do not ' work for ' maori .", "by this we mean that , for some of the traditional categories , the definition of , or guidelines for , an existing category is not suitable for maori .", "they do not have an existing category for certain word classes of maori .", "they do not reflect a maori worldview of the maori language ."], "relation": "used for", "id": "2022.computel-1.12", "year": 2022, "rel_sent": "Developing a Part - Of - Speech tagger for te reo Maori.", "forward": false, "src_ids": "2022.computel-1.12_4062"}
+{"input": "part - of - speech tagger is used for Material| context: prior to the development of this tagger , there was no pos tagger for maori from aotearoa . pos taggers tag words according to their syntactic or grammatical category . however , many traditional syntactic categories , and by consequence pos labels , do not ' work for ' maori . by this we mean that , for some of the traditional categories , the definition of , or guidelines for , an existing category is not suitable for maori . they do not have an existing category for certain word classes of maori . they do not reflect a maori worldview of the maori language .", "entity": "part - of - speech tagger", "output": "te reo maori", "neg_sample": ["part - of - speech tagger is used for Material", "prior to the development of this tagger , there was no pos tagger for maori from aotearoa .", "pos taggers tag words according to their syntactic or grammatical category .", "however , many traditional syntactic categories , and by consequence pos labels , do not ' work for ' maori .", "by this we mean that , for some of the traditional categories , the definition of , or guidelines for , an existing category is not suitable for maori .", "they do not have an existing category for certain word classes of maori .", "they do not reflect a maori worldview of the maori language ."], "relation": "used for", "id": "2022.computel-1.12", "year": 2022, "rel_sent": "Developing a Part - Of - Speech tagger for te reo Maori.", "forward": true, "src_ids": "2022.computel-1.12_4063"}
+{"input": "product attribute value extraction is done by using Method| context: although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values . furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary . a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced .", "entity": "product attribute value extraction", "output": "extreme multi - label classification ( xmc )", "neg_sample": ["product attribute value extraction is done by using Method", "although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values .", "furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary .", "a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced ."], "relation": "used for", "id": "2022.ecnlp-1.16", "year": 2022, "rel_sent": "Extreme Multi - Label Classification with Label Masking for Product Attribute Value Extraction.", "forward": false, "src_ids": "2022.ecnlp-1.16_4064"}
+{"input": "label masking is used for Task| context: although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values . furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary . a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced .", "entity": "label masking", "output": "product attribute value extraction", "neg_sample": ["label masking is used for Task", "although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values .", "furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary .", "a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced ."], "relation": "used for", "id": "2022.ecnlp-1.16", "year": 2022, "rel_sent": "Extreme Multi - Label Classification with Label Masking for Product Attribute Value Extraction.", "forward": true, "src_ids": "2022.ecnlp-1.16_4065"}
+{"input": "extreme multi - label classification ( xmc ) is used for Task| context: although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values . furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary . a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced .", "entity": "extreme multi - label classification ( xmc )", "output": "product attribute value extraction", "neg_sample": ["extreme multi - label classification ( xmc ) is used for Task", "although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values .", "furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary .", "a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced ."], "relation": "used for", "id": "2022.ecnlp-1.16", "year": 2022, "rel_sent": "Extreme Multi - Label Classification with Label Masking for Product Attribute Value Extraction.", "forward": true, "src_ids": "2022.ecnlp-1.16_4066"}
+{"input": "e - commerce platforms is done by using Method| context: although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values . furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary . a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced .", "entity": "e - commerce platforms", "output": "attribute taxonomy", "neg_sample": ["e - commerce platforms is done by using Method", "although most studies have treated attribute value extraction ( ave ) as named entity recognition , these approaches are not practical in real - world e - commerce platforms because they perform poorly , and require canonicalization of extracted values .", "furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary .", "a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced ."], "relation": "used for", "id": "2022.ecnlp-1.16", "year": 2022, "rel_sent": "We exploit attribute taxonomy designed for e - commerce platforms to determine which labels are negative for products .", "forward": false, "src_ids": "2022.ecnlp-1.16_4067"}
+{"input": "attribute taxonomy is used for Material| context: furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary . a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced .", "entity": "attribute taxonomy", "output": "e - commerce platforms", "neg_sample": ["attribute taxonomy is used for Material", "furthermore , since values needed for actual services is static in many attributes , extraction of new values is not always necessary .", "a major problem in solving ave as xmc is that the distribution between positive and negative labels for products is heavily imbalanced ."], "relation": "used for", "id": "2022.ecnlp-1.16", "year": 2022, "rel_sent": "We exploit attribute taxonomy designed for e - commerce platforms to determine which labels are negative for products .", "forward": true, "src_ids": "2022.ecnlp-1.16_4068"}
+{"input": "document - level information extraction is done by using Task| context: document - level information extraction ( ie ) tasks have recently begun to be revisited in earnest using the end - to - end neural network techniques that have been successful on their sentence - level ie counterparts . evaluation of the approaches , however , has been limited in a number of dimensions . in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make .", "entity": "document - level information extraction", "output": "automatic error analysis", "neg_sample": ["document - level information extraction is done by using Task", "document - level information extraction ( ie ) tasks have recently begun to be revisited in earnest using the end - to - end neural network techniques that have been successful on their sentence - level ie counterparts .", "evaluation of the approaches , however , has been limited in a number of dimensions .", "in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make ."], "relation": "used for", "id": "2022.acl-long.274", "year": 2022, "rel_sent": "Automatic Error Analysis for Document - level Information Extraction.", "forward": false, "src_ids": "2022.acl-long.274_4069"}
+{"input": "automatic error analysis is used for Task| context: evaluation of the approaches , however , has been limited in a number of dimensions . in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make .", "entity": "automatic error analysis", "output": "document - level information extraction", "neg_sample": ["automatic error analysis is used for Task", "evaluation of the approaches , however , has been limited in a number of dimensions .", "in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make ."], "relation": "used for", "id": "2022.acl-long.274", "year": 2022, "rel_sent": "Automatic Error Analysis for Document - level Information Extraction.", "forward": true, "src_ids": "2022.acl-long.274_4070"}
+{"input": "error analysis is done by using Method| context: document - level information extraction ( ie ) tasks have recently begun to be revisited in earnest using the end - to - end neural network techniques that have been successful on their sentence - level ie counterparts . evaluation of the approaches , however , has been limited in a number of dimensions . in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make .", "entity": "error analysis", "output": "transformation - based framework", "neg_sample": ["error analysis is done by using Method", "document - level information extraction ( ie ) tasks have recently begun to be revisited in earnest using the end - to - end neural network techniques that have been successful on their sentence - level ie counterparts .", "evaluation of the approaches , however , has been limited in a number of dimensions .", "in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make ."], "relation": "used for", "id": "2022.acl-long.274", "year": 2022, "rel_sent": "We build on the work of Kummerfeld and Klein ( 2013 ) to propose a transformation - based framework for automating error analysis in document - level event and ( N - ary ) relation extraction .", "forward": false, "src_ids": "2022.acl-long.274_4071"}
+{"input": "transformation - based framework is used for Method| context: document - level information extraction ( ie ) tasks have recently begun to be revisited in earnest using the end - to - end neural network techniques that have been successful on their sentence - level ie counterparts . evaluation of the approaches , however , has been limited in a number of dimensions . in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make .", "entity": "transformation - based framework", "output": "error analysis", "neg_sample": ["transformation - based framework is used for Method", "document - level information extraction ( ie ) tasks have recently begun to be revisited in earnest using the end - to - end neural network techniques that have been successful on their sentence - level ie counterparts .", "evaluation of the approaches , however , has been limited in a number of dimensions .", "in particular , the precision / recall / f1 scores typically reported provide few insights on the range of errors the models make ."], "relation": "used for", "id": "2022.acl-long.274", "year": 2022, "rel_sent": "We build on the work of Kummerfeld and Klein ( 2013 ) to propose a transformation - based framework for automating error analysis in document - level event and ( N - ary ) relation extraction .", "forward": true, "src_ids": "2022.acl-long.274_4072"}
+{"input": "dependency parsing is done by using Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "dependency parsing", "output": "semi - supervised domain adaptation", "neg_sample": ["dependency parsing is done by using Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "Semi - supervised Domain Adaptation for Dependency Parsing with Dynamic Matching Network.", "forward": false, "src_ids": "2022.acl-long.74_4073"}
+{"input": "semi - supervised domain adaptation is used for Task| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "semi - supervised domain adaptation", "output": "dependency parsing", "neg_sample": ["semi - supervised domain adaptation is used for Task", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "Semi - supervised Domain Adaptation for Dependency Parsing with Dynamic Matching Network.", "forward": true, "src_ids": "2022.acl-long.74_4074"}
+{"input": "dynamic matching network is used for Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "entity": "dynamic matching network", "output": "shared - private model", "neg_sample": ["dynamic matching network is used for Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "To address this issue , we for the first time apply a dynamic matching network on the shared - private model for semi - supervised cross - domain dependency parsing .", "forward": true, "src_ids": "2022.acl-long.74_4075"}
+{"input": "semi - supervised cross - domain dependency parsing is done by using Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "semi - supervised cross - domain dependency parsing", "output": "dynamic matching network", "neg_sample": ["semi - supervised cross - domain dependency parsing is done by using Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "To address this issue , we for the first time apply a dynamic matching network on the shared - private model for semi - supervised cross - domain dependency parsing .", "forward": false, "src_ids": "2022.acl-long.74_4076"}
+{"input": "shared - private model is done by using Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "shared - private model", "output": "dynamic matching network", "neg_sample": ["shared - private model is done by using Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "To address this issue , we for the first time apply a dynamic matching network on the shared - private model for semi - supervised cross - domain dependency parsing .", "forward": false, "src_ids": "2022.acl-long.74_4077"}
+{"input": "training strategy is used for Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "training strategy", "output": "dynamic matching network", "neg_sample": ["training strategy is used for Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "Meanwhile , considering the scarcity of target - domain labeled data , we leverage unlabeled data from two aspects , i.e. , designing a new training strategy to improve the capability of the dynamic matching network and fine - tuning BERT to obtain domain - related contextualized representations .", "forward": true, "src_ids": "2022.acl-long.74_4078"}
+{"input": "dynamic matching network is used for Task| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "dynamic matching network", "output": "semi - supervised cross - domain dependency parsing", "neg_sample": ["dynamic matching network is used for Task", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "To address this issue , we for the first time apply a dynamic matching network on the shared - private model for semi - supervised cross - domain dependency parsing .", "forward": true, "src_ids": "2022.acl-long.74_4079"}
+{"input": "dynamic matching network is done by using Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "dynamic matching network", "output": "training strategy", "neg_sample": ["dynamic matching network is done by using Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "Meanwhile , considering the scarcity of target - domain labeled data , we leverage unlabeled data from two aspects , i.e. , designing a new training strategy to improve the capability of the dynamic matching network and fine - tuning BERT to obtain domain - related contextualized representations .", "forward": false, "src_ids": "2022.acl-long.74_4080"}
+{"input": "domain - related contextualized representations is done by using Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "domain - related contextualized representations", "output": "training strategy", "neg_sample": ["domain - related contextualized representations is done by using Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "Meanwhile , considering the scarcity of target - domain labeled data , we leverage unlabeled data from two aspects , i.e. , designing a new training strategy to improve the capability of the dynamic matching network and fine - tuning BERT to obtain domain - related contextualized representations .", "forward": false, "src_ids": "2022.acl-long.74_4081"}
+{"input": "fine - tuning bert is used for Method| context: supervised parsing models have achieved impressive results on in - domain texts . however , their performances drop drastically on out - of - domain texts due to the data distribution shift . the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones .", "entity": "fine - tuning bert", "output": "domain - related contextualized representations", "neg_sample": ["fine - tuning bert is used for Method", "supervised parsing models have achieved impressive results on in - domain texts .", "however , their performances drop drastically on out - of - domain texts due to the data distribution shift .", "the shared - private model has shown its promising advantages for alleviating this problem via feature separation , whereas prior works pay more attention to enhance shared features but neglect the in - depth relevance of specific ones ."], "relation": "used for", "id": "2022.acl-long.74", "year": 2022, "rel_sent": "Meanwhile , considering the scarcity of target - domain labeled data , we leverage unlabeled data from two aspects , i.e. , designing a new training strategy to improve the capability of the dynamic matching network and fine - tuning BERT to obtain domain - related contextualized representations .", "forward": true, "src_ids": "2022.acl-long.74_4082"}
+{"input": "machine translation ( mt ) engines is used for Material| context: it is often a challenging task to build machine translation ( mt ) engines for a specific domain due to the lack of parallel data in that area .", "entity": "machine translation ( mt ) engines", "output": "multilingual participatory spaces", "neg_sample": ["machine translation ( mt ) engines is used for Material", "it is often a challenging task to build machine translation ( mt ) engines for a specific domain due to the lack of parallel data in that area ."], "relation": "used for", "id": "2022.eamt-1.69", "year": 2022, "rel_sent": "Developing Machine Translation Engines for Multilingual Participatory Spaces.", "forward": true, "src_ids": "2022.eamt-1.69_4083"}
+{"input": "finer - grained input is used for Method| context: pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step . while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt .", "entity": "finer - grained input", "output": "canine", "neg_sample": ["finer - grained input is used for Method", "pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step .", "while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt ."], "relation": "used for", "id": "2022.tacl-1.5", "year": 2022, "rel_sent": "To use its finer - grained input effectively and efficiently , Canine combines downsampling , which reduces the input sequence length , with a deep transformer stack , which encodes context .", "forward": true, "src_ids": "2022.tacl-1.5_4084"}
+{"input": "language representation is done by using Method| context: pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step . while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt .", "entity": "language representation", "output": "tokenization - free encoder", "neg_sample": ["language representation is done by using Method", "pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step .", "while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt ."], "relation": "used for", "id": "2022.tacl-1.5", "year": 2022, "rel_sent": "Canine : Pre - training an Efficient Tokenization - Free Encoder for Language Representation.", "forward": false, "src_ids": "2022.tacl-1.5_4085"}
+{"input": "tokenization - free encoder is used for Method| context: pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step . while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt .", "entity": "tokenization - free encoder", "output": "language representation", "neg_sample": ["tokenization - free encoder is used for Method", "pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step .", "while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt ."], "relation": "used for", "id": "2022.tacl-1.5", "year": 2022, "rel_sent": "Canine : Pre - training an Efficient Tokenization - Free Encoder for Language Representation.", "forward": true, "src_ids": "2022.tacl-1.5_4086"}
+{"input": "soft inductive bias is done by using OtherScientificTerm| context: pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step . while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt .", "entity": "soft inductive bias", "output": "subwords", "neg_sample": ["soft inductive bias is done by using OtherScientificTerm", "pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step .", "while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt ."], "relation": "used for", "id": "2022.tacl-1.5", "year": 2022, "rel_sent": "In this paper , we present Canine , a neural encoder that operates directly on character sequences - without explicit tokenization or vocabulary - and a pre - training strategy that operates either directly on characters or optionally uses subwords as a soft inductive bias .", "forward": false, "src_ids": "2022.tacl-1.5_4087"}
+{"input": "subwords is used for OtherScientificTerm| context: pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step . while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt .", "entity": "subwords", "output": "soft inductive bias", "neg_sample": ["subwords is used for OtherScientificTerm", "pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step .", "while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt ."], "relation": "used for", "id": "2022.tacl-1.5", "year": 2022, "rel_sent": "In this paper , we present Canine , a neural encoder that operates directly on character sequences - without explicit tokenization or vocabulary - and a pre - training strategy that operates either directly on characters or optionally uses subwords as a soft inductive bias .", "forward": true, "src_ids": "2022.tacl-1.5_4088"}
+{"input": "canine is done by using OtherScientificTerm| context: pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step . while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt .", "entity": "canine", "output": "finer - grained input", "neg_sample": ["canine is done by using OtherScientificTerm", "pipelined nlp systems have largely been superseded by end - to - end neural modeling , yet nearly all commonly used models still require an explicit tokenization step .", "while recent tokenization approaches based on data - derived subword lexicons are less brittle than manually engineered tokenizers , these techniques are not equally suited to all languages , and the use of any fixed vocabulary may limit a model 's ability to adapt ."], "relation": "used for", "id": "2022.tacl-1.5", "year": 2022, "rel_sent": "To use its finer - grained input effectively and efficiently , Canine combines downsampling , which reduces the input sequence length , with a deep transformer stack , which encodes context .", "forward": false, "src_ids": "2022.tacl-1.5_4089"}
+{"input": "energy - based model is done by using Method| context: recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm .", "entity": "energy - based model", "output": "metropolis - hastings sampling scheme", "neg_sample": ["energy - based model is done by using Method", "recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm ."], "relation": "used for", "id": "2022.acl-long.31", "year": 2022, "rel_sent": "We use a Metropolis - Hastings sampling scheme to sample from this energy - based model using bidirectional context and global attribute features .", "forward": false, "src_ids": "2022.acl-long.31_4090"}
+{"input": "metropolis - hastings sampling scheme is used for Method| context: recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm .", "entity": "metropolis - hastings sampling scheme", "output": "energy - based model", "neg_sample": ["metropolis - hastings sampling scheme is used for Method", "recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm ."], "relation": "used for", "id": "2022.acl-long.31", "year": 2022, "rel_sent": "We use a Metropolis - Hastings sampling scheme to sample from this energy - based model using bidirectional context and global attribute features .", "forward": true, "src_ids": "2022.acl-long.31_4091"}
+{"input": "controllable text generation is done by using Method| context: recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm .", "entity": "controllable text generation", "output": "global score - based alternative", "neg_sample": ["controllable text generation is done by using Method", "recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm ."], "relation": "used for", "id": "2022.acl-long.31", "year": 2022, "rel_sent": "In this work , we propose Mix and Match LM , a global score - based alternative for controllable text generation that combines arbitrary pre - trained black - box models for achieving the desired attributes in the generated text without involving any fine - tuning or structural assumptions about the black - box models .", "forward": false, "src_ids": "2022.acl-long.31_4092"}
+{"input": "global score - based alternative is used for Task| context: recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm .", "entity": "global score - based alternative", "output": "controllable text generation", "neg_sample": ["global score - based alternative is used for Task", "recent work on controlled text generation has either required attribute - based fine - tuning of the base language model ( lm ) , or has restricted the parameterization of the attribute discriminator to be compatible with the base autoregressive lm ."], "relation": "used for", "id": "2022.acl-long.31", "year": 2022, "rel_sent": "In this work , we propose Mix and Match LM , a global score - based alternative for controllable text generation that combines arbitrary pre - trained black - box models for achieving the desired attributes in the generated text without involving any fine - tuning or structural assumptions about the black - box models .", "forward": true, "src_ids": "2022.acl-long.31_4093"}
+{"input": "speech translation is done by using Material| context: training speech translation ( st ) models requires large and high - quality datasets . must - c is one of the most widely used st benchmark datasets . it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions . this dataset passes several quality - control filters during creation . however , we find that must - c still suffers from three major quality issues : audiotext misalignment , inaccurate translation , and unnecessary speaker 's name . what are the impacts of these data quality issues for model development and evaluation ?", "entity": "speech translation", "output": "crowd - sourced data", "neg_sample": ["speech translation is done by using Material", "training speech translation ( st ) models requires large and high - quality datasets .", "must - c is one of the most widely used st benchmark datasets .", "it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions .", "this dataset passes several quality - control filters during creation .", "however , we find that must - c still suffers from three major quality issues : audiotext misalignment , inaccurate translation , and unnecessary speaker 's name .", "what are the impacts of these data quality issues for model development and evaluation ?"], "relation": "used for", "id": "2022.iwslt-1.9", "year": 2022, "rel_sent": "On the Impact of Noises in Crowd - Sourced Data for Speech Translation.", "forward": false, "src_ids": "2022.iwslt-1.9_4094"}
+{"input": "crowd - sourced data is used for Task| context: must - c is one of the most widely used st benchmark datasets . it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions . this dataset passes several quality - control filters during creation . however , we find that must - c still suffers from three major quality issues : audiotext misalignment , inaccurate translation , and unnecessary speaker 's name . what are the impacts of these data quality issues for model development and evaluation ?", "entity": "crowd - sourced data", "output": "speech translation", "neg_sample": ["crowd - sourced data is used for Task", "must - c is one of the most widely used st benchmark datasets .", "it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions .", "this dataset passes several quality - control filters during creation .", "however , we find that must - c still suffers from three major quality issues : audiotext misalignment , inaccurate translation , and unnecessary speaker 's name .", "what are the impacts of these data quality issues for model development and evaluation ?"], "relation": "used for", "id": "2022.iwslt-1.9", "year": 2022, "rel_sent": "On the Impact of Noises in Crowd - Sourced Data for Speech Translation.", "forward": true, "src_ids": "2022.iwslt-1.9_4095"}
+{"input": "automatic method is used for Generic| context: training speech translation ( st ) models requires large and high - quality datasets . must - c is one of the most widely used st benchmark datasets . it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions . this dataset passes several quality - control filters during creation .", "entity": "automatic method", "output": "quality issues", "neg_sample": ["automatic method is used for Generic", "training speech translation ( st ) models requires large and high - quality datasets .", "must - c is one of the most widely used st benchmark datasets .", "it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions .", "this dataset passes several quality - control filters during creation ."], "relation": "used for", "id": "2022.iwslt-1.9", "year": 2022, "rel_sent": "In this paper , we propose an automatic method tofix or filter the above quality issues , using English - German ( En - De ) translation as an example .", "forward": true, "src_ids": "2022.iwslt-1.9_4096"}
+{"input": "quality issues is done by using Method| context: training speech translation ( st ) models requires large and high - quality datasets . must - c is one of the most widely used st benchmark datasets . it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions . this dataset passes several quality - control filters during creation . however , we find that must - c still suffers from three major quality issues : audiotext misalignment , inaccurate translation , and unnecessary speaker 's name . what are the impacts of these data quality issues for model development and evaluation ?", "entity": "quality issues", "output": "automatic method", "neg_sample": ["quality issues is done by using Method", "training speech translation ( st ) models requires large and high - quality datasets .", "must - c is one of the most widely used st benchmark datasets .", "it contains around 400 hours of speech - transcript - translation data for each of the eight translation directions .", "this dataset passes several quality - control filters during creation .", "however , we find that must - c still suffers from three major quality issues : audiotext misalignment , inaccurate translation , and unnecessary speaker 's name .", "what are the impacts of these data quality issues for model development and evaluation ?"], "relation": "used for", "id": "2022.iwslt-1.9", "year": 2022, "rel_sent": "In this paper , we propose an automatic method tofix or filter the above quality issues , using English - German ( En - De ) translation as an example .", "forward": false, "src_ids": "2022.iwslt-1.9_4097"}
+{"input": "multi - step radiology report summarization is done by using Method| context: the impressions section of a radiology report about an imaging study is a summary of the radiologist 's reasoning and conclusions , and it also aids the referring physician in confirming or excluding certain diagnoses . a cascade of tasks are required to automatically generate an abstractive summary of the typical information - rich radiology report . these tasks include acquisition of salient content from the report and generation of a concise , easily consumable impressions section . prior research on radiology report summarization has focused on single - step end - to - end models - which subsume the task of salient content acquisition .", "entity": "multi - step radiology report summarization", "output": "multi - agent actor - critic", "neg_sample": ["multi - step radiology report summarization is done by using Method", "the impressions section of a radiology report about an imaging study is a summary of the radiologist 's reasoning and conclusions , and it also aids the referring physician in confirming or excluding certain diagnoses .", "a cascade of tasks are required to automatically generate an abstractive summary of the typical information - rich radiology report .", "these tasks include acquisition of salient content from the report and generation of a concise , easily consumable impressions section .", "prior research on radiology report summarization has focused on single - step end - to - end models - which subsume the task of salient content acquisition ."], "relation": "used for", "id": "2022.acl-long.109", "year": 2022, "rel_sent": "Differentiable Multi - Agent Actor - Critic for Multi - Step Radiology Report Summarization.", "forward": false, "src_ids": "2022.acl-long.109_4098"}
+{"input": "multi - agent actor - critic is used for Task| context: the impressions section of a radiology report about an imaging study is a summary of the radiologist 's reasoning and conclusions , and it also aids the referring physician in confirming or excluding certain diagnoses . a cascade of tasks are required to automatically generate an abstractive summary of the typical information - rich radiology report . these tasks include acquisition of salient content from the report and generation of a concise , easily consumable impressions section . prior research on radiology report summarization has focused on single - step end - to - end models - which subsume the task of salient content acquisition .", "entity": "multi - agent actor - critic", "output": "multi - step radiology report summarization", "neg_sample": ["multi - agent actor - critic is used for Task", "the impressions section of a radiology report about an imaging study is a summary of the radiologist 's reasoning and conclusions , and it also aids the referring physician in confirming or excluding certain diagnoses .", "a cascade of tasks are required to automatically generate an abstractive summary of the typical information - rich radiology report .", "these tasks include acquisition of salient content from the report and generation of a concise , easily consumable impressions section .", "prior research on radiology report summarization has focused on single - step end - to - end models - which subsume the task of salient content acquisition ."], "relation": "used for", "id": "2022.acl-long.109", "year": 2022, "rel_sent": "Differentiable Multi - Agent Actor - Critic for Multi - Step Radiology Report Summarization.", "forward": true, "src_ids": "2022.acl-long.109_4099"}
+{"input": "probabilistic graphical framework groupanno is used for Task| context: crowdsourcing has emerged as a popular approach for collecting annotated data to train supervised machine learning models . however , annotator bias can lead to defective annotations . though there are a few works investigating individual annotator bias , the group effects in annotators are largely overlooked .", "entity": "probabilistic graphical framework groupanno", "output": "modeling annotator group bias", "neg_sample": ["probabilistic graphical framework groupanno is used for Task", "crowdsourcing has emerged as a popular approach for collecting annotated data to train supervised machine learning models .", "however , annotator bias can lead to defective annotations .", "though there are a few works investigating individual annotator bias , the group effects in annotators are largely overlooked ."], "relation": "used for", "id": "2022.acl-long.126", "year": 2022, "rel_sent": "Then , we develop a novel probabilistic graphical framework GroupAnno to capture annotator group bias with an extended Expectation Maximization ( EM ) algorithm .", "forward": true, "src_ids": "2022.acl-long.126_4100"}
+{"input": "modeling annotator group bias is done by using Method| context: crowdsourcing has emerged as a popular approach for collecting annotated data to train supervised machine learning models . however , annotator bias can lead to defective annotations . though there are a few works investigating individual annotator bias , the group effects in annotators are largely overlooked .", "entity": "modeling annotator group bias", "output": "probabilistic graphical framework groupanno", "neg_sample": ["modeling annotator group bias is done by using Method", "crowdsourcing has emerged as a popular approach for collecting annotated data to train supervised machine learning models .", "however , annotator bias can lead to defective annotations .", "though there are a few works investigating individual annotator bias , the group effects in annotators are largely overlooked ."], "relation": "used for", "id": "2022.acl-long.126", "year": 2022, "rel_sent": "Then , we develop a novel probabilistic graphical framework GroupAnno to capture annotator group bias with an extended Expectation Maximization ( EM ) algorithm .", "forward": false, "src_ids": "2022.acl-long.126_4101"}
+{"input": "answerer is done by using Method| context: most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) .", "entity": "answerer", "output": "adaptive chain visual reasoning model ( acvrm )", "neg_sample": ["answerer is done by using Method", "most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) ."], "relation": "used for", "id": "2022.findings-acl.188", "year": 2022, "rel_sent": "An Adaptive Chain Visual Reasoning Model ( ACVRM ) for Answerer is also proposed , where the question - answer pair is used to update the visual representation sequentially .", "forward": false, "src_ids": "2022.findings-acl.188_4102"}
+{"input": "adaptive chain visual reasoning model ( acvrm ) is used for OtherScientificTerm| context: most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) .", "entity": "adaptive chain visual reasoning model ( acvrm )", "output": "answerer", "neg_sample": ["adaptive chain visual reasoning model ( acvrm ) is used for OtherScientificTerm", "most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) ."], "relation": "used for", "id": "2022.findings-acl.188", "year": 2022, "rel_sent": "An Adaptive Chain Visual Reasoning Model ( ACVRM ) for Answerer is also proposed , where the question - answer pair is used to update the visual representation sequentially .", "forward": true, "src_ids": "2022.findings-acl.188_4103"}
+{"input": "direct semantic connections is done by using Method| context: most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) .", "entity": "direct semantic connections", "output": "sqss", "neg_sample": ["direct semantic connections is done by using Method", "most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) ."], "relation": "used for", "id": "2022.findings-acl.188", "year": 2022, "rel_sent": "Further analyses show that SQSs help build direct semantic connections between questions and images , provide question - adaptive variable - length reasoning chains , and with explicit interpretability as well as error traceability .", "forward": false, "src_ids": "2022.findings-acl.188_4104"}
+{"input": "question - adaptive variable - length reasoning chains is done by using Method| context: most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) .", "entity": "question - adaptive variable - length reasoning chains", "output": "sqss", "neg_sample": ["question - adaptive variable - length reasoning chains is done by using Method", "most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) ."], "relation": "used for", "id": "2022.findings-acl.188", "year": 2022, "rel_sent": "Further analyses show that SQSs help build direct semantic connections between questions and images , provide question - adaptive variable - length reasoning chains , and with explicit interpretability as well as error traceability .", "forward": false, "src_ids": "2022.findings-acl.188_4105"}
+{"input": "sqss is used for OtherScientificTerm| context: most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) .", "entity": "sqss", "output": "direct semantic connections", "neg_sample": ["sqss is used for OtherScientificTerm", "most existing approaches to visual question answering ( vqa ) answer questions directly , however , people usually decompose a complex question into a sequence of simple sub questions and finally obtain the answer to the original question after answering the sub question sequence(sqs ) ."], "relation": "used for", "id": "2022.findings-acl.188", "year": 2022, "rel_sent": "Further analyses show that SQSs help build direct semantic connections between questions and images , provide question - adaptive variable - length reasoning chains , and with explicit interpretability as well as error traceability .", "forward": true, "src_ids": "2022.findings-acl.188_4106"}
+{"input": "predicting instance difficulty is done by using Method| context: early exiting allows instances to exit at different layers according to the estimation of difficulty . previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning . in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way . though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned .", "entity": "predicting instance difficulty", "output": "neural models", "neg_sample": ["predicting instance difficulty is done by using Method", "early exiting allows instances to exit at different layers according to the estimation of difficulty .", "previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning .", "in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way .", "though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned ."], "relation": "used for", "id": "2022.findings-acl.189", "year": 2022, "rel_sent": "As a response , we first conduct experiments on the learnability of instance difficulty , which demonstrates that modern neural models perform poorly on predicting instance difficulty .", "forward": false, "src_ids": "2022.findings-acl.189_4107"}
+{"input": "neural models is used for Task| context: early exiting allows instances to exit at different layers according to the estimation of difficulty . previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning . in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way . though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned .", "entity": "neural models", "output": "predicting instance difficulty", "neg_sample": ["neural models is used for Task", "early exiting allows instances to exit at different layers according to the estimation of difficulty .", "previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning .", "in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way .", "though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned ."], "relation": "used for", "id": "2022.findings-acl.189", "year": 2022, "rel_sent": "As a response , we first conduct experiments on the learnability of instance difficulty , which demonstrates that modern neural models perform poorly on predicting instance difficulty .", "forward": true, "src_ids": "2022.findings-acl.189_4108"}
+{"input": "model architectures is done by using Method| context: early exiting allows instances to exit at different layers according to the estimation of difficulty . previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning . in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way . though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned .", "entity": "model architectures", "output": "hashee", "neg_sample": ["model architectures is done by using Method", "early exiting allows instances to exit at different layers according to the estimation of difficulty .", "previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning .", "in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way .", "though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned ."], "relation": "used for", "id": "2022.findings-acl.189", "year": 2022, "rel_sent": "Different from previous methods , HashEE requires no internal classifiers nor extra parameters , and therefore is more efficient . HashEE can be used in various tasks ( including language understanding and generation ) and model architectures such as seq2seq models .", "forward": false, "src_ids": "2022.findings-acl.189_4109"}
+{"input": "hashee is used for Method| context: early exiting allows instances to exit at different layers according to the estimation of difficulty . previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning . in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way . though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned .", "entity": "hashee", "output": "model architectures", "neg_sample": ["hashee is used for Method", "early exiting allows instances to exit at different layers according to the estimation of difficulty .", "previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty , which suffers from generalization and threshold - tuning .", "in contrast , learning to exit , or learning to predict instance difficulty is a more appealing way .", "though some effort has been devoted to employing such ' learn - to - exit ' modules , it is still unknown whether and how well the instance difficulty can be learned ."], "relation": "used for", "id": "2022.findings-acl.189", "year": 2022, "rel_sent": "Different from previous methods , HashEE requires no internal classifiers nor extra parameters , and therefore is more efficient . HashEE can be used in various tasks ( including language understanding and generation ) and model architectures such as seq2seq models .", "forward": true, "src_ids": "2022.findings-acl.189_4110"}
+{"input": "neural models is used for Generic| context: charts are commonly used for exploring data and communicating insights . generating natural language summaries from charts can be very helpful for people in inferring key insights that would otherwise require a lot of cognitive and perceptual efforts .", "entity": "neural models", "output": "problem variations", "neg_sample": ["neural models is used for Generic", "charts are commonly used for exploring data and communicating insights .", "generating natural language summaries from charts can be very helpful for people in inferring key insights that would otherwise require a lot of cognitive and perceptual efforts ."], "relation": "used for", "id": "2022.acl-long.277", "year": 2022, "rel_sent": "We also introduce a number of state - of - the - art neural models as baselines that utilize image captioning and data - to - text generation techniques to tackle two problem variations : one assumes the underlying data table of the chart is available while the other needs to extract data from chart images .", "forward": true, "src_ids": "2022.acl-long.277_4111"}
+{"input": "problem variations is done by using Method| context: charts are commonly used for exploring data and communicating insights . generating natural language summaries from charts can be very helpful for people in inferring key insights that would otherwise require a lot of cognitive and perceptual efforts .", "entity": "problem variations", "output": "neural models", "neg_sample": ["problem variations is done by using Method", "charts are commonly used for exploring data and communicating insights .", "generating natural language summaries from charts can be very helpful for people in inferring key insights that would otherwise require a lot of cognitive and perceptual efforts ."], "relation": "used for", "id": "2022.acl-long.277", "year": 2022, "rel_sent": "We also introduce a number of state - of - the - art neural models as baselines that utilize image captioning and data - to - text generation techniques to tackle two problem variations : one assumes the underlying data table of the chart is available while the other needs to extract data from chart images .", "forward": false, "src_ids": "2022.acl-long.277_4112"}
+{"input": "end - to - end framework is used for Task| context: while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content .", "entity": "end - to - end framework", "output": "machine reading", "neg_sample": ["end - to - end framework is used for Task", "while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content ."], "relation": "used for", "id": "2022.findings-acl.61", "year": 2022, "rel_sent": "We propose a new end - to - end framework that jointly models answer generation and machine reading .", "forward": true, "src_ids": "2022.findings-acl.61_4113"}
+{"input": "paragraph - length answer is done by using Task| context: while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content .", "entity": "paragraph - length answer", "output": "long - form question answering ( lfqa )", "neg_sample": ["paragraph - length answer is done by using Task", "while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content ."], "relation": "used for", "id": "2022.findings-acl.61", "year": 2022, "rel_sent": "Long - form question answering ( LFQA ) aims to generate a paragraph - length answer for a given question .", "forward": false, "src_ids": "2022.findings-acl.61_4114"}
+{"input": "long - form question answering ( lfqa ) is used for OtherScientificTerm| context: while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content .", "entity": "long - form question answering ( lfqa )", "output": "paragraph - length answer", "neg_sample": ["long - form question answering ( lfqa ) is used for OtherScientificTerm", "while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content ."], "relation": "used for", "id": "2022.findings-acl.61", "year": 2022, "rel_sent": "Long - form question answering ( LFQA ) aims to generate a paragraph - length answer for a given question .", "forward": true, "src_ids": "2022.findings-acl.61_4115"}
+{"input": "answer generation is done by using Method| context: while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content .", "entity": "answer generation", "output": "end - to - end framework", "neg_sample": ["answer generation is done by using Method", "while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content ."], "relation": "used for", "id": "2022.findings-acl.61", "year": 2022, "rel_sent": "We propose a new end - to - end framework that jointly models answer generation and machine reading .", "forward": false, "src_ids": "2022.findings-acl.61_4116"}
+{"input": "machine reading is done by using Method| context: while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content .", "entity": "machine reading", "output": "end - to - end framework", "neg_sample": ["machine reading is done by using Method", "while current work on lfqa using large pre - trained model for generation are effective at producing fluent and somewhat relevant content , one primary challenge lies in how to generate a faithful answer that has less hallucinated content ."], "relation": "used for", "id": "2022.findings-acl.61", "year": 2022, "rel_sent": "We propose a new end - to - end framework that jointly models answer generation and machine reading .", "forward": false, "src_ids": "2022.findings-acl.61_4117"}
+{"input": "substructure distribution projection ( subdp ) is used for Task| context: we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately .", "entity": "substructure distribution projection ( subdp )", "output": "zero - shot cross - lingual dependency parsing", "neg_sample": ["substructure distribution projection ( subdp ) is used for Task", "we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately ."], "relation": "used for", "id": "2022.acl-long.452", "year": 2022, "rel_sent": "Substructure Distribution Projection for Zero - Shot Cross - Lingual Dependency Parsing.", "forward": true, "src_ids": "2022.acl-long.452_4118"}
+{"input": "substructure distribution projection ( subdp ) is used for Task| context: we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately .", "entity": "substructure distribution projection ( subdp )", "output": "zero - shot cross - lingual dependency parsing", "neg_sample": ["substructure distribution projection ( subdp ) is used for Task", "we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately ."], "relation": "used for", "id": "2022.acl-long.452", "year": 2022, "rel_sent": "In addition , SubDP improves zero shot cross - lingual dependency parsing with very few ( e.g. , 50 ) supervised bitext pairs , across a broader range of target languages .", "forward": true, "src_ids": "2022.acl-long.452_4119"}
+{"input": "soft silver labels is done by using OtherScientificTerm| context: we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately .", "entity": "soft silver labels", "output": "projected distributions", "neg_sample": ["soft silver labels is done by using OtherScientificTerm", "we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately ."], "relation": "used for", "id": "2022.acl-long.452", "year": 2022, "rel_sent": "Models for the target domain can then be trained , using the projected distributions as soft silver labels .", "forward": false, "src_ids": "2022.acl-long.452_4120"}
+{"input": "projected distributions is used for OtherScientificTerm| context: we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately .", "entity": "projected distributions", "output": "soft silver labels", "neg_sample": ["projected distributions is used for OtherScientificTerm", "we present substructure distribution projection ( subdp ) , a technique that projects a distribution over structures in one domain to another , by projecting substructure distributions separately ."], "relation": "used for", "id": "2022.acl-long.452", "year": 2022, "rel_sent": "Models for the target domain can then be trained , using the projected distributions as soft silver labels .", "forward": true, "src_ids": "2022.acl-long.452_4121"}
+{"input": "machine learning classifiers is done by using Material| context: recent research has highlighted that natural language processing ( nlp ) systems exhibit a bias againstafrican american speakers . these errors are often caused by poor representation of linguistic features unique to african american english ( aae ) , which is due to the relatively low probability of occurrence for many such features . habitual ' be ' is isomorphic , and therefore ambiguous , with other forms of uninflected ' be ' found in both aae and general american english ( gae ) . this creates a clear challenge for bias in nlp technologies .", "entity": "machine learning classifiers", "output": "balanced corpus", "neg_sample": ["machine learning classifiers is done by using Material", "recent research has highlighted that natural language processing ( nlp ) systems exhibit a bias againstafrican american speakers .", "these errors are often caused by poor representation of linguistic features unique to african american english ( aae ) , which is due to the relatively low probability of occurrence for many such features .", "habitual ' be ' is isomorphic , and therefore ambiguous , with other forms of uninflected ' be ' found in both aae and general american english ( gae ) .", "this creates a clear challenge for bias in nlp technologies ."], "relation": "used for", "id": "2022.ltedi-1.9", "year": 2022, "rel_sent": "This balanced corpus trains unbiased machine learning classifiers , as demonstrated on a corpus of AAE transcribed texts , achieving .65 F1 score at classifying habitual ' be ' .", "forward": false, "src_ids": "2022.ltedi-1.9_4122"}
+{"input": "balanced corpus is used for Method| context: recent research has highlighted that natural language processing ( nlp ) systems exhibit a bias againstafrican american speakers . these errors are often caused by poor representation of linguistic features unique to african american english ( aae ) , which is due to the relatively low probability of occurrence for many such features . habitual ' be ' is isomorphic , and therefore ambiguous , with other forms of uninflected ' be ' found in both aae and general american english ( gae ) . this creates a clear challenge for bias in nlp technologies .", "entity": "balanced corpus", "output": "machine learning classifiers", "neg_sample": ["balanced corpus is used for Method", "recent research has highlighted that natural language processing ( nlp ) systems exhibit a bias againstafrican american speakers .", "these errors are often caused by poor representation of linguistic features unique to african american english ( aae ) , which is due to the relatively low probability of occurrence for many such features .", "habitual ' be ' is isomorphic , and therefore ambiguous , with other forms of uninflected ' be ' found in both aae and general american english ( gae ) .", "this creates a clear challenge for bias in nlp technologies ."], "relation": "used for", "id": "2022.ltedi-1.9", "year": 2022, "rel_sent": "This balanced corpus trains unbiased machine learning classifiers , as demonstrated on a corpus of AAE transcribed texts , achieving .65 F1 score at classifying habitual ' be ' .", "forward": true, "src_ids": "2022.ltedi-1.9_4123"}
+{"input": "low - resource machine translation is done by using Method| context: we leverage embedding duplication between aligned sub - words to extend the parent - child transfer learning method , so as to improve low - resource machine translation .", "entity": "low - resource machine translation", "output": "vest - pocket method", "neg_sample": ["low - resource machine translation is done by using Method", "we leverage embedding duplication between aligned sub - words to extend the parent - child transfer learning method , so as to improve low - resource machine translation ."], "relation": "used for", "id": "2022.acl-short.68", "year": 2022, "rel_sent": "Sub - Word Alignment is Still Useful : A Vest - Pocket Method for Enhancing Low - Resource Machine Translation.", "forward": false, "src_ids": "2022.acl-short.68_4124"}
+{"input": "language models is done by using Method| context: lm - bff ( citation ) achieves significant few - shot performance by using auto - generated prompts and adding demonstrations similar to an input example .", "entity": "language models", "output": "lm - bff - ms", "neg_sample": ["language models is done by using Method", "lm - bff ( citation ) achieves significant few - shot performance by using auto - generated prompts and adding demonstrations similar to an input example ."], "relation": "used for", "id": "2022.acl-short.34", "year": 2022, "rel_sent": "To improve the approach of LM - BFF , this paper proposes LM - BFF - MS - better few - shot fine - tuning of language models with multiple soft demonstrations by making its further extensions , which include 1 ) prompts with multiple demonstrations based on automatic generation of multiple label words ; and 2 ) soft demonstration memory which consists of multiple sequences of globally shared word embeddings for a similar context .", "forward": false, "src_ids": "2022.acl-short.34_4125"}
+{"input": "lm - bff - ms is used for Method| context: lm - bff ( citation ) achieves significant few - shot performance by using auto - generated prompts and adding demonstrations similar to an input example .", "entity": "lm - bff - ms", "output": "language models", "neg_sample": ["lm - bff - ms is used for Method", "lm - bff ( citation ) achieves significant few - shot performance by using auto - generated prompts and adding demonstrations similar to an input example ."], "relation": "used for", "id": "2022.acl-short.34", "year": 2022, "rel_sent": "To improve the approach of LM - BFF , this paper proposes LM - BFF - MS - better few - shot fine - tuning of language models with multiple soft demonstrations by making its further extensions , which include 1 ) prompts with multiple demonstrations based on automatic generation of multiple label words ; and 2 ) soft demonstration memory which consists of multiple sequences of globally shared word embeddings for a similar context .", "forward": true, "src_ids": "2022.acl-short.34_4126"}
+{"input": "language tasks is done by using OtherScientificTerm| context: pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g.", "entity": "language tasks", "output": "visual knowledge", "neg_sample": ["language tasks is done by using OtherScientificTerm", "pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g."], "relation": "used for", "id": "2022.acl-long.196", "year": 2022, "rel_sent": "Leveraging Visual Knowledge in Language Tasks : An Empirical Study on Intermediate Pre - training for Cross - Modal Knowledge Transfer.", "forward": false, "src_ids": "2022.acl-long.196_4127"}
+{"input": "visual knowledge is used for Task| context: pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g.", "entity": "visual knowledge", "output": "language tasks", "neg_sample": ["visual knowledge is used for Task", "pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g."], "relation": "used for", "id": "2022.acl-long.196", "year": 2022, "rel_sent": "Leveraging Visual Knowledge in Language Tasks : An Empirical Study on Intermediate Pre - training for Cross - Modal Knowledge Transfer.", "forward": true, "src_ids": "2022.acl-long.196_4128"}
+{"input": "cross - modal knowledge transfer is done by using Method| context: pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g.", "entity": "cross - modal knowledge transfer", "output": "intermediate pre - training", "neg_sample": ["cross - modal knowledge transfer is done by using Method", "pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g."], "relation": "used for", "id": "2022.acl-long.196", "year": 2022, "rel_sent": "Leveraging Visual Knowledge in Language Tasks : An Empirical Study on Intermediate Pre - training for Cross - Modal Knowledge Transfer.", "forward": false, "src_ids": "2022.acl-long.196_4129"}
+{"input": "intermediate pre - training is used for Task| context: pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g.", "entity": "intermediate pre - training", "output": "cross - modal knowledge transfer", "neg_sample": ["intermediate pre - training is used for Task", "pre - trained language models are still far from human performance in tasks that need understanding of properties ( e.g."], "relation": "used for", "id": "2022.acl-long.196", "year": 2022, "rel_sent": "Leveraging Visual Knowledge in Language Tasks : An Empirical Study on Intermediate Pre - training for Cross - Modal Knowledge Transfer.", "forward": true, "src_ids": "2022.acl-long.196_4130"}
+{"input": "nonsense words is done by using Task| context: people associate affective meanings to words - ' death ' is scary and sad while ' party ' is connotated with surprise and joy . this raises the question if the association is purely a product of the learned affective imports inherent to semantic meanings , or is also an effect of other features of words , e.g. , morphological and phonological patterns .", "entity": "nonsense words", "output": "emotion intensity analysis", "neg_sample": ["nonsense words is done by using Task", "people associate affective meanings to words - ' death ' is scary and sad while ' party ' is connotated with surprise and joy .", "this raises the question if the association is purely a product of the learned affective imports inherent to semantic meanings , or is also an effect of other features of words , e.g.", ", morphological and phonological patterns ."], "relation": "used for", "id": "2022.wassa-1.4", "year": 2022, "rel_sent": "' splink ' is happy and ' phrouth ' is scary : Emotion Intensity Analysis for Nonsense Words.", "forward": false, "src_ids": "2022.wassa-1.4_4131"}
+{"input": "emotion intensity analysis is used for Material| context: people associate affective meanings to words - ' death ' is scary and sad while ' party ' is connotated with surprise and joy . this raises the question if the association is purely a product of the learned affective imports inherent to semantic meanings , or is also an effect of other features of words , e.g. , morphological and phonological patterns .", "entity": "emotion intensity analysis", "output": "nonsense words", "neg_sample": ["emotion intensity analysis is used for Material", "people associate affective meanings to words - ' death ' is scary and sad while ' party ' is connotated with surprise and joy .", "this raises the question if the association is purely a product of the learned affective imports inherent to semantic meanings , or is also an effect of other features of words , e.g.", ", morphological and phonological patterns ."], "relation": "used for", "id": "2022.wassa-1.4", "year": 2022, "rel_sent": "' splink ' is happy and ' phrouth ' is scary : Emotion Intensity Analysis for Nonsense Words.", "forward": true, "src_ids": "2022.wassa-1.4_4132"}
+{"input": "embedding - based ea methods is done by using Method| context: embedding - based methods have attracted increasing attention in recent entity alignment ( ea ) studies . although great promise they can offer , there are still several limitations . the most notable is that they identify the aligned entities based on cosine similarity , ignoring the semantics underlying the embeddings themselves . furthermore , these methods are shortsighted , heuristically selecting the closest entity as the target and allowing multiple entities to match the same candidate .", "entity": "embedding - based ea methods", "output": "reinforcement learning ( rl)-based entity alignment framework", "neg_sample": ["embedding - based ea methods is done by using Method", "embedding - based methods have attracted increasing attention in recent entity alignment ( ea ) studies .", "although great promise they can offer , there are still several limitations .", "the most notable is that they identify the aligned entities based on cosine similarity , ignoring the semantics underlying the embeddings themselves .", "furthermore , these methods are shortsighted , heuristically selecting the closest entity as the target and allowing multiple entities to match the same candidate ."], "relation": "used for", "id": "2022.findings-acl.217", "year": 2022, "rel_sent": "The proposed reinforcement learning ( RL)-based entity alignment framework can be flexibly adapted to most embedding - based EA methods .", "forward": false, "src_ids": "2022.findings-acl.217_4133"}
+{"input": "reinforcement learning ( rl)-based entity alignment framework is used for Method| context: embedding - based methods have attracted increasing attention in recent entity alignment ( ea ) studies . although great promise they can offer , there are still several limitations . the most notable is that they identify the aligned entities based on cosine similarity , ignoring the semantics underlying the embeddings themselves . furthermore , these methods are shortsighted , heuristically selecting the closest entity as the target and allowing multiple entities to match the same candidate .", "entity": "reinforcement learning ( rl)-based entity alignment framework", "output": "embedding - based ea methods", "neg_sample": ["reinforcement learning ( rl)-based entity alignment framework is used for Method", "embedding - based methods have attracted increasing attention in recent entity alignment ( ea ) studies .", "although great promise they can offer , there are still several limitations .", "the most notable is that they identify the aligned entities based on cosine similarity , ignoring the semantics underlying the embeddings themselves .", "furthermore , these methods are shortsighted , heuristically selecting the closest entity as the target and allowing multiple entities to match the same candidate ."], "relation": "used for", "id": "2022.findings-acl.217", "year": 2022, "rel_sent": "The proposed reinforcement learning ( RL)-based entity alignment framework can be flexibly adapted to most embedding - based EA methods .", "forward": true, "src_ids": "2022.findings-acl.217_4134"}
+{"input": "prompt tuning is used for Task| context: prompt tuning has recently emerged as an effective method for adapting pre - trained language models to a number of language understanding and generation tasks .", "entity": "prompt tuning", "output": "low - resource semantic parsing", "neg_sample": ["prompt tuning is used for Task", "prompt tuning has recently emerged as an effective method for adapting pre - trained language models to a number of language understanding and generation tasks ."], "relation": "used for", "id": "2022.acl-short.17", "year": 2022, "rel_sent": "The Power of Prompt Tuning for Low - Resource Semantic Parsing.", "forward": true, "src_ids": "2022.acl-short.17_4135"}
+{"input": "nlp classifiers is done by using Method| context: modern nlp classifiers are known to return uncalibrated estimations of class posteriors . existing methods for posterior calibration rescale the predicted probabilities but often have an adverse impact on final classification accuracy , thus leading to poorer generalization .", "entity": "nlp classifiers", "output": "posterior calibrated training", "neg_sample": ["nlp classifiers is done by using Method", "modern nlp classifiers are known to return uncalibrated estimations of class posteriors .", "existing methods for posterior calibration rescale the predicted probabilities but often have an adverse impact on final classification accuracy , thus leading to poorer generalization ."], "relation": "used for", "id": "2022.findings-acl.290", "year": 2022, "rel_sent": "Platt - Bin : Efficient Posterior Calibrated Training for NLP Classifiers.", "forward": false, "src_ids": "2022.findings-acl.290_4136"}
+{"input": "posterior calibrated training is used for Method| context: existing methods for posterior calibration rescale the predicted probabilities but often have an adverse impact on final classification accuracy , thus leading to poorer generalization .", "entity": "posterior calibrated training", "output": "nlp classifiers", "neg_sample": ["posterior calibrated training is used for Method", "existing methods for posterior calibration rescale the predicted probabilities but often have an adverse impact on final classification accuracy , thus leading to poorer generalization ."], "relation": "used for", "id": "2022.findings-acl.290", "year": 2022, "rel_sent": "Platt - Bin : Efficient Posterior Calibrated Training for NLP Classifiers.", "forward": true, "src_ids": "2022.findings-acl.290_4137"}
+{"input": "word alignment is done by using Method| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "word alignment", "output": "structural supervision", "neg_sample": ["word alignment is done by using Method", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Structural Supervision for Word Alignment and Machine Translation.", "forward": false, "src_ids": "2022.findings-acl.322_4138"}
+{"input": "machine translation is done by using Method| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "machine translation", "output": "structural supervision", "neg_sample": ["machine translation is done by using Method", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Structural Supervision for Word Alignment and Machine Translation.", "forward": false, "src_ids": "2022.findings-acl.322_4139"}
+{"input": "structural supervision is used for Task| context: unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "structural supervision", "output": "word alignment", "neg_sample": ["structural supervision is used for Task", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Structural Supervision for Word Alignment and Machine Translation.", "forward": true, "src_ids": "2022.findings-acl.322_4140"}
+{"input": "dynamic graph convolution networks ( dgcn ) is used for Task| context: unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "dynamic graph convolution networks ( dgcn )", "output": "word alignment", "neg_sample": ["dynamic graph convolution networks ( dgcn ) is used for Task", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Particularly , we wo n't leverage any annotated syntactic graph of the target side during training , so we introduce Dynamic Graph Convolution Networks ( DGCN ) on observed target tokens to sequentially and simultaneously generate the target tokens and the corresponding syntactic graphs , and further guide the word alignment .", "forward": true, "src_ids": "2022.findings-acl.322_4141"}
+{"input": "multi - task learning is done by using Task| context: unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "multi - task learning", "output": "machine translation", "neg_sample": ["multi - task learning is done by using Task", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "In this work , we propose to incorporate the syntactic structure of both source and target tokens into the encoder - decoder framework , tightly correlating the internal logic of word alignment and machine translation for multi - task learning .", "forward": false, "src_ids": "2022.findings-acl.322_4142"}
+{"input": "machine translation is used for Method| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "machine translation", "output": "multi - task learning", "neg_sample": ["machine translation is used for Method", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "In this work , we propose to incorporate the syntactic structure of both source and target tokens into the encoder - decoder framework , tightly correlating the internal logic of word alignment and machine translation for multi - task learning .", "forward": true, "src_ids": "2022.findings-acl.322_4143"}
+{"input": "training is done by using OtherScientificTerm| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "training", "output": "annotated syntactic graph", "neg_sample": ["training is done by using OtherScientificTerm", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Particularly , we wo n't leverage any annotated syntactic graph of the target side during training , so we introduce Dynamic Graph Convolution Networks ( DGCN ) on observed target tokens to sequentially and simultaneously generate the target tokens and the corresponding syntactic graphs , and further guide the word alignment .", "forward": false, "src_ids": "2022.findings-acl.322_4144"}
+{"input": "annotated syntactic graph is used for Task| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "annotated syntactic graph", "output": "training", "neg_sample": ["annotated syntactic graph is used for Task", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Particularly , we wo n't leverage any annotated syntactic graph of the target side during training , so we introduce Dynamic Graph Convolution Networks ( DGCN ) on observed target tokens to sequentially and simultaneously generate the target tokens and the corresponding syntactic graphs , and further guide the word alignment .", "forward": true, "src_ids": "2022.findings-acl.322_4145"}
+{"input": "word alignment is done by using Method| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "word alignment", "output": "dynamic graph convolution networks ( dgcn )", "neg_sample": ["word alignment is done by using Method", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Particularly , we wo n't leverage any annotated syntactic graph of the target side during training , so we introduce Dynamic Graph Convolution Networks ( DGCN ) on observed target tokens to sequentially and simultaneously generate the target tokens and the corresponding syntactic graphs , and further guide the word alignment .", "forward": false, "src_ids": "2022.findings-acl.322_4146"}
+{"input": "syntactic graphs is done by using Method| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "syntactic graphs", "output": "dynamic graph convolution networks ( dgcn )", "neg_sample": ["syntactic graphs is done by using Method", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Particularly , we wo n't leverage any annotated syntactic graph of the target side during training , so we introduce Dynamic Graph Convolution Networks ( DGCN ) on observed target tokens to sequentially and simultaneously generate the target tokens and the corresponding syntactic graphs , and further guide the word alignment .", "forward": false, "src_ids": "2022.findings-acl.322_4147"}
+{"input": "dynamic graph convolution networks ( dgcn ) is used for OtherScientificTerm| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "dynamic graph convolution networks ( dgcn )", "output": "syntactic graphs", "neg_sample": ["dynamic graph convolution networks ( dgcn ) is used for OtherScientificTerm", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "Particularly , we wo n't leverage any annotated syntactic graph of the target side during training , so we introduce Dynamic Graph Convolution Networks ( DGCN ) on observed target tokens to sequentially and simultaneously generate the target tokens and the corresponding syntactic graphs , and further guide the word alignment .", "forward": true, "src_ids": "2022.findings-acl.322_4148"}
+{"input": "structured constraints is done by using Method| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "structured constraints", "output": "hierarchical graph random walks ( hgrw )", "neg_sample": ["structured constraints is done by using Method", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "On this basis , Hierarchical Graph Random Walks ( HGRW ) are performed on the syntactic graphs of both source and target sides , for incorporating structured constraints on machine translation outputs .", "forward": false, "src_ids": "2022.findings-acl.322_4149"}
+{"input": "hierarchical graph random walks ( hgrw ) is used for OtherScientificTerm| context: syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation . unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences .", "entity": "hierarchical graph random walks ( hgrw )", "output": "structured constraints", "neg_sample": ["hierarchical graph random walks ( hgrw ) is used for OtherScientificTerm", "syntactic structure has long been argued to be potentially useful for enforcing accurate word alignment and improving generalization performance of machine translation .", "unfortunately , existing wisdom demonstrates its significance by considering only the syntactic structure of source tokens , neglecting the rich structural information from target tokens and the structural similarity between the source and target sentences ."], "relation": "used for", "id": "2022.findings-acl.322", "year": 2022, "rel_sent": "On this basis , Hierarchical Graph Random Walks ( HGRW ) are performed on the syntactic graphs of both source and target sides , for incorporating structured constraints on machine translation outputs .", "forward": true, "src_ids": "2022.findings-acl.322_4150"}
+{"input": "word - level qe is done by using Method| context: recent quality estimation ( qe ) models based on multilingual pre - trained representations have achieved very competitive results in predicting the overall quality of translated sentences . however , detecting specifically which translated words are incorrect is a more challenging task , especially when dealing with limited amounts of training data . we hypothesize that , not unlike humans , successful qe models rely on translation errors to predict overall sentence quality .", "entity": "word - level qe", "output": "semi - supervised method", "neg_sample": ["word - level qe is done by using Method", "recent quality estimation ( qe ) models based on multilingual pre - trained representations have achieved very competitive results in predicting the overall quality of translated sentences .", "however , detecting specifically which translated words are incorrect is a more challenging task , especially when dealing with limited amounts of training data .", "we hypothesize that , not unlike humans , successful qe models rely on translation errors to predict overall sentence quality ."], "relation": "used for", "id": "2022.findings-acl.327", "year": 2022, "rel_sent": "We therefore ( i ) introduce a novel semi - supervised method for word - level QE ; and ( ii ) propose to use the QE task as a new benchmark for evaluating the plausibility of feature attribution , i.e.", "forward": false, "src_ids": "2022.findings-acl.327_4151"}
+{"input": "semi - supervised method is used for Task| context: recent quality estimation ( qe ) models based on multilingual pre - trained representations have achieved very competitive results in predicting the overall quality of translated sentences . however , detecting specifically which translated words are incorrect is a more challenging task , especially when dealing with limited amounts of training data . we hypothesize that , not unlike humans , successful qe models rely on translation errors to predict overall sentence quality .", "entity": "semi - supervised method", "output": "word - level qe", "neg_sample": ["semi - supervised method is used for Task", "recent quality estimation ( qe ) models based on multilingual pre - trained representations have achieved very competitive results in predicting the overall quality of translated sentences .", "however , detecting specifically which translated words are incorrect is a more challenging task , especially when dealing with limited amounts of training data .", "we hypothesize that , not unlike humans , successful qe models rely on translation errors to predict overall sentence quality ."], "relation": "used for", "id": "2022.findings-acl.327", "year": 2022, "rel_sent": "We therefore ( i ) introduce a novel semi - supervised method for word - level QE ; and ( ii ) propose to use the QE task as a new benchmark for evaluating the plausibility of feature attribution , i.e.", "forward": true, "src_ids": "2022.findings-acl.327_4152"}
+{"input": "relaxed multi - document summarization is done by using Metric| context: multi - document summarization ( mds ) has made significant progress in recent years , in part facilitated by the availability of new , dedicated datasets and capacious language models . however , a standing limitation of these models is that they are trained against limited references and with plain maximum - likelihood objectives . as for many other generative tasks , reinforcement learning ( rl ) offers the potential to improve the training of mds models ; yet , it requires a carefully - designed reward that can ensure appropriate leverage of both the reference summaries and the input documents .", "entity": "relaxed multi - document summarization", "output": "multi - document coverage reward", "neg_sample": ["relaxed multi - document summarization is done by using Metric", "multi - document summarization ( mds ) has made significant progress in recent years , in part facilitated by the availability of new , dedicated datasets and capacious language models .", "however , a standing limitation of these models is that they are trained against limited references and with plain maximum - likelihood objectives .", "as for many other generative tasks , reinforcement learning ( rl ) offers the potential to improve the training of mds models ; yet , it requires a carefully - designed reward that can ensure appropriate leverage of both the reference summaries and the input documents ."], "relation": "used for", "id": "2022.acl-long.351", "year": 2022, "rel_sent": "A Multi - Document Coverage Reward for RELAXed Multi - Document Summarization.", "forward": false, "src_ids": "2022.acl-long.351_4153"}
+{"input": "multi - document coverage reward is used for Task| context: multi - document summarization ( mds ) has made significant progress in recent years , in part facilitated by the availability of new , dedicated datasets and capacious language models . however , a standing limitation of these models is that they are trained against limited references and with plain maximum - likelihood objectives . as for many other generative tasks , reinforcement learning ( rl ) offers the potential to improve the training of mds models ; yet , it requires a carefully - designed reward that can ensure appropriate leverage of both the reference summaries and the input documents .", "entity": "multi - document coverage reward", "output": "relaxed multi - document summarization", "neg_sample": ["multi - document coverage reward is used for Task", "multi - document summarization ( mds ) has made significant progress in recent years , in part facilitated by the availability of new , dedicated datasets and capacious language models .", "however , a standing limitation of these models is that they are trained against limited references and with plain maximum - likelihood objectives .", "as for many other generative tasks , reinforcement learning ( rl ) offers the potential to improve the training of mds models ; yet , it requires a carefully - designed reward that can ensure appropriate leverage of both the reference summaries and the input documents ."], "relation": "used for", "id": "2022.acl-long.351", "year": 2022, "rel_sent": "A Multi - Document Coverage Reward for RELAXed Multi - Document Summarization.", "forward": true, "src_ids": "2022.acl-long.351_4154"}
+{"input": "diverse predictions is done by using OtherScientificTerm| context: ensembling is a popular method used to improve performance as a last resort . however , ensembling multiple models finetuned from a single pretrained model has been not very effective ; this could be due to the lack of diversity among ensemble members .", "entity": "diverse predictions", "output": "winning - ticket subnetworks", "neg_sample": ["diverse predictions is done by using OtherScientificTerm", "ensembling is a popular method used to improve performance as a last resort .", "however , ensembling multiple models finetuned from a single pretrained model has been not very effective ; this could be due to the lack of diversity among ensemble members ."], "relation": "used for", "id": "2022.bigscience-1.4", "year": 2022, "rel_sent": "We empirically demonstrated that winning - ticket subnetworks produced more diverse predictions than dense networks and their ensemble outperformed the standard ensemble in some tasks when accurate lottery tickets are found on the tasks .", "forward": false, "src_ids": "2022.bigscience-1.4_4155"}
+{"input": "winning - ticket subnetworks is used for OtherScientificTerm| context: ensembling is a popular method used to improve performance as a last resort . however , ensembling multiple models finetuned from a single pretrained model has been not very effective ; this could be due to the lack of diversity among ensemble members .", "entity": "winning - ticket subnetworks", "output": "diverse predictions", "neg_sample": ["winning - ticket subnetworks is used for OtherScientificTerm", "ensembling is a popular method used to improve performance as a last resort .", "however , ensembling multiple models finetuned from a single pretrained model has been not very effective ; this could be due to the lack of diversity among ensemble members ."], "relation": "used for", "id": "2022.bigscience-1.4", "year": 2022, "rel_sent": "We empirically demonstrated that winning - ticket subnetworks produced more diverse predictions than dense networks and their ensemble outperformed the standard ensemble in some tasks when accurate lottery tickets are found on the tasks .", "forward": true, "src_ids": "2022.bigscience-1.4_4156"}
+{"input": "subword regularized models is done by using Method| context: subword regularizations use multiple subword segmentations during training to improve the robustness of neural machine translation models . in previous subword regularizations , we use multiple segmentations in the training process but use only one segmentation in the inference . in this study , we propose an inference strategy to address this discrepancy .", "entity": "subword regularized models", "output": "single model ensemble", "neg_sample": ["subword regularized models is done by using Method", "subword regularizations use multiple subword segmentations during training to improve the robustness of neural machine translation models .", "in previous subword regularizations , we use multiple segmentations in the training process but use only one segmentation in the inference .", "in this study , we propose an inference strategy to address this discrepancy ."], "relation": "used for", "id": "2022.findings-acl.199", "year": 2022, "rel_sent": "Single Model Ensemble for Subword Regularized Models in Low - Resource Machine Translation.", "forward": false, "src_ids": "2022.findings-acl.199_4157"}
+{"input": "single model ensemble is used for Method| context: subword regularizations use multiple subword segmentations during training to improve the robustness of neural machine translation models . in previous subword regularizations , we use multiple segmentations in the training process but use only one segmentation in the inference . in this study , we propose an inference strategy to address this discrepancy .", "entity": "single model ensemble", "output": "subword regularized models", "neg_sample": ["single model ensemble is used for Method", "subword regularizations use multiple subword segmentations during training to improve the robustness of neural machine translation models .", "in previous subword regularizations , we use multiple segmentations in the training process but use only one segmentation in the inference .", "in this study , we propose an inference strategy to address this discrepancy ."], "relation": "used for", "id": "2022.findings-acl.199", "year": 2022, "rel_sent": "Single Model Ensemble for Subword Regularized Models in Low - Resource Machine Translation.", "forward": true, "src_ids": "2022.findings-acl.199_4158"}
+{"input": "social dialogs is done by using Method| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "social dialogs", "output": "taxonomy of empathetic questions", "neg_sample": ["social dialogs is done by using Method", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "A Taxonomy of Empathetic Questions in Social Dialogs.", "forward": false, "src_ids": "2022.acl-long.211_4159"}
+{"input": "taxonomy of empathetic questions is used for Task| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "taxonomy of empathetic questions", "output": "social dialogs", "neg_sample": ["taxonomy of empathetic questions is used for Task", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "A Taxonomy of Empathetic Questions in Social Dialogs.", "forward": true, "src_ids": "2022.acl-long.211_4160"}
+{"input": "question - asking strategies is used for Task| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "question - asking strategies", "output": "social dialogs", "neg_sample": ["question - asking strategies is used for Task", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "These results reveal important question - asking strategies in social dialogs .", "forward": true, "src_ids": "2022.acl-long.211_4161"}
+{"input": "empatheticdialogues dataset is done by using Task| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "empatheticdialogues dataset", "output": "crowd - sourcing task", "neg_sample": ["empatheticdialogues dataset is done by using Task", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "We further design a crowd - sourcing task to annotate a large subset of the EmpatheticDialogues dataset with the established labels .", "forward": false, "src_ids": "2022.acl-long.211_4162"}
+{"input": "crowd - sourcing task is used for Material| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "crowd - sourcing task", "output": "empatheticdialogues dataset", "neg_sample": ["crowd - sourcing task is used for Material", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "We further design a crowd - sourcing task to annotate a large subset of the EmpatheticDialogues dataset with the established labels .", "forward": true, "src_ids": "2022.acl-long.211_4163"}
+{"input": "automatic labeling tools is done by using Material| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "automatic labeling tools", "output": "crowd - annotated data", "neg_sample": ["automatic labeling tools is done by using Material", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "We use the crowd - annotated data to develop automatic labeling tools and produce labels for the whole dataset .", "forward": false, "src_ids": "2022.acl-long.211_4164"}
+{"input": "crowd - annotated data is used for Method| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "crowd - annotated data", "output": "automatic labeling tools", "neg_sample": ["crowd - annotated data is used for Method", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "We use the crowd - annotated data to develop automatic labeling tools and produce labels for the whole dataset .", "forward": true, "src_ids": "2022.acl-long.211_4165"}
+{"input": "co - occurrences of question acts and intents is done by using Method| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "co - occurrences of question acts and intents", "output": "information visualization techniques", "neg_sample": ["co - occurrences of question acts and intents is done by using Method", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "Finally , we employ information visualization techniques to summarize co - occurrences of question acts and intents and their role in regulating interlocutor 's emotion .", "forward": false, "src_ids": "2022.acl-long.211_4166"}
+{"input": "information visualization techniques is used for OtherScientificTerm| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "information visualization techniques", "output": "co - occurrences of question acts and intents", "neg_sample": ["information visualization techniques is used for OtherScientificTerm", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "Finally , we employ information visualization techniques to summarize co - occurrences of question acts and intents and their role in regulating interlocutor 's emotion .", "forward": true, "src_ids": "2022.acl-long.211_4167"}
+{"input": "social dialogs is done by using Method| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "social dialogs", "output": "question - asking strategies", "neg_sample": ["social dialogs is done by using Method", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "These results reveal important question - asking strategies in social dialogs .", "forward": false, "src_ids": "2022.acl-long.211_4168"}
+{"input": "computational analysis of questions in datasets is done by using Method| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "computational analysis of questions in datasets", "output": "eqt classification scheme", "neg_sample": ["computational analysis of questions in datasets is done by using Method", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "The EQT classification scheme can facilitate computational analysis of questions in datasets .", "forward": false, "src_ids": "2022.acl-long.211_4169"}
+{"input": "eqt classification scheme is used for Task| context: effective question - asking is a crucial component of a successful conversational chatbot . it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions . however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat .", "entity": "eqt classification scheme", "output": "computational analysis of questions in datasets", "neg_sample": ["eqt classification scheme is used for Task", "effective question - asking is a crucial component of a successful conversational chatbot .", "it could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker 's emotions .", "however , current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat ."], "relation": "used for", "id": "2022.acl-long.211", "year": 2022, "rel_sent": "The EQT classification scheme can facilitate computational analysis of questions in datasets .", "forward": true, "src_ids": "2022.acl-long.211_4170"}
+{"input": "social meaning detection is done by using Material| context: masked language models ( mlms ) are pre - trained with a denoising objective that is in a mismatch with the objective of downstream fine - tuning .", "entity": "social meaning detection", "output": "twitter datasets", "neg_sample": ["social meaning detection is done by using Material", "masked language models ( mlms ) are pre - trained with a denoising objective that is in a mismatch with the objective of downstream fine - tuning ."], "relation": "used for", "id": "2022.wassa-1.14", "year": 2022, "rel_sent": "We test our models on 15 different Twitter datasets for social meaning detection .", "forward": false, "src_ids": "2022.wassa-1.14_4171"}
+{"input": "twitter datasets is used for Task| context: masked language models ( mlms ) are pre - trained with a denoising objective that is in a mismatch with the objective of downstream fine - tuning .", "entity": "twitter datasets", "output": "social meaning detection", "neg_sample": ["twitter datasets is used for Task", "masked language models ( mlms ) are pre - trained with a denoising objective that is in a mismatch with the objective of downstream fine - tuning ."], "relation": "used for", "id": "2022.wassa-1.14", "year": 2022, "rel_sent": "We test our models on 15 different Twitter datasets for social meaning detection .", "forward": true, "src_ids": "2022.wassa-1.14_4172"}
+{"input": "neural language models is done by using Task| context: this work connects language model adaptation with concepts of machine learning theory . we consider a training setup with a large out - of - domain set and a small in - domain set .", "entity": "neural language models", "output": "domain adaptation", "neg_sample": ["neural language models is done by using Task", "this work connects language model adaptation with concepts of machine learning theory .", "we consider a training setup with a large out - of - domain set and a small in - domain set ."], "relation": "used for", "id": "2022.acl-long.264", "year": 2022, "rel_sent": "The Trade - offs of Domain Adaptation for Neural Language Models.", "forward": false, "src_ids": "2022.acl-long.264_4173"}
+{"input": "domain adaptation is used for Method| context: this work connects language model adaptation with concepts of machine learning theory . we consider a training setup with a large out - of - domain set and a small in - domain set .", "entity": "domain adaptation", "output": "neural language models", "neg_sample": ["domain adaptation is used for Method", "this work connects language model adaptation with concepts of machine learning theory .", "we consider a training setup with a large out - of - domain set and a small in - domain set ."], "relation": "used for", "id": "2022.acl-long.264", "year": 2022, "rel_sent": "The Trade - offs of Domain Adaptation for Neural Language Models.", "forward": true, "src_ids": "2022.acl-long.264_4174"}
+{"input": "radiology report generation is done by using Method| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "radiology report generation", "output": "reinforced cross - modal alignment", "neg_sample": ["radiology report generation is done by using Method", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "Reinforced Cross - modal Alignment for Radiology Report Generation.", "forward": false, "src_ids": "2022.findings-acl.38_4175"}
+{"input": "reinforced cross - modal alignment is used for Task| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "reinforced cross - modal alignment", "output": "radiology report generation", "neg_sample": ["reinforced cross - modal alignment is used for Task", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "Reinforced Cross - modal Alignment for Radiology Report Generation.", "forward": true, "src_ids": "2022.findings-acl.38_4176"}
+{"input": "visual and textual features is used for Task| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "visual and textual features", "output": "radiology report generation", "neg_sample": ["visual and textual features is used for Task", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "In this paper , we propose an approach with reinforcement learning ( RL ) over a cross - modal memory ( CMM ) to better align visual and textual features for radiology report generation .", "forward": true, "src_ids": "2022.findings-acl.38_4177"}
+{"input": "visual and textual features is done by using OtherScientificTerm| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "visual and textual features", "output": "cross - modal memory ( cmm )", "neg_sample": ["visual and textual features is done by using OtherScientificTerm", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "In this paper , we propose an approach with reinforcement learning ( RL ) over a cross - modal memory ( CMM ) to better align visual and textual features for radiology report generation .", "forward": false, "src_ids": "2022.findings-acl.38_4178"}
+{"input": "radiology report generation is done by using OtherScientificTerm| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "radiology report generation", "output": "visual and textual features", "neg_sample": ["radiology report generation is done by using OtherScientificTerm", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "In this paper , we propose an approach with reinforcement learning ( RL ) over a cross - modal memory ( CMM ) to better align visual and textual features for radiology report generation .", "forward": false, "src_ids": "2022.findings-acl.38_4179"}
+{"input": "cross - modal memory ( cmm ) is used for OtherScientificTerm| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "cross - modal memory ( cmm )", "output": "visual and textual features", "neg_sample": ["cross - modal memory ( cmm ) is used for OtherScientificTerm", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "In this paper , we propose an approach with reinforcement learning ( RL ) over a cross - modal memory ( CMM ) to better align visual and textual features for radiology report generation .", "forward": true, "src_ids": "2022.findings-acl.38_4180"}
+{"input": "cross - modal alignment is done by using Method| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "cross - modal alignment", "output": "reinforced algorithm", "neg_sample": ["cross - modal alignment is done by using Method", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "In detail , a shared memory is used to record the mappings between visual and textual information , and the proposed reinforced algorithm is performed to learn the signal from the reports to guide the cross - modal alignment even though such reports are not directly related to how images and texts are mapped .", "forward": false, "src_ids": "2022.findings-acl.38_4181"}
+{"input": "reinforced algorithm is used for Task| context: medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload . in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports . although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment .", "entity": "reinforced algorithm", "output": "cross - modal alignment", "neg_sample": ["reinforced algorithm is used for Task", "medical images are widely used in clinical decision - making , where writing radiology reports is a potential application that can be enhanced by automatic solutions to alleviate physicians ' workload .", "in general , radiology report generation is an image - text task , where cross - modal mappings between images and texts play an important role in generating high - quality reports .", "although previous studies attempt tofacilitate the alignment via the co - attention mechanism under supervised settings , they suffer from lacking valid and accurate correspondences due to no annotation of such alignment ."], "relation": "used for", "id": "2022.findings-acl.38", "year": 2022, "rel_sent": "In detail , a shared memory is used to record the mappings between visual and textual information , and the proposed reinforced algorithm is performed to learn the signal from the reports to guide the cross - modal alignment even though such reports are not directly related to how images and texts are mapped .", "forward": true, "src_ids": "2022.findings-acl.38_4182"}
+{"input": "xbrl tagging is done by using Task| context: publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags . manually tagging the reports is tedious and costly .", "entity": "xbrl tagging", "output": "financial numeric entity recognition", "neg_sample": ["xbrl tagging is done by using Task", "publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags .", "manually tagging the reports is tedious and costly ."], "relation": "used for", "id": "2022.acl-long.303", "year": 2022, "rel_sent": "FiNER : Financial Numeric Entity Recognition for XBRL Tagging.", "forward": false, "src_ids": "2022.acl-long.303_4183"}
+{"input": "financial numeric entity recognition is used for Task| context: publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags . manually tagging the reports is tedious and costly .", "entity": "financial numeric entity recognition", "output": "xbrl tagging", "neg_sample": ["financial numeric entity recognition is used for Task", "publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags .", "manually tagging the reports is tedious and costly ."], "relation": "used for", "id": "2022.acl-long.303", "year": 2022, "rel_sent": "FiNER : Financial Numeric Entity Recognition for XBRL Tagging.", "forward": true, "src_ids": "2022.acl-long.303_4184"}
+{"input": "data and error analysis is used for Task| context: publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags . manually tagging the reports is tedious and costly .", "entity": "data and error analysis", "output": "xbrl tagging", "neg_sample": ["data and error analysis is used for Task", "publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags .", "manually tagging the reports is tedious and costly ."], "relation": "used for", "id": "2022.acl-long.303", "year": 2022, "rel_sent": "Through data and error analysis , we finally identify possible limitations to inspire future work on XBRL tagging .", "forward": true, "src_ids": "2022.acl-long.303_4185"}
+{"input": "financial domain is done by using Task| context: publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags . manually tagging the reports is tedious and costly .", "entity": "financial domain", "output": "entity extraction task", "neg_sample": ["financial domain is done by using Task", "publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags .", "manually tagging the reports is tedious and costly ."], "relation": "used for", "id": "2022.acl-long.303", "year": 2022, "rel_sent": "We , therefore , introduce XBRL tagging as a new entity extraction task for the financial domain and release FiNER-139 , a dataset of 1.1 M sentences with gold XBRL tags .", "forward": false, "src_ids": "2022.acl-long.303_4186"}
+{"input": "entity extraction task is used for Material| context: publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags . manually tagging the reports is tedious and costly .", "entity": "entity extraction task", "output": "financial domain", "neg_sample": ["entity extraction task is used for Material", "publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags .", "manually tagging the reports is tedious and costly ."], "relation": "used for", "id": "2022.acl-long.303", "year": 2022, "rel_sent": "We , therefore , introduce XBRL tagging as a new entity extraction task for the financial domain and release FiNER-139 , a dataset of 1.1 M sentences with gold XBRL tags .", "forward": true, "src_ids": "2022.acl-long.303_4187"}
+{"input": "xbrl tagging is done by using Method| context: publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags . manually tagging the reports is tedious and costly .", "entity": "xbrl tagging", "output": "data and error analysis", "neg_sample": ["xbrl tagging is done by using Method", "publicly traded companies are required to submit periodic reports with extensive business reporting language ( xbrl ) word - level tags .", "manually tagging the reports is tedious and costly ."], "relation": "used for", "id": "2022.acl-long.303", "year": 2022, "rel_sent": "Through data and error analysis , we finally identify possible limitations to inspire future work on XBRL tagging .", "forward": false, "src_ids": "2022.acl-long.303_4188"}
+{"input": "active learning is done by using Method| context: in this paper , we argue that these lms are not adapted effectively to the downstream task during al and we explore ways to address this issue .", "entity": "active learning", "output": "pretrained language models", "neg_sample": ["active learning is done by using Method", "in this paper , we argue that these lms are not adapted effectively to the downstream task during al and we explore ways to address this issue ."], "relation": "used for", "id": "2022.acl-short.93", "year": 2022, "rel_sent": "On the Importance of Effectively Adapting Pretrained Language Models for Active Learning.", "forward": false, "src_ids": "2022.acl-short.93_4189"}
+{"input": "lm is done by using Method| context: recent active learning ( al ) approaches in natural language processing ( nlp ) proposed using off - the - shelf pretrained language models ( lms ) . in this paper , we argue that these lms are not adapted effectively to the downstream task during al and we explore ways to address this issue .", "entity": "lm", "output": "fine - tuning approach", "neg_sample": ["lm is done by using Method", "recent active learning ( al ) approaches in natural language processing ( nlp ) proposed using off - the - shelf pretrained language models ( lms ) .", "in this paper , we argue that these lms are not adapted effectively to the downstream task during al and we explore ways to address this issue ."], "relation": "used for", "id": "2022.acl-short.93", "year": 2022, "rel_sent": "We also propose a simple yet effective fine - tuning method to ensure that the adapted LM is properly trained in both low and high resource scenarios during AL .", "forward": false, "src_ids": "2022.acl-short.93_4190"}
+{"input": "question - answering ( qa ) model is used for Task| context: while entity retrieval models continue to advance their capabilities , our understanding of their wide - ranging applications is limited , especially in domain - specific settings .", "entity": "question - answering ( qa ) model", "output": "financial qa task", "neg_sample": ["question - answering ( qa ) model is used for Task", "while entity retrieval models continue to advance their capabilities , our understanding of their wide - ranging applications is limited , especially in domain - specific settings ."], "relation": "used for", "id": "2022.deelio-1.6", "year": 2022, "rel_sent": "We highlighted this issue by using recent general - domain entity - linking models , LUKE and GENRE , to inject external knowledge into a question - answering ( QA ) model for a financial QA task with a hybrid tabular - textual dataset .", "forward": true, "src_ids": "2022.deelio-1.6_4191"}
+{"input": "financial qa task is done by using Method| context: while entity retrieval models continue to advance their capabilities , our understanding of their wide - ranging applications is limited , especially in domain - specific settings .", "entity": "financial qa task", "output": "question - answering ( qa ) model", "neg_sample": ["financial qa task is done by using Method", "while entity retrieval models continue to advance their capabilities , our understanding of their wide - ranging applications is limited , especially in domain - specific settings ."], "relation": "used for", "id": "2022.deelio-1.6", "year": 2022, "rel_sent": "We highlighted this issue by using recent general - domain entity - linking models , LUKE and GENRE , to inject external knowledge into a question - answering ( QA ) model for a financial QA task with a hybrid tabular - textual dataset .", "forward": false, "src_ids": "2022.deelio-1.6_4192"}
+{"input": "guarani - spanish machine translation is done by using OtherScientificTerm| context: machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data .", "entity": "guarani - spanish machine translation", "output": "word embeddings", "neg_sample": ["guarani - spanish machine translation is done by using OtherScientificTerm", "machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data ."], "relation": "used for", "id": "2022.computel-1.16", "year": 2022, "rel_sent": "Can We Use Word Embeddings for Enhancing Guarani - Spanish Machine Translation ?.", "forward": false, "src_ids": "2022.computel-1.16_4193"}
+{"input": "word embeddings is used for Task| context: machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data . one way of tackling it is using pretrained word embeddings for model initialization .", "entity": "word embeddings", "output": "guarani - spanish machine translation", "neg_sample": ["word embeddings is used for Task", "machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data .", "one way of tackling it is using pretrained word embeddings for model initialization ."], "relation": "used for", "id": "2022.computel-1.16", "year": 2022, "rel_sent": "Can We Use Word Embeddings for Enhancing Guarani - Spanish Machine Translation ?.", "forward": true, "src_ids": "2022.computel-1.16_4194"}
+{"input": "mt is done by using Method| context: machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data . one way of tackling it is using pretrained word embeddings for model initialization .", "entity": "mt", "output": "rich embeddings", "neg_sample": ["mt is done by using Method", "machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data .", "one way of tackling it is using pretrained word embeddings for model initialization ."], "relation": "used for", "id": "2022.computel-1.16", "year": 2022, "rel_sent": "In this work we try to check if currently available data is enough to train rich embeddings for enhancing MT for Guarani and Spanish , by building a set of word embedding collections and training MT systems using them .", "forward": false, "src_ids": "2022.computel-1.16_4195"}
+{"input": "rich embeddings is used for Task| context: machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data . one way of tackling it is using pretrained word embeddings for model initialization .", "entity": "rich embeddings", "output": "mt", "neg_sample": ["rich embeddings is used for Task", "machine translation for low - resource languages , such as guarani , is a challenging task due to the lack of data .", "one way of tackling it is using pretrained word embeddings for model initialization ."], "relation": "used for", "id": "2022.computel-1.16", "year": 2022, "rel_sent": "In this work we try to check if currently available data is enough to train rich embeddings for enhancing MT for Guarani and Spanish , by building a set of word embedding collections and training MT systems using them .", "forward": true, "src_ids": "2022.computel-1.16_4196"}
+{"input": "grammatical error correction is done by using Method| context: in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference . previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks . first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections .", "entity": "grammatical error correction", "output": "type - driven multi - turn corrections approach", "neg_sample": ["grammatical error correction is done by using Method", "in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference .", "previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks .", "first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections ."], "relation": "used for", "id": "2022.findings-acl.254", "year": 2022, "rel_sent": "Type - Driven Multi - Turn Corrections for Grammatical Error Correction.", "forward": false, "src_ids": "2022.findings-acl.254_4197"}
+{"input": "grammatical error correction is done by using Method| context: in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference . previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks . first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections .", "entity": "grammatical error correction", "output": "type - driven multi - turn corrections approach", "neg_sample": ["grammatical error correction is done by using Method", "in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference .", "previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks .", "first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections ."], "relation": "used for", "id": "2022.findings-acl.254", "year": 2022, "rel_sent": "Second , they ignore the interdependence between different types of corrections . In this paper , we propose a Type - Driven Multi - Turn Corrections approach for GEC .", "forward": false, "src_ids": "2022.findings-acl.254_4198"}
+{"input": "type - driven multi - turn corrections approach is used for Task| context: in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference . previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks . first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections .", "entity": "type - driven multi - turn corrections approach", "output": "grammatical error correction", "neg_sample": ["type - driven multi - turn corrections approach is used for Task", "in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference .", "previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks .", "first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections ."], "relation": "used for", "id": "2022.findings-acl.254", "year": 2022, "rel_sent": "Type - Driven Multi - Turn Corrections for Grammatical Error Correction.", "forward": true, "src_ids": "2022.findings-acl.254_4199"}
+{"input": "type - driven multi - turn corrections approach is used for Task| context: in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference . previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks . first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections .", "entity": "type - driven multi - turn corrections approach", "output": "grammatical error correction", "neg_sample": ["type - driven multi - turn corrections approach is used for Task", "in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference .", "previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks .", "first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections ."], "relation": "used for", "id": "2022.findings-acl.254", "year": 2022, "rel_sent": "Second , they ignore the interdependence between different types of corrections . In this paper , we propose a Type - Driven Multi - Turn Corrections approach for GEC .", "forward": true, "src_ids": "2022.findings-acl.254_4200"}
+{"input": "vae representations of text is done by using OtherScientificTerm| context: injecting desired geometric properties into text representations has attracted a lot of attention . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "vae representations of text", "output": "isotropy", "neg_sample": ["vae representations of text is done by using OtherScientificTerm", "injecting desired geometric properties into text representations has attracted a lot of attention .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "On the Effect of Isotropy on VAE Representations of Text.", "forward": false, "src_ids": "2022.acl-short.78_4201"}
+{"input": "training of vaes is done by using OtherScientificTerm| context: injecting desired geometric properties into text representations has attracted a lot of attention . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "training of vaes", "output": "isotropy", "neg_sample": ["training of vaes is done by using OtherScientificTerm", "injecting desired geometric properties into text representations has attracted a lot of attention .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "To address an aspect of this , we investigate the impact of injecting isotropy during training of VAEs .", "forward": false, "src_ids": "2022.acl-short.78_4202"}
+{"input": "isotropic gaussian posterior is used for OtherScientificTerm| context: injecting desired geometric properties into text representations has attracted a lot of attention . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "isotropic gaussian posterior", "output": "isotropy", "neg_sample": ["isotropic gaussian posterior is used for OtherScientificTerm", "injecting desired geometric properties into text representations has attracted a lot of attention .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "We illustrate that IGP effectively encourages isotropy in the representations , inducing a more discriminative latent space .", "forward": true, "src_ids": "2022.acl-short.78_4203"}
+{"input": "isotropy is used for Task| context: injecting desired geometric properties into text representations has attracted a lot of attention . a property that has been argued for , due to its better utilisation of representation space , is isotropy . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "isotropy", "output": "vae representations of text", "neg_sample": ["isotropy is used for Task", "injecting desired geometric properties into text representations has attracted a lot of attention .", "a property that has been argued for , due to its better utilisation of representation space , is isotropy .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "On the Effect of Isotropy on VAE Representations of Text.", "forward": true, "src_ids": "2022.acl-short.78_4204"}
+{"input": "isotropy is used for Task| context: injecting desired geometric properties into text representations has attracted a lot of attention . a property that has been argued for , due to its better utilisation of representation space , is isotropy . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "isotropy", "output": "training of vaes", "neg_sample": ["isotropy is used for Task", "injecting desired geometric properties into text representations has attracted a lot of attention .", "a property that has been argued for , due to its better utilisation of representation space , is isotropy .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "To address an aspect of this , we investigate the impact of injecting isotropy during training of VAEs .", "forward": true, "src_ids": "2022.acl-short.78_4205"}
+{"input": "discriminative latent space is done by using Method| context: injecting desired geometric properties into text representations has attracted a lot of attention . a property that has been argued for , due to its better utilisation of representation space , is isotropy . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "discriminative latent space", "output": "isotropic gaussian posterior", "neg_sample": ["discriminative latent space is done by using Method", "injecting desired geometric properties into text representations has attracted a lot of attention .", "a property that has been argued for , due to its better utilisation of representation space , is isotropy .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "We illustrate that IGP effectively encourages isotropy in the representations , inducing a more discriminative latent space .", "forward": false, "src_ids": "2022.acl-short.78_4206"}
+{"input": "isotropy is done by using Method| context: injecting desired geometric properties into text representations has attracted a lot of attention . a property that has been argued for , due to its better utilisation of representation space , is isotropy . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "isotropy", "output": "isotropic gaussian posterior", "neg_sample": ["isotropy is done by using Method", "injecting desired geometric properties into text representations has attracted a lot of attention .", "a property that has been argued for , due to its better utilisation of representation space , is isotropy .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "We illustrate that IGP effectively encourages isotropy in the representations , inducing a more discriminative latent space .", "forward": false, "src_ids": "2022.acl-short.78_4207"}
+{"input": "isotropic gaussian posterior is used for OtherScientificTerm| context: injecting desired geometric properties into text representations has attracted a lot of attention . a property that has been argued for , due to its better utilisation of representation space , is isotropy . in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space .", "entity": "isotropic gaussian posterior", "output": "discriminative latent space", "neg_sample": ["isotropic gaussian posterior is used for OtherScientificTerm", "injecting desired geometric properties into text representations has attracted a lot of attention .", "a property that has been argued for , due to its better utilisation of representation space , is isotropy .", "in parallel , vaes have been successful in areas of nlp , but are known for their sub - optimal utilisation of the representation space ."], "relation": "used for", "id": "2022.acl-short.78", "year": 2022, "rel_sent": "We illustrate that IGP effectively encourages isotropy in the representations , inducing a more discriminative latent space .", "forward": true, "src_ids": "2022.acl-short.78_4208"}
+{"input": "machine translation is done by using Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "machine translation", "output": "sequence - to - sequence models", "neg_sample": ["machine translation is done by using Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "Rethinking the Design of Sequence - to - Sequence Models for Efficient Machine Translation.", "forward": false, "src_ids": "2022.eamt-1.1_4209"}
+{"input": "read / write alternation is done by using Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "read / write alternation", "output": "deterministic agents", "neg_sample": ["read / write alternation is done by using Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": false, "src_ids": "2022.eamt-1.1_4210"}
+{"input": "deterministic agents is used for OtherScientificTerm| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "deterministic agents", "output": "read / write alternation", "neg_sample": ["deterministic agents is used for OtherScientificTerm", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": true, "src_ids": "2022.eamt-1.1_4211"}
+{"input": "decoding path is done by using Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "decoding path", "output": "dynamic agents", "neg_sample": ["decoding path is done by using Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": false, "src_ids": "2022.eamt-1.1_4212"}
+{"input": "dynamic agents is used for OtherScientificTerm| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "dynamic agents", "output": "decoding path", "neg_sample": ["dynamic agents is used for OtherScientificTerm", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": true, "src_ids": "2022.eamt-1.1_4213"}
+{"input": "low cost predictions is done by using Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "low cost predictions", "output": "depth - adaptive transformer decoders", "neg_sample": ["low cost predictions is done by using Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": false, "src_ids": "2022.eamt-1.1_4214"}
+{"input": "anytime prediction is done by using Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "anytime prediction", "output": "depth - adaptive transformer decoders", "neg_sample": ["anytime prediction is done by using Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": false, "src_ids": "2022.eamt-1.1_4215"}
+{"input": "sample - adaptive halting mechanisms is done by using Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "sample - adaptive halting mechanisms", "output": "depth - adaptive transformer decoders", "neg_sample": ["sample - adaptive halting mechanisms is done by using Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": false, "src_ids": "2022.eamt-1.1_4216"}
+{"input": "depth - adaptive transformer decoders is used for Method| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "depth - adaptive transformer decoders", "output": "anytime prediction", "neg_sample": ["depth - adaptive transformer decoders is used for Method", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": true, "src_ids": "2022.eamt-1.1_4217"}
+{"input": "depth - adaptive transformer decoders is used for OtherScientificTerm| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "depth - adaptive transformer decoders", "output": "low cost predictions", "neg_sample": ["depth - adaptive transformer decoders is used for OtherScientificTerm", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": true, "src_ids": "2022.eamt-1.1_4218"}
+{"input": "sample - adaptive halting mechanisms is used for OtherScientificTerm| context: in recent years , deep learning has enabled impressive achievements in machine translation . neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another . one crucial factor to the success of nmt is the design of new powerful and efficient architectures .", "entity": "sample - adaptive halting mechanisms", "output": "low cost predictions", "neg_sample": ["sample - adaptive halting mechanisms is used for OtherScientificTerm", "in recent years , deep learning has enabled impressive achievements in machine translation .", "neural machine translation ( nmt ) relies on training deep neural networks with large number of parameters on vast amounts of parallel data to learn how to translate from one language to another .", "one crucial factor to the success of nmt is the design of new powerful and efficient architectures ."], "relation": "used for", "id": "2022.eamt-1.1", "year": 2022, "rel_sent": "We investigate deterministic agents that guide the read / write alternation through a rigid decoding path , and introduce new dynamic agents to estimate a decoding path for each sample . We also address the resource - efficiency of encoder - decoder models and posit that going deeper in a neural network is not required for all instances . We design depth - adaptive Transformer decoders that allow for anytime prediction and sample - adaptive halting mechanisms tofavor low cost predictions for low complexity instances and save deeper predictions for complex scenarios .", "forward": true, "src_ids": "2022.eamt-1.1_4219"}
+{"input": "language technologies is used for Task| context: plains cree ( nehiyawewin ) is an indigenous language that is spoken in canada and the usa . it is the most widely spoken dialect of cree and a morphologically complex language that is polysynthetic , highly inflective , and agglutinative . it is an extremely low resource language , with no existing corpus that is both available and prepared for supporting the development of language technologies .", "entity": "language technologies", "output": "nehiyawewin revitalization and preservation", "neg_sample": ["language technologies is used for Task", "plains cree ( nehiyawewin ) is an indigenous language that is spoken in canada and the usa .", "it is the most widely spoken dialect of cree and a morphologically complex language that is polysynthetic , highly inflective , and agglutinative .", "it is an extremely low resource language , with no existing corpus that is both available and prepared for supporting the development of language technologies ."], "relation": "used for", "id": "2022.acl-long.440", "year": 2022, "rel_sent": "The data has been verified and cleaned ; it is ready for use in developing language technologies for nehiyawewin .", "forward": true, "src_ids": "2022.acl-long.440_4220"}
+{"input": "nehiyawewin revitalization and preservation is done by using Method| context: plains cree ( nehiyawewin ) is an indigenous language that is spoken in canada and the usa . it is the most widely spoken dialect of cree and a morphologically complex language that is polysynthetic , highly inflective , and agglutinative .", "entity": "nehiyawewin revitalization and preservation", "output": "language technologies", "neg_sample": ["nehiyawewin revitalization and preservation is done by using Method", "plains cree ( nehiyawewin ) is an indigenous language that is spoken in canada and the usa .", "it is the most widely spoken dialect of cree and a morphologically complex language that is polysynthetic , highly inflective , and agglutinative ."], "relation": "used for", "id": "2022.acl-long.440", "year": 2022, "rel_sent": "The data has been verified and cleaned ; it is ready for use in developing language technologies for nehiyawewin .", "forward": false, "src_ids": "2022.acl-long.440_4221"}
+{"input": "latent dirichlet allocation models is done by using Task| context: traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences . previous studies show that representing bigrams collocations in the input can improve topic coherence in english . however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai .", "entity": "latent dirichlet allocation models", "output": "collocation retokenization", "neg_sample": ["latent dirichlet allocation models is done by using Task", "traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences .", "previous studies show that representing bigrams collocations in the input can improve topic coherence in english .", "however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai ."], "relation": "used for", "id": "2022.findings-acl.212", "year": 2022, "rel_sent": "More Than Words : Collocation Retokenization for Latent Dirichlet Allocation Models.", "forward": false, "src_ids": "2022.findings-acl.212_4222"}
+{"input": "collocation retokenization is used for Method| context: traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences . previous studies show that representing bigrams collocations in the input can improve topic coherence in english . however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai .", "entity": "collocation retokenization", "output": "latent dirichlet allocation models", "neg_sample": ["collocation retokenization is used for Method", "traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences .", "previous studies show that representing bigrams collocations in the input can improve topic coherence in english .", "however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai ."], "relation": "used for", "id": "2022.findings-acl.212", "year": 2022, "rel_sent": "More Than Words : Collocation Retokenization for Latent Dirichlet Allocation Models.", "forward": true, "src_ids": "2022.findings-acl.212_4223"}
+{"input": "frequent token ngrams is done by using OtherScientificTerm| context: traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences . previous studies show that representing bigrams collocations in the input can improve topic coherence in english . however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai .", "entity": "frequent token ngrams", "output": "raw frequency", "neg_sample": ["frequent token ngrams is done by using OtherScientificTerm", "traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences .", "previous studies show that representing bigrams collocations in the input can improve topic coherence in english .", "however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai ."], "relation": "used for", "id": "2022.findings-acl.212", "year": 2022, "rel_sent": "Here , we explore the use of retokenization based on chi - squared measures , t - statistics , and raw frequency to merge frequent token ngrams into collocations when preparing input to the LDA model .", "forward": false, "src_ids": "2022.findings-acl.212_4224"}
+{"input": "raw frequency is used for OtherScientificTerm| context: traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences . previous studies show that representing bigrams collocations in the input can improve topic coherence in english . however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai .", "entity": "raw frequency", "output": "frequent token ngrams", "neg_sample": ["raw frequency is used for OtherScientificTerm", "traditionally , latent dirichlet allocation ( lda ) ingests words in a collection of documents to discover their latent topics using word - document co - occurrences .", "previous studies show that representing bigrams collocations in the input can improve topic coherence in english .", "however , it is unclear how to achieve the best results for languages without marked word boundaries such as chinese and thai ."], "relation": "used for", "id": "2022.findings-acl.212", "year": 2022, "rel_sent": "Here , we explore the use of retokenization based on chi - squared measures , t - statistics , and raw frequency to merge frequent token ngrams into collocations when preparing input to the LDA model .", "forward": true, "src_ids": "2022.findings-acl.212_4225"}
+{"input": "static senseful embeddings is done by using Method| context: we present a word - sense induction method based on pre - trained masked language models ( mlms ) , which can cheaply scale to large vocabularies and large corpora .", "entity": "static senseful embeddings", "output": "static word embeddings algorithm", "neg_sample": ["static senseful embeddings is done by using Method", "we present a word - sense induction method based on pre - trained masked language models ( mlms ) , which can cheaply scale to large vocabularies and large corpora ."], "relation": "used for", "id": "2022.acl-long.325", "year": 2022, "rel_sent": "Furthermore , by training a static word embeddings algorithm on the sense - tagged corpus , we obtain high - quality static senseful embeddings .", "forward": false, "src_ids": "2022.acl-long.325_4226"}
+{"input": "static word embeddings algorithm is used for OtherScientificTerm| context: we present a word - sense induction method based on pre - trained masked language models ( mlms ) , which can cheaply scale to large vocabularies and large corpora .", "entity": "static word embeddings algorithm", "output": "static senseful embeddings", "neg_sample": ["static word embeddings algorithm is used for OtherScientificTerm", "we present a word - sense induction method based on pre - trained masked language models ( mlms ) , which can cheaply scale to large vocabularies and large corpora ."], "relation": "used for", "id": "2022.acl-long.325", "year": 2022, "rel_sent": "Furthermore , by training a static word embeddings algorithm on the sense - tagged corpus , we obtain high - quality static senseful embeddings .", "forward": true, "src_ids": "2022.acl-long.325_4227"}
+{"input": "social talk is done by using Method| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "social talk", "output": "linguistically - driven reasoning dialog model", "neg_sample": ["social talk is done by using Method", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "Sketching a Linguistically - Driven Reasoning Dialog Model for Social Talk.", "forward": false, "src_ids": "2022.acl-srw.14_4228"}
+{"input": "linguistically - driven reasoning dialog model is used for Material| context: heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "linguistically - driven reasoning dialog model", "output": "social talk", "neg_sample": ["linguistically - driven reasoning dialog model is used for Material", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "Sketching a Linguistically - Driven Reasoning Dialog Model for Social Talk.", "forward": true, "src_ids": "2022.acl-srw.14_4229"}
+{"input": "linguistically - informed architecture is used for Material| context: heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "linguistically - informed architecture", "output": "social talk", "neg_sample": ["linguistically - informed architecture is used for Material", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "This work sketches out a more linguistically - informed architecture to handle social talk in English , in which corpus - based methods form the backbone of the relatively context - insensitive components ( e.g.", "forward": true, "src_ids": "2022.acl-srw.14_4230"}
+{"input": "social talk is done by using Method| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "social talk", "output": "linguistically - informed architecture", "neg_sample": ["social talk is done by using Method", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "This work sketches out a more linguistically - informed architecture to handle social talk in English , in which corpus - based methods form the backbone of the relatively context - insensitive components ( e.g.", "forward": false, "src_ids": "2022.acl-srw.14_4231"}
+{"input": "context - insensitive components is done by using Method| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "context - insensitive components", "output": "corpus - based methods", "neg_sample": ["context - insensitive components is done by using Method", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "This work sketches out a more linguistically - informed architecture to handle social talk in English , in which corpus - based methods form the backbone of the relatively context - insensitive components ( e.g.", "forward": false, "src_ids": "2022.acl-srw.14_4232"}
+{"input": "corpus - based methods is used for Method| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "corpus - based methods", "output": "context - insensitive components", "neg_sample": ["corpus - based methods is used for Method", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "This work sketches out a more linguistically - informed architecture to handle social talk in English , in which corpus - based methods form the backbone of the relatively context - insensitive components ( e.g.", "forward": true, "src_ids": "2022.acl-srw.14_4233"}
+{"input": "symbolic modeling is used for Method| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "symbolic modeling", "output": "context - insensitive components", "neg_sample": ["symbolic modeling is used for Method", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "part - of - speech tagging , approximation of lexical meaning and constituent chunking ) , while symbolic modeling is used for reasoning out the context - sensitive components , which do not have any consistent mapping to linguistic forms .", "forward": true, "src_ids": "2022.acl-srw.14_4234"}
+{"input": "context - insensitive components is done by using Task| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "context - insensitive components", "output": "symbolic modeling", "neg_sample": ["context - insensitive components is done by using Task", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "part - of - speech tagging , approximation of lexical meaning and constituent chunking ) , while symbolic modeling is used for reasoning out the context - sensitive components , which do not have any consistent mapping to linguistic forms .", "forward": false, "src_ids": "2022.acl-srw.14_4235"}
+{"input": "interactive and rational aspects of conversation is done by using Method| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "interactive and rational aspects of conversation", "output": "bayesian game - theoretic model", "neg_sample": ["interactive and rational aspects of conversation is done by using Method", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "All components are fitted into a Bayesian game - theoretic model to address the interactive and rational aspects of conversation .", "forward": false, "src_ids": "2022.acl-srw.14_4236"}
+{"input": "bayesian game - theoretic model is used for Task| context: the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots . heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context .", "entity": "bayesian game - theoretic model", "output": "interactive and rational aspects of conversation", "neg_sample": ["bayesian game - theoretic model is used for Task", "the capability of holding social talk ( or casual conversation ) and making sense of conversational content requires context - sensitive natural language understanding and reasoning , which can not be handled efficiently by the current popular open - domain dialog systems and chatbots .", "heavily relying on corpus - based machine learning techniques to encode and decode context - sensitive meanings , these systems focus on fitting a particular training dataset , but not tracking what is actually happening in a conversation , and therefore easily derail in a new context ."], "relation": "used for", "id": "2022.acl-srw.14", "year": 2022, "rel_sent": "All components are fitted into a Bayesian game - theoretic model to address the interactive and rational aspects of conversation .", "forward": true, "src_ids": "2022.acl-srw.14_4237"}
+{"input": "fact - oriented oie evaluation benchmarks is done by using Method| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "fact - oriented oie evaluation benchmarks", "output": "annie", "neg_sample": ["fact - oriented oie evaluation benchmarks is done by using Method", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We propose AnnIE : an interactive annotation platform that facilitates such challenging annotation tasks and supports creation of complete fact - oriented OIE evaluation benchmarks .", "forward": false, "src_ids": "2022.acl-demo.5_4238"}
+{"input": "annotation tasks is done by using Method| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "annotation tasks", "output": "annie", "neg_sample": ["annotation tasks is done by using Method", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We propose AnnIE : an interactive annotation platform that facilitates such challenging annotation tasks and supports creation of complete fact - oriented OIE evaluation benchmarks .", "forward": false, "src_ids": "2022.acl-demo.5_4239"}
+{"input": "use case scenarios is done by using Method| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "use case scenarios", "output": "annie", "neg_sample": ["use case scenarios is done by using Method", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "AnnIE is modular and flexible in order to support different use case scenarios ( i.e. , benchmarks covering different types of facts ) and different languages .", "forward": false, "src_ids": "2022.acl-demo.5_4240"}
+{"input": "oie benchmarks is done by using Method| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "oie benchmarks", "output": "annie", "neg_sample": ["oie benchmarks is done by using Method", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We use AnnIE to build two complete OIE benchmarks : one with verb - mediated facts and another with facts encompassing named entities .", "forward": false, "src_ids": "2022.acl-demo.5_4241"}
+{"input": "annie is used for Material| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "annie", "output": "oie benchmarks", "neg_sample": ["annie is used for Material", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We use AnnIE to build two complete OIE benchmarks : one with verb - mediated facts and another with facts encompassing named entities .", "forward": true, "src_ids": "2022.acl-demo.5_4242"}
+{"input": "interactive annotation platform is used for Task| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "interactive annotation platform", "output": "annotation tasks", "neg_sample": ["interactive annotation platform is used for Task", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We propose AnnIE : an interactive annotation platform that facilitates such challenging annotation tasks and supports creation of complete fact - oriented OIE evaluation benchmarks .", "forward": true, "src_ids": "2022.acl-demo.5_4243"}
+{"input": "annie is used for Task| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "annie", "output": "annotation tasks", "neg_sample": ["annie is used for Task", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We propose AnnIE : an interactive annotation platform that facilitates such challenging annotation tasks and supports creation of complete fact - oriented OIE evaluation benchmarks .", "forward": true, "src_ids": "2022.acl-demo.5_4244"}
+{"input": "annie is used for Material| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "annie", "output": "fact - oriented oie evaluation benchmarks", "neg_sample": ["annie is used for Material", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We propose AnnIE : an interactive annotation platform that facilitates such challenging annotation tasks and supports creation of complete fact - oriented OIE evaluation benchmarks .", "forward": true, "src_ids": "2022.acl-demo.5_4245"}
+{"input": "interactive annotation platform is used for Material| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "interactive annotation platform", "output": "fact - oriented oie evaluation benchmarks", "neg_sample": ["interactive annotation platform is used for Material", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "We propose AnnIE : an interactive annotation platform that facilitates such challenging annotation tasks and supports creation of complete fact - oriented OIE evaluation benchmarks .", "forward": true, "src_ids": "2022.acl-demo.5_4246"}
+{"input": "annie is used for OtherScientificTerm| context: open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner . intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence . to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e. , clusters of all acceptable surface realizations of the same fact ) from input sentences .", "entity": "annie", "output": "use case scenarios", "neg_sample": ["annie is used for OtherScientificTerm", "open information extraction ( oie ) is the task of extracting facts from sentences in the form of relations and their corresponding arguments in schema - free manner .", "intrinsic performance of oie systems is difficult to measure due to the incompleteness of existing oie benchmarks : ground truth extractions do not group all acceptable surface realizations of the same fact that can be extracted from a sentence .", "to measure performance of oie systems more realistically , it is necessary to manually annotate complete facts ( i.e.", ", clusters of all acceptable surface realizations of the same fact ) from input sentences ."], "relation": "used for", "id": "2022.acl-demo.5", "year": 2022, "rel_sent": "AnnIE is modular and flexible in order to support different use case scenarios ( i.e. , benchmarks covering different types of facts ) and different languages .", "forward": true, "src_ids": "2022.acl-demo.5_4247"}
+{"input": "long - form text generation tasks is done by using Task| context: despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow .", "entity": "long - form text generation tasks", "output": "dynamic content planning", "neg_sample": ["long - form text generation tasks is done by using Task", "despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow ."], "relation": "used for", "id": "2022.acl-long.163", "year": 2022, "rel_sent": "PLANET : Dynamic Content Planning in Autoregressive Transformers for Long - form Text Generation.", "forward": false, "src_ids": "2022.acl-long.163_4248"}
+{"input": "long - form text generation tasks is done by using Method| context: despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow .", "entity": "long - form text generation tasks", "output": "autoregressive transformers", "neg_sample": ["long - form text generation tasks is done by using Method", "despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow ."], "relation": "used for", "id": "2022.acl-long.163", "year": 2022, "rel_sent": "PLANET : Dynamic Content Planning in Autoregressive Transformers for Long - form Text Generation.", "forward": false, "src_ids": "2022.acl-long.163_4249"}
+{"input": "content planning is done by using Method| context: despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow .", "entity": "content planning", "output": "autoregressive self - attention mechanism", "neg_sample": ["content planning is done by using Method", "despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow ."], "relation": "used for", "id": "2022.acl-long.163", "year": 2022, "rel_sent": "In this work , we propose PLANET , a novel generation framework leveraging autoregressive self - attention mechanism to conduct content planning and surface realization dynamically .", "forward": false, "src_ids": "2022.acl-long.163_4250"}
+{"input": "surface realization is done by using Method| context: despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow .", "entity": "surface realization", "output": "autoregressive self - attention mechanism", "neg_sample": ["surface realization is done by using Method", "despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow ."], "relation": "used for", "id": "2022.acl-long.163", "year": 2022, "rel_sent": "In this work , we propose PLANET , a novel generation framework leveraging autoregressive self - attention mechanism to conduct content planning and surface realization dynamically .", "forward": false, "src_ids": "2022.acl-long.163_4251"}
+{"input": "autoregressive self - attention mechanism is used for Task| context: despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow .", "entity": "autoregressive self - attention mechanism", "output": "content planning", "neg_sample": ["autoregressive self - attention mechanism is used for Task", "despite recent progress of pre - trained language models on generating fluent text , existing methods still suffer from incoherence problems in long - form text generation tasks that require proper content control and planning toform a coherent high - level logical flow ."], "relation": "used for", "id": "2022.acl-long.163", "year": 2022, "rel_sent": "In this work , we propose PLANET , a novel generation framework leveraging autoregressive self - attention mechanism to conduct content planning and surface realization dynamically .", "forward": true, "src_ids": "2022.acl-long.163_4252"}
+{"input": "conditional random field is done by using Method| context: style transfer is the task of paraphrasing text into a target - style domain while retaining the content . unsupervised approaches mainly focus on training a generator to rewrite input sentences . in this work , we assume that text styles are determined by only a small proportion of words ; therefore , rewriting sentences via generative models may be unnecessary .", "entity": "conditional random field", "output": "language model", "neg_sample": ["conditional random field is done by using Method", "style transfer is the task of paraphrasing text into a target - style domain while retaining the content .", "unsupervised approaches mainly focus on training a generator to rewrite input sentences .", "in this work , we assume that text styles are determined by only a small proportion of words ; therefore , rewriting sentences via generative models may be unnecessary ."], "relation": "used for", "id": "2022.wassa-1.33", "year": 2022, "rel_sent": "We train a classifier and a language model to score tagged sequences and build a conditional random field .", "forward": false, "src_ids": "2022.wassa-1.33_4253"}
+{"input": "language model is used for Method| context: style transfer is the task of paraphrasing text into a target - style domain while retaining the content . unsupervised approaches mainly focus on training a generator to rewrite input sentences . in this work , we assume that text styles are determined by only a small proportion of words ; therefore , rewriting sentences via generative models may be unnecessary .", "entity": "language model", "output": "conditional random field", "neg_sample": ["language model is used for Method", "style transfer is the task of paraphrasing text into a target - style domain while retaining the content .", "unsupervised approaches mainly focus on training a generator to rewrite input sentences .", "in this work , we assume that text styles are determined by only a small proportion of words ; therefore , rewriting sentences via generative models may be unnecessary ."], "relation": "used for", "id": "2022.wassa-1.33", "year": 2022, "rel_sent": "We train a classifier and a language model to score tagged sequences and build a conditional random field .", "forward": true, "src_ids": "2022.wassa-1.33_4254"}
+{"input": "speaker to dialogue attribution is done by using Method| context: although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made .", "entity": "speaker to dialogue attribution", "output": "bert based dst style approach", "neg_sample": ["speaker to dialogue attribution is done by using Method", "although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made ."], "relation": "used for", "id": "2022.acl-long.400", "year": 2022, "rel_sent": "What does the sea say to the shore ? A BERT based DST style approach for speaker to dialogue attribution in novels.", "forward": false, "src_ids": "2022.acl-long.400_4255"}
+{"input": "bert based dst style approach is used for Task| context: although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made .", "entity": "bert based dst style approach", "output": "speaker to dialogue attribution", "neg_sample": ["bert based dst style approach is used for Task", "although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made ."], "relation": "used for", "id": "2022.acl-long.400", "year": 2022, "rel_sent": "What does the sea say to the shore ? A BERT based DST style approach for speaker to dialogue attribution in novels.", "forward": true, "src_ids": "2022.acl-long.400_4256"}
+{"input": "speaker attribution is done by using Method| context: although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made .", "entity": "speaker attribution", "output": "deep learning", "neg_sample": ["speaker attribution is done by using Method", "although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made ."], "relation": "used for", "id": "2022.acl-long.400", "year": 2022, "rel_sent": "This is the first application of deep learning to speaker attribution , and it shows that is possible to overcome the need for the hand - crafted features and rules used in the past .", "forward": false, "src_ids": "2022.acl-long.400_4257"}
+{"input": "deep learning is used for Task| context: although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made .", "entity": "deep learning", "output": "speaker attribution", "neg_sample": ["deep learning is used for Task", "although we find that existing systems can perform the first two tasks accurately , attributing characters to direct speech is a challenging problem due to the narrator 's lack of explicit character mentions , and the frequent use of nominal and pronominal coreference when such explicit mentions are made ."], "relation": "used for", "id": "2022.acl-long.400", "year": 2022, "rel_sent": "This is the first application of deep learning to speaker attribution , and it shows that is possible to overcome the need for the hand - crafted features and rules used in the past .", "forward": true, "src_ids": "2022.acl-long.400_4258"}
+{"input": "online dictionaries is done by using Method| context: many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format . however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian .", "entity": "online dictionaries", "output": "graph - based methods", "neg_sample": ["online dictionaries is done by using Method", "many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format .", "however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian ."], "relation": "used for", "id": "2022.computel-1.18", "year": 2022, "rel_sent": "Using Graph - Based Methods to Augment Online Dictionaries of Endangered Languages.", "forward": false, "src_ids": "2022.computel-1.18_4259"}
+{"input": "graph - based methods is used for Material| context: many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format . however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian .", "entity": "graph - based methods", "output": "online dictionaries", "neg_sample": ["graph - based methods is used for Material", "many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format .", "however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian ."], "relation": "used for", "id": "2022.computel-1.18", "year": 2022, "rel_sent": "Using Graph - Based Methods to Augment Online Dictionaries of Endangered Languages.", "forward": true, "src_ids": "2022.computel-1.18_4260"}
+{"input": "dictionaries is done by using Method| context: many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format . however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian .", "entity": "dictionaries", "output": "graph - based approaches", "neg_sample": ["dictionaries is done by using Method", "many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format .", "however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian ."], "relation": "used for", "id": "2022.computel-1.18", "year": 2022, "rel_sent": "We utilize graph - based approaches to augment such dictionaries by predicting new translations to existing and new languages based on different dictionaries for endangered languages and Wiktionaries .", "forward": false, "src_ids": "2022.computel-1.18_4261"}
+{"input": "dictionaries is used for Material| context: many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format . however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian .", "entity": "dictionaries", "output": "endangered languages", "neg_sample": ["dictionaries is used for Material", "many endangered uralic languages have multilingual machine readable dictionaries saved in an xml format .", "however , the dictionaries cover translations very inconsistently between language pairs , for instance , the livonian dictionary has some translations tofinnish , latvian and estonian , and the komi - zyrian dictionary has some translations tofinnish , english and russian ."], "relation": "used for", "id": "2022.computel-1.18", "year": 2022, "rel_sent": "We utilize graph - based approaches to augment such dictionaries by predicting new translations to existing and new languages based on different dictionaries for endangered languages and Wiktionaries .", "forward": true, "src_ids": "2022.computel-1.18_4262"}
+{"input": "vocabulary alignment is done by using OtherScientificTerm| context: recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process .", "entity": "vocabulary alignment", "output": "biomedical word embeddings", "neg_sample": ["vocabulary alignment is done by using OtherScientificTerm", "recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process ."], "relation": "used for", "id": "2022.insights-1.11", "year": 2022, "rel_sent": "Evaluating Biomedical Word Embeddings for Vocabulary Alignment at Scale in the UMLS Metathesaurus Using Siamese Networks.", "forward": false, "src_ids": "2022.insights-1.11_4263"}
+{"input": "biomedical word embeddings is used for Task| context: recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process .", "entity": "biomedical word embeddings", "output": "vocabulary alignment", "neg_sample": ["biomedical word embeddings is used for Task", "recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process ."], "relation": "used for", "id": "2022.insights-1.11", "year": 2022, "rel_sent": "Evaluating Biomedical Word Embeddings for Vocabulary Alignment at Scale in the UMLS Metathesaurus Using Siamese Networks.", "forward": true, "src_ids": "2022.insights-1.11_4264"}
+{"input": "synonym prediction is done by using Method| context: recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process .", "entity": "synonym prediction", "output": "contextualized word embeddings", "neg_sample": ["synonym prediction is done by using Method", "recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process ."], "relation": "used for", "id": "2022.insights-1.11", "year": 2022, "rel_sent": "We evaluate the use of contextualized word embeddings extracted from nine different biomedical BERT - based models for synonym prediction in the UMLS by replacing BioWordVec embeddings with embeddings extracted from each biomedical BERT model using different feature extraction methods .", "forward": false, "src_ids": "2022.insights-1.11_4265"}
+{"input": "embeddings is used for Method| context: recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process .", "entity": "embeddings", "output": "biomedical bert - based models", "neg_sample": ["embeddings is used for Method", "recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process ."], "relation": "used for", "id": "2022.insights-1.11", "year": 2022, "rel_sent": "We evaluate the use of contextualized word embeddings extracted from nine different biomedical BERT - based models for synonym prediction in the UMLS by replacing BioWordVec embeddings with embeddings extracted from each biomedical BERT model using different feature extraction methods .", "forward": true, "src_ids": "2022.insights-1.11_4266"}
+{"input": "biomedical bert - based models is used for Task| context: recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process .", "entity": "biomedical bert - based models", "output": "synonym prediction", "neg_sample": ["biomedical bert - based models is used for Task", "recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process ."], "relation": "used for", "id": "2022.insights-1.11", "year": 2022, "rel_sent": "We evaluate the use of contextualized word embeddings extracted from nine different biomedical BERT - based models for synonym prediction in the UMLS by replacing BioWordVec embeddings with embeddings extracted from each biomedical BERT model using different feature extraction methods .", "forward": true, "src_ids": "2022.insights-1.11_4267"}
+{"input": "contextualized word embeddings is used for Task| context: recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process .", "entity": "contextualized word embeddings", "output": "synonym prediction", "neg_sample": ["contextualized word embeddings is used for Task", "recent work uses a siamese network , initialized with biowordvec embeddings ( distributed word embeddings ) , for predicting synonymy among biomedical terms to automate a part of the umls ( unified medical language system ) metathesaurus construction process ."], "relation": "used for", "id": "2022.insights-1.11", "year": 2022, "rel_sent": "We evaluate the use of contextualized word embeddings extracted from nine different biomedical BERT - based models for synonym prediction in the UMLS by replacing BioWordVec embeddings with embeddings extracted from each biomedical BERT model using different feature extraction methods .", "forward": true, "src_ids": "2022.insights-1.11_4268"}
+{"input": "distantly supervised relation extraction is done by using Material| context: research in multi ds - re has remained limited due to the absence of a reliable benchmarking dataset . the only available dataset for this task , relx - distant ( koksal and ozgur , 2020 ) , displays several unrealistic characteristics , leading to a systematic overestimation of model performance .", "entity": "distantly supervised relation extraction", "output": "dis - rex", "neg_sample": ["distantly supervised relation extraction is done by using Material", "research in multi ds - re has remained limited due to the absence of a reliable benchmarking dataset .", "the only available dataset for this task , relx - distant ( koksal and ozgur , 2020 ) , displays several unrealistic characteristics , leading to a systematic overestimation of model performance ."], "relation": "used for", "id": "2022.acl-short.95", "year": 2022, "rel_sent": "DiS - ReX : A Multilingual Dataset for Distantly Supervised Relation Extraction.", "forward": false, "src_ids": "2022.acl-short.95_4269"}
+{"input": "multilingual dataset is used for Task| context: research in multi ds - re has remained limited due to the absence of a reliable benchmarking dataset . the only available dataset for this task , relx - distant ( koksal and ozgur , 2020 ) , displays several unrealistic characteristics , leading to a systematic overestimation of model performance .", "entity": "multilingual dataset", "output": "distantly supervised relation extraction", "neg_sample": ["multilingual dataset is used for Task", "research in multi ds - re has remained limited due to the absence of a reliable benchmarking dataset .", "the only available dataset for this task , relx - distant ( koksal and ozgur , 2020 ) , displays several unrealistic characteristics , leading to a systematic overestimation of model performance ."], "relation": "used for", "id": "2022.acl-short.95", "year": 2022, "rel_sent": "DiS - ReX : A Multilingual Dataset for Distantly Supervised Relation Extraction.", "forward": true, "src_ids": "2022.acl-short.95_4270"}
+{"input": "dis - rex is used for Task| context: research in multi ds - re has remained limited due to the absence of a reliable benchmarking dataset . the only available dataset for this task , relx - distant ( koksal and ozgur , 2020 ) , displays several unrealistic characteristics , leading to a systematic overestimation of model performance .", "entity": "dis - rex", "output": "distantly supervised relation extraction", "neg_sample": ["dis - rex is used for Task", "research in multi ds - re has remained limited due to the absence of a reliable benchmarking dataset .", "the only available dataset for this task , relx - distant ( koksal and ozgur , 2020 ) , displays several unrealistic characteristics , leading to a systematic overestimation of model performance ."], "relation": "used for", "id": "2022.acl-short.95", "year": 2022, "rel_sent": "DiS - ReX : A Multilingual Dataset for Distantly Supervised Relation Extraction.", "forward": true, "src_ids": "2022.acl-short.95_4271"}
+{"input": "recall is done by using Method| context: recall and ranking are two critical steps in personalized news recommendation . most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "recall", "output": "unified model learning", "neg_sample": ["recall is done by using Method", "recall and ranking are two critical steps in personalized news recommendation .", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "Two Birds with One Stone : Unified Model Learning for Both Recall and Ranking in News Recommendation.", "forward": false, "src_ids": "2022.findings-acl.274_4272"}
+{"input": "unified model learning is used for Metric| context: however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "unified model learning", "output": "recall", "neg_sample": ["unified model learning is used for Metric", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "Two Birds with One Stone : Unified Model Learning for Both Recall and Ranking in News Recommendation.", "forward": true, "src_ids": "2022.findings-acl.274_4273"}
+{"input": "user embedding is used for Metric| context: however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "user embedding", "output": "recall", "neg_sample": ["user embedding is used for Metric", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "Then we derive the user embedding for recall from the obtained user embedding for ranking by using it as the attention query to select a set of basis user embeddings which encode different general user interests and synthesize them into a user embedding for recall .", "forward": true, "src_ids": "2022.findings-acl.274_4274"}
+{"input": "unified method is used for Task| context: most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "unified method", "output": "news recommendation", "neg_sample": ["unified method is used for Task", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "In order to handle this problem , in this paper we propose UniRec , a unified method for recall and ranking in news recommendation .", "forward": true, "src_ids": "2022.findings-acl.274_4275"}
+{"input": "unified method is used for Task| context: however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "unified method", "output": "ranking", "neg_sample": ["unified method is used for Task", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "In order to handle this problem , in this paper we propose UniRec , a unified method for recall and ranking in news recommendation .", "forward": true, "src_ids": "2022.findings-acl.274_4276"}
+{"input": "user embedding is used for Task| context: however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "user embedding", "output": "ranking", "neg_sample": ["user embedding is used for Task", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "In our method , we first infer user embedding for ranking from the historical news click behaviors of a user using a user encoder model .", "forward": true, "src_ids": "2022.findings-acl.274_4277"}
+{"input": "user embedding is used for Task| context: however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "user embedding", "output": "ranking", "neg_sample": ["user embedding is used for Task", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "Then we derive the user embedding for recall from the obtained user embedding for ranking by using it as the attention query to select a set of basis user embeddings which encode different general user interests and synthesize them into a user embedding for recall .", "forward": true, "src_ids": "2022.findings-acl.274_4278"}
+{"input": "ranking is done by using Generic| context: recall and ranking are two critical steps in personalized news recommendation . most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "ranking", "output": "unified method", "neg_sample": ["ranking is done by using Generic", "recall and ranking are two critical steps in personalized news recommendation .", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "In order to handle this problem , in this paper we propose UniRec , a unified method for recall and ranking in news recommendation .", "forward": false, "src_ids": "2022.findings-acl.274_4279"}
+{"input": "news recommendation is done by using Generic| context: recall and ranking are two critical steps in personalized news recommendation . most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "news recommendation", "output": "unified method", "neg_sample": ["news recommendation is done by using Generic", "recall and ranking are two critical steps in personalized news recommendation .", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "In order to handle this problem , in this paper we propose UniRec , a unified method for recall and ranking in news recommendation .", "forward": false, "src_ids": "2022.findings-acl.274_4280"}
+{"input": "ranking is done by using OtherScientificTerm| context: recall and ranking are two critical steps in personalized news recommendation . most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "ranking", "output": "user embedding", "neg_sample": ["ranking is done by using OtherScientificTerm", "recall and ranking are two critical steps in personalized news recommendation .", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "In our method , we first infer user embedding for ranking from the historical news click behaviors of a user using a user encoder model .", "forward": false, "src_ids": "2022.findings-acl.274_4281"}
+{"input": "ranking is done by using OtherScientificTerm| context: recall and ranking are two critical steps in personalized news recommendation . most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "ranking", "output": "user embedding", "neg_sample": ["ranking is done by using OtherScientificTerm", "recall and ranking are two critical steps in personalized news recommendation .", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "Then we derive the user embedding for recall from the obtained user embedding for ranking by using it as the attention query to select a set of basis user embeddings which encode different general user interests and synthesize them into a user embedding for recall .", "forward": false, "src_ids": "2022.findings-acl.274_4282"}
+{"input": "recall is done by using OtherScientificTerm| context: recall and ranking are two critical steps in personalized news recommendation . most existing news recommender systems conduct personalized news recall and ranking separately with different models . however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems .", "entity": "recall", "output": "user embedding", "neg_sample": ["recall is done by using OtherScientificTerm", "recall and ranking are two critical steps in personalized news recommendation .", "most existing news recommender systems conduct personalized news recall and ranking separately with different models .", "however , maintaining multiple models leads to high computational cost and poses great challenges to meeting the online latency requirement of news recommender systems ."], "relation": "used for", "id": "2022.findings-acl.274", "year": 2022, "rel_sent": "Then we derive the user embedding for recall from the obtained user embedding for ranking by using it as the attention query to select a set of basis user embeddings which encode different general user interests and synthesize them into a user embedding for recall .", "forward": false, "src_ids": "2022.findings-acl.274_4283"}
+{"input": "lot identification problem is done by using Method| context: the term lot in is defined to mean an offering that contains a collection of multiple identical items for sale . in a large online marketplace , lot offerings play an important role , allowing buyers and sellers to set price levels to optimally balance supply and demand needs . in spite of their central role , platforms often struggle to identify lot offerings , since explicit lot status identification is frequently not provided by sellers . the ability to identify lot offerings plays a key role in many fundamental tasks , from matching offerings to catalog products , through ranking search results , to providing effective pricing guidance .", "entity": "lot identification problem", "output": "natural language techniques", "neg_sample": ["lot identification problem is done by using Method", "the term lot in is defined to mean an offering that contains a collection of multiple identical items for sale .", "in a large online marketplace , lot offerings play an important role , allowing buyers and sellers to set price levels to optimally balance supply and demand needs .", "in spite of their central role , platforms often struggle to identify lot offerings , since explicit lot status identification is frequently not provided by sellers .", "the ability to identify lot offerings plays a key role in many fundamental tasks , from matching offerings to catalog products , through ranking search results , to providing effective pricing guidance ."], "relation": "used for", "id": "2022.ecnlp-1.29", "year": 2022, "rel_sent": "We demonstrate experimentally the ability to accurately classify offerings as lots and predict their lot size using only the offer title , by adapting state - of - the - art natural language techniques to the lot identification problem .", "forward": false, "src_ids": "2022.ecnlp-1.29_4284"}
+{"input": "natural language techniques is used for Task| context: the term lot in is defined to mean an offering that contains a collection of multiple identical items for sale . in a large online marketplace , lot offerings play an important role , allowing buyers and sellers to set price levels to optimally balance supply and demand needs . in spite of their central role , platforms often struggle to identify lot offerings , since explicit lot status identification is frequently not provided by sellers . the ability to identify lot offerings plays a key role in many fundamental tasks , from matching offerings to catalog products , through ranking search results , to providing effective pricing guidance .", "entity": "natural language techniques", "output": "lot identification problem", "neg_sample": ["natural language techniques is used for Task", "the term lot in is defined to mean an offering that contains a collection of multiple identical items for sale .", "in a large online marketplace , lot offerings play an important role , allowing buyers and sellers to set price levels to optimally balance supply and demand needs .", "in spite of their central role , platforms often struggle to identify lot offerings , since explicit lot status identification is frequently not provided by sellers .", "the ability to identify lot offerings plays a key role in many fundamental tasks , from matching offerings to catalog products , through ranking search results , to providing effective pricing guidance ."], "relation": "used for", "id": "2022.ecnlp-1.29", "year": 2022, "rel_sent": "We demonstrate experimentally the ability to accurately classify offerings as lots and predict their lot size using only the offer title , by adapting state - of - the - art natural language techniques to the lot identification problem .", "forward": true, "src_ids": "2022.ecnlp-1.29_4285"}
+{"input": "deep semantic roles is done by using Method| context: in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure . argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms . in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information .", "entity": "deep semantic roles", "output": "hierarchical model", "neg_sample": ["deep semantic roles is done by using Method", "in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure .", "argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms .", "in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information ."], "relation": "used for", "id": "2022.scil-1.8", "year": 2022, "rel_sent": "In ad - dition , we probed incremental sentence processing by LMs through the lens of surprisal , and suggested that the hierarchical model may capture deep semantic roles that verbs assign to arguments , while the sequential models seem to be influenced by surface case alignments .", "forward": false, "src_ids": "2022.scil-1.8_4286"}
+{"input": "hierarchical model is used for OtherScientificTerm| context: in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure . argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms . in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information .", "entity": "hierarchical model", "output": "deep semantic roles", "neg_sample": ["hierarchical model is used for OtherScientificTerm", "in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure .", "argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms .", "in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information ."], "relation": "used for", "id": "2022.scil-1.8", "year": 2022, "rel_sent": "In ad - dition , we probed incremental sentence processing by LMs through the lens of surprisal , and suggested that the hierarchical model may capture deep semantic roles that verbs assign to arguments , while the sequential models seem to be influenced by surface case alignments .", "forward": true, "src_ids": "2022.scil-1.8_4287"}
+{"input": "language models is done by using OtherScientificTerm| context: in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure . argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms . in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information .", "entity": "language models", "output": "hierarchical bias", "neg_sample": ["language models is done by using OtherScientificTerm", "in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure .", "argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms .", "in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information ."], "relation": "used for", "id": "2022.scil-1.8", "year": 2022, "rel_sent": "We conclude that the explicit hierarchical bias is essential for LMs to learn argument structures like humans .", "forward": false, "src_ids": "2022.scil-1.8_4288"}
+{"input": "argument structures is done by using OtherScientificTerm| context: in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure . argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms . in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information .", "entity": "argument structures", "output": "hierarchical bias", "neg_sample": ["argument structures is done by using OtherScientificTerm", "in targeted syntactic evaluations , the syntactic competence of language models ( lms ) has been investigated through various syntactic phenomena , among which one of the important domains has been argument structure .", "argument structures in head - initial languages have been exclusively tested in the previous literature , but may be readily predicted from lexical information of verbs , potentially overestimating the syntactic competence of lms .", "in this paper , we explore whether argument structures can be learned by lms in head - final languages , which could be more challenging given that argument structures must be predicted before encountering verbs during incremental sentence processing , so that the relative weight of syntactic information should be heavier than lexical information ."], "relation": "used for", "id": "2022.scil-1.8", "year": 2022, "rel_sent": "We conclude that the explicit hierarchical bias is essential for LMs to learn argument structures like humans .", "forward": false, "src_ids": "2022.scil-1.8_4289"}
+{"input": "event - centric summary generation is used for Task| context: generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability . however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness .", "entity": "event - centric summary generation", "output": "educational question generation", "neg_sample": ["event - centric summary generation is used for Task", "generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability .", "however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness ."], "relation": "used for", "id": "2022.acl-long.348", "year": 2022, "rel_sent": "Our work indicates the necessity of decomposing question type distribution learning and event - centric summary generation for educational question generation .", "forward": true, "src_ids": "2022.acl-long.348_4290"}
+{"input": "educational question generation is done by using Method| context: generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability . however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness .", "entity": "educational question generation", "output": "event - centric summary generation", "neg_sample": ["educational question generation is done by using Method", "generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability .", "however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness ."], "relation": "used for", "id": "2022.acl-long.348", "year": 2022, "rel_sent": "Our work indicates the necessity of decomposing question type distribution learning and event - centric summary generation for educational question generation .", "forward": false, "src_ids": "2022.acl-long.348_4291"}
+{"input": "question type distribution is done by using Method| context: generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability . however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness .", "entity": "question type distribution", "output": "question generation method", "neg_sample": ["question type distribution is done by using Method", "generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability .", "however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness ."], "relation": "used for", "id": "2022.acl-long.348", "year": 2022, "rel_sent": "In this paper , we propose a novel question generation method that first learns the question type distribution of an input story paragraph , and then summarizes salient events which can be used to generate high - cognitive - demand questions .", "forward": false, "src_ids": "2022.acl-long.348_4292"}
+{"input": "question generation method is used for OtherScientificTerm| context: generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability . however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness .", "entity": "question generation method", "output": "question type distribution", "neg_sample": ["question generation method is used for OtherScientificTerm", "generating educational questions of fairytales or storybooks is vital for improving children 's literacy ability .", "however , it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness ."], "relation": "used for", "id": "2022.acl-long.348", "year": 2022, "rel_sent": "In this paper , we propose a novel question generation method that first learns the question type distribution of an input story paragraph , and then summarizes salient events which can be used to generate high - cognitive - demand questions .", "forward": true, "src_ids": "2022.acl-long.348_4293"}
+{"input": "prompt - based techniques is used for Task| context: as a recent development in few - shot learning , prompt - based techniques have demonstrated promising potential in a variety of natural language processing tasks .", "entity": "prompt - based techniques", "output": "semantic distinction", "neg_sample": ["prompt - based techniques is used for Task", "as a recent development in few - shot learning , prompt - based techniques have demonstrated promising potential in a variety of natural language processing tasks ."], "relation": "used for", "id": "2022.acl-short.36", "year": 2022, "rel_sent": "Our simple adaptation shows that the failure of existing prompt - based techniques in semantic distinction is due to their improper configuration , rather than lack of relevant knowledge in the representations .", "forward": true, "src_ids": "2022.acl-short.36_4294"}
+{"input": "relational triple extraction is done by using Task| context: relational triple extraction is a critical task for constructing knowledge graphs . existing methods focused on learning text patterns from explicit relational mentions . however , they usually suffered from ignoring relational reasoning patterns , thus failed to extract the implicitly implied triples . fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths . moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations .", "entity": "relational triple extraction", "output": "learning reasoning patterns", "neg_sample": ["relational triple extraction is done by using Task", "relational triple extraction is a critical task for constructing knowledge graphs .", "existing methods focused on learning text patterns from explicit relational mentions .", "however , they usually suffered from ignoring relational reasoning patterns , thus failed to extract the implicitly implied triples .", "fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths .", "moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations ."], "relation": "used for", "id": "2022.findings-acl.129", "year": 2022, "rel_sent": "Learning Reasoning Patterns for Relational Triple Extraction with Mutual Generation of Text and Graph.", "forward": false, "src_ids": "2022.findings-acl.129_4295"}
+{"input": "relational reasoning patterns is used for Task| context: relational triple extraction is a critical task for constructing knowledge graphs . existing methods focused on learning text patterns from explicit relational mentions . however , they usually suffered from ignoring relational reasoning patterns , thus failed to extract the implicitly implied triples . fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths . moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations .", "entity": "relational reasoning patterns", "output": "learning reasoning patterns", "neg_sample": ["relational reasoning patterns is used for Task", "relational triple extraction is a critical task for constructing knowledge graphs .", "existing methods focused on learning text patterns from explicit relational mentions .", "however , they usually suffered from ignoring relational reasoning patterns , thus failed to extract the implicitly implied triples .", "fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths .", "moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations ."], "relation": "used for", "id": "2022.findings-acl.129", "year": 2022, "rel_sent": "In this paper , we propose a unified framework to learn the relational reasoning patterns for this task .", "forward": true, "src_ids": "2022.findings-acl.129_4296"}
+{"input": "learning reasoning patterns is used for Task| context: existing methods focused on learning text patterns from explicit relational mentions . however , they usually suffered from ignoring relational reasoning patterns , thus failed to extract the implicitly implied triples . fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths . moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations .", "entity": "learning reasoning patterns", "output": "relational triple extraction", "neg_sample": ["learning reasoning patterns is used for Task", "existing methods focused on learning text patterns from explicit relational mentions .", "however , they usually suffered from ignoring relational reasoning patterns , thus failed to extract the implicitly implied triples .", "fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths .", "moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations ."], "relation": "used for", "id": "2022.findings-acl.129", "year": 2022, "rel_sent": "Learning Reasoning Patterns for Relational Triple Extraction with Mutual Generation of Text and Graph.", "forward": true, "src_ids": "2022.findings-acl.129_4297"}
+{"input": "learning reasoning patterns is done by using OtherScientificTerm| context: relational triple extraction is a critical task for constructing knowledge graphs . existing methods focused on learning text patterns from explicit relational mentions . fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths . moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations .", "entity": "learning reasoning patterns", "output": "relational reasoning patterns", "neg_sample": ["learning reasoning patterns is done by using OtherScientificTerm", "relational triple extraction is a critical task for constructing knowledge graphs .", "existing methods focused on learning text patterns from explicit relational mentions .", "fortunately , the graph structure of a sentence 's relational triples can help find multi - hop reasoning paths .", "moreover , the type inference logic through the paths can be captured with the sentence 's supplementary relational expressions that represent the real - world conceptual meanings of the paths ' composite relations ."], "relation": "used for", "id": "2022.findings-acl.129", "year": 2022, "rel_sent": "In this paper , we propose a unified framework to learn the relational reasoning patterns for this task .", "forward": false, "src_ids": "2022.findings-acl.129_4298"}
+{"input": "neural machine translation is done by using Task| context: neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions . this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks . in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training . neural networks , especially neural machine translation models , suffer from catastrophic forgetting even if they learn from a static training set . to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples .", "entity": "neural machine translation", "output": "balanced training", "neg_sample": ["neural machine translation is done by using Task", "neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions .", "this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks .", "in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training .", "neural networks , especially neural machine translation models , suffer from catastrophic forgetting even if they learn from a static training set .", "to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples ."], "relation": "used for", "id": "2022.acl-long.143", "year": 2022, "rel_sent": "Overcoming Catastrophic Forgetting beyond Continual Learning : Balanced Training for Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.143_4299"}
+{"input": "balanced training is used for Task| context: neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions . this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks . in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training . to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples . the underlying cause is that training samples do not get balanced training in each model update , so we name this problem imbalanced training .", "entity": "balanced training", "output": "neural machine translation", "neg_sample": ["balanced training is used for Task", "neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions .", "this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks .", "in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training .", "to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples .", "the underlying cause is that training samples do not get balanced training in each model update , so we name this problem imbalanced training ."], "relation": "used for", "id": "2022.acl-long.143", "year": 2022, "rel_sent": "Overcoming Catastrophic Forgetting beyond Continual Learning : Balanced Training for Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.143_4300"}
+{"input": "complementary knowledge is done by using Method| context: neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions . this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks . in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training . neural networks , especially neural machine translation models , suffer from catastrophic forgetting even if they learn from a static training set . to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples . the underlying cause is that training samples do not get balanced training in each model update , so we name this problem imbalanced training .", "entity": "complementary knowledge", "output": "dynamically updated teacher models", "neg_sample": ["complementary knowledge is done by using Method", "neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions .", "this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks .", "in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training .", "neural networks , especially neural machine translation models , suffer from catastrophic forgetting even if they learn from a static training set .", "to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples .", "the underlying cause is that training samples do not get balanced training in each model update , so we name this problem imbalanced training ."], "relation": "used for", "id": "2022.acl-long.143", "year": 2022, "rel_sent": "To alleviate this problem , we propose Complementary Online Knowledge Distillation ( COKD ) , which uses dynamically updated teacher models trained on specific data orders to iteratively provide complementary knowledge to the student model .", "forward": false, "src_ids": "2022.acl-long.143_4301"}
+{"input": "dynamically updated teacher models is used for OtherScientificTerm| context: neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions . this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks . in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training . neural networks , especially neural machine translation models , suffer from catastrophic forgetting even if they learn from a static training set . to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples . the underlying cause is that training samples do not get balanced training in each model update , so we name this problem imbalanced training .", "entity": "dynamically updated teacher models", "output": "complementary knowledge", "neg_sample": ["dynamically updated teacher models is used for OtherScientificTerm", "neural networks tend to gradually forget the previously learned knowledge when learning multiple tasks sequentially from dynamic data distributions .", "this problem is called catastrophic forgetting , which is a fundamental challenge in the continual learning of neural networks .", "in this work , we observe that catastrophic forgetting not only occurs in continual learning but also affects the traditional static training .", "neural networks , especially neural machine translation models , suffer from catastrophic forgetting even if they learn from a static training set .", "to be specific , the final model pays imbalanced attention to training samples , where recently exposed samples attract more attention than earlier samples .", "the underlying cause is that training samples do not get balanced training in each model update , so we name this problem imbalanced training ."], "relation": "used for", "id": "2022.acl-long.143", "year": 2022, "rel_sent": "To alleviate this problem , we propose Complementary Online Knowledge Distillation ( COKD ) , which uses dynamically updated teacher models trained on specific data orders to iteratively provide complementary knowledge to the student model .", "forward": true, "src_ids": "2022.acl-long.143_4302"}
+{"input": "triaffine mechanism is used for Task| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition . tofuse these heterogeneous factors , we propose a novel triaffine mechanism including triaffine attention and scoring . triaffine attention uses boundaries and labels as queries and uses inside tokens and related spans as keys and values for span representations . triaffine scoring interacts with boundaries and span representations for classification . experiments show that our proposed method outperforms previous span - based methods , achieves the state - of - the - art f1 scores on nested ner datasets genia and kbp2017 , and shows comparable results on ace2004 and ace2005 .", "entity": "triaffine mechanism", "output": "nested named entity recognition", "neg_sample": ["triaffine mechanism is used for Task", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "tofuse these heterogeneous factors , we propose a novel triaffine mechanism including triaffine attention and scoring .", "triaffine attention uses boundaries and labels as queries and uses inside tokens and related spans as keys and values for span representations .", "triaffine scoring interacts with boundaries and span representations for classification .", "experiments show that our proposed method outperforms previous span - based methods , achieves the state - of - the - art f1 scores on nested ner datasets genia and kbp2017 , and shows comparable results on ace2004 and ace2005 ."], "relation": "used for", "id": "2022.findings-acl.250", "year": 2022, "rel_sent": "Fusing Heterogeneous Factors with Triaffine Mechanism for Nested Named Entity Recognition.", "forward": true, "src_ids": "2022.findings-acl.250_4303"}
+{"input": "heterogeneous factors is used for Task| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition . tofuse these heterogeneous factors , we propose a novel triaffine mechanism including triaffine attention and scoring . triaffine attention uses boundaries and labels as queries and uses inside tokens and related spans as keys and values for span representations . triaffine scoring interacts with boundaries and span representations for classification . experiments show that our proposed method outperforms previous span - based methods , achieves the state - of - the - art f1 scores on nested ner datasets genia and kbp2017 , and shows comparable results on ace2004 and ace2005 .", "entity": "heterogeneous factors", "output": "nested named entity recognition", "neg_sample": ["heterogeneous factors is used for Task", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "tofuse these heterogeneous factors , we propose a novel triaffine mechanism including triaffine attention and scoring .", "triaffine attention uses boundaries and labels as queries and uses inside tokens and related spans as keys and values for span representations .", "triaffine scoring interacts with boundaries and span representations for classification .", "experiments show that our proposed method outperforms previous span - based methods , achieves the state - of - the - art f1 scores on nested ner datasets genia and kbp2017 , and shows comparable results on ace2004 and ace2005 ."], "relation": "used for", "id": "2022.findings-acl.250", "year": 2022, "rel_sent": "Fusing Heterogeneous Factors with Triaffine Mechanism for Nested Named Entity Recognition.", "forward": true, "src_ids": "2022.findings-acl.250_4304"}
+{"input": "rich syntactic knowledge is done by using Method| context: previous studies often rely on additional syntax - guided attention components to enhance the transformer , which require more parameters and additional syntactic parsing in downstream tasks . this increase in complexity severely limits the application of syntax - enhanced language model in a wide range of scenarios .", "entity": "rich syntactic knowledge", "output": "pre - trained language models", "neg_sample": ["rich syntactic knowledge is done by using Method", "previous studies often rely on additional syntax - guided attention components to enhance the transformer , which require more parameters and additional syntactic parsing in downstream tasks .", "this increase in complexity severely limits the application of syntax - enhanced language model in a wide range of scenarios ."], "relation": "used for", "id": "2022.findings-acl.191", "year": 2022, "rel_sent": "Based on constituency and dependency structures of syntax trees , we design phrase - guided and tree - guided contrastive objectives , and optimize them in the pre - training stage , so as to help the pre - trained language model to capture rich syntactic knowledge in its representations .", "forward": false, "src_ids": "2022.findings-acl.191_4305"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: syntactic information has been proved to be useful for transformer - based pre - trained language models . previous studies often rely on additional syntax - guided attention components to enhance the transformer , which require more parameters and additional syntactic parsing in downstream tasks . this increase in complexity severely limits the application of syntax - enhanced language model in a wide range of scenarios .", "entity": "pre - trained language models", "output": "rich syntactic knowledge", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "syntactic information has been proved to be useful for transformer - based pre - trained language models .", "previous studies often rely on additional syntax - guided attention components to enhance the transformer , which require more parameters and additional syntactic parsing in downstream tasks .", "this increase in complexity severely limits the application of syntax - enhanced language model in a wide range of scenarios ."], "relation": "used for", "id": "2022.findings-acl.191", "year": 2022, "rel_sent": "Based on constituency and dependency structures of syntax trees , we design phrase - guided and tree - guided contrastive objectives , and optimize them in the pre - training stage , so as to help the pre - trained language model to capture rich syntactic knowledge in its representations .", "forward": true, "src_ids": "2022.findings-acl.191_4306"}
+{"input": "mrc capability assessment framework is used for OtherScientificTerm| context: machine reading comprehension ( mrc ) reveals the ability to understand a given text passage and answer questions based on it . existing research works in mrc rely heavily on large - size models and corpus to improve the performance evaluated by metrics such as exact match ( em ) and f1 . however , such a paradigm lacks sufficient interpretation to model capability and can not efficiently train a model with a large corpus .", "entity": "mrc capability assessment framework", "output": "model capabilities", "neg_sample": ["mrc capability assessment framework is used for OtherScientificTerm", "machine reading comprehension ( mrc ) reveals the ability to understand a given text passage and answer questions based on it .", "existing research works in mrc rely heavily on large - size models and corpus to improve the performance evaluated by metrics such as exact match ( em ) and f1 .", "however , such a paradigm lacks sufficient interpretation to model capability and can not efficiently train a model with a large corpus ."], "relation": "used for", "id": "2022.acl-long.403", "year": 2022, "rel_sent": "Specifically , we design an MRC capability assessment framework that assesses model capabilities in an explainable and multi - dimensional manner .", "forward": true, "src_ids": "2022.acl-long.403_4307"}
+{"input": "model capabilities is done by using Method| context: machine reading comprehension ( mrc ) reveals the ability to understand a given text passage and answer questions based on it . existing research works in mrc rely heavily on large - size models and corpus to improve the performance evaluated by metrics such as exact match ( em ) and f1 . however , such a paradigm lacks sufficient interpretation to model capability and can not efficiently train a model with a large corpus .", "entity": "model capabilities", "output": "mrc capability assessment framework", "neg_sample": ["model capabilities is done by using Method", "machine reading comprehension ( mrc ) reveals the ability to understand a given text passage and answer questions based on it .", "existing research works in mrc rely heavily on large - size models and corpus to improve the performance evaluated by metrics such as exact match ( em ) and f1 .", "however , such a paradigm lacks sufficient interpretation to model capability and can not efficiently train a model with a large corpus ."], "relation": "used for", "id": "2022.acl-long.403", "year": 2022, "rel_sent": "Specifically , we design an MRC capability assessment framework that assesses model capabilities in an explainable and multi - dimensional manner .", "forward": false, "src_ids": "2022.acl-long.403_4308"}
+{"input": "troll - based memes classification is done by using Generic| context: the spread of fake news , propaganda , misinformation , disinformation , and harmful content online raised concerns among social mediaplatforms , government agencies , policymakers , and society as a whole . this is because such harmful or abusive content leads to several consequences to people such as physical , emotional , relational , and financial . among different harmful content trolling - based online content is one of them , where the idea is to post a message that is provocative , offensive , or menacing with an intent to mislead the audience . the content can be textual , visual , a combination of both , or a meme .", "entity": "troll - based memes classification", "output": "comparative analysis", "neg_sample": ["troll - based memes classification is done by using Generic", "the spread of fake news , propaganda , misinformation , disinformation , and harmful content online raised concerns among social mediaplatforms , government agencies , policymakers , and society as a whole .", "this is because such harmful or abusive content leads to several consequences to people such as physical , emotional , relational , and financial .", "among different harmful content trolling - based online content is one of them , where the idea is to post a message that is provocative , offensive , or menacing with an intent to mislead the audience .", "the content can be textual , visual , a combination of both , or a meme ."], "relation": "used for", "id": "2022.dravidianlangtech-1.13", "year": 2022, "rel_sent": "TeamX@DravidianLangTech - ACL2022 : A Comparative Analysis for Troll - Based Meme Classification.", "forward": false, "src_ids": "2022.dravidianlangtech-1.13_4309"}
+{"input": "comparative analysis is used for Task| context: the spread of fake news , propaganda , misinformation , disinformation , and harmful content online raised concerns among social mediaplatforms , government agencies , policymakers , and society as a whole . this is because such harmful or abusive content leads to several consequences to people such as physical , emotional , relational , and financial . among different harmful content trolling - based online content is one of them , where the idea is to post a message that is provocative , offensive , or menacing with an intent to mislead the audience . the content can be textual , visual , a combination of both , or a meme .", "entity": "comparative analysis", "output": "troll - based memes classification", "neg_sample": ["comparative analysis is used for Task", "the spread of fake news , propaganda , misinformation , disinformation , and harmful content online raised concerns among social mediaplatforms , government agencies , policymakers , and society as a whole .", "this is because such harmful or abusive content leads to several consequences to people such as physical , emotional , relational , and financial .", "among different harmful content trolling - based online content is one of them , where the idea is to post a message that is provocative , offensive , or menacing with an intent to mislead the audience .", "the content can be textual , visual , a combination of both , or a meme ."], "relation": "used for", "id": "2022.dravidianlangtech-1.13", "year": 2022, "rel_sent": "TeamX@DravidianLangTech - ACL2022 : A Comparative Analysis for Troll - Based Meme Classification.", "forward": true, "src_ids": "2022.dravidianlangtech-1.13_4310"}
+{"input": "sign language generation is done by using Task| context: end - to - end sign language generation models do not accurately represent the prosody in sign language . a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters .", "entity": "sign language generation", "output": "modeling intensification", "neg_sample": ["sign language generation is done by using Task", "end - to - end sign language generation models do not accurately represent the prosody in sign language .", "a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters ."], "relation": "used for", "id": "2022.findings-acl.228", "year": 2022, "rel_sent": "Modeling Intensification for Sign Language Generation : A Computational Approach.", "forward": false, "src_ids": "2022.findings-acl.228_4311"}
+{"input": "modeling intensification is used for Task| context: a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters .", "entity": "modeling intensification", "output": "sign language generation", "neg_sample": ["modeling intensification is used for Task", "a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters ."], "relation": "used for", "id": "2022.findings-acl.228", "year": 2022, "rel_sent": "Modeling Intensification for Sign Language Generation : A Computational Approach.", "forward": true, "src_ids": "2022.findings-acl.228_4312"}
+{"input": "transformer models is used for Task| context: a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters .", "entity": "transformer models", "output": "sign language generation", "neg_sample": ["transformer models is used for Task", "a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters ."], "relation": "used for", "id": "2022.findings-acl.228", "year": 2022, "rel_sent": "This enhanced dataset is then used to train state - of - the - art transformer models for sign language generation .", "forward": true, "src_ids": "2022.findings-acl.228_4313"}
+{"input": "annotated dataset is done by using Method| context: end - to - end sign language generation models do not accurately represent the prosody in sign language . a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters .", "entity": "annotated dataset", "output": "supervised intensity tagger", "neg_sample": ["annotated dataset is done by using Method", "end - to - end sign language generation models do not accurately represent the prosody in sign language .", "a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters ."], "relation": "used for", "id": "2022.findings-acl.228", "year": 2022, "rel_sent": "We then use a supervised intensity tagger to extend the annotated dataset and obtain labels for the remaining portion of it .", "forward": false, "src_ids": "2022.findings-acl.228_4314"}
+{"input": "supervised intensity tagger is used for Material| context: end - to - end sign language generation models do not accurately represent the prosody in sign language . a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters .", "entity": "supervised intensity tagger", "output": "annotated dataset", "neg_sample": ["supervised intensity tagger is used for Material", "end - to - end sign language generation models do not accurately represent the prosody in sign language .", "a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters ."], "relation": "used for", "id": "2022.findings-acl.228", "year": 2022, "rel_sent": "We then use a supervised intensity tagger to extend the annotated dataset and obtain labels for the remaining portion of it .", "forward": true, "src_ids": "2022.findings-acl.228_4315"}
+{"input": "sign language generation is done by using Method| context: end - to - end sign language generation models do not accurately represent the prosody in sign language . a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters .", "entity": "sign language generation", "output": "transformer models", "neg_sample": ["sign language generation is done by using Method", "end - to - end sign language generation models do not accurately represent the prosody in sign language .", "a lack of temporal and spatial variations leads to poor - quality generated presentations that confuse human interpreters ."], "relation": "used for", "id": "2022.findings-acl.228", "year": 2022, "rel_sent": "This enhanced dataset is then used to train state - of - the - art transformer models for sign language generation .", "forward": false, "src_ids": "2022.findings-acl.228_4316"}
+{"input": "event detection is done by using Method| context: event detection ( ed ) is a critical subtask of event extraction that seeks to identify event triggers of certain types in texts . despite significant advances in ed , existing methods typically follow a ' one model fits all types ' approach , which sees no differences between event types and often results in a quite skewed performance . finding the causes of skewed performance is crucial for the robustness of an ed model , but to date there has been little exploration of this problem . this research examines the issue in depth and presents a new concept termed trigger salience attribution , which can explicitly quantify the underlying patterns of events .", "entity": "event detection", "output": "training mechanism", "neg_sample": ["event detection is done by using Method", "event detection ( ed ) is a critical subtask of event extraction that seeks to identify event triggers of certain types in texts .", "despite significant advances in ed , existing methods typically follow a ' one model fits all types ' approach , which sees no differences between event types and often results in a quite skewed performance .", "finding the causes of skewed performance is crucial for the robustness of an ed model , but to date there has been little exploration of this problem .", "this research examines the issue in depth and presents a new concept termed trigger salience attribution , which can explicitly quantify the underlying patterns of events ."], "relation": "used for", "id": "2022.acl-long.313", "year": 2022, "rel_sent": "On this foundation , we develop a new training mechanism for ED , which can distinguish between trigger - dependent and context - dependent types and achieve promising performance on two benchmarks . Finally , by highlighting many distinct characteristics of trigger - dependent and context - dependent types , our work may promote more research into this problem .", "forward": false, "src_ids": "2022.acl-long.313_4317"}
+{"input": "grammatical error correction ( gec ) is done by using Material| context: the grammar error correction corpus for czech ( geccc ) offers a variety of four domains , covering error distributions ranging from high error density essays written by non - native speakers , to website texts , where errors are expected to be much less common .", "entity": "grammatical error correction ( gec )", "output": "czech corpus", "neg_sample": ["grammatical error correction ( gec ) is done by using Material", "the grammar error correction corpus for czech ( geccc ) offers a variety of four domains , covering error distributions ranging from high error density essays written by non - native speakers , to website texts , where errors are expected to be much less common ."], "relation": "used for", "id": "2022.tacl-1.26", "year": 2022, "rel_sent": "We introduce a large and diverse Czech corpus annotated for grammatical error correction ( GEC ) with the aim to contribute to the still scarce data resources in this domain for languages other than English .", "forward": false, "src_ids": "2022.tacl-1.26_4318"}
+{"input": "czech corpus is used for Task| context: the grammar error correction corpus for czech ( geccc ) offers a variety of four domains , covering error distributions ranging from high error density essays written by non - native speakers , to website texts , where errors are expected to be much less common .", "entity": "czech corpus", "output": "grammatical error correction ( gec )", "neg_sample": ["czech corpus is used for Task", "the grammar error correction corpus for czech ( geccc ) offers a variety of four domains , covering error distributions ranging from high error density essays written by non - native speakers , to website texts , where errors are expected to be much less common ."], "relation": "used for", "id": "2022.tacl-1.26", "year": 2022, "rel_sent": "We introduce a large and diverse Czech corpus annotated for grammatical error correction ( GEC ) with the aim to contribute to the still scarce data resources in this domain for languages other than English .", "forward": true, "src_ids": "2022.tacl-1.26_4319"}
+{"input": "neural chat translation ( nct ) is done by using Method| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "neural chat translation ( nct )", "output": "scheduled multi - task learning", "neg_sample": ["neural chat translation ( nct ) is done by using Method", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "Scheduled Multi - task Learning for Neural Chat Translation.", "forward": false, "src_ids": "2022.acl-long.300_4320"}
+{"input": "scheduled multi - task learning is used for Task| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "scheduled multi - task learning", "output": "neural chat translation ( nct )", "neg_sample": ["scheduled multi - task learning is used for Task", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "Scheduled Multi - task Learning for Neural Chat Translation.", "forward": true, "src_ids": "2022.acl-long.300_4321"}
+{"input": "scheduled multi - task learning framework is used for Task| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "scheduled multi - task learning framework", "output": "neural chat translation ( nct )", "neg_sample": ["scheduled multi - task learning framework is used for Task", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "To address the above issues , we propose a scheduled multi - task learning framework for NCT .", "forward": true, "src_ids": "2022.acl-long.300_4322"}
+{"input": "neural chat translation ( nct ) is done by using Method| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "neural chat translation ( nct )", "output": "scheduled multi - task learning framework", "neg_sample": ["neural chat translation ( nct ) is done by using Method", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "To address the above issues , we propose a scheduled multi - task learning framework for NCT .", "forward": false, "src_ids": "2022.acl-long.300_4323"}
+{"input": "training is done by using Method| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "training", "output": "three - stage training framework", "neg_sample": ["training is done by using Method", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "Specifically , we devise a three - stage training framework to incorporate the large - scale in - domain chat translation data into training by adding a second pre - training stage between the original pre - training and fine - tuning stages .", "forward": false, "src_ids": "2022.acl-long.300_4324"}
+{"input": "three - stage training framework is used for Task| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "entity": "three - stage training framework", "output": "training", "neg_sample": ["three - stage training framework is used for Task", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "Specifically , we devise a three - stage training framework to incorporate the large - scale in - domain chat translation data into training by adding a second pre - training stage between the original pre - training and fine - tuning stages .", "forward": true, "src_ids": "2022.acl-long.300_4325"}
+{"input": "main chat translation task is done by using OtherScientificTerm| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "main chat translation task", "output": "dialogue - related auxiliary tasks", "neg_sample": ["main chat translation task is done by using OtherScientificTerm", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "Further , we investigate where and how to schedule the dialogue - related auxiliary tasks in multiple training stages to effectively enhance the main chat translation task .", "forward": false, "src_ids": "2022.acl-long.300_4326"}
+{"input": "dialogue - related auxiliary tasks is used for Task| context: existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g. , coherence ) to improve chat translation via multi - task learning on small - scale chat translation data . although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners .", "entity": "dialogue - related auxiliary tasks", "output": "main chat translation task", "neg_sample": ["dialogue - related auxiliary tasks is used for Task", "existing methods mainly focus on modeling the bilingual dialogue characteristics ( e.g.", ", coherence ) to improve chat translation via multi - task learning on small - scale chat translation data .", "although the nct models have achieved impressive success , it is still far from satisfactory due to insufficient chat translation data and simple joint training manners ."], "relation": "used for", "id": "2022.acl-long.300", "year": 2022, "rel_sent": "Further , we investigate where and how to schedule the dialogue - related auxiliary tasks in multiple training stages to effectively enhance the main chat translation task .", "forward": true, "src_ids": "2022.acl-long.300_4327"}
+{"input": "text style transfer is done by using Method| context: text style transfer is an important task in natural language generation , which aims to control certain attributes in the generated text , such as politeness , emotion , humor , and many others . it has a long history in the field of natural language processing , and recently has re - gained significant attention thanks to the promising performance brought by deep neural models .", "entity": "text style transfer", "output": "deep learning", "neg_sample": ["text style transfer is done by using Method", "text style transfer is an important task in natural language generation , which aims to control certain attributes in the generated text , such as politeness , emotion , humor , and many others .", "it has a long history in the field of natural language processing , and recently has re - gained significant attention thanks to the promising performance brought by deep neural models ."], "relation": "used for", "id": "2022.cl-1.6", "year": 2022, "rel_sent": "Deep Learning for Text Style Transfer : A Survey.", "forward": false, "src_ids": "2022.cl-1.6_4328"}
+{"input": "deep learning is used for Task| context: it has a long history in the field of natural language processing , and recently has re - gained significant attention thanks to the promising performance brought by deep neural models .", "entity": "deep learning", "output": "text style transfer", "neg_sample": ["deep learning is used for Task", "it has a long history in the field of natural language processing , and recently has re - gained significant attention thanks to the promising performance brought by deep neural models ."], "relation": "used for", "id": "2022.cl-1.6", "year": 2022, "rel_sent": "Deep Learning for Text Style Transfer : A Survey.", "forward": true, "src_ids": "2022.cl-1.6_4329"}
+{"input": "infusing multi - source knowledge is done by using Method| context: knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses . however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source . to this end , infusing knowledge from multiple sources becomes a trend .", "entity": "infusing multi - source knowledge", "output": "ksam", "neg_sample": ["infusing multi - source knowledge is done by using Method", "knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses .", "however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source .", "to this end , infusing knowledge from multiple sources becomes a trend ."], "relation": "used for", "id": "2022.findings-acl.30", "year": 2022, "rel_sent": "Rather than following the traditional single decoder paradigm , KSAM uses multiple independent source - aware decoder heads to alleviate three challenging problems in infusing multi - source knowledge , namely , the diversity among different knowledge sources , the indefinite knowledge alignment issue , and the insufficient flexibility / scalability in knowledge usage .", "forward": false, "src_ids": "2022.findings-acl.30_4330"}
+{"input": "dialogue generation is done by using OtherScientificTerm| context: knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses . however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source . to this end , infusing knowledge from multiple sources becomes a trend .", "entity": "dialogue generation", "output": "multi - source knowledge", "neg_sample": ["dialogue generation is done by using OtherScientificTerm", "knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses .", "however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source .", "to this end , infusing knowledge from multiple sources becomes a trend ."], "relation": "used for", "id": "2022.findings-acl.30", "year": 2022, "rel_sent": "KSAM : Infusing Multi - Source Knowledge into Dialogue Generation via Knowledge Source Aware Multi - Head Decoding.", "forward": false, "src_ids": "2022.findings-acl.30_4331"}
+{"input": "dialogue generation is done by using OtherScientificTerm| context: knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses . however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source . to this end , infusing knowledge from multiple sources becomes a trend .", "entity": "dialogue generation", "output": "multi - source knowledge", "neg_sample": ["dialogue generation is done by using OtherScientificTerm", "knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses .", "however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source .", "to this end , infusing knowledge from multiple sources becomes a trend ."], "relation": "used for", "id": "2022.findings-acl.30", "year": 2022, "rel_sent": "This paper proposes a novel approach Knowledge Source Aware Multi - Head Decoding , KSAM , to infuse multi - source knowledge into dialogue generation more efficiently .", "forward": false, "src_ids": "2022.findings-acl.30_4332"}
+{"input": "multi - source knowledge is used for Task| context: knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses . however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source . to this end , infusing knowledge from multiple sources becomes a trend .", "entity": "multi - source knowledge", "output": "dialogue generation", "neg_sample": ["multi - source knowledge is used for Task", "knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses .", "however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source .", "to this end , infusing knowledge from multiple sources becomes a trend ."], "relation": "used for", "id": "2022.findings-acl.30", "year": 2022, "rel_sent": "KSAM : Infusing Multi - Source Knowledge into Dialogue Generation via Knowledge Source Aware Multi - Head Decoding.", "forward": true, "src_ids": "2022.findings-acl.30_4333"}
+{"input": "multi - source knowledge is used for Task| context: knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses . however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source . to this end , infusing knowledge from multiple sources becomes a trend .", "entity": "multi - source knowledge", "output": "dialogue generation", "neg_sample": ["multi - source knowledge is used for Task", "knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses .", "however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source .", "to this end , infusing knowledge from multiple sources becomes a trend ."], "relation": "used for", "id": "2022.findings-acl.30", "year": 2022, "rel_sent": "This paper proposes a novel approach Knowledge Source Aware Multi - Head Decoding , KSAM , to infuse multi - source knowledge into dialogue generation more efficiently .", "forward": true, "src_ids": "2022.findings-acl.30_4334"}
+{"input": "ksam is used for Task| context: knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses . however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source . to this end , infusing knowledge from multiple sources becomes a trend .", "entity": "ksam", "output": "infusing multi - source knowledge", "neg_sample": ["ksam is used for Task", "knowledge - enhanced methods have bridged the gap between human beings and machines in generating dialogue responses .", "however , most previous works solely seek knowledge from a single source , and thus they often fail to obtain available knowledge because of the insufficient coverage of a single knowledge source .", "to this end , infusing knowledge from multiple sources becomes a trend ."], "relation": "used for", "id": "2022.findings-acl.30", "year": 2022, "rel_sent": "Rather than following the traditional single decoder paradigm , KSAM uses multiple independent source - aware decoder heads to alleviate three challenging problems in infusing multi - source knowledge , namely , the diversity among different knowledge sources , the indefinite knowledge alignment issue , and the insufficient flexibility / scalability in knowledge usage .", "forward": true, "src_ids": "2022.findings-acl.30_4335"}
+{"input": "neural machine translation is done by using Task| context: the filtering and/or selection of training data is one of the core aspects to be considered when building a strong machine translation system . in their influential work , khayrallah and koehn ( 2018 ) investigated the impact of different types of noise on the performance of machine translation systems . in the same year the wmt introduced a shared task on parallel corpus filtering , which went on to be repeated in the following years , and resulted in many different filtering approaches being proposed . in this work we aim to combine the recent achievements in data filtering with the original analysis of khayrallah and koehn ( 2018 ) and investigate whether state - of - the - art filtering systems are capable of removing all the suggested noise types . we observe that most of these types of noise can be detected with an accuracy of over 90 % by modern filtering systems when operating in a well studied high resource setting . however , we alsofind that when confronted with more refined noise categories or when working with a less common language pair , the performance of the filtering systems is far from optimal , showing that there is still room for improvement in this area of research .", "entity": "neural machine translation", "output": "detecting various types of noise", "neg_sample": ["neural machine translation is done by using Task", "the filtering and/or selection of training data is one of the core aspects to be considered when building a strong machine translation system .", "in their influential work , khayrallah and koehn ( 2018 ) investigated the impact of different types of noise on the performance of machine translation systems .", "in the same year the wmt introduced a shared task on parallel corpus filtering , which went on to be repeated in the following years , and resulted in many different filtering approaches being proposed .", "in this work we aim to combine the recent achievements in data filtering with the original analysis of khayrallah and koehn ( 2018 ) and investigate whether state - of - the - art filtering systems are capable of removing all the suggested noise types .", "we observe that most of these types of noise can be detected with an accuracy of over 90 % by modern filtering systems when operating in a well studied high resource setting .", "however , we alsofind that when confronted with more refined noise categories or when working with a less common language pair , the performance of the filtering systems is far from optimal , showing that there is still room for improvement in this area of research ."], "relation": "used for", "id": "2022.findings-acl.200", "year": 2022, "rel_sent": "Detecting Various Types of Noise for Neural Machine Translation.", "forward": false, "src_ids": "2022.findings-acl.200_4336"}
+{"input": "detecting various types of noise is used for Task| context: the filtering and/or selection of training data is one of the core aspects to be considered when building a strong machine translation system . in their influential work , khayrallah and koehn ( 2018 ) investigated the impact of different types of noise on the performance of machine translation systems . in the same year the wmt introduced a shared task on parallel corpus filtering , which went on to be repeated in the following years , and resulted in many different filtering approaches being proposed . in this work we aim to combine the recent achievements in data filtering with the original analysis of khayrallah and koehn ( 2018 ) and investigate whether state - of - the - art filtering systems are capable of removing all the suggested noise types . we observe that most of these types of noise can be detected with an accuracy of over 90 % by modern filtering systems when operating in a well studied high resource setting . however , we alsofind that when confronted with more refined noise categories or when working with a less common language pair , the performance of the filtering systems is far from optimal , showing that there is still room for improvement in this area of research .", "entity": "detecting various types of noise", "output": "neural machine translation", "neg_sample": ["detecting various types of noise is used for Task", "the filtering and/or selection of training data is one of the core aspects to be considered when building a strong machine translation system .", "in their influential work , khayrallah and koehn ( 2018 ) investigated the impact of different types of noise on the performance of machine translation systems .", "in the same year the wmt introduced a shared task on parallel corpus filtering , which went on to be repeated in the following years , and resulted in many different filtering approaches being proposed .", "in this work we aim to combine the recent achievements in data filtering with the original analysis of khayrallah and koehn ( 2018 ) and investigate whether state - of - the - art filtering systems are capable of removing all the suggested noise types .", "we observe that most of these types of noise can be detected with an accuracy of over 90 % by modern filtering systems when operating in a well studied high resource setting .", "however , we alsofind that when confronted with more refined noise categories or when working with a less common language pair , the performance of the filtering systems is far from optimal , showing that there is still room for improvement in this area of research ."], "relation": "used for", "id": "2022.findings-acl.200", "year": 2022, "rel_sent": "Detecting Various Types of Noise for Neural Machine Translation.", "forward": true, "src_ids": "2022.findings-acl.200_4337"}
+{"input": "document understanding tasks is done by using Method| context: multimodal pre - training with text , layout , and image has made significant progress for visually rich document understanding ( vrdu ) , especially the fixed - layout documents such as scanned document images . while , there are still a large number of digital documents where the layout information is not fixed and needs to be interactively and dynamically rendered for visualization , making existing layout - based pre - training approaches not easy to apply .", "entity": "document understanding tasks", "output": "markuplm", "neg_sample": ["document understanding tasks is done by using Method", "multimodal pre - training with text , layout , and image has made significant progress for visually rich document understanding ( vrdu ) , especially the fixed - layout documents such as scanned document images .", "while , there are still a large number of digital documents where the layout information is not fixed and needs to be interactively and dynamically rendered for visualization , making existing layout - based pre - training approaches not easy to apply ."], "relation": "used for", "id": "2022.acl-long.420", "year": 2022, "rel_sent": "In this paper , we propose MarkupLM for document understanding tasks with markup languages as the backbone , such as HTML / XML - based documents , where text and markup information is jointly pre - trained .", "forward": false, "src_ids": "2022.acl-long.420_4338"}
+{"input": "markuplm is used for Task| context: multimodal pre - training with text , layout , and image has made significant progress for visually rich document understanding ( vrdu ) , especially the fixed - layout documents such as scanned document images . while , there are still a large number of digital documents where the layout information is not fixed and needs to be interactively and dynamically rendered for visualization , making existing layout - based pre - training approaches not easy to apply .", "entity": "markuplm", "output": "document understanding tasks", "neg_sample": ["markuplm is used for Task", "multimodal pre - training with text , layout , and image has made significant progress for visually rich document understanding ( vrdu ) , especially the fixed - layout documents such as scanned document images .", "while , there are still a large number of digital documents where the layout information is not fixed and needs to be interactively and dynamically rendered for visualization , making existing layout - based pre - training approaches not easy to apply ."], "relation": "used for", "id": "2022.acl-long.420", "year": 2022, "rel_sent": "In this paper , we propose MarkupLM for document understanding tasks with markup languages as the backbone , such as HTML / XML - based documents , where text and markup information is jointly pre - trained .", "forward": true, "src_ids": "2022.acl-long.420_4339"}
+{"input": "translating streaming speech content is done by using Method| context: how tofind proper moments to generate partial sentence translation given a streaming speech input ? existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even .", "entity": "translating streaming speech content", "output": "mosst", "neg_sample": ["translating streaming speech content is done by using Method", "how tofind proper moments to generate partial sentence translation given a streaming speech input ?", "existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even ."], "relation": "used for", "id": "2022.acl-long.50", "year": 2022, "rel_sent": "In this paper , we propose MoSST , a simple yet effective method for translating streaming speech content .", "forward": false, "src_ids": "2022.acl-long.50_4340"}
+{"input": "mosst is used for Task| context: how tofind proper moments to generate partial sentence translation given a streaming speech input ? existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even .", "entity": "mosst", "output": "translating streaming speech content", "neg_sample": ["mosst is used for Task", "how tofind proper moments to generate partial sentence translation given a streaming speech input ?", "existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even ."], "relation": "used for", "id": "2022.acl-long.50", "year": 2022, "rel_sent": "In this paper , we propose MoSST , a simple yet effective method for translating streaming speech content .", "forward": true, "src_ids": "2022.acl-long.50_4341"}
+{"input": "acoustic information is done by using Method| context: how tofind proper moments to generate partial sentence translation given a streaming speech input ? existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even .", "entity": "acoustic information", "output": "monotonic segmentation module", "neg_sample": ["acoustic information is done by using Method", "how tofind proper moments to generate partial sentence translation given a streaming speech input ?", "existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even ."], "relation": "used for", "id": "2022.acl-long.50", "year": 2022, "rel_sent": "Given a usually long speech sequence , we develop an efficient monotonic segmentation module inside an encoder - decoder model to accumulate acoustic information incrementally and detect proper speech unit boundaries for the input in speech translation task .", "forward": false, "src_ids": "2022.acl-long.50_4342"}
+{"input": "speech unit boundaries is done by using Method| context: how tofind proper moments to generate partial sentence translation given a streaming speech input ? existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even .", "entity": "speech unit boundaries", "output": "monotonic segmentation module", "neg_sample": ["speech unit boundaries is done by using Method", "how tofind proper moments to generate partial sentence translation given a streaming speech input ?", "existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even ."], "relation": "used for", "id": "2022.acl-long.50", "year": 2022, "rel_sent": "Given a usually long speech sequence , we develop an efficient monotonic segmentation module inside an encoder - decoder model to accumulate acoustic information incrementally and detect proper speech unit boundaries for the input in speech translation task .", "forward": false, "src_ids": "2022.acl-long.50_4343"}
+{"input": "monotonic segmentation module is used for OtherScientificTerm| context: how tofind proper moments to generate partial sentence translation given a streaming speech input ? existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even .", "entity": "monotonic segmentation module", "output": "acoustic information", "neg_sample": ["monotonic segmentation module is used for OtherScientificTerm", "how tofind proper moments to generate partial sentence translation given a streaming speech input ?", "existing approaches waiting - and - translating for a fixed duration often break the acoustic units in speech , since the boundaries between acoustic units in speech are not even ."], "relation": "used for", "id": "2022.acl-long.50", "year": 2022, "rel_sent": "Given a usually long speech sequence , we develop an efficient monotonic segmentation module inside an encoder - decoder model to accumulate acoustic information incrementally and detect proper speech unit boundaries for the input in speech translation task .", "forward": true, "src_ids": "2022.acl-long.50_4344"}
+{"input": "arabic is done by using Task| context: when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance .", "entity": "arabic", "output": "responsible natural language annotation", "neg_sample": ["arabic is done by using Task", "when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance ."], "relation": "used for", "id": "2022.findings-acl.31", "year": 2022, "rel_sent": "Towards Responsible Natural Language Annotation for the Varieties of Arabic.", "forward": false, "src_ids": "2022.findings-acl.31_4345"}
+{"input": "responsible natural language annotation is used for Material| context: when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance .", "entity": "responsible natural language annotation", "output": "arabic", "neg_sample": ["responsible natural language annotation is used for Material", "when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance ."], "relation": "used for", "id": "2022.findings-acl.31", "year": 2022, "rel_sent": "Towards Responsible Natural Language Annotation for the Varieties of Arabic.", "forward": true, "src_ids": "2022.findings-acl.31_4346"}
+{"input": "polyglossic , multidialectal languages is done by using Task| context: when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance .", "entity": "polyglossic , multidialectal languages", "output": "responsible dataset creation", "neg_sample": ["polyglossic , multidialectal languages is done by using Task", "when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance ."], "relation": "used for", "id": "2022.findings-acl.31", "year": 2022, "rel_sent": "We present a playbook for responsible dataset creation for polyglossic , multidialectal languages .", "forward": false, "src_ids": "2022.findings-acl.31_4347"}
+{"input": "responsible dataset creation is used for Material| context: when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance .", "entity": "responsible dataset creation", "output": "polyglossic , multidialectal languages", "neg_sample": ["responsible dataset creation is used for Material", "when building nlp models , there is a tendency to aim for broader coverage , often overlooking cultural and ( socio)linguistic nuance ."], "relation": "used for", "id": "2022.findings-acl.31", "year": 2022, "rel_sent": "We present a playbook for responsible dataset creation for polyglossic , multidialectal languages .", "forward": true, "src_ids": "2022.findings-acl.31_4348"}
+{"input": "temporal order is done by using Method| context: natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis . nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives . to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics . however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed .", "entity": "temporal order", "output": "inference system", "neg_sample": ["temporal order is done by using Method", "natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis .", "nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives .", "to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics .", "however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed ."], "relation": "used for", "id": "2022.acl-srw.10", "year": 2022, "rel_sent": "Compositional Semantics and Inference System for Temporal Order based on Japanese CCG.", "forward": false, "src_ids": "2022.acl-srw.10_4349"}
+{"input": "inference is done by using Method| context: natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis . nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives . to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics . however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed .", "entity": "inference", "output": "inference system", "neg_sample": ["inference is done by using Method", "natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis .", "nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives .", "to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics .", "however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed ."], "relation": "used for", "id": "2022.acl-srw.10", "year": 2022, "rel_sent": "Our system performs inference involving temporal order by using axioms for temporal relations and automated theorem provers .", "forward": false, "src_ids": "2022.acl-srw.10_4350"}
+{"input": "inference system is used for OtherScientificTerm| context: natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis . to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics .", "entity": "inference system", "output": "temporal order", "neg_sample": ["inference system is used for OtherScientificTerm", "natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis .", "to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics ."], "relation": "used for", "id": "2022.acl-srw.10", "year": 2022, "rel_sent": "Compositional Semantics and Inference System for Temporal Order based on Japanese CCG.", "forward": true, "src_ids": "2022.acl-srw.10_4351"}
+{"input": "inference system is used for Task| context: nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives . however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed .", "entity": "inference system", "output": "inference", "neg_sample": ["inference system is used for Task", "nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives .", "however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed ."], "relation": "used for", "id": "2022.acl-srw.10", "year": 2022, "rel_sent": "Our system performs inference involving temporal order by using axioms for temporal relations and automated theorem provers .", "forward": true, "src_ids": "2022.acl-srw.10_4352"}
+{"input": "temporal relations is done by using OtherScientificTerm| context: natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis . nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives . to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics . however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed .", "entity": "temporal relations", "output": "axioms", "neg_sample": ["temporal relations is done by using OtherScientificTerm", "natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis .", "nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives .", "to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics .", "however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed ."], "relation": "used for", "id": "2022.acl-srw.10", "year": 2022, "rel_sent": "Our system performs inference involving temporal order by using axioms for temporal relations and automated theorem provers .", "forward": false, "src_ids": "2022.acl-srw.10_4353"}
+{"input": "axioms is used for OtherScientificTerm| context: natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis . nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives . to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics . however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed .", "entity": "axioms", "output": "temporal relations", "neg_sample": ["axioms is used for OtherScientificTerm", "natural language inference ( nli ) is the task of determining whether a premise entails a hypothesis .", "nli with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives .", "to tackle this , temporal and aspectual inference has been analyzed in various ways in the field of formal semantics .", "however , a japanese nli system for temporal order based on the analysis of formal semantics has not been sufficiently developed ."], "relation": "used for", "id": "2022.acl-srw.10", "year": 2022, "rel_sent": "Our system performs inference involving temporal order by using axioms for temporal relations and automated theorem provers .", "forward": true, "src_ids": "2022.acl-srw.10_4354"}
+{"input": "nlp tasks is done by using Task| context: however , which approaches work best across tasks or even if they consistently outperform the simplest baseline maxprob remains to be explored .", "entity": "nlp tasks", "output": "selective prediction", "neg_sample": ["nlp tasks is done by using Task", "however , which approaches work best across tasks or even if they consistently outperform the simplest baseline maxprob remains to be explored ."], "relation": "used for", "id": "2022.findings-acl.158", "year": 2022, "rel_sent": "To this end , we systematically study selective prediction in a large - scale setup of 17 datasets across several NLP tasks .", "forward": false, "src_ids": "2022.findings-acl.158_4355"}
+{"input": "selective prediction is used for Task| context: in order to equip nlp systems with ' selective prediction ' capability , several task - specific approaches have been proposed . however , which approaches work best across tasks or even if they consistently outperform the simplest baseline maxprob remains to be explored .", "entity": "selective prediction", "output": "nlp tasks", "neg_sample": ["selective prediction is used for Task", "in order to equip nlp systems with ' selective prediction ' capability , several task - specific approaches have been proposed .", "however , which approaches work best across tasks or even if they consistently outperform the simplest baseline maxprob remains to be explored ."], "relation": "used for", "id": "2022.findings-acl.158", "year": 2022, "rel_sent": "To this end , we systematically study selective prediction in a large - scale setup of 17 datasets across several NLP tasks .", "forward": true, "src_ids": "2022.findings-acl.158_4356"}
+{"input": "sign language translation is done by using Method| context: we examine methods and techniques , proven to be helpful for the text - to - text translation of spoken languages in the context of gloss - to - text translation systems , where the glosses are the written representation of the signs .", "entity": "sign language translation", "output": "neural machine translation methods", "neg_sample": ["sign language translation is done by using Method", "we examine methods and techniques , proven to be helpful for the text - to - text translation of spoken languages in the context of gloss - to - text translation systems , where the glosses are the written representation of the signs ."], "relation": "used for", "id": "2022.acl-srw.21", "year": 2022, "rel_sent": "Using Neural Machine Translation Methods for Sign Language Translation.", "forward": false, "src_ids": "2022.acl-srw.21_4357"}
+{"input": "neural machine translation methods is used for Task| context: we examine methods and techniques , proven to be helpful for the text - to - text translation of spoken languages in the context of gloss - to - text translation systems , where the glosses are the written representation of the signs .", "entity": "neural machine translation methods", "output": "sign language translation", "neg_sample": ["neural machine translation methods is used for Task", "we examine methods and techniques , proven to be helpful for the text - to - text translation of spoken languages in the context of gloss - to - text translation systems , where the glosses are the written representation of the signs ."], "relation": "used for", "id": "2022.acl-srw.21", "year": 2022, "rel_sent": "Using Neural Machine Translation Methods for Sign Language Translation.", "forward": true, "src_ids": "2022.acl-srw.21_4358"}
+{"input": "classifier learning is done by using OtherScientificTerm| context: supervised learning has traditionally focused on inductive learning by observing labeled examples of a task . in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language . here , we explore training zero - shot classifiers for structured data purely from language .", "entity": "classifier learning", "output": "clues", "neg_sample": ["classifier learning is done by using OtherScientificTerm", "supervised learning has traditionally focused on inductive learning by observing labeled examples of a task .", "in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language .", "here , we explore training zero - shot classifiers for structured data purely from language ."], "relation": "used for", "id": "2022.acl-long.451", "year": 2022, "rel_sent": "For this , we introduce CLUES , a benchmark for Classifier Learning Using natural language ExplanationS , consisting of a range of classification tasks over structured data along with natural language supervision in the form of explanations .", "forward": false, "src_ids": "2022.acl-long.451_4359"}
+{"input": "entailment - based model is used for Method| context: supervised learning has traditionally focused on inductive learning by observing labeled examples of a task . in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language .", "entity": "entailment - based model", "output": "classifiers", "neg_sample": ["entailment - based model is used for Method", "supervised learning has traditionally focused on inductive learning by observing labeled examples of a task .", "in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language ."], "relation": "used for", "id": "2022.acl-long.451", "year": 2022, "rel_sent": "To model the influence of explanations in classifying an example , we develop ExEnt , an entailment - based model that learns classifiers using explanations .", "forward": true, "src_ids": "2022.acl-long.451_4360"}
+{"input": "clues is used for Task| context: supervised learning has traditionally focused on inductive learning by observing labeled examples of a task . in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language . here , we explore training zero - shot classifiers for structured data purely from language .", "entity": "clues", "output": "classifier learning", "neg_sample": ["clues is used for Task", "supervised learning has traditionally focused on inductive learning by observing labeled examples of a task .", "in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language .", "here , we explore training zero - shot classifiers for structured data purely from language ."], "relation": "used for", "id": "2022.acl-long.451", "year": 2022, "rel_sent": "For this , we introduce CLUES , a benchmark for Classifier Learning Using natural language ExplanationS , consisting of a range of classification tasks over structured data along with natural language supervision in the form of explanations .", "forward": true, "src_ids": "2022.acl-long.451_4361"}
+{"input": "real - world tasks is done by using OtherScientificTerm| context: supervised learning has traditionally focused on inductive learning by observing labeled examples of a task . in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language . here , we explore training zero - shot classifiers for structured data purely from language .", "entity": "real - world tasks", "output": "crowdsourced explanations", "neg_sample": ["real - world tasks is done by using OtherScientificTerm", "supervised learning has traditionally focused on inductive learning by observing labeled examples of a task .", "in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language .", "here , we explore training zero - shot classifiers for structured data purely from language ."], "relation": "used for", "id": "2022.acl-long.451", "year": 2022, "rel_sent": "It contains crowdsourced explanations describing real - world tasks from multiple teachers and programmatically generated explanations for the synthetic tasks .", "forward": false, "src_ids": "2022.acl-long.451_4362"}
+{"input": "crowdsourced explanations is used for Task| context: supervised learning has traditionally focused on inductive learning by observing labeled examples of a task . in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language . here , we explore training zero - shot classifiers for structured data purely from language .", "entity": "crowdsourced explanations", "output": "real - world tasks", "neg_sample": ["crowdsourced explanations is used for Task", "supervised learning has traditionally focused on inductive learning by observing labeled examples of a task .", "in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language .", "here , we explore training zero - shot classifiers for structured data purely from language ."], "relation": "used for", "id": "2022.acl-long.451", "year": 2022, "rel_sent": "It contains crowdsourced explanations describing real - world tasks from multiple teachers and programmatically generated explanations for the synthetic tasks .", "forward": true, "src_ids": "2022.acl-long.451_4363"}
+{"input": "classifiers is done by using Method| context: supervised learning has traditionally focused on inductive learning by observing labeled examples of a task . in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language . here , we explore training zero - shot classifiers for structured data purely from language .", "entity": "classifiers", "output": "entailment - based model", "neg_sample": ["classifiers is done by using Method", "supervised learning has traditionally focused on inductive learning by observing labeled examples of a task .", "in contrast , a hallmark of human intelligence is the ability to learn new concepts purely from language .", "here , we explore training zero - shot classifiers for structured data purely from language ."], "relation": "used for", "id": "2022.acl-long.451", "year": 2022, "rel_sent": "To model the influence of explanations in classifying an example , we develop ExEnt , an entailment - based model that learns classifiers using explanations .", "forward": false, "src_ids": "2022.acl-long.451_4364"}
+{"input": "knowledge - grounded dialogue task is done by using Method| context: to diversify and enrich generated dialogue responses , knowledge - grounded dialogue has been investigated in recent years . the existing methods tackle the knowledge grounding challenge by retrieving the relevant sentences over a large corpus and augmenting the dialogues with explicit extra information . despite their success , however , the existing works have drawbacks on the inference efficiency .", "entity": "knowledge - grounded dialogue task", "output": "pre - trained language models", "neg_sample": ["knowledge - grounded dialogue task is done by using Method", "to diversify and enrich generated dialogue responses , knowledge - grounded dialogue has been investigated in recent years .", "the existing methods tackle the knowledge grounding challenge by retrieving the relevant sentences over a large corpus and augmenting the dialogues with explicit extra information .", "despite their success , however , the existing works have drawbacks on the inference efficiency ."], "relation": "used for", "id": "2022.dialdoc-1.10", "year": 2022, "rel_sent": "This paper proposes KnowExpert , an end - to - end framework to bypass the explicit retrieval process and inject knowledge into the pre - trained language models with lightweight adapters and adapt to the knowledge - grounded dialogue task .", "forward": false, "src_ids": "2022.dialdoc-1.10_4365"}
+{"input": "pre - trained language models is used for Task| context: to diversify and enrich generated dialogue responses , knowledge - grounded dialogue has been investigated in recent years . the existing methods tackle the knowledge grounding challenge by retrieving the relevant sentences over a large corpus and augmenting the dialogues with explicit extra information . despite their success , however , the existing works have drawbacks on the inference efficiency .", "entity": "pre - trained language models", "output": "knowledge - grounded dialogue task", "neg_sample": ["pre - trained language models is used for Task", "to diversify and enrich generated dialogue responses , knowledge - grounded dialogue has been investigated in recent years .", "the existing methods tackle the knowledge grounding challenge by retrieving the relevant sentences over a large corpus and augmenting the dialogues with explicit extra information .", "despite their success , however , the existing works have drawbacks on the inference efficiency ."], "relation": "used for", "id": "2022.dialdoc-1.10", "year": 2022, "rel_sent": "This paper proposes KnowExpert , an end - to - end framework to bypass the explicit retrieval process and inject knowledge into the pre - trained language models with lightweight adapters and adapt to the knowledge - grounded dialogue task .", "forward": true, "src_ids": "2022.dialdoc-1.10_4366"}
+{"input": "assistive interaction is done by using Task| context: conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle . language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support .", "entity": "assistive interaction", "output": "cue - controlled response generation", "neg_sample": ["assistive interaction is done by using Task", "conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle .", "language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support ."], "relation": "used for", "id": "2022.slpat-1.9", "year": 2022, "rel_sent": "CueBot : Cue - Controlled Response Generation for Assistive Interaction Usages.", "forward": false, "src_ids": "2022.slpat-1.9_4367"}
+{"input": "cue - controlled response generation is used for Task| context: conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle . language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support .", "entity": "cue - controlled response generation", "output": "assistive interaction", "neg_sample": ["cue - controlled response generation is used for Task", "conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle .", "language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support ."], "relation": "used for", "id": "2022.slpat-1.9", "year": 2022, "rel_sent": "CueBot : Cue - Controlled Response Generation for Assistive Interaction Usages.", "forward": true, "src_ids": "2022.slpat-1.9_4368"}
+{"input": "model response output is done by using OtherScientificTerm| context: conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle . language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support .", "entity": "model response output", "output": "keyword - loss", "neg_sample": ["model response output is done by using OtherScientificTerm", "conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle .", "language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support ."], "relation": "used for", "id": "2022.slpat-1.9", "year": 2022, "rel_sent": "We also introduce a keyword - loss to lexically constrain the model response output .", "forward": false, "src_ids": "2022.slpat-1.9_4369"}
+{"input": "keyword - loss is used for OtherScientificTerm| context: conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle . language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support .", "entity": "keyword - loss", "output": "model response output", "neg_sample": ["keyword - loss is used for OtherScientificTerm", "conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle .", "language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support ."], "relation": "used for", "id": "2022.slpat-1.9", "year": 2022, "rel_sent": "We also introduce a keyword - loss to lexically constrain the model response output .", "forward": true, "src_ids": "2022.slpat-1.9_4370"}
+{"input": "end - to - end response generation models is done by using Method| context: conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle . language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support .", "entity": "end - to - end response generation models", "output": "keyword - control", "neg_sample": ["end - to - end response generation models is done by using Method", "conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle .", "language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support ."], "relation": "used for", "id": "2022.slpat-1.9", "year": 2022, "rel_sent": "Our evaluation and user study shows that keyword - control on end - to - end response generation models is powerful and can enable and empower users with degenerative disorders to carry out their day - to - day communication .", "forward": false, "src_ids": "2022.slpat-1.9_4371"}
+{"input": "keyword - control is used for Method| context: conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle . language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support .", "entity": "keyword - control", "output": "end - to - end response generation models", "neg_sample": ["keyword - control is used for Method", "conversational assistants are ubiquitous among the general population , however , these systems have not had an impact on people with disabilities , or speech and language disorders , for whom basic day - to - day communication and social interaction is a huge struggle .", "language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support ."], "relation": "used for", "id": "2022.slpat-1.9", "year": 2022, "rel_sent": "Our evaluation and user study shows that keyword - control on end - to - end response generation models is powerful and can enable and empower users with degenerative disorders to carry out their day - to - day communication .", "forward": true, "src_ids": "2022.slpat-1.9_4372"}
+{"input": "bias signals is done by using Method| context: weighted decoding methods composed of the pretrained language model ( lm ) and the controller have achieved promising results for controllable text generation . however , these models often suffer from a control strength / fluency trade - off problem as higher control strength is more likely to generate incoherent and repetitive text . in this paper , we illustrate this trade - off is arisen by the controller imposing the target attribute on the lm at improper positions .", "entity": "bias signals", "output": "lightweight regulator", "neg_sample": ["bias signals is done by using Method", "weighted decoding methods composed of the pretrained language model ( lm ) and the controller have achieved promising results for controllable text generation .", "however , these models often suffer from a control strength / fluency trade - off problem as higher control strength is more likely to generate incoherent and repetitive text .", "in this paper , we illustrate this trade - off is arisen by the controller imposing the target attribute on the lm at improper positions ."], "relation": "used for", "id": "2022.findings-acl.272", "year": 2022, "rel_sent": "And we propose a novel framework based on existing weighted decoding methods called CAT - PAW , which introduces a lightweight regulator to adjust bias signals from the controller at different decoding positions .", "forward": false, "src_ids": "2022.findings-acl.272_4373"}
+{"input": "lightweight regulator is used for OtherScientificTerm| context: weighted decoding methods composed of the pretrained language model ( lm ) and the controller have achieved promising results for controllable text generation . however , these models often suffer from a control strength / fluency trade - off problem as higher control strength is more likely to generate incoherent and repetitive text . in this paper , we illustrate this trade - off is arisen by the controller imposing the target attribute on the lm at improper positions .", "entity": "lightweight regulator", "output": "bias signals", "neg_sample": ["lightweight regulator is used for OtherScientificTerm", "weighted decoding methods composed of the pretrained language model ( lm ) and the controller have achieved promising results for controllable text generation .", "however , these models often suffer from a control strength / fluency trade - off problem as higher control strength is more likely to generate incoherent and repetitive text .", "in this paper , we illustrate this trade - off is arisen by the controller imposing the target attribute on the lm at improper positions ."], "relation": "used for", "id": "2022.findings-acl.272", "year": 2022, "rel_sent": "And we propose a novel framework based on existing weighted decoding methods called CAT - PAW , which introduces a lightweight regulator to adjust bias signals from the controller at different decoding positions .", "forward": true, "src_ids": "2022.findings-acl.272_4374"}
+{"input": "event structures is done by using Method| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "event structures", "output": "beeds", "neg_sample": ["event structures is done by using Method", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "We propose BEEDS , a new approach on how to mine event structures from PubMed based on a question - answering paradigm .", "forward": false, "src_ids": "2022.bionlp-1.28_4375"}
+{"input": "event triples is done by using Method| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "event triples", "output": "beeds", "neg_sample": ["event triples is done by using Method", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "Using a three - step pipeline comprising a document retriever , a document reader , and an entity normalizer , BEEDS is able tofully automatically extract event triples involving a query protein or gene and to store this information directly in a knowledge base .", "forward": false, "src_ids": "2022.bionlp-1.28_4376"}
+{"input": "three - step pipeline is used for Method| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "three - step pipeline", "output": "beeds", "neg_sample": ["three - step pipeline is used for Method", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "Using a three - step pipeline comprising a document retriever , a document reader , and an entity normalizer , BEEDS is able tofully automatically extract event triples involving a query protein or gene and to store this information directly in a knowledge base .", "forward": true, "src_ids": "2022.bionlp-1.28_4377"}
+{"input": "beeds is done by using Generic| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "beeds", "output": "three - step pipeline", "neg_sample": ["beeds is done by using Generic", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "Using a three - step pipeline comprising a document retriever , a document reader , and an entity normalizer , BEEDS is able tofully automatically extract event triples involving a query protein or gene and to store this information directly in a knowledge base .", "forward": false, "src_ids": "2022.bionlp-1.28_4378"}
+{"input": "beeds is used for OtherScientificTerm| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "beeds", "output": "event triples", "neg_sample": ["beeds is used for OtherScientificTerm", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "Using a three - step pipeline comprising a document retriever , a document reader , and an entity normalizer , BEEDS is able tofully automatically extract event triples involving a query protein or gene and to store this information directly in a knowledge base .", "forward": true, "src_ids": "2022.bionlp-1.28_4379"}
+{"input": "event extraction is done by using Method| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "event extraction", "output": "transformer - based architecture", "neg_sample": ["event extraction is done by using Method", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "BEEDS applies a transformer - based architecture for event extraction and uses distant supervision to augment the scarce training data in event mining .", "forward": false, "src_ids": "2022.bionlp-1.28_4380"}
+{"input": "transformer - based architecture is used for Task| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "transformer - based architecture", "output": "event extraction", "neg_sample": ["transformer - based architecture is used for Task", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "BEEDS applies a transformer - based architecture for event extraction and uses distant supervision to augment the scarce training data in event mining .", "forward": true, "src_ids": "2022.bionlp-1.28_4381"}
+{"input": "event mining is done by using Method| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "event mining", "output": "distant supervision", "neg_sample": ["event mining is done by using Method", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "BEEDS applies a transformer - based architecture for event extraction and uses distant supervision to augment the scarce training data in event mining .", "forward": false, "src_ids": "2022.bionlp-1.28_4382"}
+{"input": "distant supervision is used for Task| context: automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature .", "entity": "distant supervision", "output": "event mining", "neg_sample": ["distant supervision is used for Task", "automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature ."], "relation": "used for", "id": "2022.bionlp-1.28", "year": 2022, "rel_sent": "BEEDS applies a transformer - based architecture for event extraction and uses distant supervision to augment the scarce training data in event mining .", "forward": true, "src_ids": "2022.bionlp-1.28_4383"}
+{"input": "dialogue disentanglement is done by using Task| context: when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately . this task is referred as dialogue disentanglement . a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling .", "entity": "dialogue disentanglement", "output": "conversation- and tree - structure losses", "neg_sample": ["dialogue disentanglement is done by using Task", "when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately .", "this task is referred as dialogue disentanglement .", "a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling ."], "relation": "used for", "id": "2022.dialdoc-1.6", "year": 2022, "rel_sent": "Conversation- and Tree - Structure Losses for Dialogue Disentanglement.", "forward": false, "src_ids": "2022.dialdoc-1.6_4384"}
+{"input": "conversation- and tree - structure losses is used for Task| context: when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately . a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling .", "entity": "conversation- and tree - structure losses", "output": "dialogue disentanglement", "neg_sample": ["conversation- and tree - structure losses is used for Task", "when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately .", "a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling ."], "relation": "used for", "id": "2022.dialdoc-1.6", "year": 2022, "rel_sent": "Conversation- and Tree - Structure Losses for Dialogue Disentanglement.", "forward": true, "src_ids": "2022.dialdoc-1.6_4385"}
+{"input": "bilstm is used for Method| context: when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately . this task is referred as dialogue disentanglement . a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling .", "entity": "bilstm", "output": "bert", "neg_sample": ["bilstm is used for Method", "when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately .", "this task is referred as dialogue disentanglement .", "a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling ."], "relation": "used for", "id": "2022.dialdoc-1.6", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical model , named Dialogue BERT ( DIALBERT ) , which integrates the local and global semantics in the context range by using BERT to encode each message - pair and using BiLSTM to aggregate the chronological context information into the output of BERT .", "forward": true, "src_ids": "2022.dialdoc-1.6_4386"}
+{"input": "bert is done by using Method| context: when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately . this task is referred as dialogue disentanglement . a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling .", "entity": "bert", "output": "bilstm", "neg_sample": ["bert is done by using Method", "when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately .", "this task is referred as dialogue disentanglement .", "a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling ."], "relation": "used for", "id": "2022.dialdoc-1.6", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical model , named Dialogue BERT ( DIALBERT ) , which integrates the local and global semantics in the context range by using BERT to encode each message - pair and using BiLSTM to aggregate the chronological context information into the output of BERT .", "forward": false, "src_ids": "2022.dialdoc-1.6_4387"}
+{"input": "chronological context information is done by using Method| context: when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately . this task is referred as dialogue disentanglement . a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling .", "entity": "chronological context information", "output": "bilstm", "neg_sample": ["chronological context information is done by using Method", "when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately .", "this task is referred as dialogue disentanglement .", "a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling ."], "relation": "used for", "id": "2022.dialdoc-1.6", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical model , named Dialogue BERT ( DIALBERT ) , which integrates the local and global semantics in the context range by using BERT to encode each message - pair and using BiLSTM to aggregate the chronological context information into the output of BERT .", "forward": false, "src_ids": "2022.dialdoc-1.6_4388"}
+{"input": "bilstm is used for OtherScientificTerm| context: when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately . this task is referred as dialogue disentanglement . a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling .", "entity": "bilstm", "output": "chronological context information", "neg_sample": ["bilstm is used for OtherScientificTerm", "when multiple conversations occur simultaneously , a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately .", "this task is referred as dialogue disentanglement .", "a significant drawback of previous studies on disentanglement lies in that they only focus on pair - wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling ."], "relation": "used for", "id": "2022.dialdoc-1.6", "year": 2022, "rel_sent": "In this paper , we propose a hierarchical model , named Dialogue BERT ( DIALBERT ) , which integrates the local and global semantics in the context range by using BERT to encode each message - pair and using BiLSTM to aggregate the chronological context information into the output of BERT .", "forward": true, "src_ids": "2022.dialdoc-1.6_4389"}
+{"input": "global and equitable language technologies is done by using Task| context: natural language processing ( nlp ) systems have become a central technology in communication , education , medicine , artificial intelligence , and many other domains of research and development . while the performance of nlp methods has grown enormously over the last decade , this progress has been restricted to a minuscule subset of the world 's ~6,500 languages .", "entity": "global and equitable language technologies", "output": "evidence - based policy making", "neg_sample": ["global and equitable language technologies is done by using Task", "natural language processing ( nlp ) systems have become a central technology in communication , education , medicine , artificial intelligence , and many other domains of research and development .", "while the performance of nlp methods has grown enormously over the last decade , this progress has been restricted to a minuscule subset of the world 's ~6,500 languages ."], "relation": "used for", "id": "2022.acl-long.376", "year": 2022, "rel_sent": "In the process , we ( 1 ) quantify disparities in the current state of NLP research , ( 2 ) explore some of its associated societal and academic factors , and ( 3 ) produce tailored recommendations for evidence - based policy making aimed at promoting more global and equitable language technologies .", "forward": false, "src_ids": "2022.acl-long.376_4390"}
+{"input": "evidence - based policy making is used for OtherScientificTerm| context: natural language processing ( nlp ) systems have become a central technology in communication , education , medicine , artificial intelligence , and many other domains of research and development . while the performance of nlp methods has grown enormously over the last decade , this progress has been restricted to a minuscule subset of the world 's ~6,500 languages .", "entity": "evidence - based policy making", "output": "global and equitable language technologies", "neg_sample": ["evidence - based policy making is used for OtherScientificTerm", "natural language processing ( nlp ) systems have become a central technology in communication , education , medicine , artificial intelligence , and many other domains of research and development .", "while the performance of nlp methods has grown enormously over the last decade , this progress has been restricted to a minuscule subset of the world 's ~6,500 languages ."], "relation": "used for", "id": "2022.acl-long.376", "year": 2022, "rel_sent": "In the process , we ( 1 ) quantify disparities in the current state of NLP research , ( 2 ) explore some of its associated societal and academic factors , and ( 3 ) produce tailored recommendations for evidence - based policy making aimed at promoting more global and equitable language technologies .", "forward": true, "src_ids": "2022.acl-long.376_4391"}
+{"input": "word order information is done by using Method| context: somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text .", "entity": "word order information", "output": "language models", "neg_sample": ["word order information is done by using Method", "somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text ."], "relation": "used for", "id": "2022.acl-long.476", "year": 2022, "rel_sent": "We probe these language models for word order information and investigate what position embeddings learned from shuffled text encode , showing that these models retain a notion of word order information .", "forward": false, "src_ids": "2022.acl-long.476_4392"}
+{"input": "language models is used for OtherScientificTerm| context: somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text .", "entity": "language models", "output": "word order information", "neg_sample": ["language models is used for OtherScientificTerm", "somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text ."], "relation": "used for", "id": "2022.acl-long.476", "year": 2022, "rel_sent": "We probe these language models for word order information and investigate what position embeddings learned from shuffled text encode , showing that these models retain a notion of word order information .", "forward": true, "src_ids": "2022.acl-long.476_4393"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: many works show the plms ' ability tofill in the missing factual words in cloze - style prompts such as ' dante was born in [ mask ] . ' however , it is still a mystery how plms generate the results correctly : relying on effective clues or shortcut patterns ?", "entity": "pre - trained language models", "output": "factual knowledge", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "many works show the plms ' ability tofill in the missing factual words in cloze - style prompts such as ' dante was born in [ mask ] . '", "however , it is still a mystery how plms generate the results correctly : relying on effective clues or shortcut patterns ?"], "relation": "used for", "id": "2022.findings-acl.136", "year": 2022, "rel_sent": "Accordingly , we conclude that the PLMs capture the factual knowledge ineffectively because of depending on the inadequate associations .", "forward": true, "src_ids": "2022.findings-acl.136_4394"}
+{"input": "pre - trained language models is used for OtherScientificTerm| context: recently , there has been a trend to investigate the factual knowledge captured by pre - trained language models ( plms ) . however , it is still a mystery how plms generate the results correctly : relying on effective clues or shortcut patterns ?", "entity": "pre - trained language models", "output": "missing factual words", "neg_sample": ["pre - trained language models is used for OtherScientificTerm", "recently , there has been a trend to investigate the factual knowledge captured by pre - trained language models ( plms ) .", "however , it is still a mystery how plms generate the results correctly : relying on effective clues or shortcut patterns ?"], "relation": "used for", "id": "2022.findings-acl.136", "year": 2022, "rel_sent": "Our analysis shows : ( 1 ) PLMs generate the missing factual words more by the positionally close and highly co - occurred words than the knowledge - dependent words ; ( 2 ) the dependence on the knowledge - dependent words is more effective than the positionally close and highly co - occurred words .", "forward": true, "src_ids": "2022.findings-acl.136_4395"}
+{"input": "non - autoregressive translation is done by using Method| context: we participate in all three language pairs ( english - german , english - french , english - spanish ) under the constrained setting , and submit an english - german result under the unconstrained setting .", "entity": "non - autoregressive translation", "output": "length control decoding", "neg_sample": ["non - autoregressive translation is done by using Method", "we participate in all three language pairs ( english - german , english - french , english - spanish ) under the constrained setting , and submit an english - german result under the unconstrained setting ."], "relation": "used for", "id": "2022.iwslt-1.33", "year": 2022, "rel_sent": "We investigate three methods for biasing the output length : i ) conditioning the output to a given target - source length - ratio class ; ii ) enriching the transformer positional embedding with length information and iii ) length control decoding for non - autoregressive translation etc .", "forward": false, "src_ids": "2022.iwslt-1.33_4396"}
+{"input": "length control decoding is used for Task| context: we participate in all three language pairs ( english - german , english - french , english - spanish ) under the constrained setting , and submit an english - german result under the unconstrained setting .", "entity": "length control decoding", "output": "non - autoregressive translation", "neg_sample": ["length control decoding is used for Task", "we participate in all three language pairs ( english - german , english - french , english - spanish ) under the constrained setting , and submit an english - german result under the unconstrained setting ."], "relation": "used for", "id": "2022.iwslt-1.33", "year": 2022, "rel_sent": "We investigate three methods for biasing the output length : i ) conditioning the output to a given target - source length - ratio class ; ii ) enriching the transformer positional embedding with length information and iii ) length control decoding for non - autoregressive translation etc .", "forward": true, "src_ids": "2022.iwslt-1.33_4397"}
+{"input": "nlp preprints is done by using Material| context: twitter has slowly but surely established itself as a forum for disseminating , analysing and promoting nlp research . the trend of researchers promoting work not yet peer - reviewed ( preprints ) by posting concise summaries presented itself as an opportunity to collect and combine multiple modalities of data .", "entity": "nlp preprints", "output": "twitter threads", "neg_sample": ["nlp preprints is done by using Material", "twitter has slowly but surely established itself as a forum for disseminating , analysing and promoting nlp research .", "the trend of researchers promoting work not yet peer - reviewed ( preprints ) by posting concise summaries presented itself as an opportunity to collect and combine multiple modalities of data ."], "relation": "used for", "id": "2022.wassa-1.32", "year": 2022, "rel_sent": "In scope of this paper , we ( 1 ) construct a dataset of Twitter threads in which researchers promote NLP preprints and ( 2 ) evaluate whether it is possible to predict the popularity of a thread based on the content of the Twitter thread , paper content and user metadata .", "forward": false, "src_ids": "2022.wassa-1.32_4398"}
+{"input": "twitter threads is used for Material| context: twitter has slowly but surely established itself as a forum for disseminating , analysing and promoting nlp research . the trend of researchers promoting work not yet peer - reviewed ( preprints ) by posting concise summaries presented itself as an opportunity to collect and combine multiple modalities of data .", "entity": "twitter threads", "output": "nlp preprints", "neg_sample": ["twitter threads is used for Material", "twitter has slowly but surely established itself as a forum for disseminating , analysing and promoting nlp research .", "the trend of researchers promoting work not yet peer - reviewed ( preprints ) by posting concise summaries presented itself as an opportunity to collect and combine multiple modalities of data ."], "relation": "used for", "id": "2022.wassa-1.32", "year": 2022, "rel_sent": "In scope of this paper , we ( 1 ) construct a dataset of Twitter threads in which researchers promote NLP preprints and ( 2 ) evaluate whether it is possible to predict the popularity of a thread based on the content of the Twitter thread , paper content and user metadata .", "forward": true, "src_ids": "2022.wassa-1.32_4399"}
+{"input": "case markers is done by using Material| context: we introduce camel ( case marker extraction without labels ) , a novel and challenging task in computational morphology that is especially relevant for low - resource languages .", "entity": "case markers", "output": "massively multilingual corpus", "neg_sample": ["case markers is done by using Material", "we introduce camel ( case marker extraction without labels ) , a novel and challenging task in computational morphology that is especially relevant for low - resource languages ."], "relation": "used for", "id": "2022.acl-long.377", "year": 2022, "rel_sent": "We propose a first model for CaMEL that uses a massively multilingual corpus to extract case markers in 83 languages based only on a noun phrase chunker and an alignment system .", "forward": false, "src_ids": "2022.acl-long.377_4400"}
+{"input": "massively multilingual corpus is used for OtherScientificTerm| context: we introduce camel ( case marker extraction without labels ) , a novel and challenging task in computational morphology that is especially relevant for low - resource languages .", "entity": "massively multilingual corpus", "output": "case markers", "neg_sample": ["massively multilingual corpus is used for OtherScientificTerm", "we introduce camel ( case marker extraction without labels ) , a novel and challenging task in computational morphology that is especially relevant for low - resource languages ."], "relation": "used for", "id": "2022.acl-long.377", "year": 2022, "rel_sent": "We propose a first model for CaMEL that uses a massively multilingual corpus to extract case markers in 83 languages based only on a noun phrase chunker and an alignment system .", "forward": true, "src_ids": "2022.acl-long.377_4401"}
+{"input": "deep neural network is used for Task| context: noisy labels in large e - commerce product data ( i.e. , product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly .", "entity": "deep neural network", "output": "product title classification", "neg_sample": ["deep neural network is used for Task", "noisy labels in large e - commerce product data ( i.e.", ", product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly ."], "relation": "used for", "id": "2022.ecnlp-1.20", "year": 2022, "rel_sent": "We develop a simple yet effective Deep Neural Network for product title classification to use as a base classifier .", "forward": true, "src_ids": "2022.ecnlp-1.20_4402"}
+{"input": "classifier model is done by using Method| context: noisy labels in large e - commerce product data ( i.e. , product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly . training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical .", "entity": "classifier model", "output": "noise - resistance training algorithms", "neg_sample": ["classifier model is done by using Method", "noisy labels in large e - commerce product data ( i.e.", ", product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly .", "training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical ."], "relation": "used for", "id": "2022.ecnlp-1.20", "year": 2022, "rel_sent": "In this paper , we study the impact of instance - dependent noise to performance of product title classification by comparing our data denoising algorithm and different noise - resistance training algorithms which were designed to prevent a classifier model from over - fitting to noise .", "forward": false, "src_ids": "2022.ecnlp-1.20_4403"}
+{"input": "noise - resistance training algorithms is used for Method| context: noisy labels in large e - commerce product data ( i.e. , product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly . training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical .", "entity": "noise - resistance training algorithms", "output": "classifier model", "neg_sample": ["noise - resistance training algorithms is used for Method", "noisy labels in large e - commerce product data ( i.e.", ", product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly .", "training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical ."], "relation": "used for", "id": "2022.ecnlp-1.20", "year": 2022, "rel_sent": "In this paper , we study the impact of instance - dependent noise to performance of product title classification by comparing our data denoising algorithm and different noise - resistance training algorithms which were designed to prevent a classifier model from over - fitting to noise .", "forward": true, "src_ids": "2022.ecnlp-1.20_4404"}
+{"input": "base classifier is done by using Method| context: noisy labels in large e - commerce product data ( i.e. , product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly . training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical .", "entity": "base classifier", "output": "deep neural network", "neg_sample": ["base classifier is done by using Method", "noisy labels in large e - commerce product data ( i.e.", ", product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly .", "training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical ."], "relation": "used for", "id": "2022.ecnlp-1.20", "year": 2022, "rel_sent": "We develop a simple yet effective Deep Neural Network for product title classification to use as a base classifier .", "forward": false, "src_ids": "2022.ecnlp-1.20_4405"}
+{"input": "product title classification is done by using Method| context: noisy labels in large e - commerce product data ( i.e. , product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly . training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical .", "entity": "product title classification", "output": "deep neural network", "neg_sample": ["product title classification is done by using Method", "noisy labels in large e - commerce product data ( i.e.", ", product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly .", "training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical ."], "relation": "used for", "id": "2022.ecnlp-1.20", "year": 2022, "rel_sent": "We develop a simple yet effective Deep Neural Network for product title classification to use as a base classifier .", "forward": false, "src_ids": "2022.ecnlp-1.20_4406"}
+{"input": "deep neural network is used for Method| context: noisy labels in large e - commerce product data ( i.e. , product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly . training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical .", "entity": "deep neural network", "output": "base classifier", "neg_sample": ["deep neural network is used for Method", "noisy labels in large e - commerce product data ( i.e.", ", product items are placed into incorrect categories ) is a critical issue for product categorization task because they are unavoidable , non - trivial to remove and degrade prediction performance significantly .", "training a product title classification model which is robust to noisy labels in the data is very important to make product classification applications more practical ."], "relation": "used for", "id": "2022.ecnlp-1.20", "year": 2022, "rel_sent": "We develop a simple yet effective Deep Neural Network for product title classification to use as a base classifier .", "forward": true, "src_ids": "2022.ecnlp-1.20_4407"}
+{"input": "linguists engagement is used for Material| context: language technologies , particularly speech technologies , are becoming more pervasive for access to digital platforms and resources . this brings to the forefront concerns of their inclusivity , first in terms of language diversity . additionally , research shows speech recognition to be more accurate for men than for women and more accurate for individuals younger than 30 years of age than those older . in the global south where languages are low resource , these same issues should be taken into consideration in data collection efforts to not replicate these mistakes . it is also important to note that in varying contexts within the global south , this work presents additional nuance and potential for bias based on accents , related dialects and variants of a language .", "entity": "linguists engagement", "output": "kiswahili speech recognition dataset", "neg_sample": ["linguists engagement is used for Material", "language technologies , particularly speech technologies , are becoming more pervasive for access to digital platforms and resources .", "this brings to the forefront concerns of their inclusivity , first in terms of language diversity .", "additionally , research shows speech recognition to be more accurate for men than for women and more accurate for individuals younger than 30 years of age than those older .", "in the global south where languages are low resource , these same issues should be taken into consideration in data collection efforts to not replicate these mistakes .", "it is also important to note that in varying contexts within the global south , this work presents additional nuance and potential for bias based on accents , related dialects and variants of a language ."], "relation": "used for", "id": "2022.computel-1.3", "year": 2022, "rel_sent": "This paper documents i ) the designing and execution of a Linguists Engagement for purposes of building an inclusive Kiswahili Speech Recognition dataset , representative of the diversity among speakers of the language ii ) the unexpected yet key learning in terms of socio - linguistcs which demonstrate the importance of multi - disciplinarity in teams developing datasets and NLP technologies iii ) the creation of a test dataset intended to be used for evaluating the performance of Speech Recognition models on demographic groups that are likely to be underrepresented .", "forward": true, "src_ids": "2022.computel-1.3_4408"}
+{"input": "kiswahili speech recognition dataset is done by using Method| context: language technologies , particularly speech technologies , are becoming more pervasive for access to digital platforms and resources . this brings to the forefront concerns of their inclusivity , first in terms of language diversity . additionally , research shows speech recognition to be more accurate for men than for women and more accurate for individuals younger than 30 years of age than those older . in the global south where languages are low resource , these same issues should be taken into consideration in data collection efforts to not replicate these mistakes . it is also important to note that in varying contexts within the global south , this work presents additional nuance and potential for bias based on accents , related dialects and variants of a language .", "entity": "kiswahili speech recognition dataset", "output": "linguists engagement", "neg_sample": ["kiswahili speech recognition dataset is done by using Method", "language technologies , particularly speech technologies , are becoming more pervasive for access to digital platforms and resources .", "this brings to the forefront concerns of their inclusivity , first in terms of language diversity .", "additionally , research shows speech recognition to be more accurate for men than for women and more accurate for individuals younger than 30 years of age than those older .", "in the global south where languages are low resource , these same issues should be taken into consideration in data collection efforts to not replicate these mistakes .", "it is also important to note that in varying contexts within the global south , this work presents additional nuance and potential for bias based on accents , related dialects and variants of a language ."], "relation": "used for", "id": "2022.computel-1.3", "year": 2022, "rel_sent": "This paper documents i ) the designing and execution of a Linguists Engagement for purposes of building an inclusive Kiswahili Speech Recognition dataset , representative of the diversity among speakers of the language ii ) the unexpected yet key learning in terms of socio - linguistcs which demonstrate the importance of multi - disciplinarity in teams developing datasets and NLP technologies iii ) the creation of a test dataset intended to be used for evaluating the performance of Speech Recognition models on demographic groups that are likely to be underrepresented .", "forward": false, "src_ids": "2022.computel-1.3_4409"}
+{"input": "online text sequences is done by using Method| context: analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities .", "entity": "online text sequences", "output": "hyperbolic hawkes attention network ( hyphen )", "neg_sample": ["online text sequences is done by using Method", "analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities ."], "relation": "used for", "id": "2022.acl-short.69", "year": 2022, "rel_sent": "Through quantitative and exploratory experiments over financial NLP , suicide ideation detection , and political debate analysis we demonstrate HYPHEN 's practical applicability for modeling online text sequences in a geometry agnostic manner .", "forward": false, "src_ids": "2022.acl-short.69_4410"}
+{"input": "text streams is done by using Method| context: analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities .", "entity": "text streams", "output": "hyperbolic hawkes attention", "neg_sample": ["text streams is done by using Method", "analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities ."], "relation": "used for", "id": "2022.acl-short.69", "year": 2022, "rel_sent": "HYPHEN : Hyperbolic Hawkes Attention For Text Streams.", "forward": false, "src_ids": "2022.acl-short.69_4411"}
+{"input": "hyperbolic hawkes attention is used for Material| context: analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities .", "entity": "hyperbolic hawkes attention", "output": "text streams", "neg_sample": ["hyperbolic hawkes attention is used for Material", "analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities ."], "relation": "used for", "id": "2022.acl-short.69", "year": 2022, "rel_sent": "HYPHEN : Hyperbolic Hawkes Attention For Text Streams.", "forward": true, "src_ids": "2022.acl-short.69_4412"}
+{"input": "hyperbolic hawkes attention network ( hyphen ) is used for Material| context: analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities .", "entity": "hyperbolic hawkes attention network ( hyphen )", "output": "online text sequences", "neg_sample": ["hyperbolic hawkes attention network ( hyphen ) is used for Material", "analyzing the temporal sequence of texts from sources such as social media , news , and parliamentary debates is a challenging problem as it exhibits time - varying scale - free properties and fine - grained timing irregularities ."], "relation": "used for", "id": "2022.acl-short.69", "year": 2022, "rel_sent": "Through quantitative and exploratory experiments over financial NLP , suicide ideation detection , and political debate analysis we demonstrate HYPHEN 's practical applicability for modeling online text sequences in a geometry agnostic manner .", "forward": true, "src_ids": "2022.acl-short.69_4413"}
+{"input": "transfer learning is done by using Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "transfer learning", "output": "cogtaskonomy", "neg_sample": ["transfer learning is done by using Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "CogTaskonomy : Cognitively Inspired Task Taxonomy Is Beneficial to Transfer Learning in NLP.", "forward": false, "src_ids": "2022.acl-long.64_4414"}
+{"input": "transfer learning is done by using Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "transfer learning", "output": "cogtaskonomy", "neg_sample": ["transfer learning is done by using Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "Experiments on 12 NLP tasks , where BERT / TinyBERT are used as the underlying models for transfer learning , demonstrate that the proposed CogTaxonomy is able to guide transfer learning , achieving performance competitive to the Analytic Hierarchy Process ( Saaty , 1987 ) used in visual Taskonomy ( Zamir et al . , 2018 ) but without requiring exhaustive pairwise O(m^2 ) task transferring .", "forward": false, "src_ids": "2022.acl-long.64_4415"}
+{"input": "transfer learning is used for Task| context: taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "transfer learning", "output": "natural language processing", "neg_sample": ["transfer learning is used for Task", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "CogTaskonomy : Cognitively Inspired Task Taxonomy Is Beneficial to Transfer Learning in NLP.", "forward": true, "src_ids": "2022.acl-long.64_4416"}
+{"input": "model - agnostic task taxonomy is done by using Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "model - agnostic task taxonomy", "output": "cognitive - neural mapping ( cnm )", "neg_sample": ["model - agnostic task taxonomy is done by using Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "Analyses further discover that CNM is capable of learning model - agnostic task taxonomy .", "forward": false, "src_ids": "2022.acl-long.64_4417"}
+{"input": "computational neuroscience is done by using Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "computational neuroscience", "output": "representational similarity analysis", "neg_sample": ["computational neuroscience is done by using Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "The former employs Representational Similarity Analysis , which is commonly used in computational neuroscience tofind a correlation between brain - activity measurement and computational modeling , to estimate task similarity with task - specific sentence representations .", "forward": false, "src_ids": "2022.acl-long.64_4418"}
+{"input": "task similarity is done by using Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "task similarity", "output": "representational similarity analysis", "neg_sample": ["task similarity is done by using Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "The former employs Representational Similarity Analysis , which is commonly used in computational neuroscience tofind a correlation between brain - activity measurement and computational modeling , to estimate task similarity with task - specific sentence representations .", "forward": false, "src_ids": "2022.acl-long.64_4419"}
+{"input": "representational similarity analysis is used for Task| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "representational similarity analysis", "output": "computational neuroscience", "neg_sample": ["representational similarity analysis is used for Task", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "The former employs Representational Similarity Analysis , which is commonly used in computational neuroscience tofind a correlation between brain - activity measurement and computational modeling , to estimate task similarity with task - specific sentence representations .", "forward": true, "src_ids": "2022.acl-long.64_4420"}
+{"input": "representational similarity analysis is used for OtherScientificTerm| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "representational similarity analysis", "output": "task similarity", "neg_sample": ["representational similarity analysis is used for OtherScientificTerm", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "The former employs Representational Similarity Analysis , which is commonly used in computational neuroscience tofind a correlation between brain - activity measurement and computational modeling , to estimate task similarity with task - specific sentence representations .", "forward": true, "src_ids": "2022.acl-long.64_4421"}
+{"input": "cognitive signals is done by using Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "cognitive signals", "output": "neural representations", "neg_sample": ["cognitive signals is done by using Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "The latter learns to detect task relations by projecting neural representations from NLP models to cognitive signals ( i.e. , fMRI voxels ) .", "forward": false, "src_ids": "2022.acl-long.64_4422"}
+{"input": "neural representations is used for Material| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "neural representations", "output": "cognitive signals", "neg_sample": ["neural representations is used for Material", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "The latter learns to detect task relations by projecting neural representations from NLP models to cognitive signals ( i.e. , fMRI voxels ) .", "forward": true, "src_ids": "2022.acl-long.64_4423"}
+{"input": "cognitive - neural mapping ( cnm ) is used for Method| context: is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ? taxonomy ( zamir et al . , 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them .", "entity": "cognitive - neural mapping ( cnm )", "output": "model - agnostic task taxonomy", "neg_sample": ["cognitive - neural mapping ( cnm ) is used for Method", "is there a principle to guide transfer learning across tasks in natural language processing ( nlp ) ?", "taxonomy ( zamir et al .", ", 2018 ) finds that a structure exists among visual tasks , as a principle underlying transfer learning for them ."], "relation": "used for", "id": "2022.acl-long.64", "year": 2022, "rel_sent": "Analyses further discover that CNM is capable of learning model - agnostic task taxonomy .", "forward": true, "src_ids": "2022.acl-long.64_4424"}
+{"input": "risk score elements is done by using Method| context: clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories . although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort . previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score .", "entity": "risk score elements", "output": "qa models", "neg_sample": ["risk score elements is done by using Method", "clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories .", "although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort .", "previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score ."], "relation": "used for", "id": "2022.bionlp-1.42", "year": 2022, "rel_sent": "We show that QA models can achieve comparable or better performance for certain risk score elements as compared to heuristic - based methods , and demonstrate the potential for more scalable risk score automation without the need for expert - curated heuristics .", "forward": false, "src_ids": "2022.bionlp-1.42_4425"}
+{"input": "risk score automation is done by using Method| context: clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories . although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort . previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score .", "entity": "risk score automation", "output": "qa models", "neg_sample": ["risk score automation is done by using Method", "clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories .", "although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort .", "previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score ."], "relation": "used for", "id": "2022.bionlp-1.42", "year": 2022, "rel_sent": "We show that QA models can achieve comparable or better performance for certain risk score elements as compared to heuristic - based methods , and demonstrate the potential for more scalable risk score automation without the need for expert - curated heuristics .", "forward": false, "src_ids": "2022.bionlp-1.42_4426"}
+{"input": "qa models is used for OtherScientificTerm| context: clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories . although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort . previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score .", "entity": "qa models", "output": "risk score elements", "neg_sample": ["qa models is used for OtherScientificTerm", "clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories .", "although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort .", "previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score ."], "relation": "used for", "id": "2022.bionlp-1.42", "year": 2022, "rel_sent": "We show that QA models can achieve comparable or better performance for certain risk score elements as compared to heuristic - based methods , and demonstrate the potential for more scalable risk score automation without the need for expert - curated heuristics .", "forward": true, "src_ids": "2022.bionlp-1.42_4427"}
+{"input": "qa models is used for Task| context: clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories . although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort . previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score .", "entity": "qa models", "output": "risk score automation", "neg_sample": ["qa models is used for Task", "clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories .", "although risk scores are widely used to inform decision - making at the point - of - care , collecting the information necessary to calculate such scores requires considerable time and effort .", "previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score ."], "relation": "used for", "id": "2022.bionlp-1.42", "year": 2022, "rel_sent": "We show that QA models can achieve comparable or better performance for certain risk score elements as compared to heuristic - based methods , and demonstrate the potential for more scalable risk score automation without the need for expert - curated heuristics .", "forward": true, "src_ids": "2022.bionlp-1.42_4428"}
+{"input": "extracting of - fensive spans is done by using Method| context: identifying offensive speech is an exciting andessential area of research , with ample tractionin recent times .", "entity": "extracting of - fensive spans", "output": "supervised approaches", "neg_sample": ["extracting of - fensive spans is done by using Method", "identifying offensive speech is an exciting andessential area of research , with ample tractionin recent times ."], "relation": "used for", "id": "2022.dravidianlangtech-1.38", "year": 2022, "rel_sent": "This paper presents our sys - tem submission to the subtask 1 , focusing onusing supervised approaches for extracting Of - fensive spans from code - mixed Tamil - Englishcomments .", "forward": false, "src_ids": "2022.dravidianlangtech-1.38_4429"}
+{"input": "supervised approaches is used for Task| context: identifying offensive speech is an exciting andessential area of research , with ample tractionin recent times .", "entity": "supervised approaches", "output": "extracting of - fensive spans", "neg_sample": ["supervised approaches is used for Task", "identifying offensive speech is an exciting andessential area of research , with ample tractionin recent times ."], "relation": "used for", "id": "2022.dravidianlangtech-1.38", "year": 2022, "rel_sent": "This paper presents our sys - tem submission to the subtask 1 , focusing onusing supervised approaches for extracting Of - fensive spans from code - mixed Tamil - Englishcomments .", "forward": true, "src_ids": "2022.dravidianlangtech-1.38_4430"}
+{"input": "linguistic property is done by using Method| context: however , recent studies have demonstrated various methodological limitations of this approach .", "entity": "linguistic property", "output": "classifier", "neg_sample": ["linguistic property is done by using Method", "however , recent studies have demonstrated various methodological limitations of this approach ."], "relation": "used for", "id": "2022.cl-1.7", "year": 2022, "rel_sent": "The basic idea is simple - a classifier is trained to predict some linguistic property from a model 's representations - and has been used to examine a wide variety of models and properties .", "forward": false, "src_ids": "2022.cl-1.7_4431"}
+{"input": "classifier is used for OtherScientificTerm| context: probing classifiers have emerged as one of the prominent methodologies for interpreting and analyzing deep neural network models of natural language processing . however , recent studies have demonstrated various methodological limitations of this approach .", "entity": "classifier", "output": "linguistic property", "neg_sample": ["classifier is used for OtherScientificTerm", "probing classifiers have emerged as one of the prominent methodologies for interpreting and analyzing deep neural network models of natural language processing .", "however , recent studies have demonstrated various methodological limitations of this approach ."], "relation": "used for", "id": "2022.cl-1.7", "year": 2022, "rel_sent": "The basic idea is simple - a classifier is trained to predict some linguistic property from a model 's representations - and has been used to examine a wide variety of models and properties .", "forward": true, "src_ids": "2022.cl-1.7_4432"}
+{"input": "equality is done by using Task| context: hope speech is any message or content that is positive , encouraging , reassuring , inclusive and supportive that inspires and engenders optimism in the minds of people .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "hope speech is any message or content that is positive , encouraging , reassuring , inclusive and supportive that inspires and engenders optimism in the minds of people ."], "relation": "used for", "id": "2022.ltedi-1.58", "year": 2022, "rel_sent": "Overview of the Shared Task on Hope Speech Detection for Equality , Diversity , and Inclusion.", "forward": false, "src_ids": "2022.ltedi-1.58_4433"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: hope speech detection is the task of classifying a sentence as hope speech or non - hope speech given a corpus of sentences . hope speech is any message or content that is positive , encouraging , reassuring , inclusive and supportive that inspires and engenders optimism in the minds of people . in contrast to identifying and censoring negative speech patterns , hope speech detection is focussed on recognising and promoting positive speech patterns online .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "hope speech detection is the task of classifying a sentence as hope speech or non - hope speech given a corpus of sentences .", "hope speech is any message or content that is positive , encouraging , reassuring , inclusive and supportive that inspires and engenders optimism in the minds of people .", "in contrast to identifying and censoring negative speech patterns , hope speech detection is focussed on recognising and promoting positive speech patterns online ."], "relation": "used for", "id": "2022.ltedi-1.58", "year": 2022, "rel_sent": "Overview of the Shared Task on Hope Speech Detection for Equality , Diversity , and Inclusion.", "forward": true, "src_ids": "2022.ltedi-1.58_4434"}
+{"input": "joint model is done by using Method| context: a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain . although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) . these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming . additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results .", "entity": "joint model", "output": "wisdom", "neg_sample": ["joint model is done by using Method", "a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain .", "although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) .", "these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming .", "additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results ."], "relation": "used for", "id": "2022.findings-acl.94", "year": 2022, "rel_sent": "WISDOM learns a joint model on the ( same ) labeled dataset used for LF induction along with any unlabeled data in a semi - supervised manner , and more critically , reweighs each LF according to its goodness , influencing its contribution to the semi - supervised loss using a robust bi - level optimization algorithm .", "forward": false, "src_ids": "2022.findings-acl.94_4435"}
+{"input": "lf induction is done by using Method| context: a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain . although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) . these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming . additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results .", "entity": "lf induction", "output": "joint model", "neg_sample": ["lf induction is done by using Method", "a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain .", "although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) .", "these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming .", "additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results ."], "relation": "used for", "id": "2022.findings-acl.94", "year": 2022, "rel_sent": "WISDOM learns a joint model on the ( same ) labeled dataset used for LF induction along with any unlabeled data in a semi - supervised manner , and more critically , reweighs each LF according to its goodness , influencing its contribution to the semi - supervised loss using a robust bi - level optimization algorithm .", "forward": false, "src_ids": "2022.findings-acl.94_4436"}
+{"input": "wisdom is used for Method| context: a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain . although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) . these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming . additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results .", "entity": "wisdom", "output": "joint model", "neg_sample": ["wisdom is used for Method", "a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain .", "although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) .", "these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming .", "additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results ."], "relation": "used for", "id": "2022.findings-acl.94", "year": 2022, "rel_sent": "WISDOM learns a joint model on the ( same ) labeled dataset used for LF induction along with any unlabeled data in a semi - supervised manner , and more critically , reweighs each LF according to its goodness , influencing its contribution to the semi - supervised loss using a robust bi - level optimization algorithm .", "forward": true, "src_ids": "2022.findings-acl.94_4437"}
+{"input": "lf induction is done by using Material| context: a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain . although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) . these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming . additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results .", "entity": "lf induction", "output": "labeled dataset", "neg_sample": ["lf induction is done by using Material", "a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain .", "although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) .", "these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming .", "additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results ."], "relation": "used for", "id": "2022.findings-acl.94", "year": 2022, "rel_sent": "WISDOM learns a joint model on the ( same ) labeled dataset used for LF induction along with any unlabeled data in a semi - supervised manner , and more critically , reweighs each LF according to its goodness , influencing its contribution to the semi - supervised loss using a robust bi - level optimization algorithm .", "forward": false, "src_ids": "2022.findings-acl.94_4438"}
+{"input": "labeled dataset is used for Task| context: a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain . although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) . these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming . additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results .", "entity": "labeled dataset", "output": "lf induction", "neg_sample": ["labeled dataset is used for Task", "a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain .", "although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) .", "these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming .", "additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results ."], "relation": "used for", "id": "2022.findings-acl.94", "year": 2022, "rel_sent": "WISDOM learns a joint model on the ( same ) labeled dataset used for LF induction along with any unlabeled data in a semi - supervised manner , and more critically , reweighs each LF according to its goodness , influencing its contribution to the semi - supervised loss using a robust bi - level optimization algorithm .", "forward": true, "src_ids": "2022.findings-acl.94_4439"}
+{"input": "joint model is used for Task| context: a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain . although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) . these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming . additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results .", "entity": "joint model", "output": "lf induction", "neg_sample": ["joint model is used for Task", "a critical bottleneck in supervised machine learning is the need for large amounts of labeled data which is expensive and time - consuming to obtain .", "although a small amount of labeled data can not be used to train a model , it can be used effectively for the generation of humaninterpretable labeling functions ( lfs ) .", "these lfs , in turn , have been used to generate a large amount of additional noisy labeled data in a paradigm that is now commonly referred to as data programming .", "additionally , since the lfs are generated automatically , they are likely to be noisy , and naively aggregating these lfs can lead to suboptimal results ."], "relation": "used for", "id": "2022.findings-acl.94", "year": 2022, "rel_sent": "WISDOM learns a joint model on the ( same ) labeled dataset used for LF induction along with any unlabeled data in a semi - supervised manner , and more critically , reweighs each LF according to its goodness , influencing its contribution to the semi - supervised loss using a robust bi - level optimization algorithm .", "forward": true, "src_ids": "2022.findings-acl.94_4440"}
+{"input": "human trafficking is done by using Task| context: online escort advertisement websites are widely used for advertising victims of human trafficking . domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking . thus , extracting person names from the text of these ads can provide valuable clues for further analysis . most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task .", "entity": "human trafficking", "output": "named - entity recognition", "neg_sample": ["human trafficking is done by using Task", "online escort advertisement websites are widely used for advertising victims of human trafficking .", "domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking .", "thus , extracting person names from the text of these ads can provide valuable clues for further analysis .", "most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task ."], "relation": "used for", "id": "2022.findings-acl.225", "year": 2022, "rel_sent": "Extracting Person Names from User Generated Text : Named - Entity Recognition for Combating Human Trafficking.", "forward": false, "src_ids": "2022.findings-acl.225_4441"}
+{"input": "named - entity recognition is used for Task| context: domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking . thus , extracting person names from the text of these ads can provide valuable clues for further analysis . however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation . most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task .", "entity": "named - entity recognition", "output": "human trafficking", "neg_sample": ["named - entity recognition is used for Task", "domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking .", "thus , extracting person names from the text of these ads can provide valuable clues for further analysis .", "however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation .", "most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task ."], "relation": "used for", "id": "2022.findings-acl.225", "year": 2022, "rel_sent": "Extracting Person Names from User Generated Text : Named - Entity Recognition for Combating Human Trafficking.", "forward": true, "src_ids": "2022.findings-acl.225_4442"}
+{"input": "ambiguous names is done by using Method| context: online escort advertisement websites are widely used for advertising victims of human trafficking . domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking . thus , extracting person names from the text of these ads can provide valuable clues for further analysis . however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation . most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task .", "entity": "ambiguous names", "output": "contextualized language model", "neg_sample": ["ambiguous names is done by using Method", "online escort advertisement websites are widely used for advertising victims of human trafficking .", "domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking .", "thus , extracting person names from the text of these ads can provide valuable clues for further analysis .", "however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation .", "most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task ."], "relation": "used for", "id": "2022.findings-acl.225", "year": 2022, "rel_sent": "It effectively combines classic rule - based and dictionary extractors with a contextualized language model to capture ambiguous names ( e.g penny , hazel ) and adapts to adversarial changes in the text by expanding its dictionary .", "forward": false, "src_ids": "2022.findings-acl.225_4443"}
+{"input": "contextualized language model is used for OtherScientificTerm| context: online escort advertisement websites are widely used for advertising victims of human trafficking . domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking . thus , extracting person names from the text of these ads can provide valuable clues for further analysis . however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation . most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task .", "entity": "contextualized language model", "output": "ambiguous names", "neg_sample": ["contextualized language model is used for OtherScientificTerm", "online escort advertisement websites are widely used for advertising victims of human trafficking .", "domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking .", "thus , extracting person names from the text of these ads can provide valuable clues for further analysis .", "however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation .", "most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task ."], "relation": "used for", "id": "2022.findings-acl.225", "year": 2022, "rel_sent": "It effectively combines classic rule - based and dictionary extractors with a contextualized language model to capture ambiguous names ( e.g penny , hazel ) and adapts to adversarial changes in the text by expanding its dictionary .", "forward": true, "src_ids": "2022.findings-acl.225_4444"}
+{"input": "name extraction is done by using Method| context: online escort advertisement websites are widely used for advertising victims of human trafficking . domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking . thus , extracting person names from the text of these ads can provide valuable clues for further analysis . however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation . most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task .", "entity": "name extraction", "output": "name extraction against trafficking", "neg_sample": ["name extraction is done by using Method", "online escort advertisement websites are widely used for advertising victims of human trafficking .", "domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking .", "thus , extracting person names from the text of these ads can provide valuable clues for further analysis .", "however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation .", "most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task ."], "relation": "used for", "id": "2022.findings-acl.225", "year": 2022, "rel_sent": "NEAT shows 19 % improvement on average in the F1 classification score for name extraction compared to previous state - of - the - art in two domain - specific datasets .", "forward": false, "src_ids": "2022.findings-acl.225_4445"}
+{"input": "name extraction against trafficking is used for Task| context: online escort advertisement websites are widely used for advertising victims of human trafficking . domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking . thus , extracting person names from the text of these ads can provide valuable clues for further analysis . however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation . most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task .", "entity": "name extraction against trafficking", "output": "name extraction", "neg_sample": ["name extraction against trafficking is used for Task", "online escort advertisement websites are widely used for advertising victims of human trafficking .", "domain experts agree that advertising multiple people in the same ad is a strong indicator of trafficking .", "thus , extracting person names from the text of these ads can provide valuable clues for further analysis .", "however , named - entity recognition ( ner ) on escort ads is challenging because the text can be noisy , colloquial and often lacking proper grammar and punctuation .", "most existing state - of - the - art ner models fail to demonstrate satisfactory performance in this task ."], "relation": "used for", "id": "2022.findings-acl.225", "year": 2022, "rel_sent": "NEAT shows 19 % improvement on average in the F1 classification score for name extraction compared to previous state - of - the - art in two domain - specific datasets .", "forward": true, "src_ids": "2022.findings-acl.225_4446"}
+{"input": "rct abstract is done by using OtherScientificTerm| context: we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) .", "entity": "rct abstract", "output": "structured objects", "neg_sample": ["rct abstract is done by using OtherScientificTerm", "we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) ."], "relation": "used for", "id": "2022.bionlp-1.18", "year": 2022, "rel_sent": "As main benefit , our approach yields a structured object for every RCT abstract that supports the aggregation and summarization of clinical trial results across published studies and can facilitate the task of creating a systematic review or meta - analysis .", "forward": false, "src_ids": "2022.bionlp-1.18_4447"}
+{"input": "systematic review is done by using OtherScientificTerm| context: we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) .", "entity": "systematic review", "output": "structured objects", "neg_sample": ["systematic review is done by using OtherScientificTerm", "we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) ."], "relation": "used for", "id": "2022.bionlp-1.18", "year": 2022, "rel_sent": "As main benefit , our approach yields a structured object for every RCT abstract that supports the aggregation and summarization of clinical trial results across published studies and can facilitate the task of creating a systematic review or meta - analysis .", "forward": false, "src_ids": "2022.bionlp-1.18_4448"}
+{"input": "glaucoma is done by using Material| context: we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) . in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects .", "entity": "glaucoma", "output": "manually annotated abstracts", "neg_sample": ["glaucoma is done by using Material", "we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) .", "in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects ."], "relation": "used for", "id": "2022.bionlp-1.18", "year": 2022, "rel_sent": "We evaluate the approach on a dataset of 211 manually annotated abstracts for type 2 Diabetes and Glaucoma , showing the positive impact of modelling intra - template entity compatibility .", "forward": false, "src_ids": "2022.bionlp-1.18_4449"}
+{"input": "manually annotated abstracts is used for Material| context: we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) . in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects .", "entity": "manually annotated abstracts", "output": "glaucoma", "neg_sample": ["manually annotated abstracts is used for Material", "we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) .", "in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects ."], "relation": "used for", "id": "2022.bionlp-1.18", "year": 2022, "rel_sent": "We evaluate the approach on a dataset of 211 manually annotated abstracts for type 2 Diabetes and Glaucoma , showing the positive impact of modelling intra - template entity compatibility .", "forward": true, "src_ids": "2022.bionlp-1.18_4450"}
+{"input": "structured objects is used for OtherScientificTerm| context: we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) . in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects .", "entity": "structured objects", "output": "rct abstract", "neg_sample": ["structured objects is used for OtherScientificTerm", "we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) .", "in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects ."], "relation": "used for", "id": "2022.bionlp-1.18", "year": 2022, "rel_sent": "As main benefit , our approach yields a structured object for every RCT abstract that supports the aggregation and summarization of clinical trial results across published studies and can facilitate the task of creating a systematic review or meta - analysis .", "forward": true, "src_ids": "2022.bionlp-1.18_4451"}
+{"input": "structured objects is used for Task| context: we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) . in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects .", "entity": "structured objects", "output": "systematic review", "neg_sample": ["structured objects is used for Task", "we present a deep learning based information extraction system that can extract the design and results of a published abstract describing a randomized controlled trial ( rct ) .", "in contrast to other approaches , our system does not regard the pico elements as flat objects or labels but as structured objects ."], "relation": "used for", "id": "2022.bionlp-1.18", "year": 2022, "rel_sent": "As main benefit , our approach yields a structured object for every RCT abstract that supports the aggregation and summarization of clinical trial results across published studies and can facilitate the task of creating a systematic review or meta - analysis .", "forward": true, "src_ids": "2022.bionlp-1.18_4452"}
+{"input": "graphical models is used for Method| context: many methods now exist for conditioning models on task instructions and user - provided explanations for individual data points . these methods show great promise for improving task performance of language models beyond what can be achieved by learning from individual ( x , y ) pairs .", "entity": "graphical models", "output": "modeling approaches", "neg_sample": ["graphical models is used for Method", "many methods now exist for conditioning models on task instructions and user - provided explanations for individual data points .", "these methods show great promise for improving task performance of language models beyond what can be achieved by learning from individual ( x , y ) pairs ."], "relation": "used for", "id": "2022.lnls-1.4", "year": 2022, "rel_sent": "In the first direction , we give graphical models for the available modeling approaches , in which explanation data can be used as model inputs , as targets , or as a prior .", "forward": true, "src_ids": "2022.lnls-1.4_4453"}
+{"input": "modeling approaches is done by using Method| context: many methods now exist for conditioning models on task instructions and user - provided explanations for individual data points . these methods show great promise for improving task performance of language models beyond what can be achieved by learning from individual ( x , y ) pairs .", "entity": "modeling approaches", "output": "graphical models", "neg_sample": ["modeling approaches is done by using Method", "many methods now exist for conditioning models on task instructions and user - provided explanations for individual data points .", "these methods show great promise for improving task performance of language models beyond what can be achieved by learning from individual ( x , y ) pairs ."], "relation": "used for", "id": "2022.lnls-1.4", "year": 2022, "rel_sent": "In the first direction , we give graphical models for the available modeling approaches , in which explanation data can be used as model inputs , as targets , or as a prior .", "forward": false, "src_ids": "2022.lnls-1.4_4454"}
+{"input": "zerorte task is done by using Method| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "zerorte task", "output": "relationprompt", "neg_sample": ["zerorte task is done by using Method", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Experiments on FewRel and Wiki - ZSL datasets show the efficacy of RelationPrompt for the ZeroRTE task and zero - shot relation classification .", "forward": false, "src_ids": "2022.findings-acl.5_4455"}
+{"input": "zero - shot relation classification is done by using Method| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "zero - shot relation classification", "output": "relationprompt", "neg_sample": ["zero - shot relation classification is done by using Method", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Experiments on FewRel and Wiki - ZSL datasets show the efficacy of RelationPrompt for the ZeroRTE task and zero - shot relation classification .", "forward": false, "src_ids": "2022.findings-acl.5_4456"}
+{"input": "zero - shot relation triplet extraction is done by using Material| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "zero - shot relation triplet extraction", "output": "synthetic data", "neg_sample": ["zero - shot relation triplet extraction is done by using Material", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "RelationPrompt : Leveraging Prompts to Generate Synthetic Data for Zero - Shot Relation Triplet Extraction.", "forward": false, "src_ids": "2022.findings-acl.5_4457"}
+{"input": "synthetic data is used for Task| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "synthetic data", "output": "zero - shot relation triplet extraction", "neg_sample": ["synthetic data is used for Task", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "RelationPrompt : Leveraging Prompts to Generate Synthetic Data for Zero - Shot Relation Triplet Extraction.", "forward": true, "src_ids": "2022.findings-acl.5_4458"}
+{"input": "structured texts is done by using Method| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "structured texts", "output": "language models", "neg_sample": ["structured texts is done by using Method", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "To solve ZeroRTE , we propose to synthesize relation examples by prompting language models to generate structured texts .", "forward": false, "src_ids": "2022.findings-acl.5_4459"}
+{"input": "language models is used for Material| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "language models", "output": "structured texts", "neg_sample": ["language models is used for Material", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "To solve ZeroRTE , we propose to synthesize relation examples by prompting language models to generate structured texts .", "forward": true, "src_ids": "2022.findings-acl.5_4460"}
+{"input": "structured prompt template is done by using OtherScientificTerm| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "structured prompt template", "output": "language model prompts", "neg_sample": ["structured prompt template is done by using OtherScientificTerm", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Concretely , we unify language model prompts and structured text approaches to design a structured prompt template for generating synthetic relation samples when conditioning on relation label prompts ( RelationPrompt ) .", "forward": false, "src_ids": "2022.findings-acl.5_4461"}
+{"input": "structured prompt template is done by using Method| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "structured prompt template", "output": "structured text approaches", "neg_sample": ["structured prompt template is done by using Method", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Concretely , we unify language model prompts and structured text approaches to design a structured prompt template for generating synthetic relation samples when conditioning on relation label prompts ( RelationPrompt ) .", "forward": false, "src_ids": "2022.findings-acl.5_4462"}
+{"input": "synthetic relation samples is done by using OtherScientificTerm| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "synthetic relation samples", "output": "structured prompt template", "neg_sample": ["synthetic relation samples is done by using OtherScientificTerm", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Concretely , we unify language model prompts and structured text approaches to design a structured prompt template for generating synthetic relation samples when conditioning on relation label prompts ( RelationPrompt ) .", "forward": false, "src_ids": "2022.findings-acl.5_4463"}
+{"input": "structured text approaches is used for OtherScientificTerm| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "structured text approaches", "output": "structured prompt template", "neg_sample": ["structured text approaches is used for OtherScientificTerm", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Concretely , we unify language model prompts and structured text approaches to design a structured prompt template for generating synthetic relation samples when conditioning on relation label prompts ( RelationPrompt ) .", "forward": true, "src_ids": "2022.findings-acl.5_4464"}
+{"input": "language model prompts is used for OtherScientificTerm| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "language model prompts", "output": "structured prompt template", "neg_sample": ["language model prompts is used for OtherScientificTerm", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Concretely , we unify language model prompts and structured text approaches to design a structured prompt template for generating synthetic relation samples when conditioning on relation label prompts ( RelationPrompt ) .", "forward": true, "src_ids": "2022.findings-acl.5_4465"}
+{"input": "structured prompt template is used for OtherScientificTerm| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "structured prompt template", "output": "synthetic relation samples", "neg_sample": ["structured prompt template is used for OtherScientificTerm", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Concretely , we unify language model prompts and structured text approaches to design a structured prompt template for generating synthetic relation samples when conditioning on relation label prompts ( RelationPrompt ) .", "forward": true, "src_ids": "2022.findings-acl.5_4466"}
+{"input": "relationprompt is used for Task| context: despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types .", "entity": "relationprompt", "output": "zerorte task", "neg_sample": ["relationprompt is used for Task", "despite the importance of relation extraction in building and representing knowledge , less research is focused on generalizing to unseen relations types ."], "relation": "used for", "id": "2022.findings-acl.5", "year": 2022, "rel_sent": "Experiments on FewRel and Wiki - ZSL datasets show the efficacy of RelationPrompt for the ZeroRTE task and zero - shot relation classification .", "forward": true, "src_ids": "2022.findings-acl.5_4467"}
+{"input": "features is done by using Method| context: depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide . therefore , it is important to recognize the signs of depression early . in the last decade , social media has become one of the most common places to express one 's feelings . hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression .", "entity": "features", "output": "automl", "neg_sample": ["features is done by using Method", "depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide .", "therefore , it is important to recognize the signs of depression early .", "in the last decade , social media has become one of the most common places to express one 's feelings .", "hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression ."], "relation": "used for", "id": "2022.ltedi-1.36", "year": 2022, "rel_sent": "We explore three different approaches to solve the challenge : fine - tuning BERT model , leveraging AutoML for the construction of features and classifier selection and finally , we explore latent spaces derived from the combination of textual and knowledge - based representations .", "forward": false, "src_ids": "2022.ltedi-1.36_4468"}
+{"input": "classifier selection is done by using Method| context: depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide . therefore , it is important to recognize the signs of depression early . in the last decade , social media has become one of the most common places to express one 's feelings . hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression .", "entity": "classifier selection", "output": "automl", "neg_sample": ["classifier selection is done by using Method", "depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide .", "therefore , it is important to recognize the signs of depression early .", "in the last decade , social media has become one of the most common places to express one 's feelings .", "hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression ."], "relation": "used for", "id": "2022.ltedi-1.36", "year": 2022, "rel_sent": "We explore three different approaches to solve the challenge : fine - tuning BERT model , leveraging AutoML for the construction of features and classifier selection and finally , we explore latent spaces derived from the combination of textual and knowledge - based representations .", "forward": false, "src_ids": "2022.ltedi-1.36_4469"}
+{"input": "automl is used for OtherScientificTerm| context: depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide . therefore , it is important to recognize the signs of depression early . in the last decade , social media has become one of the most common places to express one 's feelings . hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression .", "entity": "automl", "output": "features", "neg_sample": ["automl is used for OtherScientificTerm", "depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide .", "therefore , it is important to recognize the signs of depression early .", "in the last decade , social media has become one of the most common places to express one 's feelings .", "hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression ."], "relation": "used for", "id": "2022.ltedi-1.36", "year": 2022, "rel_sent": "We explore three different approaches to solve the challenge : fine - tuning BERT model , leveraging AutoML for the construction of features and classifier selection and finally , we explore latent spaces derived from the combination of textual and knowledge - based representations .", "forward": true, "src_ids": "2022.ltedi-1.36_4470"}
+{"input": "automl is used for Task| context: depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide . therefore , it is important to recognize the signs of depression early . in the last decade , social media has become one of the most common places to express one 's feelings . hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression .", "entity": "automl", "output": "classifier selection", "neg_sample": ["automl is used for Task", "depression is a mental illness that negatively affects a person 's well - being and can , if left untreated , lead to serious consequences such as suicide .", "therefore , it is important to recognize the signs of depression early .", "in the last decade , social media has become one of the most common places to express one 's feelings .", "hence , there is a possibility of text processing and applying machine learning techniques to detect possible signs of depression ."], "relation": "used for", "id": "2022.ltedi-1.36", "year": 2022, "rel_sent": "We explore three different approaches to solve the challenge : fine - tuning BERT model , leveraging AutoML for the construction of features and classifier selection and finally , we explore latent spaces derived from the combination of textual and knowledge - based representations .", "forward": true, "src_ids": "2022.ltedi-1.36_4471"}
+{"input": "classifiers is done by using Method| context: we demonstrate that knowledge distillation can be used not only to reduce model size , but to simultaneously adapt a contextual language model to a specific domain .", "entity": "classifiers", "output": "encoder", "neg_sample": ["classifiers is done by using Method", "we demonstrate that knowledge distillation can be used not only to reduce model size , but to simultaneously adapt a contextual language model to a specific domain ."], "relation": "used for", "id": "2022.ecnlp-1.18", "year": 2022, "rel_sent": "Whereas much previous work with BERT has fine - tuned the encoder weights during task training , we show that the model improvements from distillation on in - domain data persist even when the encoder weights are frozen during task training , allowing a single encoder to support classifiers for multiple tasks and languages .", "forward": false, "src_ids": "2022.ecnlp-1.18_4472"}
+{"input": "encoder is used for Method| context: we demonstrate that knowledge distillation can be used not only to reduce model size , but to simultaneously adapt a contextual language model to a specific domain .", "entity": "encoder", "output": "classifiers", "neg_sample": ["encoder is used for Method", "we demonstrate that knowledge distillation can be used not only to reduce model size , but to simultaneously adapt a contextual language model to a specific domain ."], "relation": "used for", "id": "2022.ecnlp-1.18", "year": 2022, "rel_sent": "Whereas much previous work with BERT has fine - tuned the encoder weights during task training , we show that the model improvements from distillation on in - domain data persist even when the encoder weights are frozen during task training , allowing a single encoder to support classifiers for multiple tasks and languages .", "forward": true, "src_ids": "2022.ecnlp-1.18_4473"}
+{"input": "nli models is done by using Method| context: natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding . however , the ability of nli models to perform inferences requiring understanding of figurative language such as idioms and metaphors remains understudied .", "entity": "nli models", "output": "idiomatic and metaphoric paired language inference", "neg_sample": ["nli models is done by using Method", "natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding .", "however , the ability of nli models to perform inferences requiring understanding of figurative language such as idioms and metaphors remains understudied ."], "relation": "used for", "id": "2022.acl-long.369", "year": 2022, "rel_sent": "We use IMPLI to evaluate NLI models based on RoBERTa fine - tuned on the widely used MNLI dataset .", "forward": false, "src_ids": "2022.acl-long.369_4474"}
+{"input": "understanding figurative language is done by using Method| context: natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding .", "entity": "understanding figurative language", "output": "nli models", "neg_sample": ["understanding figurative language is done by using Method", "natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding ."], "relation": "used for", "id": "2022.acl-long.369", "year": 2022, "rel_sent": "This suggests the limits of current NLI models with regard to understanding figurative language and this dataset serves as a benchmark for future improvements in this direction .", "forward": false, "src_ids": "2022.acl-long.369_4475"}
+{"input": "idiomatic and metaphoric paired language inference is used for Method| context: natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding .", "entity": "idiomatic and metaphoric paired language inference", "output": "nli models", "neg_sample": ["idiomatic and metaphoric paired language inference is used for Method", "natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding ."], "relation": "used for", "id": "2022.acl-long.369", "year": 2022, "rel_sent": "We use IMPLI to evaluate NLI models based on RoBERTa fine - tuned on the widely used MNLI dataset .", "forward": true, "src_ids": "2022.acl-long.369_4476"}
+{"input": "nli models is used for Task| context: natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding . however , the ability of nli models to perform inferences requiring understanding of figurative language such as idioms and metaphors remains understudied .", "entity": "nli models", "output": "understanding figurative language", "neg_sample": ["nli models is used for Task", "natural language inference ( nli ) has been widely used as a task to train and evaluate models for language understanding .", "however , the ability of nli models to perform inferences requiring understanding of figurative language such as idioms and metaphors remains understudied ."], "relation": "used for", "id": "2022.acl-long.369", "year": 2022, "rel_sent": "This suggests the limits of current NLI models with regard to understanding figurative language and this dataset serves as a benchmark for future improvements in this direction .", "forward": true, "src_ids": "2022.acl-long.369_4477"}
+{"input": "dependency structures is done by using Method| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "dependency structures", "output": "unsupervised dependency graph network ( udgn )", "neg_sample": ["dependency structures is done by using Method", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "We introduce a new model , the Unsupervised Dependency Graph Network ( UDGN ) , that can induce dependency structures from raw corpora and the masked language modeling task .", "forward": false, "src_ids": "2022.acl-long.327_4478"}
+{"input": "masked language modeling task is done by using Method| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "masked language modeling task", "output": "unsupervised dependency graph network ( udgn )", "neg_sample": ["masked language modeling task is done by using Method", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "Furthermore , the UDGN can also achieve competitive performance on masked language modeling and sentence textual similarity tasks .", "forward": false, "src_ids": "2022.acl-long.327_4479"}
+{"input": "dependency relations is done by using Method| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "dependency relations", "output": "competitive mechanism", "neg_sample": ["dependency relations is done by using Method", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "Inspired by these developments , we propose a new competitive mechanism that encourages these attention heads to model different dependency relations .", "forward": false, "src_ids": "2022.acl-long.327_4480"}
+{"input": "attention heads is done by using Method| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "attention heads", "output": "competitive mechanism", "neg_sample": ["attention heads is done by using Method", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "Inspired by these developments , we propose a new competitive mechanism that encourages these attention heads to model different dependency relations .", "forward": false, "src_ids": "2022.acl-long.327_4481"}
+{"input": "competitive mechanism is used for OtherScientificTerm| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "entity": "competitive mechanism", "output": "attention heads", "neg_sample": ["competitive mechanism is used for OtherScientificTerm", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "Inspired by these developments , we propose a new competitive mechanism that encourages these attention heads to model different dependency relations .", "forward": true, "src_ids": "2022.acl-long.327_4482"}
+{"input": "competitive mechanism is used for OtherScientificTerm| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "competitive mechanism", "output": "dependency relations", "neg_sample": ["competitive mechanism is used for OtherScientificTerm", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "Inspired by these developments , we propose a new competitive mechanism that encourages these attention heads to model different dependency relations .", "forward": true, "src_ids": "2022.acl-long.327_4483"}
+{"input": "unsupervised dependency graph network ( udgn ) is used for OtherScientificTerm| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "unsupervised dependency graph network ( udgn )", "output": "dependency structures", "neg_sample": ["unsupervised dependency graph network ( udgn ) is used for OtherScientificTerm", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "We introduce a new model , the Unsupervised Dependency Graph Network ( UDGN ) , that can induce dependency structures from raw corpora and the masked language modeling task .", "forward": true, "src_ids": "2022.acl-long.327_4484"}
+{"input": "unsupervised dependency graph network ( udgn ) is used for Task| context: recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures . in particular , some self - attention heads correspond well to individual dependency types .", "entity": "unsupervised dependency graph network ( udgn )", "output": "masked language modeling task", "neg_sample": ["unsupervised dependency graph network ( udgn ) is used for Task", "recent work has identified properties of pretrained self - attention models that mirror those of dependency parse structures .", "in particular , some self - attention heads correspond well to individual dependency types ."], "relation": "used for", "id": "2022.acl-long.327", "year": 2022, "rel_sent": "Furthermore , the UDGN can also achieve competitive performance on masked language modeling and sentence textual similarity tasks .", "forward": true, "src_ids": "2022.acl-long.327_4485"}
+{"input": "data augmentation is used for Task| context: we study the effect of seven data augmentation ( da ) methods in factoid question answering , focusing on the biomedical domain , where obtaining training instances is particularly difficult .", "entity": "data augmentation", "output": "biomedical factoid question answering", "neg_sample": ["data augmentation is used for Task", "we study the effect of seven data augmentation ( da ) methods in factoid question answering , focusing on the biomedical domain , where obtaining training instances is particularly difficult ."], "relation": "used for", "id": "2022.bionlp-1.6", "year": 2022, "rel_sent": "Data Augmentation for Biomedical Factoid Question Answering.", "forward": true, "src_ids": "2022.bionlp-1.6_4486"}
+{"input": "word substitution is done by using Method| context: we study the effect of seven data augmentation ( da ) methods in factoid question answering , focusing on the biomedical domain , where obtaining training instances is particularly difficult .", "entity": "word substitution", "output": "back - translation", "neg_sample": ["word substitution is done by using Method", "we study the effect of seven data augmentation ( da ) methods in factoid question answering , focusing on the biomedical domain , where obtaining training instances is particularly difficult ."], "relation": "used for", "id": "2022.bionlp-1.6", "year": 2022, "rel_sent": "We experiment with data from the BIOASQ challenge , which we augment with training instances obtained from an artificial biomedical machine reading comprehension dataset , or via back - translation , information retrieval , word substitution based on WORD2VEC embeddings , or masked language modeling , question generation , or extending the given passage with additional context .", "forward": false, "src_ids": "2022.bionlp-1.6_4487"}
+{"input": "back - translation is used for Task| context: we study the effect of seven data augmentation ( da ) methods in factoid question answering , focusing on the biomedical domain , where obtaining training instances is particularly difficult .", "entity": "back - translation", "output": "word substitution", "neg_sample": ["back - translation is used for Task", "we study the effect of seven data augmentation ( da ) methods in factoid question answering , focusing on the biomedical domain , where obtaining training instances is particularly difficult ."], "relation": "used for", "id": "2022.bionlp-1.6", "year": 2022, "rel_sent": "We experiment with data from the BIOASQ challenge , which we augment with training instances obtained from an artificial biomedical machine reading comprehension dataset , or via back - translation , information retrieval , word substitution based on WORD2VEC embeddings , or masked language modeling , question generation , or extending the given passage with additional context .", "forward": true, "src_ids": "2022.bionlp-1.6_4488"}
+{"input": "coreferent mentions is done by using Method| context: coreference resolution - which is a crucial task for understanding discourse and language at large - has yet to witness widespread benefits from large language models ( llms ) . moreover , coreference resolution systems largely rely on supervised labels , which are highly expensive and difficult to annotate , thus making it ripe for prompt engineering .", "entity": "coreferent mentions", "output": "gpt - neo", "neg_sample": ["coreferent mentions is done by using Method", "coreference resolution - which is a crucial task for understanding discourse and language at large - has yet to witness widespread benefits from large language models ( llms ) .", "moreover , coreference resolution systems largely rely on supervised labels , which are highly expensive and difficult to annotate , thus making it ripe for prompt engineering ."], "relation": "used for", "id": "2022.insights-1.10", "year": 2022, "rel_sent": "Our experiments show that GPT-2 and GPT - Neo can return valid answers , but that their capabilities to identify coreferent mentions are limited and prompt - sensitive , leading to inconsistent results .", "forward": false, "src_ids": "2022.insights-1.10_4489"}
+{"input": "gpt-2 is used for OtherScientificTerm| context: coreference resolution - which is a crucial task for understanding discourse and language at large - has yet to witness widespread benefits from large language models ( llms ) . moreover , coreference resolution systems largely rely on supervised labels , which are highly expensive and difficult to annotate , thus making it ripe for prompt engineering .", "entity": "gpt-2", "output": "coreferent mentions", "neg_sample": ["gpt-2 is used for OtherScientificTerm", "coreference resolution - which is a crucial task for understanding discourse and language at large - has yet to witness widespread benefits from large language models ( llms ) .", "moreover , coreference resolution systems largely rely on supervised labels , which are highly expensive and difficult to annotate , thus making it ripe for prompt engineering ."], "relation": "used for", "id": "2022.insights-1.10", "year": 2022, "rel_sent": "Our experiments show that GPT-2 and GPT - Neo can return valid answers , but that their capabilities to identify coreferent mentions are limited and prompt - sensitive , leading to inconsistent results .", "forward": true, "src_ids": "2022.insights-1.10_4490"}
+{"input": "gpt - neo is used for OtherScientificTerm| context: coreference resolution - which is a crucial task for understanding discourse and language at large - has yet to witness widespread benefits from large language models ( llms ) . moreover , coreference resolution systems largely rely on supervised labels , which are highly expensive and difficult to annotate , thus making it ripe for prompt engineering .", "entity": "gpt - neo", "output": "coreferent mentions", "neg_sample": ["gpt - neo is used for OtherScientificTerm", "coreference resolution - which is a crucial task for understanding discourse and language at large - has yet to witness widespread benefits from large language models ( llms ) .", "moreover , coreference resolution systems largely rely on supervised labels , which are highly expensive and difficult to annotate , thus making it ripe for prompt engineering ."], "relation": "used for", "id": "2022.insights-1.10", "year": 2022, "rel_sent": "Our experiments show that GPT-2 and GPT - Neo can return valid answers , but that their capabilities to identify coreferent mentions are limited and prompt - sensitive , leading to inconsistent results .", "forward": true, "src_ids": "2022.insights-1.10_4491"}
+{"input": "chit - chat and task - oriented dialogues is done by using Method| context: with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems . however , these two systems are often tackled separately in current methods . to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks .", "entity": "chit - chat and task - oriented dialogues", "output": "unified dialogue system", "neg_sample": ["chit - chat and task - oriented dialogues is done by using Method", "with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems .", "however , these two systems are often tackled separately in current methods .", "to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks ."], "relation": "used for", "id": "2022.dialdoc-1.2", "year": 2022, "rel_sent": "UniDS : A Unified Dialogue System for Chit - Chat and Task - oriented Dialogues.", "forward": false, "src_ids": "2022.dialdoc-1.2_4492"}
+{"input": "two - stage training method is used for Method| context: with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems . however , these two systems are often tackled separately in current methods . to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks .", "entity": "two - stage training method", "output": "unified dialogue system", "neg_sample": ["two - stage training method is used for Method", "with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems .", "however , these two systems are often tackled separately in current methods .", "to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks ."], "relation": "used for", "id": "2022.dialdoc-1.2", "year": 2022, "rel_sent": "Besides , we propose a two - stage training method to train UniDS based on the unified dialogue data schema .", "forward": true, "src_ids": "2022.dialdoc-1.2_4493"}
+{"input": "unified dialogue system is used for Task| context: with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems . however , these two systems are often tackled separately in current methods . to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks .", "entity": "unified dialogue system", "output": "chit - chat and task - oriented dialogues", "neg_sample": ["unified dialogue system is used for Task", "with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems .", "however , these two systems are often tackled separately in current methods .", "to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks ."], "relation": "used for", "id": "2022.dialdoc-1.2", "year": 2022, "rel_sent": "UniDS : A Unified Dialogue System for Chit - Chat and Task - oriented Dialogues.", "forward": true, "src_ids": "2022.dialdoc-1.2_4494"}
+{"input": "unified dialogue data schema is used for Task| context: with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems . however , these two systems are often tackled separately in current methods . to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks .", "entity": "unified dialogue data schema", "output": "chit - chat and task - oriented dialogues", "neg_sample": ["unified dialogue data schema is used for Task", "with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems .", "however , these two systems are often tackled separately in current methods .", "to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks ."], "relation": "used for", "id": "2022.dialdoc-1.2", "year": 2022, "rel_sent": "In particular , we design a unified dialogue data schema , compatible for both chit - chat and task - oriented dialogues .", "forward": true, "src_ids": "2022.dialdoc-1.2_4495"}
+{"input": "chit - chat and task - oriented dialogues is done by using Method| context: with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems . however , these two systems are often tackled separately in current methods . to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks .", "entity": "chit - chat and task - oriented dialogues", "output": "unified dialogue data schema", "neg_sample": ["chit - chat and task - oriented dialogues is done by using Method", "with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems .", "however , these two systems are often tackled separately in current methods .", "to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks ."], "relation": "used for", "id": "2022.dialdoc-1.2", "year": 2022, "rel_sent": "In particular , we design a unified dialogue data schema , compatible for both chit - chat and task - oriented dialogues .", "forward": false, "src_ids": "2022.dialdoc-1.2_4496"}
+{"input": "unified dialogue system is done by using Method| context: with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems . however , these two systems are often tackled separately in current methods . to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks .", "entity": "unified dialogue system", "output": "two - stage training method", "neg_sample": ["unified dialogue system is done by using Method", "with the advances in deep learning , tremendous progress has been made with chit - chat dialogue systems and task - oriented dialogue systems .", "however , these two systems are often tackled separately in current methods .", "to achieve more natural interaction with humans , dialogue systems need to be capable of both chatting and accomplishing tasks ."], "relation": "used for", "id": "2022.dialdoc-1.2", "year": 2022, "rel_sent": "Besides , we propose a two - stage training method to train UniDS based on the unified dialogue data schema .", "forward": false, "src_ids": "2022.dialdoc-1.2_4497"}
+{"input": "language model compression is done by using Method| context: transferring the knowledge to a small model through distillation has raised great interest in recent years . prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g. , token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge . besides , these methods form the knowledge as individual representations or their simple dependencies , neglecting abundant structural relations among intermediate representations .", "entity": "language model compression", "output": "multi - granularity structural knowledge distillation", "neg_sample": ["language model compression is done by using Method", "transferring the knowledge to a small model through distillation has raised great interest in recent years .", "prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g.", ", token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge .", "besides , these methods form the knowledge as individual representations or their simple dependencies , neglecting abundant structural relations among intermediate representations ."], "relation": "used for", "id": "2022.acl-long.71", "year": 2022, "rel_sent": "Multi - Granularity Structural Knowledge Distillation for Language Model Compression.", "forward": false, "src_ids": "2022.acl-long.71_4498"}
+{"input": "multi - granularity structural knowledge distillation is used for Task| context: transferring the knowledge to a small model through distillation has raised great interest in recent years . prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g. , token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge . besides , these methods form the knowledge as individual representations or their simple dependencies , neglecting abundant structural relations among intermediate representations .", "entity": "multi - granularity structural knowledge distillation", "output": "language model compression", "neg_sample": ["multi - granularity structural knowledge distillation is used for Task", "transferring the knowledge to a small model through distillation has raised great interest in recent years .", "prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g.", ", token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge .", "besides , these methods form the knowledge as individual representations or their simple dependencies , neglecting abundant structural relations among intermediate representations ."], "relation": "used for", "id": "2022.acl-long.71", "year": 2022, "rel_sent": "Multi - Granularity Structural Knowledge Distillation for Language Model Compression.", "forward": true, "src_ids": "2022.acl-long.71_4499"}
+{"input": "intermediate representations is done by using Method| context: transferring the knowledge to a small model through distillation has raised great interest in recent years . prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g. , token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge . besides , these methods form the knowledge as individual representations or their simple dependencies , neglecting abundant structural relations among intermediate representations .", "entity": "intermediate representations", "output": "knowledge distillation framework", "neg_sample": ["intermediate representations is done by using Method", "transferring the knowledge to a small model through distillation has raised great interest in recent years .", "prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g.", ", token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge .", "besides , these methods form the knowledge as individual representations or their simple dependencies , neglecting abundant structural relations among intermediate representations ."], "relation": "used for", "id": "2022.acl-long.71", "year": 2022, "rel_sent": "To overcome the problems , we present a novel knowledge distillation framework that gathers intermediate representations from multiple semantic granularities ( e.g. , tokens , spans and samples ) and forms the knowledge as more sophisticated structural relations specified as the pair - wise interactions and the triplet - wise geometric angles based on multi - granularity representations .", "forward": false, "src_ids": "2022.acl-long.71_4500"}
+{"input": "knowledge distillation framework is used for Method| context: transferring the knowledge to a small model through distillation has raised great interest in recent years . prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g. , token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge .", "entity": "knowledge distillation framework", "output": "intermediate representations", "neg_sample": ["knowledge distillation framework is used for Method", "transferring the knowledge to a small model through distillation has raised great interest in recent years .", "prevailing methods transfer the knowledge derived from mono - granularity language units ( e.g.", ", token - level or sample - level ) , which is not enough to represent the rich semantics of a text and may lose some vital knowledge ."], "relation": "used for", "id": "2022.acl-long.71", "year": 2022, "rel_sent": "To overcome the problems , we present a novel knowledge distillation framework that gathers intermediate representations from multiple semantic granularities ( e.g. , tokens , spans and samples ) and forms the knowledge as more sophisticated structural relations specified as the pair - wise interactions and the triplet - wise geometric angles based on multi - granularity representations .", "forward": true, "src_ids": "2022.acl-long.71_4501"}
+{"input": "human and machine - generated text is done by using OtherScientificTerm| context: when generating natural language from neural probabilistic models , high probability does not always coincide with high quality . rather , mode - seeking decoding methods can lead to incredibly unnatural language , while stochastic methods produce text perceived as much more human - like .", "entity": "human and machine - generated text", "output": "quality ratings", "neg_sample": ["human and machine - generated text is done by using OtherScientificTerm", "when generating natural language from neural probabilistic models , high probability does not always coincide with high quality .", "rather , mode - seeking decoding methods can lead to incredibly unnatural language , while stochastic methods produce text perceived as much more human - like ."], "relation": "used for", "id": "2022.acl-short.5", "year": 2022, "rel_sent": "We provide preliminary empirical evidence for this hypothesis using quality ratings for both human and machine - generated text , covering multiple tasks and common decoding schemes .", "forward": false, "src_ids": "2022.acl-short.5_4502"}
+{"input": "quality ratings is used for Material| context: when generating natural language from neural probabilistic models , high probability does not always coincide with high quality . rather , mode - seeking decoding methods can lead to incredibly unnatural language , while stochastic methods produce text perceived as much more human - like .", "entity": "quality ratings", "output": "human and machine - generated text", "neg_sample": ["quality ratings is used for Material", "when generating natural language from neural probabilistic models , high probability does not always coincide with high quality .", "rather , mode - seeking decoding methods can lead to incredibly unnatural language , while stochastic methods produce text perceived as much more human - like ."], "relation": "used for", "id": "2022.acl-short.5", "year": 2022, "rel_sent": "We provide preliminary empirical evidence for this hypothesis using quality ratings for both human and machine - generated text , covering multiple tasks and common decoding schemes .", "forward": true, "src_ids": "2022.acl-short.5_4503"}
+{"input": "prediction consistency regularizer is used for Method| context: model ensemble is a popular approach to produce a low - variance and well - generalized model . however , it induces large memory and inference costs , which is often not affordable for real - world deployment . existing work has resorted to sharing weights among models . however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish .", "entity": "prediction consistency regularizer", "output": "perturbed models", "neg_sample": ["prediction consistency regularizer is used for Method", "model ensemble is a popular approach to produce a low - variance and well - generalized model .", "however , it induces large memory and inference costs , which is often not affordable for real - world deployment .", "existing work has resorted to sharing weights among models .", "however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish ."], "relation": "used for", "id": "2022.acl-long.495", "year": 2022, "rel_sent": "Meanwhile , we apply a prediction consistency regularizer across the perturbed models to control the variance due to the model diversity .", "forward": true, "src_ids": "2022.acl-long.495_4504"}
+{"input": "hidden representations is done by using OtherScientificTerm| context: model ensemble is a popular approach to produce a low - variance and well - generalized model . however , it induces large memory and inference costs , which is often not affordable for real - world deployment . existing work has resorted to sharing weights among models . however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish .", "entity": "hidden representations", "output": "perturbations", "neg_sample": ["hidden representations is done by using OtherScientificTerm", "model ensemble is a popular approach to produce a low - variance and well - generalized model .", "however , it induces large memory and inference costs , which is often not affordable for real - world deployment .", "existing work has resorted to sharing weights among models .", "however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish ."], "relation": "used for", "id": "2022.acl-long.495", "year": 2022, "rel_sent": "Specifically , we share the weights of bottom layers across all models and apply different perturbations to the hidden representations for different models , which can effectively promote the model diversity .", "forward": false, "src_ids": "2022.acl-long.495_4505"}
+{"input": "perturbations is used for Method| context: model ensemble is a popular approach to produce a low - variance and well - generalized model . however , it induces large memory and inference costs , which is often not affordable for real - world deployment . existing work has resorted to sharing weights among models . however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish .", "entity": "perturbations", "output": "hidden representations", "neg_sample": ["perturbations is used for Method", "model ensemble is a popular approach to produce a low - variance and well - generalized model .", "however , it induces large memory and inference costs , which is often not affordable for real - world deployment .", "existing work has resorted to sharing weights among models .", "however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish ."], "relation": "used for", "id": "2022.acl-long.495", "year": 2022, "rel_sent": "Specifically , we share the weights of bottom layers across all models and apply different perturbations to the hidden representations for different models , which can effectively promote the model diversity .", "forward": true, "src_ids": "2022.acl-long.495_4506"}
+{"input": "perturbed models is done by using Method| context: model ensemble is a popular approach to produce a low - variance and well - generalized model . however , it induces large memory and inference costs , which is often not affordable for real - world deployment . existing work has resorted to sharing weights among models . however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish .", "entity": "perturbed models", "output": "prediction consistency regularizer", "neg_sample": ["perturbed models is done by using Method", "model ensemble is a popular approach to produce a low - variance and well - generalized model .", "however , it induces large memory and inference costs , which is often not affordable for real - world deployment .", "existing work has resorted to sharing weights among models .", "however , when increasing the proportion of the shared weights , the resulting models tend to be similar , and the benefits of using model ensemble diminish ."], "relation": "used for", "id": "2022.acl-long.495", "year": 2022, "rel_sent": "Meanwhile , we apply a prediction consistency regularizer across the perturbed models to control the variance due to the model diversity .", "forward": false, "src_ids": "2022.acl-long.495_4507"}
+{"input": "simultaneous machine translation is done by using Method| context: simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating . the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control .", "entity": "simultaneous machine translation", "output": "gaussian multi - head attention ( gma )", "neg_sample": ["simultaneous machine translation is done by using Method", "simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating .", "the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control ."], "relation": "used for", "id": "2022.findings-acl.238", "year": 2022, "rel_sent": "Gaussian Multi - head Attention for Simultaneous Machine Translation.", "forward": false, "src_ids": "2022.findings-acl.238_4508"}
+{"input": "simt policy is done by using Method| context: simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating . the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control .", "entity": "simt policy", "output": "gaussian multi - head attention ( gma )", "neg_sample": ["simt policy is done by using Method", "simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating .", "the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control ."], "relation": "used for", "id": "2022.findings-acl.238", "year": 2022, "rel_sent": "In this paper , we propose Gaussian Multi - head Attention ( GMA ) to develop a new SiMT policy by modeling alignment and translation in a unified manner .", "forward": false, "src_ids": "2022.findings-acl.238_4509"}
+{"input": "gaussian multi - head attention ( gma ) is used for Task| context: the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control .", "entity": "gaussian multi - head attention ( gma )", "output": "simultaneous machine translation", "neg_sample": ["gaussian multi - head attention ( gma ) is used for Task", "the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control ."], "relation": "used for", "id": "2022.findings-acl.238", "year": 2022, "rel_sent": "Gaussian Multi - head Attention for Simultaneous Machine Translation.", "forward": true, "src_ids": "2022.findings-acl.238_4510"}
+{"input": "gaussian multi - head attention ( gma ) is used for Method| context: simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating . the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control .", "entity": "gaussian multi - head attention ( gma )", "output": "simt policy", "neg_sample": ["gaussian multi - head attention ( gma ) is used for Method", "simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating .", "the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control ."], "relation": "used for", "id": "2022.findings-acl.238", "year": 2022, "rel_sent": "In this paper , we propose Gaussian Multi - head Attention ( GMA ) to develop a new SiMT policy by modeling alignment and translation in a unified manner .", "forward": true, "src_ids": "2022.findings-acl.238_4511"}
+{"input": "alignment - related prior is done by using Method| context: simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating . the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control .", "entity": "alignment - related prior", "output": "gaussian distribution", "neg_sample": ["alignment - related prior is done by using Method", "simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating .", "the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control ."], "relation": "used for", "id": "2022.findings-acl.238", "year": 2022, "rel_sent": "To integrate the learning of alignment into the translation model , a Gaussian distribution centered on predicted aligned position is introduced as an alignment - related prior , which cooperates with translation - related soft attention to determine the final attention .", "forward": false, "src_ids": "2022.findings-acl.238_4512"}
+{"input": "gaussian distribution is used for OtherScientificTerm| context: simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating . the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control .", "entity": "gaussian distribution", "output": "alignment - related prior", "neg_sample": ["gaussian distribution is used for OtherScientificTerm", "simultaneous machine translation ( simt ) outputs translation while receiving the streaming source inputs , and hence needs a policy to determine where to start translating .", "the alignment between target and source words often implies the most informative source word for each target word , and hence provides the unified control over translation quality and latency , but unfortunately the existing simt methods do not explicitly model the alignment to perform the control ."], "relation": "used for", "id": "2022.findings-acl.238", "year": 2022, "rel_sent": "To integrate the learning of alignment into the translation model , a Gaussian distribution centered on predicted aligned position is introduced as an alignment - related prior , which cooperates with translation - related soft attention to determine the final attention .", "forward": true, "src_ids": "2022.findings-acl.238_4513"}
+{"input": "leakage is done by using Metric| context: recent work by sogaard ( 2020 ) showed that , treebank size aside , overlap between training and test graphs ( termed leakage ) explains more of the observed variation in dependency parsing performance than other explanations .", "entity": "leakage", "output": "fine - grained measure", "neg_sample": ["leakage is done by using Metric", "recent work by sogaard ( 2020 ) showed that , treebank size aside , overlap between training and test graphs ( termed leakage ) explains more of the observed variation in dependency parsing performance than other explanations ."], "relation": "used for", "id": "2022.findings-acl.230", "year": 2022, "rel_sent": "We then propose a more fine - grained measure of such leakage which , unlike the original measure , not only explains but also correlates with observed performance variation .", "forward": false, "src_ids": "2022.findings-acl.230_4514"}
+{"input": "local entailment relations is done by using Method| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "local entailment relations", "output": "egt2", "neg_sample": ["local entailment relations is done by using Method", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "EGT2 learns the local entailment relations by recognizing the textual entailment between template sentences formed by typed CCG - parsed predicates .", "forward": false, "src_ids": "2022.acl-long.406_4515"}
+{"input": "sparsity is done by using Method| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "sparsity", "output": "egt2", "neg_sample": ["sparsity is done by using Method", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Experiments on benchmark datasets show that EGT2 can well model the transitivity in entailment graph to alleviate the sparsity , and leads to signifcant improvement over current state - of - the - art methods .", "forward": false, "src_ids": "2022.acl-long.406_4516"}
+{"input": "transitivity is done by using Method| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "transitivity", "output": "egt2", "neg_sample": ["transitivity is done by using Method", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Experiments on benchmark datasets show that EGT2 can well model the transitivity in entailment graph to alleviate the sparsity , and leads to signifcant improvement over current state - of - the - art methods .", "forward": false, "src_ids": "2022.acl-long.406_4517"}
+{"input": "local graph is used for Method| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "local graph", "output": "egt2", "neg_sample": ["local graph is used for Method", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Based on the generated local graph , EGT2 then uses three novel soft transitivity constraints to consider the logical transitivity in entailment structures .", "forward": true, "src_ids": "2022.acl-long.406_4518"}
+{"input": "egt2 is used for OtherScientificTerm| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "egt2", "output": "local entailment relations", "neg_sample": ["egt2 is used for OtherScientificTerm", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "EGT2 learns the local entailment relations by recognizing the textual entailment between template sentences formed by typed CCG - parsed predicates .", "forward": true, "src_ids": "2022.acl-long.406_4519"}
+{"input": "egt2 is done by using OtherScientificTerm| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "egt2", "output": "local graph", "neg_sample": ["egt2 is done by using OtherScientificTerm", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Based on the generated local graph , EGT2 then uses three novel soft transitivity constraints to consider the logical transitivity in entailment structures .", "forward": false, "src_ids": "2022.acl-long.406_4520"}
+{"input": "logical transitivity is done by using OtherScientificTerm| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "logical transitivity", "output": "soft transitivity constraints", "neg_sample": ["logical transitivity is done by using OtherScientificTerm", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Based on the generated local graph , EGT2 then uses three novel soft transitivity constraints to consider the logical transitivity in entailment structures .", "forward": false, "src_ids": "2022.acl-long.406_4521"}
+{"input": "soft transitivity constraints is used for OtherScientificTerm| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "soft transitivity constraints", "output": "logical transitivity", "neg_sample": ["soft transitivity constraints is used for OtherScientificTerm", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Based on the generated local graph , EGT2 then uses three novel soft transitivity constraints to consider the logical transitivity in entailment structures .", "forward": true, "src_ids": "2022.acl-long.406_4522"}
+{"input": "egt2 is used for OtherScientificTerm| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes . the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity .", "entity": "egt2", "output": "transitivity", "neg_sample": ["egt2 is used for OtherScientificTerm", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "the construction of entailment graphs usually suffers from severe sparsity and unreliability of distributional similarity ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Experiments on benchmark datasets show that EGT2 can well model the transitivity in entailment graph to alleviate the sparsity , and leads to signifcant improvement over current state - of - the - art methods .", "forward": true, "src_ids": "2022.acl-long.406_4523"}
+{"input": "egt2 is used for OtherScientificTerm| context: typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes .", "entity": "egt2", "output": "sparsity", "neg_sample": ["egt2 is used for OtherScientificTerm", "typed entailment graphs try to learn the entailment relations between predicates from text and model them as edges between predicate nodes ."], "relation": "used for", "id": "2022.acl-long.406", "year": 2022, "rel_sent": "Experiments on benchmark datasets show that EGT2 can well model the transitivity in entailment graph to alleviate the sparsity , and leads to signifcant improvement over current state - of - the - art methods .", "forward": true, "src_ids": "2022.acl-long.406_4524"}
+{"input": "coherence is done by using Method| context: to understand a story with multiple events , it is important to capture the proper relations across these events . however , existing event relation extraction ( ere ) framework regards it as a multi - class classification task and do not guarantee any coherence between different relation types , such as anti - symmetry . if a phone line ' died ' after ' storm ' , then it is obvious that the ' storm ' happened before the ' died ' . current framework of event relation extraction do not guarantee this coherence and thus enforces it via constraint loss function ( wang et al . , 2020 ) .", "entity": "coherence", "output": "ere model", "neg_sample": ["coherence is done by using Method", "to understand a story with multiple events , it is important to capture the proper relations across these events .", "however , existing event relation extraction ( ere ) framework regards it as a multi - class classification task and do not guarantee any coherence between different relation types , such as anti - symmetry .", "if a phone line ' died ' after ' storm ' , then it is obvious that the ' storm ' happened before the ' died ' .", "current framework of event relation extraction do not guarantee this coherence and thus enforces it via constraint loss function ( wang et al .", ", 2020 ) ."], "relation": "used for", "id": "2022.acl-short.26", "year": 2022, "rel_sent": "In this work , we propose to modify the underlying ERE model to guarantee coherence by representing each event as a box representation ( BERE ) without applying explicit constraints .", "forward": false, "src_ids": "2022.acl-short.26_4525"}
+{"input": "ere model is used for OtherScientificTerm| context: to understand a story with multiple events , it is important to capture the proper relations across these events . if a phone line ' died ' after ' storm ' , then it is obvious that the ' storm ' happened before the ' died ' .", "entity": "ere model", "output": "coherence", "neg_sample": ["ere model is used for OtherScientificTerm", "to understand a story with multiple events , it is important to capture the proper relations across these events .", "if a phone line ' died ' after ' storm ' , then it is obvious that the ' storm ' happened before the ' died ' ."], "relation": "used for", "id": "2022.acl-short.26", "year": 2022, "rel_sent": "In this work , we propose to modify the underlying ERE model to guarantee coherence by representing each event as a box representation ( BERE ) without applying explicit constraints .", "forward": true, "src_ids": "2022.acl-short.26_4526"}
+{"input": "features is used for Task| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "entity": "features", "output": "response generation", "neg_sample": ["features is used for Task", "personality traits influence human actions and thoughts , which is manifested in day to day conversations ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "Since most of the existing large scale open domain chat corpora do not include Big-5 personality traits and discourse intent , we employ automatic annotation schemes to enrich the corpora with noisy estimates of personality and intent annotations , and further assess the impact of using such features as control codes for response generation using automatic evaluation metrics , ablation studies and human judgement .", "forward": true, "src_ids": "2022.nlp4convai-1.16_4527"}
+{"input": "control codes is used for Task| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "entity": "control codes", "output": "response generation", "neg_sample": ["control codes is used for Task", "personality traits influence human actions and thoughts , which is manifested in day to day conversations ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "Since most of the existing large scale open domain chat corpora do not include Big-5 personality traits and discourse intent , we employ automatic annotation schemes to enrich the corpora with noisy estimates of personality and intent annotations , and further assess the impact of using such features as control codes for response generation using automatic evaluation metrics , ablation studies and human judgement .", "forward": true, "src_ids": "2022.nlp4convai-1.16_4528"}
+{"input": "neural response generators is done by using Method| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations . although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses .", "entity": "neural response generators", "output": "end - to - end mechanisms", "neg_sample": ["neural response generators is done by using Method", "personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "With the motivation of enabling stylistically configurable response generators , in this paper we experiment with end - to - end mechanisms to ground neural response generators based on both ( i ) interlocutor Big-5 personality traits , and ( ii ) discourse intent as stylistic control codes .", "forward": false, "src_ids": "2022.nlp4convai-1.16_4529"}
+{"input": "end - to - end mechanisms is used for Method| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations . although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses .", "entity": "end - to - end mechanisms", "output": "neural response generators", "neg_sample": ["end - to - end mechanisms is used for Method", "personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "With the motivation of enabling stylistically configurable response generators , in this paper we experiment with end - to - end mechanisms to ground neural response generators based on both ( i ) interlocutor Big-5 personality traits , and ( ii ) discourse intent as stylistic control codes .", "forward": true, "src_ids": "2022.nlp4convai-1.16_4530"}
+{"input": "response generation is done by using OtherScientificTerm| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations . although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses .", "entity": "response generation", "output": "features", "neg_sample": ["response generation is done by using OtherScientificTerm", "personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "Since most of the existing large scale open domain chat corpora do not include Big-5 personality traits and discourse intent , we employ automatic annotation schemes to enrich the corpora with noisy estimates of personality and intent annotations , and further assess the impact of using such features as control codes for response generation using automatic evaluation metrics , ablation studies and human judgement .", "forward": false, "src_ids": "2022.nlp4convai-1.16_4531"}
+{"input": "control codes is done by using OtherScientificTerm| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations . although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses .", "entity": "control codes", "output": "features", "neg_sample": ["control codes is done by using OtherScientificTerm", "personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "Since most of the existing large scale open domain chat corpora do not include Big-5 personality traits and discourse intent , we employ automatic annotation schemes to enrich the corpora with noisy estimates of personality and intent annotations , and further assess the impact of using such features as control codes for response generation using automatic evaluation metrics , ablation studies and human judgement .", "forward": false, "src_ids": "2022.nlp4convai-1.16_4532"}
+{"input": "features is used for OtherScientificTerm| context: personality traits influence human actions and thoughts , which is manifested in day to day conversations . although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses .", "entity": "features", "output": "control codes", "neg_sample": ["features is used for OtherScientificTerm", "personality traits influence human actions and thoughts , which is manifested in day to day conversations .", "although glimpses of personality traits are observable in existing open domain conversation corpora , leveraging generic language modelling for response generation overlooks the interlocutor idiosyncrasies , resulting in non - customizable personality agnostic responses ."], "relation": "used for", "id": "2022.nlp4convai-1.16", "year": 2022, "rel_sent": "Since most of the existing large scale open domain chat corpora do not include Big-5 personality traits and discourse intent , we employ automatic annotation schemes to enrich the corpora with noisy estimates of personality and intent annotations , and further assess the impact of using such features as control codes for response generation using automatic evaluation metrics , ablation studies and human judgement .", "forward": true, "src_ids": "2022.nlp4convai-1.16_4533"}
+{"input": "natural language understanding is done by using Method| context: pretrained language models ( plms ) have achieved superhuman performance on many benchmarks , creating a need for harder tasks .", "entity": "natural language understanding", "output": "coda21", "neg_sample": ["natural language understanding is done by using Method", "pretrained language models ( plms ) have achieved superhuman performance on many benchmarks , creating a need for harder tasks ."], "relation": "used for", "id": "2022.acl-short.92", "year": 2022, "rel_sent": "We find that there is a large gap between human and PLM performance , suggesting that CoDA21 measures an aspect of NLU that is not sufficiently covered in existing benchmarks .", "forward": false, "src_ids": "2022.acl-short.92_4534"}
+{"input": "coda21 is used for Task| context: pretrained language models ( plms ) have achieved superhuman performance on many benchmarks , creating a need for harder tasks .", "entity": "coda21", "output": "natural language understanding", "neg_sample": ["coda21 is used for Task", "pretrained language models ( plms ) have achieved superhuman performance on many benchmarks , creating a need for harder tasks ."], "relation": "used for", "id": "2022.acl-short.92", "year": 2022, "rel_sent": "We find that there is a large gap between human and PLM performance , suggesting that CoDA21 measures an aspect of NLU that is not sufficiently covered in existing benchmarks .", "forward": true, "src_ids": "2022.acl-short.92_4535"}
+{"input": "unification task is done by using Method| context: discourse information is difficult to represent and annotate . among the major frameworks for annotating discourse information , rst , pdtb and sdrt are widely discussed and used , each having its own theoretical foundation and focus . corpora annotated under different frameworks vary considerably . although the issue of framework unification has been a topic of discussion for a long time , there is currently no comprehensive approach which considers unifying both discourse structure and discourse relations and evaluates the unified framework intrinsically and extrinsically .", "entity": "unification task", "output": "automatic means", "neg_sample": ["unification task is done by using Method", "discourse information is difficult to represent and annotate .", "among the major frameworks for annotating discourse information , rst , pdtb and sdrt are widely discussed and used , each having its own theoretical foundation and focus .", "corpora annotated under different frameworks vary considerably .", "although the issue of framework unification has been a topic of discussion for a long time , there is currently no comprehensive approach which considers unifying both discourse structure and discourse relations and evaluates the unified framework intrinsically and extrinsically ."], "relation": "used for", "id": "2022.acl-srw.12", "year": 2022, "rel_sent": "We plan to use automatic means for the unification task and evaluate the result with structural complexity and downstream tasks .", "forward": false, "src_ids": "2022.acl-srw.12_4536"}
+{"input": "automatic means is used for Task| context: discourse information is difficult to represent and annotate . among the major frameworks for annotating discourse information , rst , pdtb and sdrt are widely discussed and used , each having its own theoretical foundation and focus . corpora annotated under different frameworks vary considerably . although the issue of framework unification has been a topic of discussion for a long time , there is currently no comprehensive approach which considers unifying both discourse structure and discourse relations and evaluates the unified framework intrinsically and extrinsically .", "entity": "automatic means", "output": "unification task", "neg_sample": ["automatic means is used for Task", "discourse information is difficult to represent and annotate .", "among the major frameworks for annotating discourse information , rst , pdtb and sdrt are widely discussed and used , each having its own theoretical foundation and focus .", "corpora annotated under different frameworks vary considerably .", "although the issue of framework unification has been a topic of discussion for a long time , there is currently no comprehensive approach which considers unifying both discourse structure and discourse relations and evaluates the unified framework intrinsically and extrinsically ."], "relation": "used for", "id": "2022.acl-srw.12", "year": 2022, "rel_sent": "We plan to use automatic means for the unification task and evaluate the result with structural complexity and downstream tasks .", "forward": true, "src_ids": "2022.acl-srw.12_4537"}
+{"input": "east asian languages is done by using Method| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "entity": "east asian languages", "output": "separation inference", "neg_sample": ["east asian languages is done by using Method", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Word Segmentation by Separation Inference for East Asian Languages.", "forward": false, "src_ids": "2022.findings-acl.309_4538"}
+{"input": "separation inference is used for Material| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words . in such a way , cws is reformed as a separation inference task in every adjacent character pair .", "entity": "separation inference", "output": "east asian languages", "neg_sample": ["separation inference is used for Material", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "in such a way , cws is reformed as a separation inference task in every adjacent character pair ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Word Segmentation by Separation Inference for East Asian Languages.", "forward": true, "src_ids": "2022.findings-acl.309_4539"}
+{"input": "separation state is done by using OtherScientificTerm| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words . in such a way , cws is reformed as a separation inference task in every adjacent character pair .", "entity": "separation state", "output": "bigram", "neg_sample": ["separation state is done by using OtherScientificTerm", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "in such a way , cws is reformed as a separation inference task in every adjacent character pair ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Therefore , bigram is specially tailored for ' C - NC ' to model the separation state of every two consecutive characters .", "forward": false, "src_ids": "2022.findings-acl.309_4540"}
+{"input": "c - nc ' is done by using OtherScientificTerm| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words . in such a way , cws is reformed as a separation inference task in every adjacent character pair .", "entity": "c - nc '", "output": "bigram", "neg_sample": ["c - nc ' is done by using OtherScientificTerm", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "in such a way , cws is reformed as a separation inference task in every adjacent character pair ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Therefore , bigram is specially tailored for ' C - NC ' to model the separation state of every two consecutive characters .", "forward": false, "src_ids": "2022.findings-acl.309_4541"}
+{"input": "bigram is used for OtherScientificTerm| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words . in such a way , cws is reformed as a separation inference task in every adjacent character pair .", "entity": "bigram", "output": "c - nc '", "neg_sample": ["bigram is used for OtherScientificTerm", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "in such a way , cws is reformed as a separation inference task in every adjacent character pair ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Therefore , bigram is specially tailored for ' C - NC ' to model the separation state of every two consecutive characters .", "forward": true, "src_ids": "2022.findings-acl.309_4542"}
+{"input": "machine learning and deep learning models is done by using Method| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words . in such a way , cws is reformed as a separation inference task in every adjacent character pair .", "entity": "machine learning and deep learning models", "output": "separation inference ( spin ) framework", "neg_sample": ["machine learning and deep learning models is done by using Method", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "in such a way , cws is reformed as a separation inference task in every adjacent character pair ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Our Separation Inference ( SpIn ) framework is evaluated on five public datasets , is demonstrated to work for machine learning and deep learning models , and outperforms state - of - the - art performance for CWS in all experiments .", "forward": false, "src_ids": "2022.findings-acl.309_4543"}
+{"input": "separation inference ( spin ) framework is used for Method| context: chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling . thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words . in such a way , cws is reformed as a separation inference task in every adjacent character pair .", "entity": "separation inference ( spin ) framework", "output": "machine learning and deep learning models", "neg_sample": ["separation inference ( spin ) framework is used for Method", "chinese word segmentation ( cws ) intends to divide a raw sentence into words through sequence labeling .", "thinking in reverse , cws can also be viewed as a process of grouping a sequence of characters into a sequence of words .", "in such a way , cws is reformed as a separation inference task in every adjacent character pair ."], "relation": "used for", "id": "2022.findings-acl.309", "year": 2022, "rel_sent": "Our Separation Inference ( SpIn ) framework is evaluated on five public datasets , is demonstrated to work for machine learning and deep learning models , and outperforms state - of - the - art performance for CWS in all experiments .", "forward": true, "src_ids": "2022.findings-acl.309_4544"}
+{"input": "dialog state tracking is done by using Method| context: a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle . however , continually training a model often leads to a well - known catastrophic forgetting issue .", "entity": "dialog state tracking", "output": "continual prompt tuning", "neg_sample": ["dialog state tracking is done by using Method", "a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle .", "however , continually training a model often leads to a well - known catastrophic forgetting issue ."], "relation": "used for", "id": "2022.acl-long.80", "year": 2022, "rel_sent": "Continual Prompt Tuning for Dialog State Tracking.", "forward": false, "src_ids": "2022.acl-long.80_4545"}
+{"input": "continual prompt tuning is used for Task| context: a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle . however , continually training a model often leads to a well - known catastrophic forgetting issue .", "entity": "continual prompt tuning", "output": "dialog state tracking", "neg_sample": ["continual prompt tuning is used for Task", "a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle .", "however , continually training a model often leads to a well - known catastrophic forgetting issue ."], "relation": "used for", "id": "2022.acl-long.80", "year": 2022, "rel_sent": "Continual Prompt Tuning for Dialog State Tracking.", "forward": true, "src_ids": "2022.acl-long.80_4546"}
+{"input": "continual learning is used for Task| context: a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle . however , continually training a model often leads to a well - known catastrophic forgetting issue .", "entity": "continual learning", "output": "dialog state tracking", "neg_sample": ["continual learning is used for Task", "a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle .", "however , continually training a model often leads to a well - known catastrophic forgetting issue ."], "relation": "used for", "id": "2022.acl-long.80", "year": 2022, "rel_sent": "Extensive experiments demonstrate the effectiveness and efficiency of our proposed method on continual learning for dialog state tracking , compared with state - of - the - art baselines .", "forward": true, "src_ids": "2022.acl-long.80_4547"}
+{"input": "dialog state tracking is done by using Method| context: a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle . however , continually training a model often leads to a well - known catastrophic forgetting issue .", "entity": "dialog state tracking", "output": "continual learning", "neg_sample": ["dialog state tracking is done by using Method", "a desirable dialog system should be able to continually learn new skills without forgetting old ones , and thereby adapt to new domains or tasks in its life cycle .", "however , continually training a model often leads to a well - known catastrophic forgetting issue ."], "relation": "used for", "id": "2022.acl-long.80", "year": 2022, "rel_sent": "Extensive experiments demonstrate the effectiveness and efficiency of our proposed method on continual learning for dialog state tracking , compared with state - of - the - art baselines .", "forward": false, "src_ids": "2022.acl-long.80_4548"}
+{"input": "probability of zero missampling rate is done by using OtherScientificTerm| context: negative sampling is highly effective in handling missing annotations for named entity recognition ( ner ) .", "entity": "probability of zero missampling rate", "output": "lower bound", "neg_sample": ["probability of zero missampling rate is done by using OtherScientificTerm", "negative sampling is highly effective in handling missing annotations for named entity recognition ( ner ) ."], "relation": "used for", "id": "2022.acl-long.497", "year": 2022, "rel_sent": "Based on the sparsity of named entities , we also theoretically derive a lower bound for the probability of zero missampling rate , which is only relevant to sentence length .", "forward": false, "src_ids": "2022.acl-long.497_4549"}
+{"input": "lower bound is used for Metric| context: negative sampling is highly effective in handling missing annotations for named entity recognition ( ner ) .", "entity": "lower bound", "output": "probability of zero missampling rate", "neg_sample": ["lower bound is used for Metric", "negative sampling is highly effective in handling missing annotations for named entity recognition ( ner ) ."], "relation": "used for", "id": "2022.acl-long.497", "year": 2022, "rel_sent": "Based on the sparsity of named entities , we also theoretically derive a lower bound for the probability of zero missampling rate , which is only relevant to sentence length .", "forward": true, "src_ids": "2022.acl-long.497_4550"}
+{"input": "negative sampling is done by using Method| context: negative sampling is highly effective in handling missing annotations for named entity recognition ( ner ) .", "entity": "negative sampling", "output": "adaptive and weighted sampling distribution", "neg_sample": ["negative sampling is done by using Method", "negative sampling is highly effective in handling missing annotations for named entity recognition ( ner ) ."], "relation": "used for", "id": "2022.acl-long.497", "year": 2022, "rel_sent": "The other contribution is an adaptive and weighted sampling distribution that further improves negative sampling via our former analysis .", "forward": false, "src_ids": "2022.acl-long.497_4551"}
+{"input": "zero - shot is done by using Method| context: we study the zero - shot setting for the aspect - based scientific document summarization task . summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience . however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain .", "entity": "zero - shot", "output": "self - supervised pre - training approach", "neg_sample": ["zero - shot is done by using Method", "we study the zero - shot setting for the aspect - based scientific document summarization task .", "summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience .", "however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain ."], "relation": "used for", "id": "2022.bionlp-1.5", "year": 2022, "rel_sent": "We propose a self - supervised pre - training approach to enhance the zero - shot performance .", "forward": false, "src_ids": "2022.bionlp-1.5_4552"}
+{"input": "self - supervised pre - training approach is used for OtherScientificTerm| context: summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience . however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain .", "entity": "self - supervised pre - training approach", "output": "zero - shot", "neg_sample": ["self - supervised pre - training approach is used for OtherScientificTerm", "summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience .", "however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain ."], "relation": "used for", "id": "2022.bionlp-1.5", "year": 2022, "rel_sent": "We propose a self - supervised pre - training approach to enhance the zero - shot performance .", "forward": true, "src_ids": "2022.bionlp-1.5_4553"}
+{"input": "biomedical aspect - based summarization dataset is done by using Material| context: we study the zero - shot setting for the aspect - based scientific document summarization task . summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience . however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain .", "entity": "biomedical aspect - based summarization dataset", "output": "pubmed structured abstracts", "neg_sample": ["biomedical aspect - based summarization dataset is done by using Material", "we study the zero - shot setting for the aspect - based scientific document summarization task .", "summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience .", "however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain ."], "relation": "used for", "id": "2022.bionlp-1.5", "year": 2022, "rel_sent": "We leverage the PubMed structured abstracts to create a biomedical aspect - based summarization dataset .", "forward": false, "src_ids": "2022.bionlp-1.5_4554"}
+{"input": "pubmed structured abstracts is used for Material| context: we study the zero - shot setting for the aspect - based scientific document summarization task . summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience . however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain .", "entity": "pubmed structured abstracts", "output": "biomedical aspect - based summarization dataset", "neg_sample": ["pubmed structured abstracts is used for Material", "we study the zero - shot setting for the aspect - based scientific document summarization task .", "summarizing scientific documents with respect to an aspect can remarkably improve document assistance systems and readers experience .", "however , existing large - scale datasets contain a limited variety of aspects , causing summarization models to over - fit to a small set of aspects and a specific domain ."], "relation": "used for", "id": "2022.bionlp-1.5", "year": 2022, "rel_sent": "We leverage the PubMed structured abstracts to create a biomedical aspect - based summarization dataset .", "forward": true, "src_ids": "2022.bionlp-1.5_4555"}
+{"input": "word embeddings is done by using Method| context: although contextualized embeddings generated from large - scale pre - trained models perform well in many tasks , traditional static embeddings ( e.g. , skip - gram , word2vec ) still play an important role in low - resource and lightweight settings due to their low computational cost , ease of deployment , and stability .", "entity": "word embeddings", "output": "context - to - vector with graph retrofitting", "neg_sample": ["word embeddings is done by using Method", "although contextualized embeddings generated from large - scale pre - trained models perform well in many tasks , traditional static embeddings ( e.g.", ", skip - gram , word2vec ) still play an important role in low - resource and lightweight settings due to their low computational cost , ease of deployment , and stability ."], "relation": "used for", "id": "2022.acl-long.561", "year": 2022, "rel_sent": "Using Context - to - Vector with Graph Retrofitting to Improve Word Embeddings.", "forward": false, "src_ids": "2022.acl-long.561_4556"}
+{"input": "context - to - vector with graph retrofitting is used for OtherScientificTerm| context: although contextualized embeddings generated from large - scale pre - trained models perform well in many tasks , traditional static embeddings ( e.g. , skip - gram , word2vec ) still play an important role in low - resource and lightweight settings due to their low computational cost , ease of deployment , and stability .", "entity": "context - to - vector with graph retrofitting", "output": "word embeddings", "neg_sample": ["context - to - vector with graph retrofitting is used for OtherScientificTerm", "although contextualized embeddings generated from large - scale pre - trained models perform well in many tasks , traditional static embeddings ( e.g.", ", skip - gram , word2vec ) still play an important role in low - resource and lightweight settings due to their low computational cost , ease of deployment , and stability ."], "relation": "used for", "id": "2022.acl-long.561", "year": 2022, "rel_sent": "Using Context - to - Vector with Graph Retrofitting to Improve Word Embeddings.", "forward": true, "src_ids": "2022.acl-long.561_4557"}
+{"input": "static embeddings is done by using Method| context: although contextualized embeddings generated from large - scale pre - trained models perform well in many tasks , traditional static embeddings ( e.g. , skip - gram , word2vec ) still play an important role in low - resource and lightweight settings due to their low computational cost , ease of deployment , and stability .", "entity": "static embeddings", "output": "post - processing retrofitting method", "neg_sample": ["static embeddings is done by using Method", "although contextualized embeddings generated from large - scale pre - trained models perform well in many tasks , traditional static embeddings ( e.g.", ", skip - gram , word2vec ) still play an important role in low - resource and lightweight settings due to their low computational cost , ease of deployment , and stability ."], "relation": "used for", "id": "2022.acl-long.561", "year": 2022, "rel_sent": "In this paper , we aim to improve word embeddings by 1 ) incorporating more contextual information from existing pre - trained models into the Skip - gram framework , which we call Context - to - Vec ; 2 ) proposing a post - processing retrofitting method for static embeddings independent of training by employing priori synonym knowledge and weighted vector distribution .", "forward": false, "src_ids": "2022.acl-long.561_4558"}
+{"input": "performance mining is done by using Method| context: reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic . they are easy to understand and increase empathy : this makes them powerful in argumentation . the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp .", "entity": "performance mining", "output": "linear regression", "neg_sample": ["performance mining is done by using Method", "reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic .", "they are easy to understand and increase empathy : this makes them powerful in argumentation .", "the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp ."], "relation": "used for", "id": "2022.acl-long.379", "year": 2022, "rel_sent": "Second , we employ linear regression for performance mining , identifying performance trends both for overall classification performance and individual classifier predictions .", "forward": false, "src_ids": "2022.acl-long.379_4559"}
+{"input": "performance trends is done by using Method| context: reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic . they are easy to understand and increase empathy : this makes them powerful in argumentation . the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp .", "entity": "performance trends", "output": "linear regression", "neg_sample": ["performance trends is done by using Method", "reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic .", "they are easy to understand and increase empathy : this makes them powerful in argumentation .", "the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp ."], "relation": "used for", "id": "2022.acl-long.379", "year": 2022, "rel_sent": "Second , we employ linear regression for performance mining , identifying performance trends both for overall classification performance and individual classifier predictions .", "forward": false, "src_ids": "2022.acl-long.379_4560"}
+{"input": "linear regression is used for Task| context: reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic . they are easy to understand and increase empathy : this makes them powerful in argumentation . the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp .", "entity": "linear regression", "output": "performance mining", "neg_sample": ["linear regression is used for Task", "reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic .", "they are easy to understand and increase empathy : this makes them powerful in argumentation .", "the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp ."], "relation": "used for", "id": "2022.acl-long.379", "year": 2022, "rel_sent": "Second , we employ linear regression for performance mining , identifying performance trends both for overall classification performance and individual classifier predictions .", "forward": true, "src_ids": "2022.acl-long.379_4561"}
+{"input": "overall classification performance is done by using OtherScientificTerm| context: reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic . they are easy to understand and increase empathy : this makes them powerful in argumentation . the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp .", "entity": "overall classification performance", "output": "performance trends", "neg_sample": ["overall classification performance is done by using OtherScientificTerm", "reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic .", "they are easy to understand and increase empathy : this makes them powerful in argumentation .", "the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp ."], "relation": "used for", "id": "2022.acl-long.379", "year": 2022, "rel_sent": "Second , we employ linear regression for performance mining , identifying performance trends both for overall classification performance and individual classifier predictions .", "forward": false, "src_ids": "2022.acl-long.379_4562"}
+{"input": "linear regression is used for OtherScientificTerm| context: reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic . they are easy to understand and increase empathy : this makes them powerful in argumentation . the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp .", "entity": "linear regression", "output": "performance trends", "neg_sample": ["linear regression is used for OtherScientificTerm", "reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic .", "they are easy to understand and increase empathy : this makes them powerful in argumentation .", "the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp ."], "relation": "used for", "id": "2022.acl-long.379", "year": 2022, "rel_sent": "Second , we employ linear regression for performance mining , identifying performance trends both for overall classification performance and individual classifier predictions .", "forward": true, "src_ids": "2022.acl-long.379_4563"}
+{"input": "performance trends is used for Metric| context: reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic . they are easy to understand and increase empathy : this makes them powerful in argumentation . the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp .", "entity": "performance trends", "output": "overall classification performance", "neg_sample": ["performance trends is used for Metric", "reports of personal experiences or stories can play a crucial role in argumentation , as they represent an immediate and ( often ) relatable way to back up one 's position with respect to a given topic .", "they are easy to understand and increase empathy : this makes them powerful in argumentation .", "the impact of personal reports and stories in argumentation has been studied in the social sciences , but it is still largely underexplored in nlp ."], "relation": "used for", "id": "2022.acl-long.379", "year": 2022, "rel_sent": "Second , we employ linear regression for performance mining , identifying performance trends both for overall classification performance and individual classifier predictions .", "forward": true, "src_ids": "2022.acl-long.379_4564"}
+{"input": "logical information is done by using Method| context: logical reasoning of text requires identifying critical logical structures in the text and performing inference over them . existing methods for logical reasoning mainly focus on contextual semantics of text while struggling to explicitly model the logical inference process .", "entity": "logical information", "output": "contrastive learning", "neg_sample": ["logical information is done by using Method", "logical reasoning of text requires identifying critical logical structures in the text and performing inference over them .", "existing methods for logical reasoning mainly focus on contextual semantics of text while struggling to explicitly model the logical inference process ."], "relation": "used for", "id": "2022.findings-acl.127", "year": 2022, "rel_sent": "The latter augments literally similar but logically different instances and incorporates contrastive learning to better capture logical information , especially logical negative and conditional relationships .", "forward": false, "src_ids": "2022.findings-acl.127_4565"}
+{"input": "contrastive learning is used for OtherScientificTerm| context: logical reasoning of text requires identifying critical logical structures in the text and performing inference over them . existing methods for logical reasoning mainly focus on contextual semantics of text while struggling to explicitly model the logical inference process .", "entity": "contrastive learning", "output": "logical information", "neg_sample": ["contrastive learning is used for OtherScientificTerm", "logical reasoning of text requires identifying critical logical structures in the text and performing inference over them .", "existing methods for logical reasoning mainly focus on contextual semantics of text while struggling to explicitly model the logical inference process ."], "relation": "used for", "id": "2022.findings-acl.127", "year": 2022, "rel_sent": "The latter augments literally similar but logically different instances and incorporates contrastive learning to better capture logical information , especially logical negative and conditional relationships .", "forward": true, "src_ids": "2022.findings-acl.127_4566"}
+{"input": "coreference resolution ( coref ) is done by using Method| context: we consider the task of document - level entity linking ( el ) , where it is important to make consistent decisions for entity mentions over the full document jointly .", "entity": "coreference resolution ( coref )", "output": "joint models", "neg_sample": ["coreference resolution ( coref ) is done by using Method", "we consider the task of document - level entity linking ( el ) , where it is important to make consistent decisions for entity mentions over the full document jointly ."], "relation": "used for", "id": "2022.acl-short.88", "year": 2022, "rel_sent": "Towards Consistent Document - level Entity Linking : Joint Models for Entity Linking and Coreference Resolution.", "forward": false, "src_ids": "2022.acl-short.88_4567"}
+{"input": "entity linking is done by using Method| context: we consider the task of document - level entity linking ( el ) , where it is important to make consistent decisions for entity mentions over the full document jointly .", "entity": "entity linking", "output": "joint models", "neg_sample": ["entity linking is done by using Method", "we consider the task of document - level entity linking ( el ) , where it is important to make consistent decisions for entity mentions over the full document jointly ."], "relation": "used for", "id": "2022.acl-short.88", "year": 2022, "rel_sent": "Towards Consistent Document - level Entity Linking : Joint Models for Entity Linking and Coreference Resolution.", "forward": false, "src_ids": "2022.acl-short.88_4568"}
+{"input": "joint models is used for Task| context: we consider the task of document - level entity linking ( el ) , where it is important to make consistent decisions for entity mentions over the full document jointly .", "entity": "joint models", "output": "coreference resolution ( coref )", "neg_sample": ["joint models is used for Task", "we consider the task of document - level entity linking ( el ) , where it is important to make consistent decisions for entity mentions over the full document jointly ."], "relation": "used for", "id": "2022.acl-short.88", "year": 2022, "rel_sent": "Towards Consistent Document - level Entity Linking : Joint Models for Entity Linking and Coreference Resolution.", "forward": true, "src_ids": "2022.acl-short.88_4569"}
+{"input": "data augmentation methods is done by using Task| context: before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary . smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation .", "entity": "data augmentation methods", "output": "text smoothing", "neg_sample": ["data augmentation methods is done by using Task", "before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary .", "smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation ."], "relation": "used for", "id": "2022.acl-short.97", "year": 2022, "rel_sent": "Text Smoothing : Enhance Various Data Augmentation Methods on Text Classification Tasks.", "forward": false, "src_ids": "2022.acl-short.97_4570"}
+{"input": "text classification tasks is done by using Method| context: before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary . smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation .", "entity": "text classification tasks", "output": "data augmentation methods", "neg_sample": ["text classification tasks is done by using Method", "before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary .", "smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation ."], "relation": "used for", "id": "2022.acl-short.97", "year": 2022, "rel_sent": "Text Smoothing : Enhance Various Data Augmentation Methods on Text Classification Tasks.", "forward": false, "src_ids": "2022.acl-short.97_4571"}
+{"input": "text smoothing is used for Method| context: before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary . smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation .", "entity": "text smoothing", "output": "data augmentation methods", "neg_sample": ["text smoothing is used for Method", "before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary .", "smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation ."], "relation": "used for", "id": "2022.acl-short.97", "year": 2022, "rel_sent": "Text Smoothing : Enhance Various Data Augmentation Methods on Text Classification Tasks.", "forward": true, "src_ids": "2022.acl-short.97_4572"}
+{"input": "data augmentation methods is used for Task| context: before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary . smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation .", "entity": "data augmentation methods", "output": "text classification tasks", "neg_sample": ["data augmentation methods is used for Task", "before entering the neural network , a token needs to be converted to its one - hot representation , which is a discrete distribution of the vocabulary .", "smoothed representation is the probability of candidate tokens obtained from the pre - trained masked language model , which can be seen as a more informative augmented substitution to the one - hot representation ."], "relation": "used for", "id": "2022.acl-short.97", "year": 2022, "rel_sent": "Text Smoothing : Enhance Various Data Augmentation Methods on Text Classification Tasks.", "forward": true, "src_ids": "2022.acl-short.97_4573"}
+{"input": "time - sensitive kg encoder is used for OtherScientificTerm| context: question answering over temporal knowledge graphs ( kgs ) efficiently uses facts contained in a temporal kg , which records entity relations and when they occur in time , to answer natural language questions ( e.g. , ' who was the president of the us before obama ? ' ) . these questions often involve three time - related challenges that previous work fail to adequately address : 1 ) questions often do not specify exact timestamps of interest ( e.g. , ' obama ' instead of 2000 ) ; 2 ) subtle lexical differences in time relations ( e.g. , ' before ' vs ' after ' ) ; 3 ) off - the - shelf temporal kg embeddings that previous work builds on ignore the temporal order of timestamps , which is crucial for answering temporal - order related questions .", "entity": "time - sensitive kg encoder", "output": "ordering information", "neg_sample": ["time - sensitive kg encoder is used for OtherScientificTerm", "question answering over temporal knowledge graphs ( kgs ) efficiently uses facts contained in a temporal kg , which records entity relations and when they occur in time , to answer natural language questions ( e.g.", ", ' who was the president of the us before obama ? '", ") .", "these questions often involve three time - related challenges that previous work fail to adequately address : 1 ) questions often do not specify exact timestamps of interest ( e.g.", ", ' obama ' instead of 2000 ) ; 2 ) subtle lexical differences in time relations ( e.g.", ", ' before ' vs ' after ' ) ; 3 ) off - the - shelf temporal kg embeddings that previous work builds on ignore the temporal order of timestamps , which is crucial for answering temporal - order related questions ."], "relation": "used for", "id": "2022.acl-long.552", "year": 2022, "rel_sent": "We also employ a time - sensitive KG encoder to inject ordering information into the temporal KG embeddings that TSQA is based on .", "forward": true, "src_ids": "2022.acl-long.552_4574"}
+{"input": "ordering information is done by using Method| context: question answering over temporal knowledge graphs ( kgs ) efficiently uses facts contained in a temporal kg , which records entity relations and when they occur in time , to answer natural language questions ( e.g. , ' who was the president of the us before obama ? ' ) . these questions often involve three time - related challenges that previous work fail to adequately address : 1 ) questions often do not specify exact timestamps of interest ( e.g. , ' obama ' instead of 2000 ) ; 2 ) subtle lexical differences in time relations ( e.g. , ' before ' vs ' after ' ) ; 3 ) off - the - shelf temporal kg embeddings that previous work builds on ignore the temporal order of timestamps , which is crucial for answering temporal - order related questions .", "entity": "ordering information", "output": "time - sensitive kg encoder", "neg_sample": ["ordering information is done by using Method", "question answering over temporal knowledge graphs ( kgs ) efficiently uses facts contained in a temporal kg , which records entity relations and when they occur in time , to answer natural language questions ( e.g.", ", ' who was the president of the us before obama ? '", ") .", "these questions often involve three time - related challenges that previous work fail to adequately address : 1 ) questions often do not specify exact timestamps of interest ( e.g.", ", ' obama ' instead of 2000 ) ; 2 ) subtle lexical differences in time relations ( e.g.", ", ' before ' vs ' after ' ) ; 3 ) off - the - shelf temporal kg embeddings that previous work builds on ignore the temporal order of timestamps , which is crucial for answering temporal - order related questions ."], "relation": "used for", "id": "2022.acl-long.552", "year": 2022, "rel_sent": "We also employ a time - sensitive KG encoder to inject ordering information into the temporal KG embeddings that TSQA is based on .", "forward": false, "src_ids": "2022.acl-long.552_4575"}
+{"input": "doctor recommendation is done by using Task| context: huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task .", "entity": "doctor recommendation", "output": "doctor recommendation", "neg_sample": ["doctor recommendation is done by using Task", "huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task ."], "relation": "used for", "id": "2022.acl-long.79", "year": 2022, "rel_sent": "To better help patients , this paper studies a novel task of doctor recommendation to enable automatic pairing of a patient to a doctor with relevant expertise .", "forward": false, "src_ids": "2022.acl-long.79_4576"}
+{"input": "doctor recommendation is used for Task| context: huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task .", "entity": "doctor recommendation", "output": "doctor recommendation", "neg_sample": ["doctor recommendation is used for Task", "huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task ."], "relation": "used for", "id": "2022.acl-long.79", "year": 2022, "rel_sent": "To better help patients , this paper studies a novel task of doctor recommendation to enable automatic pairing of a patient to a doctor with relevant expertise .", "forward": true, "src_ids": "2022.acl-long.79_4577"}
+{"input": "patient query is done by using OtherScientificTerm| context: huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task .", "entity": "patient query", "output": "doctor embeddings", "neg_sample": ["patient query is done by using OtherScientificTerm", "huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task ."], "relation": "used for", "id": "2022.acl-long.79", "year": 2022, "rel_sent": "The learned doctor embeddings are further employed to estimate their capabilities of handling a patient query with a multi - head attention mechanism .", "forward": false, "src_ids": "2022.acl-long.79_4578"}
+{"input": "doctor embeddings is used for OtherScientificTerm| context: huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task .", "entity": "doctor embeddings", "output": "patient query", "neg_sample": ["doctor embeddings is used for OtherScientificTerm", "huge volumes of patient queries are daily generated on online health forums , rendering manual doctor allocation a labor - intensive task ."], "relation": "used for", "id": "2022.acl-long.79", "year": 2022, "rel_sent": "The learned doctor embeddings are further employed to estimate their capabilities of handling a patient query with a multi - head attention mechanism .", "forward": true, "src_ids": "2022.acl-long.79_4579"}
+{"input": "morphological segments is done by using Method| context: large pre - trained language models ( plms ) have become ubiquitous in the development of language understanding technology and lie at the heart of many artificial intelligence advances . while advances reported for english using plms are unprecedented , reported advances using plms for hebrew are few and far between . the problem is twofold . first , sofar , hebrew resources for training large language models are not of the same magnitude as their english counterparts . second , most benchmarks available to evaluate progress in hebrew nlp require morphological boundaries which are not available in the output of standard plms . in this work we remedy both aspects .", "entity": "morphological segments", "output": "neural architecture", "neg_sample": ["morphological segments is done by using Method", "large pre - trained language models ( plms ) have become ubiquitous in the development of language understanding technology and lie at the heart of many artificial intelligence advances .", "while advances reported for english using plms are unprecedented , reported advances using plms for hebrew are few and far between .", "the problem is twofold .", "first , sofar , hebrew resources for training large language models are not of the same magnitude as their english counterparts .", "second , most benchmarks available to evaluate progress in hebrew nlp require morphological boundaries which are not available in the output of standard plms .", "in this work we remedy both aspects ."], "relation": "used for", "id": "2022.acl-long.4", "year": 2022, "rel_sent": "Moreover , we introduce a novel neural architecture that recovers the morphological segments encoded in contextualized embedding vectors .", "forward": false, "src_ids": "2022.acl-long.4_4580"}
+{"input": "neural architecture is used for OtherScientificTerm| context: large pre - trained language models ( plms ) have become ubiquitous in the development of language understanding technology and lie at the heart of many artificial intelligence advances . while advances reported for english using plms are unprecedented , reported advances using plms for hebrew are few and far between . the problem is twofold . first , sofar , hebrew resources for training large language models are not of the same magnitude as their english counterparts . second , most benchmarks available to evaluate progress in hebrew nlp require morphological boundaries which are not available in the output of standard plms . in this work we remedy both aspects .", "entity": "neural architecture", "output": "morphological segments", "neg_sample": ["neural architecture is used for OtherScientificTerm", "large pre - trained language models ( plms ) have become ubiquitous in the development of language understanding technology and lie at the heart of many artificial intelligence advances .", "while advances reported for english using plms are unprecedented , reported advances using plms for hebrew are few and far between .", "the problem is twofold .", "first , sofar , hebrew resources for training large language models are not of the same magnitude as their english counterparts .", "second , most benchmarks available to evaluate progress in hebrew nlp require morphological boundaries which are not available in the output of standard plms .", "in this work we remedy both aspects ."], "relation": "used for", "id": "2022.acl-long.4", "year": 2022, "rel_sent": "Moreover , we introduce a novel neural architecture that recovers the morphological segments encoded in contextualized embedding vectors .", "forward": true, "src_ids": "2022.acl-long.4_4581"}
+{"input": "prompttuning is done by using Method| context: tuning pre - trained language models ( plms ) with task - specific prompts has been a promising approach for text classification . particularly , previous studies suggest that prompt - tuning has remarkable superiority in the low - data scenario over the generic fine - tuning methods with extra classifiers . a verbalizer is usually handcrafted or searched by gradient descent , which may lack coverage and bring considerable bias and high variances to the results .", "entity": "prompttuning", "output": "knowledgeable prompttuning ( kpt )", "neg_sample": ["prompttuning is done by using Method", "tuning pre - trained language models ( plms ) with task - specific prompts has been a promising approach for text classification .", "particularly , previous studies suggest that prompt - tuning has remarkable superiority in the low - data scenario over the generic fine - tuning methods with extra classifiers .", "a verbalizer is usually handcrafted or searched by gradient descent , which may lack coverage and bring considerable bias and high variances to the results ."], "relation": "used for", "id": "2022.acl-long.158", "year": 2022, "rel_sent": "In this work , we focus on incorporating external knowledge into the verbalizer , forming a knowledgeable prompttuning ( KPT ) , to improve and stabilize prompttuning .", "forward": false, "src_ids": "2022.acl-long.158_4582"}
+{"input": "knowledgeable prompttuning ( kpt ) is used for Task| context: tuning pre - trained language models ( plms ) with task - specific prompts has been a promising approach for text classification . particularly , previous studies suggest that prompt - tuning has remarkable superiority in the low - data scenario over the generic fine - tuning methods with extra classifiers . a verbalizer is usually handcrafted or searched by gradient descent , which may lack coverage and bring considerable bias and high variances to the results .", "entity": "knowledgeable prompttuning ( kpt )", "output": "prompttuning", "neg_sample": ["knowledgeable prompttuning ( kpt ) is used for Task", "tuning pre - trained language models ( plms ) with task - specific prompts has been a promising approach for text classification .", "particularly , previous studies suggest that prompt - tuning has remarkable superiority in the low - data scenario over the generic fine - tuning methods with extra classifiers .", "a verbalizer is usually handcrafted or searched by gradient descent , which may lack coverage and bring considerable bias and high variances to the results ."], "relation": "used for", "id": "2022.acl-long.158", "year": 2022, "rel_sent": "In this work , we focus on incorporating external knowledge into the verbalizer , forming a knowledgeable prompttuning ( KPT ) , to improve and stabilize prompttuning .", "forward": true, "src_ids": "2022.acl-long.158_4583"}
+{"input": "arabic language generation is done by using Method| context: transfer learning with a unified transformer framework ( t5 ) that converts all language problems into a text - to - text format was recently proposed as a simple and effective transfer learning approach . although a multilingual version of the t5 model ( mt5 ) was also introduced , it is not clear how well it can fare on non - english tasks involving diverse data .", "entity": "arabic language generation", "output": "text - to - text transformers", "neg_sample": ["arabic language generation is done by using Method", "transfer learning with a unified transformer framework ( t5 ) that converts all language problems into a text - to - text format was recently proposed as a simple and effective transfer learning approach .", "although a multilingual version of the t5 model ( mt5 ) was also introduced , it is not clear how well it can fare on non - english tasks involving diverse data ."], "relation": "used for", "id": "2022.acl-long.47", "year": 2022, "rel_sent": "AraT5 : Text - to - Text Transformers for Arabic Language Generation.", "forward": false, "src_ids": "2022.acl-long.47_4584"}
+{"input": "text - to - text transformers is used for Task| context: transfer learning with a unified transformer framework ( t5 ) that converts all language problems into a text - to - text format was recently proposed as a simple and effective transfer learning approach . although a multilingual version of the t5 model ( mt5 ) was also introduced , it is not clear how well it can fare on non - english tasks involving diverse data .", "entity": "text - to - text transformers", "output": "arabic language generation", "neg_sample": ["text - to - text transformers is used for Task", "transfer learning with a unified transformer framework ( t5 ) that converts all language problems into a text - to - text format was recently proposed as a simple and effective transfer learning approach .", "although a multilingual version of the t5 model ( mt5 ) was also introduced , it is not clear how well it can fare on non - english tasks involving diverse data ."], "relation": "used for", "id": "2022.acl-long.47", "year": 2022, "rel_sent": "AraT5 : Text - to - Text Transformers for Arabic Language Generation.", "forward": true, "src_ids": "2022.acl-long.47_4585"}
+{"input": "dependency parsing is done by using Task| context: meta - learning , or learning to learn , is a technique that can help to overcome resource scarcity in cross - lingual nlp problems , by enabling fast adaptation to new tasks .", "entity": "dependency parsing", "output": "fast cross - lingual adaptation", "neg_sample": ["dependency parsing is done by using Task", "meta - learning , or learning to learn , is a technique that can help to overcome resource scarcity in cross - lingual nlp problems , by enabling fast adaptation to new tasks ."], "relation": "used for", "id": "2022.acl-long.582", "year": 2022, "rel_sent": "Meta - Learning for Fast Cross - Lingual Adaptation in Dependency Parsing.", "forward": false, "src_ids": "2022.acl-long.582_4586"}
+{"input": "meta - learning is used for Task| context: meta - learning , or learning to learn , is a technique that can help to overcome resource scarcity in cross - lingual nlp problems , by enabling fast adaptation to new tasks .", "entity": "meta - learning", "output": "fast cross - lingual adaptation", "neg_sample": ["meta - learning is used for Task", "meta - learning , or learning to learn , is a technique that can help to overcome resource scarcity in cross - lingual nlp problems , by enabling fast adaptation to new tasks ."], "relation": "used for", "id": "2022.acl-long.582", "year": 2022, "rel_sent": "Meta - Learning for Fast Cross - Lingual Adaptation in Dependency Parsing.", "forward": true, "src_ids": "2022.acl-long.582_4587"}
+{"input": "fast cross - lingual adaptation is used for Task| context: meta - learning , or learning to learn , is a technique that can help to overcome resource scarcity in cross - lingual nlp problems , by enabling fast adaptation to new tasks .", "entity": "fast cross - lingual adaptation", "output": "dependency parsing", "neg_sample": ["fast cross - lingual adaptation is used for Task", "meta - learning , or learning to learn , is a technique that can help to overcome resource scarcity in cross - lingual nlp problems , by enabling fast adaptation to new tasks ."], "relation": "used for", "id": "2022.acl-long.582", "year": 2022, "rel_sent": "Meta - Learning for Fast Cross - Lingual Adaptation in Dependency Parsing.", "forward": true, "src_ids": "2022.acl-long.582_4588"}
+{"input": "named entity recognition is done by using Method| context: named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text . as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used .", "entity": "named entity recognition", "output": "auxiliary learning", "neg_sample": ["named entity recognition is done by using Method", "named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text .", "as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used ."], "relation": "used for", "id": "2022.bionlp-1.13", "year": 2022, "rel_sent": "Auxiliary Learning for Named Entity Recognition with Multiple Auxiliary Biomedical Training Data.", "forward": false, "src_ids": "2022.bionlp-1.13_4589"}
+{"input": "auxiliary training dataset is done by using Method| context: named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text . as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used .", "entity": "auxiliary training dataset", "output": "auxiliary learning", "neg_sample": ["auxiliary training dataset is done by using Method", "named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text .", "as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used ."], "relation": "used for", "id": "2022.bionlp-1.13", "year": 2022, "rel_sent": "MUNCHABLES utilizes multiple training datasets as auxiliary training data by the following methods ; the first one is tofinetune the NER model of the target task by sequentially performing auxiliary learning for each auxiliary training dataset , and the other is to use all training datasets in one auxiliary learning .", "forward": false, "src_ids": "2022.bionlp-1.13_4590"}
+{"input": "auxiliary learning is used for Task| context: as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used .", "entity": "auxiliary learning", "output": "named entity recognition", "neg_sample": ["auxiliary learning is used for Task", "as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used ."], "relation": "used for", "id": "2022.bionlp-1.13", "year": 2022, "rel_sent": "Auxiliary Learning for Named Entity Recognition with Multiple Auxiliary Biomedical Training Data.", "forward": true, "src_ids": "2022.bionlp-1.13_4591"}
+{"input": "auxiliary blessing ( munch ables ) is done by using Material| context: named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text . as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used .", "entity": "auxiliary blessing ( munch ables )", "output": "ner corpora", "neg_sample": ["auxiliary blessing ( munch ables ) is done by using Material", "named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text .", "as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used ."], "relation": "used for", "id": "2022.bionlp-1.13", "year": 2022, "rel_sent": "We propose Multiple Utilization of NER Corpora Helpful for Auxiliary BLESsing ( MUNCH ABLES ) .", "forward": false, "src_ids": "2022.bionlp-1.13_4592"}
+{"input": "ner corpora is used for Task| context: named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text . as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used .", "entity": "ner corpora", "output": "auxiliary blessing ( munch ables )", "neg_sample": ["ner corpora is used for Task", "named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text .", "as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used ."], "relation": "used for", "id": "2022.bionlp-1.13", "year": 2022, "rel_sent": "We propose Multiple Utilization of NER Corpora Helpful for Auxiliary BLESsing ( MUNCH ABLES ) .", "forward": true, "src_ids": "2022.bionlp-1.13_4593"}
+{"input": "auxiliary learning is used for OtherScientificTerm| context: named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text . as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used .", "entity": "auxiliary learning", "output": "auxiliary training dataset", "neg_sample": ["auxiliary learning is used for OtherScientificTerm", "named entity recognition ( ner ) is one of the elemental technologies , which has been used for knowledge extraction from biomedical text .", "as one of the ner improvement approaches , multi - task learning that learns a model from multiple training data has been used ."], "relation": "used for", "id": "2022.bionlp-1.13", "year": 2022, "rel_sent": "MUNCHABLES utilizes multiple training datasets as auxiliary training data by the following methods ; the first one is tofinetune the NER model of the target task by sequentially performing auxiliary learning for each auxiliary training dataset , and the other is to use all training datasets in one auxiliary learning .", "forward": true, "src_ids": "2022.bionlp-1.13_4594"}
+{"input": "imbalanced training dataset is done by using Method| context: word sense disambiguation ( wsd ) is a crucial problem in the natural language processing ( nlp ) community . current methods achieve decent performance by utilizing supervised learning and large pre - trained language models . however , the imbalanced training dataset leads to poor performance on rare senses and zero - shot senses . there are more training instances and senses for words with top frequency ranks than those with low frequency ranks in the training dataset .", "entity": "imbalanced training dataset", "output": "z - reweighting strategy", "neg_sample": ["imbalanced training dataset is done by using Method", "word sense disambiguation ( wsd ) is a crucial problem in the natural language processing ( nlp ) community .", "current methods achieve decent performance by utilizing supervised learning and large pre - trained language models .", "however , the imbalanced training dataset leads to poor performance on rare senses and zero - shot senses .", "there are more training instances and senses for words with top frequency ranks than those with low frequency ranks in the training dataset ."], "relation": "used for", "id": "2022.acl-long.323", "year": 2022, "rel_sent": "Based on the relation , we propose a Z - reweighting method on the word level to adjust the training on the imbalanced dataset .", "forward": false, "src_ids": "2022.acl-long.323_4595"}
+{"input": "z - reweighting strategy is used for Material| context: word sense disambiguation ( wsd ) is a crucial problem in the natural language processing ( nlp ) community . current methods achieve decent performance by utilizing supervised learning and large pre - trained language models . there are more training instances and senses for words with top frequency ranks than those with low frequency ranks in the training dataset .", "entity": "z - reweighting strategy", "output": "imbalanced training dataset", "neg_sample": ["z - reweighting strategy is used for Material", "word sense disambiguation ( wsd ) is a crucial problem in the natural language processing ( nlp ) community .", "current methods achieve decent performance by utilizing supervised learning and large pre - trained language models .", "there are more training instances and senses for words with top frequency ranks than those with low frequency ranks in the training dataset ."], "relation": "used for", "id": "2022.acl-long.323", "year": 2022, "rel_sent": "Based on the relation , we propose a Z - reweighting method on the word level to adjust the training on the imbalanced dataset .", "forward": true, "src_ids": "2022.acl-long.323_4596"}
+{"input": "discriminative reading comprehension is done by using Method| context: as a broad and major category in machine reading comprehension ( mrc ) , the generalized goal of discriminative mrc is answer prediction from the given materials . however , the focuses of various discriminative mrc tasks may be diverse enough : multi - choice mrc requires model to highlight and integrate all potential critical evidence globally ; while extractive mrc focuses on higher local boundary preciseness for answer extraction . among previous works , there lacks a unified design with pertinence for the overall discriminative mrc tasks .", "entity": "discriminative reading comprehension", "output": "lite unified modeling", "neg_sample": ["discriminative reading comprehension is done by using Method", "as a broad and major category in machine reading comprehension ( mrc ) , the generalized goal of discriminative mrc is answer prediction from the given materials .", "however , the focuses of various discriminative mrc tasks may be diverse enough : multi - choice mrc requires model to highlight and integrate all potential critical evidence globally ; while extractive mrc focuses on higher local boundary preciseness for answer extraction .", "among previous works , there lacks a unified design with pertinence for the overall discriminative mrc tasks ."], "relation": "used for", "id": "2022.acl-long.594", "year": 2022, "rel_sent": "Lite Unified Modeling for Discriminative Reading Comprehension.", "forward": false, "src_ids": "2022.acl-long.594_4597"}
+{"input": "lite unified modeling is used for Task| context: as a broad and major category in machine reading comprehension ( mrc ) , the generalized goal of discriminative mrc is answer prediction from the given materials . however , the focuses of various discriminative mrc tasks may be diverse enough : multi - choice mrc requires model to highlight and integrate all potential critical evidence globally ; while extractive mrc focuses on higher local boundary preciseness for answer extraction . among previous works , there lacks a unified design with pertinence for the overall discriminative mrc tasks .", "entity": "lite unified modeling", "output": "discriminative reading comprehension", "neg_sample": ["lite unified modeling is used for Task", "as a broad and major category in machine reading comprehension ( mrc ) , the generalized goal of discriminative mrc is answer prediction from the given materials .", "however , the focuses of various discriminative mrc tasks may be diverse enough : multi - choice mrc requires model to highlight and integrate all potential critical evidence globally ; while extractive mrc focuses on higher local boundary preciseness for answer extraction .", "among previous works , there lacks a unified design with pertinence for the overall discriminative mrc tasks ."], "relation": "used for", "id": "2022.acl-long.594", "year": 2022, "rel_sent": "Lite Unified Modeling for Discriminative Reading Comprehension.", "forward": true, "src_ids": "2022.acl-long.594_4598"}
+{"input": "positive - sum retailing is done by using Task| context: retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals . in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met . in this paper , we manage the retailing problem by mixing cooperation and competition .", "entity": "positive - sum retailing", "output": "conversational ai", "neg_sample": ["positive - sum retailing is done by using Task", "retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals .", "in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met .", "in this paper , we manage the retailing problem by mixing cooperation and competition ."], "relation": "used for", "id": "2022.nlp4convai-1.3", "year": 2022, "rel_sent": "Conversational AI for Positive - sum Retailing under Falsehood Control.", "forward": false, "src_ids": "2022.nlp4convai-1.3_4599"}
+{"input": "conversational ai is used for Task| context: retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals . in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met . in this paper , we manage the retailing problem by mixing cooperation and competition .", "entity": "conversational ai", "output": "positive - sum retailing", "neg_sample": ["conversational ai is used for Task", "retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals .", "in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met .", "in this paper , we manage the retailing problem by mixing cooperation and competition ."], "relation": "used for", "id": "2022.nlp4convai-1.3", "year": 2022, "rel_sent": "Conversational AI for Positive - sum Retailing under Falsehood Control.", "forward": true, "src_ids": "2022.nlp4convai-1.3_4600"}
+{"input": "transfer learning techniques is used for OtherScientificTerm| context: retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals . in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met . in this paper , we manage the retailing problem by mixing cooperation and competition .", "entity": "transfer learning techniques", "output": "decision - making ability", "neg_sample": ["transfer learning techniques is used for OtherScientificTerm", "retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals .", "in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met .", "in this paper , we manage the retailing problem by mixing cooperation and competition ."], "relation": "used for", "id": "2022.nlp4convai-1.3", "year": 2022, "rel_sent": "Using this dataset , we leverage transfer learning techniques on a pretrained , end - to - end model and enhance its decision - making ability toward the best choice in terms of utility by means of multi - agent reinforcement learning .", "forward": true, "src_ids": "2022.nlp4convai-1.3_4601"}
+{"input": "decision - making ability is done by using Method| context: retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals . in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met . in this paper , we manage the retailing problem by mixing cooperation and competition .", "entity": "decision - making ability", "output": "transfer learning techniques", "neg_sample": ["decision - making ability is done by using Method", "retailing combines complicated communication skills and strategies to reach an agreement between buyer and seller with identical or different goals .", "in each transaction a good seller finds an optimal solution by considering his / her own profits while simultaneously considering whether the buyer 's needs have been met .", "in this paper , we manage the retailing problem by mixing cooperation and competition ."], "relation": "used for", "id": "2022.nlp4convai-1.3", "year": 2022, "rel_sent": "Using this dataset , we leverage transfer learning techniques on a pretrained , end - to - end model and enhance its decision - making ability toward the best choice in terms of utility by means of multi - agent reinforcement learning .", "forward": false, "src_ids": "2022.nlp4convai-1.3_4602"}
+{"input": "back - translation is done by using Task| context: back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available . recently , diverse approaches have been proposed to get better automatic evaluation results of nmt models using back - translation , including the use of sampling instead of beam search as decoding algorithm for creating the synthetic corpus . alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "back - translation", "output": "tagging", "neg_sample": ["back - translation is done by using Task", "back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available .", "recently , diverse approaches have been proposed to get better automatic evaluation results of nmt models using back - translation , including the use of sampling instead of beam search as decoding algorithm for creating the synthetic corpus .", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "Comparing and combining tagging with different decoding algorithms for back - translation in NMT : learnings from a low resource scenario.", "forward": false, "src_ids": "2022.eamt-1.6_4603"}
+{"input": "back - translation is done by using Method| context: back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available . alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "back - translation", "output": "decoding algorithms", "neg_sample": ["back - translation is done by using Method", "back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available .", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "Comparing and combining tagging with different decoding algorithms for back - translation in NMT : learnings from a low resource scenario.", "forward": false, "src_ids": "2022.eamt-1.6_4604"}
+{"input": "neural machine translation is done by using Method| context: alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "neural machine translation", "output": "back - translation", "neg_sample": ["neural machine translation is done by using Method", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "Comparing and combining tagging with different decoding algorithms for back - translation in NMT : learnings from a low resource scenario.", "forward": false, "src_ids": "2022.eamt-1.6_4605"}
+{"input": "decoding algorithms is used for Method| context: alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "decoding algorithms", "output": "back - translation", "neg_sample": ["decoding algorithms is used for Method", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "Comparing and combining tagging with different decoding algorithms for back - translation in NMT : learnings from a low resource scenario.", "forward": true, "src_ids": "2022.eamt-1.6_4606"}
+{"input": "tagging is used for Method| context: alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "tagging", "output": "back - translation", "neg_sample": ["tagging is used for Method", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "Comparing and combining tagging with different decoding algorithms for back - translation in NMT : learnings from a low resource scenario.", "forward": true, "src_ids": "2022.eamt-1.6_4607"}
+{"input": "back - translation is used for Task| context: recently , diverse approaches have been proposed to get better automatic evaluation results of nmt models using back - translation , including the use of sampling instead of beam search as decoding algorithm for creating the synthetic corpus . alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "back - translation", "output": "neural machine translation", "neg_sample": ["back - translation is used for Task", "recently , diverse approaches have been proposed to get better automatic evaluation results of nmt models using back - translation , including the use of sampling instead of beam search as decoding algorithm for creating the synthetic corpus .", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "Comparing and combining tagging with different decoding algorithms for back - translation in NMT : learnings from a low resource scenario.", "forward": true, "src_ids": "2022.eamt-1.6_4608"}
+{"input": "decoding is done by using Method| context: back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available . alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "decoding", "output": "sampling", "neg_sample": ["decoding is done by using Method", "back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available .", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "The results suggest that in lower resource scenarios tagging only helps when using sampling for decoding , in contradiction with the previous literature using bigger corpora from the news domain .", "forward": false, "src_ids": "2022.eamt-1.6_4609"}
+{"input": "sampling is used for Task| context: back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available . alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one . however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system .", "entity": "sampling", "output": "decoding", "neg_sample": ["sampling is used for Task", "back - translation is a well established approach to improve the performance of neural machine translation ( nmt ) systems when large monolingual corpora of the target language and domain are available .", "alternatively , it has been proposed to append a tag to the back - translated corpus for helping the nmt system to distinguish the synthetic bilingual corpus from the authentic one .", "however , not all the combinations of the previous approaches have been tested , and thus it is not clear which is the best approach for developing a given nmt system ."], "relation": "used for", "id": "2022.eamt-1.6", "year": 2022, "rel_sent": "The results suggest that in lower resource scenarios tagging only helps when using sampling for decoding , in contradiction with the previous literature using bigger corpora from the news domain .", "forward": true, "src_ids": "2022.eamt-1.6_4610"}
+{"input": "deep learning age is done by using Method| context: graph - based methods , which decompose the score of a dependency tree into scores of dependency arcs , are popular in dependency parsing for decades .", "entity": "deep learning age", "output": "second - order graph - based parsing", "neg_sample": ["deep learning age is done by using Method", "graph - based methods , which decompose the score of a dependency tree into scores of dependency arcs , are popular in dependency parsing for decades ."], "relation": "used for", "id": "2022.findings-acl.112", "year": 2022, "rel_sent": "We also confirm the effectiveness of second - order graph - based parsing in the deep learning age , however , we observe marginal or no improvement when combining second - order graph - based and headed - span - based methods .", "forward": false, "src_ids": "2022.findings-acl.112_4611"}
+{"input": "second - order graph - based parsing is used for Task| context: graph - based methods , which decompose the score of a dependency tree into scores of dependency arcs , are popular in dependency parsing for decades .", "entity": "second - order graph - based parsing", "output": "deep learning age", "neg_sample": ["second - order graph - based parsing is used for Task", "graph - based methods , which decompose the score of a dependency tree into scores of dependency arcs , are popular in dependency parsing for decades ."], "relation": "used for", "id": "2022.findings-acl.112", "year": 2022, "rel_sent": "We also confirm the effectiveness of second - order graph - based parsing in the deep learning age , however , we observe marginal or no improvement when combining second - order graph - based and headed - span - based methods .", "forward": true, "src_ids": "2022.findings-acl.112_4612"}
+{"input": "healthcare applications is done by using Task| context: although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent . as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area .", "entity": "healthcare applications", "output": "task - oriented dialogue systems", "neg_sample": ["healthcare applications is done by using Task", "although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent .", "as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area ."], "relation": "used for", "id": "2022.acl-long.458", "year": 2022, "rel_sent": "The AI Doctor Is In : A Survey of Task - Oriented Dialogue Systems for Healthcare Applications.", "forward": false, "src_ids": "2022.acl-long.458_4613"}
+{"input": "task - oriented dialogue systems is used for Task| context: task - oriented dialogue systems are increasingly prevalent in healthcare settings , and have been characterized by a diverse range of architectures and objectives . although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent . as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area .", "entity": "task - oriented dialogue systems", "output": "healthcare applications", "neg_sample": ["task - oriented dialogue systems is used for Task", "task - oriented dialogue systems are increasingly prevalent in healthcare settings , and have been characterized by a diverse range of architectures and objectives .", "although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent .", "as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area ."], "relation": "used for", "id": "2022.acl-long.458", "year": 2022, "rel_sent": "The AI Doctor Is In : A Survey of Task - Oriented Dialogue Systems for Healthcare Applications.", "forward": true, "src_ids": "2022.acl-long.458_4614"}
+{"input": "amr - to - text generation is done by using Task| context: a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) . metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards . however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference . furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting .", "entity": "amr - to - text generation", "output": "rewarding semantic similarity", "neg_sample": ["amr - to - text generation is done by using Task", "a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) .", "metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards .", "however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference .", "furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting ."], "relation": "used for", "id": "2022.acl-short.80", "year": 2022, "rel_sent": "Rewarding Semantic Similarity under Optimized Alignments for AMR - to - Text Generation.", "forward": false, "src_ids": "2022.acl-short.80_4615"}
+{"input": "amr - to - text generation is done by using OtherScientificTerm| context: a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) . metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards . however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference . furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting .", "entity": "amr - to - text generation", "output": "optimized alignments", "neg_sample": ["amr - to - text generation is done by using OtherScientificTerm", "a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) .", "metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards .", "however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference .", "furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting ."], "relation": "used for", "id": "2022.acl-short.80", "year": 2022, "rel_sent": "Rewarding Semantic Similarity under Optimized Alignments for AMR - to - Text Generation.", "forward": false, "src_ids": "2022.acl-short.80_4616"}
+{"input": "rewarding semantic similarity is used for Task| context: a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) . metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards . however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference . furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting .", "entity": "rewarding semantic similarity", "output": "amr - to - text generation", "neg_sample": ["rewarding semantic similarity is used for Task", "a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) .", "metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards .", "however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference .", "furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting ."], "relation": "used for", "id": "2022.acl-short.80", "year": 2022, "rel_sent": "Rewarding Semantic Similarity under Optimized Alignments for AMR - to - Text Generation.", "forward": true, "src_ids": "2022.acl-short.80_4617"}
+{"input": "optimized alignments is used for Task| context: a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) . metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards . however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference . furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting .", "entity": "optimized alignments", "output": "amr - to - text generation", "neg_sample": ["optimized alignments is used for Task", "a common way to combat exposure bias is by applying scores from evaluation metrics as rewards in reinforcement learning ( rl ) .", "metrics leveraging contextualized embeddings appear more flexible than their n - gram matching counterparts and thus ideal as training rewards .", "however , metrics such as bertscore greedily align candidate and reference tokens , which can allow system outputs to receive excess credit relative to a reference .", "furthermore , past approaches featuring semantic similarity rewards suffer from repetitive outputs and overfitting ."], "relation": "used for", "id": "2022.acl-short.80", "year": 2022, "rel_sent": "Rewarding Semantic Similarity under Optimized Alignments for AMR - to - Text Generation.", "forward": true, "src_ids": "2022.acl-short.80_4618"}
+{"input": "creole languages documentation is done by using Task| context: we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien . automatic language processing tools are almost non - existent for these two languages .", "entity": "creole languages documentation", "output": "automatic speech recognition", "neg_sample": ["creole languages documentation is done by using Task", "we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien .", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "Automatic Speech Recognition and Query By Example for Creole Languages Documentation.", "forward": false, "src_ids": "2022.findings-acl.197_4619"}
+{"input": "query by example is used for Material| context: we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien . automatic language processing tools are almost non - existent for these two languages .", "entity": "query by example", "output": "creole languages documentation", "neg_sample": ["query by example is used for Material", "we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien .", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "Automatic Speech Recognition and Query By Example for Creole Languages Documentation.", "forward": true, "src_ids": "2022.findings-acl.197_4620"}
+{"input": "automatic speech recognition is used for Material| context: we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien . automatic language processing tools are almost non - existent for these two languages .", "entity": "automatic speech recognition", "output": "creole languages documentation", "neg_sample": ["automatic speech recognition is used for Material", "we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien .", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "Automatic Speech Recognition and Query By Example for Creole Languages Documentation.", "forward": true, "src_ids": "2022.findings-acl.197_4621"}
+{"input": "multilingual self - supervised models is used for Material| context: automatic language processing tools are almost non - existent for these two languages .", "entity": "multilingual self - supervised models", "output": "creole languages", "neg_sample": ["multilingual self - supervised models is used for Material", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "Moreover , our experiments show that multilingual self - supervised models are not necessarily the most efficient for Creole languages .", "forward": true, "src_ids": "2022.findings-acl.197_4622"}
+{"input": "automatic speech recognition system is done by using Material| context: we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien . automatic language processing tools are almost non - existent for these two languages .", "entity": "automatic speech recognition system", "output": "annotated data", "neg_sample": ["automatic speech recognition system is done by using Material", "we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien .", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "We propose to use about one hour of annotated data to design an automatic speech recognition system for each language .", "forward": false, "src_ids": "2022.findings-acl.197_4623"}
+{"input": "annotated data is used for Method| context: we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien . automatic language processing tools are almost non - existent for these two languages .", "entity": "annotated data", "output": "automatic speech recognition system", "neg_sample": ["annotated data is used for Method", "we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien .", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "We propose to use about one hour of annotated data to design an automatic speech recognition system for each language .", "forward": true, "src_ids": "2022.findings-acl.197_4624"}
+{"input": "creole languages is done by using Method| context: we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien . automatic language processing tools are almost non - existent for these two languages .", "entity": "creole languages", "output": "multilingual self - supervised models", "neg_sample": ["creole languages is done by using Method", "we investigate the exploitation of self - supervised models for two creole languages with few resources : gwadloupeyen and morisien .", "automatic language processing tools are almost non - existent for these two languages ."], "relation": "used for", "id": "2022.findings-acl.197", "year": 2022, "rel_sent": "Moreover , our experiments show that multilingual self - supervised models are not necessarily the most efficient for Creole languages .", "forward": false, "src_ids": "2022.findings-acl.197_4625"}
+{"input": "multiword expressions detection is done by using Method| context: effective methods for multiword expressions detection are important for many technologies related to natural language processing . most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures .", "entity": "multiword expressions detection", "output": "deep neural representations", "neg_sample": ["multiword expressions detection is done by using Method", "effective methods for multiword expressions detection are important for many technologies related to natural language processing .", "most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures ."], "relation": "used for", "id": "2022.acl-srw.36", "year": 2022, "rel_sent": "Deep Neural Representations for Multiword Expressions Detection.", "forward": false, "src_ids": "2022.acl-srw.36_4626"}
+{"input": "deep neural representations is used for Task| context: most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures .", "entity": "deep neural representations", "output": "multiword expressions detection", "neg_sample": ["deep neural representations is used for Task", "most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures ."], "relation": "used for", "id": "2022.acl-srw.36", "year": 2022, "rel_sent": "Deep Neural Representations for Multiword Expressions Detection.", "forward": true, "src_ids": "2022.acl-srw.36_4627"}
+{"input": "silver dataset is done by using Method| context: effective methods for multiword expressions detection are important for many technologies related to natural language processing . most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures .", "entity": "silver dataset", "output": "deep learning architectures", "neg_sample": ["silver dataset is done by using Method", "effective methods for multiword expressions detection are important for many technologies related to natural language processing .", "most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures ."], "relation": "used for", "id": "2022.acl-srw.36", "year": 2022, "rel_sent": "We propose a method for building a silver dataset by spotting multiword expression occurrences and acquiring statistical collocations as negative samples .", "forward": false, "src_ids": "2022.acl-srw.36_4628"}
+{"input": "deep learning architectures is used for Material| context: effective methods for multiword expressions detection are important for many technologies related to natural language processing . most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures .", "entity": "deep learning architectures", "output": "silver dataset", "neg_sample": ["deep learning architectures is used for Material", "effective methods for multiword expressions detection are important for many technologies related to natural language processing .", "most contemporary methods are based on the sequence labeling scheme applied to an annotated corpus , while traditional methods use statistical measures ."], "relation": "used for", "id": "2022.acl-srw.36", "year": 2022, "rel_sent": "We propose a method for building a silver dataset by spotting multiword expression occurrences and acquiring statistical collocations as negative samples .", "forward": true, "src_ids": "2022.acl-srw.36_4629"}
+{"input": "computational language documentation is done by using Task| context: word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown . however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data .", "entity": "computational language documentation", "output": "weakly supervised word segmentation", "neg_sample": ["computational language documentation is done by using Task", "word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown .", "however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data ."], "relation": "used for", "id": "2022.acl-long.510", "year": 2022, "rel_sent": "Weakly Supervised Word Segmentation for Computational Language Documentation.", "forward": false, "src_ids": "2022.acl-long.510_4630"}
+{"input": "weakly supervised word segmentation is used for Task| context: word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown . however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data .", "entity": "weakly supervised word segmentation", "output": "computational language documentation", "neg_sample": ["weakly supervised word segmentation is used for Task", "word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown .", "however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data ."], "relation": "used for", "id": "2022.acl-long.510", "year": 2022, "rel_sent": "Weakly Supervised Word Segmentation for Computational Language Documentation.", "forward": true, "src_ids": "2022.acl-long.510_4631"}
+{"input": "weak supervision is used for Method| context: word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown . however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data .", "entity": "weak supervision", "output": "bayesian non - parametric models of segmentation", "neg_sample": ["weak supervision is used for Method", "word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown .", "however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data ."], "relation": "used for", "id": "2022.acl-long.510", "year": 2022, "rel_sent": "This paper studies how such a weak supervision can be taken advantage of in Bayesian non - parametric models of segmentation .", "forward": true, "src_ids": "2022.acl-long.510_4632"}
+{"input": "bayesian non - parametric models of segmentation is done by using OtherScientificTerm| context: word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown . however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data .", "entity": "bayesian non - parametric models of segmentation", "output": "weak supervision", "neg_sample": ["bayesian non - parametric models of segmentation is done by using OtherScientificTerm", "word and morpheme segmentation are fundamental steps of language documentation as they allow to discover lexical units in a language for which the lexicon is unknown .", "however , in most language documentation scenarios , linguists do not start from a blank page : they may already have a pre - existing dictionary or have initiated manual segmentation of a small part of their data ."], "relation": "used for", "id": "2022.acl-long.510", "year": 2022, "rel_sent": "This paper studies how such a weak supervision can be taken advantage of in Bayesian non - parametric models of segmentation .", "forward": false, "src_ids": "2022.acl-long.510_4633"}
+{"input": "argument mining is done by using OtherScientificTerm| context: inferential reasoning is an essential feature of argumentation . therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment . there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining .", "entity": "argument mining", "output": "relational propositions", "neg_sample": ["argument mining is done by using OtherScientificTerm", "inferential reasoning is an essential feature of argumentation .", "therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment .", "there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining ."], "relation": "used for", "id": "2022.scil-1.7", "year": 2022, "rel_sent": "Inferring Inferences : Relational Propositions for Argument Mining.", "forward": false, "src_ids": "2022.scil-1.7_4634"}
+{"input": "rhetorical structures is used for OtherScientificTerm| context: inferential reasoning is an essential feature of argumentation . therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment . there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining .", "entity": "rhetorical structures", "output": "relational propositions", "neg_sample": ["rhetorical structures is used for OtherScientificTerm", "inferential reasoning is an essential feature of argumentation .", "therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment .", "there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining ."], "relation": "used for", "id": "2022.scil-1.7", "year": 2022, "rel_sent": "From rhetorical structures , relational propositions are defined , and from these propositions , logical expressions are then generated .", "forward": true, "src_ids": "2022.scil-1.7_4635"}
+{"input": "relational propositions is used for Task| context: inferential reasoning is an essential feature of argumentation . therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment . there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining .", "entity": "relational propositions", "output": "argument mining", "neg_sample": ["relational propositions is used for Task", "inferential reasoning is an essential feature of argumentation .", "therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment .", "there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining ."], "relation": "used for", "id": "2022.scil-1.7", "year": 2022, "rel_sent": "Inferring Inferences : Relational Propositions for Argument Mining.", "forward": true, "src_ids": "2022.scil-1.7_4636"}
+{"input": "relational propositions is done by using OtherScientificTerm| context: inferential reasoning is an essential feature of argumentation . therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment . there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining .", "entity": "relational propositions", "output": "rhetorical structures", "neg_sample": ["relational propositions is done by using OtherScientificTerm", "inferential reasoning is an essential feature of argumentation .", "therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment .", "there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining ."], "relation": "used for", "id": "2022.scil-1.7", "year": 2022, "rel_sent": "From rhetorical structures , relational propositions are defined , and from these propositions , logical expressions are then generated .", "forward": false, "src_ids": "2022.scil-1.7_4637"}
+{"input": "logical expressions is done by using OtherScientificTerm| context: inferential reasoning is an essential feature of argumentation . therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment . there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining .", "entity": "logical expressions", "output": "propositions", "neg_sample": ["logical expressions is done by using OtherScientificTerm", "inferential reasoning is an essential feature of argumentation .", "therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment .", "there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining ."], "relation": "used for", "id": "2022.scil-1.7", "year": 2022, "rel_sent": "From rhetorical structures , relational propositions are defined , and from these propositions , logical expressions are then generated .", "forward": false, "src_ids": "2022.scil-1.7_4638"}
+{"input": "propositions is used for OtherScientificTerm| context: inferential reasoning is an essential feature of argumentation . therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment . there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining .", "entity": "propositions", "output": "logical expressions", "neg_sample": ["propositions is used for OtherScientificTerm", "inferential reasoning is an essential feature of argumentation .", "therefore , a method for mining discourse for inferential structures would be of value for argument analysis and assessment .", "there are , however , unsettled issues associated with rhetorical structure theory ( rst ) , some of which are problematic for inference mining ."], "relation": "used for", "id": "2022.scil-1.7", "year": 2022, "rel_sent": "From rhetorical structures , relational propositions are defined , and from these propositions , logical expressions are then generated .", "forward": true, "src_ids": "2022.scil-1.7_4639"}
+{"input": "event - centric narrative creation is done by using Task| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "event - centric narrative creation", "output": "news article retrieval", "neg_sample": ["event - centric narrative creation is done by using Task", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "News Article Retrieval in Context for Event - centric Narrative Creation.", "forward": false, "src_ids": "2022.in2writing-1.10_4640"}
+{"input": "news article retrieval is used for Task| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "news article retrieval", "output": "event - centric narrative creation", "neg_sample": ["news article retrieval is used for Task", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "News Article Retrieval in Context for Event - centric Narrative Creation.", "forward": true, "src_ids": "2022.in2writing-1.10_4641"}
+{"input": "news articles is used for Material| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "news articles", "output": "incomplete narratives", "neg_sample": ["news articles is used for Material", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "We formally define this task and propose a retrieval dataset construction procedure that relies on existing news articles to simulate incomplete narratives and relevant articles .", "forward": true, "src_ids": "2022.in2writing-1.10_4642"}
+{"input": "retrieval dataset construction procedure is used for Material| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "retrieval dataset construction procedure", "output": "incomplete narratives", "neg_sample": ["retrieval dataset construction procedure is used for Material", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "We formally define this task and propose a retrieval dataset construction procedure that relies on existing news articles to simulate incomplete narratives and relevant articles .", "forward": true, "src_ids": "2022.in2writing-1.10_4643"}
+{"input": "incomplete narratives is done by using Material| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "incomplete narratives", "output": "news articles", "neg_sample": ["incomplete narratives is done by using Material", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "We formally define this task and propose a retrieval dataset construction procedure that relies on existing news articles to simulate incomplete narratives and relevant articles .", "forward": false, "src_ids": "2022.in2writing-1.10_4644"}
+{"input": "relevant articles is done by using Material| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "relevant articles", "output": "news articles", "neg_sample": ["relevant articles is done by using Material", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "We formally define this task and propose a retrieval dataset construction procedure that relies on existing news articles to simulate incomplete narratives and relevant articles .", "forward": false, "src_ids": "2022.in2writing-1.10_4645"}
+{"input": "incomplete narratives is done by using Method| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "incomplete narratives", "output": "retrieval dataset construction procedure", "neg_sample": ["incomplete narratives is done by using Method", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "We formally define this task and propose a retrieval dataset construction procedure that relies on existing news articles to simulate incomplete narratives and relevant articles .", "forward": false, "src_ids": "2022.in2writing-1.10_4646"}
+{"input": "relevant articles is done by using Method| context: writers such as journalists often use automatic tools tofind relevant content to include in their narratives .", "entity": "relevant articles", "output": "retrieval dataset construction procedure", "neg_sample": ["relevant articles is done by using Method", "writers such as journalists often use automatic tools tofind relevant content to include in their narratives ."], "relation": "used for", "id": "2022.in2writing-1.10", "year": 2022, "rel_sent": "We formally define this task and propose a retrieval dataset construction procedure that relies on existing news articles to simulate incomplete narratives and relevant articles .", "forward": false, "src_ids": "2022.in2writing-1.10_4647"}
+{"input": "bert is used for Method| context: however , their generalization ability to other domains remains weak . existing findings on cross - domain constituency parsing are only made on a limited number of domains .", "entity": "bert", "output": "parsers", "neg_sample": ["bert is used for Method", "however , their generalization ability to other domains remains weak .", "existing findings on cross - domain constituency parsing are only made on a limited number of domains ."], "relation": "used for", "id": "2022.findings-acl.11", "year": 2022, "rel_sent": "Primarily , we find that 1 ) BERT significantly increases parsers ' cross - domain performance by reducing their sensitivity on the domain - variant features.2 ) Compared with single metrics such as unigram distribution and OOV rate , challenges to open - domain constituency parsing arise from complex features , including cross - domain lexical and constituent structure variations .", "forward": true, "src_ids": "2022.findings-acl.11_4648"}
+{"input": "constituency parsers is done by using OtherScientificTerm| context: neural constituency parsers have reached practical performance on news - domain benchmarks . however , their generalization ability to other domains remains weak . existing findings on cross - domain constituency parsing are only made on a limited number of domains .", "entity": "constituency parsers", "output": "linguistic features", "neg_sample": ["constituency parsers is done by using OtherScientificTerm", "neural constituency parsers have reached practical performance on news - domain benchmarks .", "however , their generalization ability to other domains remains weak .", "existing findings on cross - domain constituency parsing are only made on a limited number of domains ."], "relation": "used for", "id": "2022.findings-acl.11", "year": 2022, "rel_sent": "We analyze challenges to open - domain constituency parsing using a set of linguistic features on various strong constituency parsers .", "forward": false, "src_ids": "2022.findings-acl.11_4649"}
+{"input": "linguistic features is used for Method| context: however , their generalization ability to other domains remains weak . existing findings on cross - domain constituency parsing are only made on a limited number of domains .", "entity": "linguistic features", "output": "constituency parsers", "neg_sample": ["linguistic features is used for Method", "however , their generalization ability to other domains remains weak .", "existing findings on cross - domain constituency parsing are only made on a limited number of domains ."], "relation": "used for", "id": "2022.findings-acl.11", "year": 2022, "rel_sent": "We analyze challenges to open - domain constituency parsing using a set of linguistic features on various strong constituency parsers .", "forward": true, "src_ids": "2022.findings-acl.11_4650"}
+{"input": "parsers is done by using Method| context: neural constituency parsers have reached practical performance on news - domain benchmarks . however , their generalization ability to other domains remains weak . existing findings on cross - domain constituency parsing are only made on a limited number of domains .", "entity": "parsers", "output": "bert", "neg_sample": ["parsers is done by using Method", "neural constituency parsers have reached practical performance on news - domain benchmarks .", "however , their generalization ability to other domains remains weak .", "existing findings on cross - domain constituency parsing are only made on a limited number of domains ."], "relation": "used for", "id": "2022.findings-acl.11", "year": 2022, "rel_sent": "Primarily , we find that 1 ) BERT significantly increases parsers ' cross - domain performance by reducing their sensitivity on the domain - variant features.2 ) Compared with single metrics such as unigram distribution and OOV rate , challenges to open - domain constituency parsing arise from complex features , including cross - domain lexical and constituent structure variations .", "forward": false, "src_ids": "2022.findings-acl.11_4651"}
+{"input": "suboptimal serialization of forms is done by using Method| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "suboptimal serialization of forms", "output": "formnet", "neg_sample": ["suboptimal serialization of forms is done by using Method", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "We propose FormNet , a structure - aware sequence model to mitigate the suboptimal serialization of forms .", "forward": false, "src_ids": "2022.acl-long.260_4652"}
+{"input": "local syntactic information is done by using Method| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "local syntactic information", "output": "formnet", "neg_sample": ["local syntactic information is done by using Method", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "FormNet therefore explicitly recovers local syntactic information that may have been lost during serialization .", "forward": false, "src_ids": "2022.acl-long.260_4653"}
+{"input": "formnet is used for Task| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "formnet", "output": "suboptimal serialization of forms", "neg_sample": ["formnet is used for Task", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "We propose FormNet , a structure - aware sequence model to mitigate the suboptimal serialization of forms .", "forward": true, "src_ids": "2022.acl-long.260_4654"}
+{"input": "structure - aware sequence model is used for Task| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "structure - aware sequence model", "output": "suboptimal serialization of forms", "neg_sample": ["structure - aware sequence model is used for Task", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "We propose FormNet , a structure - aware sequence model to mitigate the suboptimal serialization of forms .", "forward": true, "src_ids": "2022.acl-long.260_4655"}
+{"input": "attention score calculation is done by using Method| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "attention score calculation", "output": "rich attention", "neg_sample": ["attention score calculation is done by using Method", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "First , we design Rich Attention that leverages the spatial relationship between tokens in a form for more precise attention score calculation .", "forward": false, "src_ids": "2022.acl-long.260_4656"}
+{"input": "rich attention is used for Task| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "rich attention", "output": "attention score calculation", "neg_sample": ["rich attention is used for Task", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "First , we design Rich Attention that leverages the spatial relationship between tokens in a form for more precise attention score calculation .", "forward": true, "src_ids": "2022.acl-long.260_4657"}
+{"input": "formnet is used for OtherScientificTerm| context: sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks . however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns .", "entity": "formnet", "output": "local syntactic information", "neg_sample": ["formnet is used for OtherScientificTerm", "sequence modeling has demonstrated state - of - the - art performance on natural language and document understanding tasks .", "however , it is challenging to correctly serialize tokens in form - like documents in practice due to their variety of layout patterns ."], "relation": "used for", "id": "2022.acl-long.260", "year": 2022, "rel_sent": "FormNet therefore explicitly recovers local syntactic information that may have been lost during serialization .", "forward": true, "src_ids": "2022.acl-long.260_4658"}
+{"input": "prosody - aware generative spoken language model is used for Material| context: speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored . generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units . unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech .", "entity": "prosody - aware generative spoken language model", "output": "natural , meaningful , and coherent speech", "neg_sample": ["prosody - aware generative spoken language model is used for Material", "speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored .", "generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units .", "unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech ."], "relation": "used for", "id": "2022.acl-long.593", "year": 2022, "rel_sent": "Experimental results show that the pGSLM can utilize prosody to improve both prosody and content modeling , and also generate natural , meaningful , and coherent speech given a spoken prompt .", "forward": true, "src_ids": "2022.acl-long.593_4659"}
+{"input": "ms - tlm outputs is done by using Method| context: speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored . generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units . unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech .", "entity": "ms - tlm outputs", "output": "hifi - gan model", "neg_sample": ["ms - tlm outputs is done by using Method", "speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored .", "generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units .", "unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech ."], "relation": "used for", "id": "2022.acl-long.593", "year": 2022, "rel_sent": "It is composed of a multi - stream transformer language model ( MS - TLM ) of speech , represented as discovered unit and prosodic feature streams , and an adapted HiFi - GAN model converting MS - TLM outputs to waveforms .", "forward": false, "src_ids": "2022.acl-long.593_4660"}
+{"input": "hifi - gan model is used for OtherScientificTerm| context: speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored . generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units . unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech .", "entity": "hifi - gan model", "output": "ms - tlm outputs", "neg_sample": ["hifi - gan model is used for OtherScientificTerm", "speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored .", "generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units .", "unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech ."], "relation": "used for", "id": "2022.acl-long.593", "year": 2022, "rel_sent": "It is composed of a multi - stream transformer language model ( MS - TLM ) of speech , represented as discovered unit and prosodic feature streams , and an adapted HiFi - GAN model converting MS - TLM outputs to waveforms .", "forward": true, "src_ids": "2022.acl-long.593_4661"}
+{"input": "natural , meaningful , and coherent speech is done by using Method| context: speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored . generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units . unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech .", "entity": "natural , meaningful , and coherent speech", "output": "prosody - aware generative spoken language model", "neg_sample": ["natural , meaningful , and coherent speech is done by using Method", "speech pre - training has primarily demonstrated efficacy on classification tasks , while its capability of generating novel speech , similar to how gpt-2 can generate coherent paragraphs , has barely been explored .", "generative spoken language modeling ( gslm ) ( citation ) is the only prior work addressing the generative aspect of speech pre - training , which builds a text - free language model using discovered units .", "unfortunately , because the units used in gslm discard most prosodic information , gslm fails to leverage prosody for better comprehension and does not generate expressive speech ."], "relation": "used for", "id": "2022.acl-long.593", "year": 2022, "rel_sent": "Experimental results show that the pGSLM can utilize prosody to improve both prosody and content modeling , and also generate natural , meaningful , and coherent speech given a spoken prompt .", "forward": false, "src_ids": "2022.acl-long.593_4662"}
+{"input": "language education is done by using Task| context: more than 43 % of the languages spoken in the world are endangered , and language loss currently occurs at an accelerated rate because of globalization and neocolonialism . saving and revitalizing endangered languages has become very important for maintaining the cultural diversity on our planet .", "entity": "language education", "output": "nlp", "neg_sample": ["language education is done by using Task", "more than 43 % of the languages spoken in the world are endangered , and language loss currently occurs at an accelerated rate because of globalization and neocolonialism .", "saving and revitalizing endangered languages has become very important for maintaining the cultural diversity on our planet ."], "relation": "used for", "id": "2022.acl-long.108", "year": 2022, "rel_sent": "We first suggest three principles that may help NLP practitioners tofoster mutual understanding and collaboration with language communities , and we discuss three ways in which NLP can potentially assist in language education .", "forward": false, "src_ids": "2022.acl-long.108_4663"}
+{"input": "nlp is used for Task| context: more than 43 % of the languages spoken in the world are endangered , and language loss currently occurs at an accelerated rate because of globalization and neocolonialism . saving and revitalizing endangered languages has become very important for maintaining the cultural diversity on our planet .", "entity": "nlp", "output": "language education", "neg_sample": ["nlp is used for Task", "more than 43 % of the languages spoken in the world are endangered , and language loss currently occurs at an accelerated rate because of globalization and neocolonialism .", "saving and revitalizing endangered languages has become very important for maintaining the cultural diversity on our planet ."], "relation": "used for", "id": "2022.acl-long.108", "year": 2022, "rel_sent": "We first suggest three principles that may help NLP practitioners tofoster mutual understanding and collaboration with language communities , and we discuss three ways in which NLP can potentially assist in language education .", "forward": true, "src_ids": "2022.acl-long.108_4664"}
+{"input": "explainable systems is used for Task| context: the healthcare domain suffers from the spread of poor quality articles on the internet . while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation . such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions .", "entity": "explainable systems", "output": "explanation generation", "neg_sample": ["explainable systems is used for Task", "the healthcare domain suffers from the spread of poor quality articles on the internet .", "while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation .", "such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions ."], "relation": "used for", "id": "2022.bionlp-1.1", "year": 2022, "rel_sent": "While current explainable systems tackle explanation generation as summarization , we propose a new approach based on question answering ( QA ) that allows us to generate explanations for multiple criteria using a single model .", "forward": true, "src_ids": "2022.bionlp-1.1_4665"}
+{"input": "summarization - based models is used for Task| context: the healthcare domain suffers from the spread of poor quality articles on the internet . while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation . such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions .", "entity": "summarization - based models", "output": "explainable quality assessment", "neg_sample": ["summarization - based models is used for Task", "the healthcare domain suffers from the spread of poor quality articles on the internet .", "while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation .", "such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions ."], "relation": "used for", "id": "2022.bionlp-1.1", "year": 2022, "rel_sent": "We show that this QA - based approach is competitive with the current state - of - the - art , and complements summarization - based models for explainable quality assessment .", "forward": true, "src_ids": "2022.bionlp-1.1_4666"}
+{"input": "explainable quality assessment is done by using Method| context: the healthcare domain suffers from the spread of poor quality articles on the internet . while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation . such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions .", "entity": "explainable quality assessment", "output": "summarization - based models", "neg_sample": ["explainable quality assessment is done by using Method", "the healthcare domain suffers from the spread of poor quality articles on the internet .", "while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation .", "such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions ."], "relation": "used for", "id": "2022.bionlp-1.1", "year": 2022, "rel_sent": "We show that this QA - based approach is competitive with the current state - of - the - art , and complements summarization - based models for explainable quality assessment .", "forward": false, "src_ids": "2022.bionlp-1.1_4667"}
+{"input": "explanation generation is done by using Method| context: the healthcare domain suffers from the spread of poor quality articles on the internet . while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation . such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions .", "entity": "explanation generation", "output": "explainable systems", "neg_sample": ["explanation generation is done by using Method", "the healthcare domain suffers from the spread of poor quality articles on the internet .", "while manual efforts exist to check the quality of online healthcare articles , they are not sufficient to assess all those in circulation .", "such quality assessment can be automated as a text classification task , however , explanations for the labels are necessary for the users to trust the model predictions ."], "relation": "used for", "id": "2022.bionlp-1.1", "year": 2022, "rel_sent": "While current explainable systems tackle explanation generation as summarization , we propose a new approach based on question answering ( QA ) that allows us to generate explanations for multiple criteria using a single model .", "forward": false, "src_ids": "2022.bionlp-1.1_4668"}
+{"input": "claim entailment is done by using Material| context: the dataset includes claims ( from speeches , interviews , social media and news articles ) , review articles published by professional fact checkers and premise articles used by those professional fact checkers to support their review and verify the veracity of the claims .", "entity": "claim entailment", "output": "watclaimcheck", "neg_sample": ["claim entailment is done by using Material", "the dataset includes claims ( from speeches , interviews , social media and news articles ) , review articles published by professional fact checkers and premise articles used by those professional fact checkers to support their review and verify the veracity of the claims ."], "relation": "used for", "id": "2022.acl-long.92", "year": 2022, "rel_sent": "WatClaimCheck : A new Dataset for Claim Entailment and Inference.", "forward": false, "src_ids": "2022.acl-long.92_4669"}
+{"input": "watclaimcheck is used for Task| context: the dataset includes claims ( from speeches , interviews , social media and news articles ) , review articles published by professional fact checkers and premise articles used by those professional fact checkers to support their review and verify the veracity of the claims .", "entity": "watclaimcheck", "output": "claim entailment", "neg_sample": ["watclaimcheck is used for Task", "the dataset includes claims ( from speeches , interviews , social media and news articles ) , review articles published by professional fact checkers and premise articles used by those professional fact checkers to support their review and verify the veracity of the claims ."], "relation": "used for", "id": "2022.acl-long.92", "year": 2022, "rel_sent": "WatClaimCheck : A new Dataset for Claim Entailment and Inference.", "forward": true, "src_ids": "2022.acl-long.92_4670"}
+{"input": "detecting homophobia is done by using Method| context: online social networks are ubiquitous and user - friendly . nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy . however , mining social media texts is a complex task since users do n't adhere to any fixed patterns .", "entity": "detecting homophobia", "output": "pretrained language models", "neg_sample": ["detecting homophobia is done by using Method", "online social networks are ubiquitous and user - friendly .", "nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy .", "however , mining social media texts is a complex task since users do n't adhere to any fixed patterns ."], "relation": "used for", "id": "2022.ltedi-1.18", "year": 2022, "rel_sent": "bitsa_nlp@LT - EDI - ACL2022 : Leveraging Pretrained Language Models for Detecting Homophobia and Transphobia in Social Media Comments.", "forward": false, "src_ids": "2022.ltedi-1.18_4671"}
+{"input": "pretrained language models is used for Task| context: online social networks are ubiquitous and user - friendly . nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy . however , mining social media texts is a complex task since users do n't adhere to any fixed patterns .", "entity": "pretrained language models", "output": "detecting homophobia", "neg_sample": ["pretrained language models is used for Task", "online social networks are ubiquitous and user - friendly .", "nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy .", "however , mining social media texts is a complex task since users do n't adhere to any fixed patterns ."], "relation": "used for", "id": "2022.ltedi-1.18", "year": 2022, "rel_sent": "bitsa_nlp@LT - EDI - ACL2022 : Leveraging Pretrained Language Models for Detecting Homophobia and Transphobia in Social Media Comments.", "forward": true, "src_ids": "2022.ltedi-1.18_4672"}
+{"input": "lt - edi shared task is used for Task| context: online social networks are ubiquitous and user - friendly . nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy . however , mining social media texts is a complex task since users do n't adhere to any fixed patterns .", "entity": "lt - edi shared task", "output": "detecting homophobia", "neg_sample": ["lt - edi shared task is used for Task", "online social networks are ubiquitous and user - friendly .", "nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy .", "however , mining social media texts is a complex task since users do n't adhere to any fixed patterns ."], "relation": "used for", "id": "2022.ltedi-1.18", "year": 2022, "rel_sent": "Comments can be written in any combination of languages and many of them may be low - resource . In this paper , we present our system for the LT - EDI shared task on detecting homophobia and transphobia in social media comments .", "forward": true, "src_ids": "2022.ltedi-1.18_4673"}
+{"input": "detecting homophobia is done by using Task| context: online social networks are ubiquitous and user - friendly . nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy . however , mining social media texts is a complex task since users do n't adhere to any fixed patterns .", "entity": "detecting homophobia", "output": "lt - edi shared task", "neg_sample": ["detecting homophobia is done by using Task", "online social networks are ubiquitous and user - friendly .", "nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy .", "however , mining social media texts is a complex task since users do n't adhere to any fixed patterns ."], "relation": "used for", "id": "2022.ltedi-1.18", "year": 2022, "rel_sent": "Comments can be written in any combination of languages and many of them may be low - resource . In this paper , we present our system for the LT - EDI shared task on detecting homophobia and transphobia in social media comments .", "forward": false, "src_ids": "2022.ltedi-1.18_4674"}
+{"input": "class imbalance is done by using Method| context: online social networks are ubiquitous and user - friendly . nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy . however , mining social media texts is a complex task since users do n't adhere to any fixed patterns .", "entity": "class imbalance", "output": "data augmentation technique", "neg_sample": ["class imbalance is done by using Method", "online social networks are ubiquitous and user - friendly .", "nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy .", "however , mining social media texts is a complex task since users do n't adhere to any fixed patterns ."], "relation": "used for", "id": "2022.ltedi-1.18", "year": 2022, "rel_sent": "We experiment with a number of monolingual and multilingual transformer based models such as mBERT along with a data augmentation technique for tackling class imbalance .", "forward": false, "src_ids": "2022.ltedi-1.18_4675"}
+{"input": "data augmentation technique is used for OtherScientificTerm| context: online social networks are ubiquitous and user - friendly . nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy . however , mining social media texts is a complex task since users do n't adhere to any fixed patterns .", "entity": "data augmentation technique", "output": "class imbalance", "neg_sample": ["data augmentation technique is used for OtherScientificTerm", "online social networks are ubiquitous and user - friendly .", "nevertheless , it is vital to detect and moderate offensive content to maintain decency and empathy .", "however , mining social media texts is a complex task since users do n't adhere to any fixed patterns ."], "relation": "used for", "id": "2022.ltedi-1.18", "year": 2022, "rel_sent": "We experiment with a number of monolingual and multilingual transformer based models such as mBERT along with a data augmentation technique for tackling class imbalance .", "forward": true, "src_ids": "2022.ltedi-1.18_4676"}
+{"input": "twitter stance detection is done by using OtherScientificTerm| context: research in stance detection has sofar focused on models which leverage purely textual input .", "entity": "twitter stance detection", "output": "stock market signals", "neg_sample": ["twitter stance detection is done by using OtherScientificTerm", "research in stance detection has sofar focused on models which leverage purely textual input ."], "relation": "used for", "id": "2022.acl-long.281", "year": 2022, "rel_sent": "Incorporating Stock Market Signals for Twitter Stance Detection.", "forward": false, "src_ids": "2022.acl-long.281_4677"}
+{"input": "stock market signals is used for Task| context: research in stance detection has sofar focused on models which leverage purely textual input .", "entity": "stock market signals", "output": "twitter stance detection", "neg_sample": ["stock market signals is used for Task", "research in stance detection has sofar focused on models which leverage purely textual input ."], "relation": "used for", "id": "2022.acl-long.281", "year": 2022, "rel_sent": "Incorporating Stock Market Signals for Twitter Stance Detection.", "forward": true, "src_ids": "2022.acl-long.281_4678"}
+{"input": "stance detection is done by using OtherScientificTerm| context: research in stance detection has sofar focused on models which leverage purely textual input .", "entity": "stance detection", "output": "textual and financial signals", "neg_sample": ["stance detection is done by using OtherScientificTerm", "research in stance detection has sofar focused on models which leverage purely textual input ."], "relation": "used for", "id": "2022.acl-long.281", "year": 2022, "rel_sent": "In this paper , we investigate the integration of textual and financial signals for stance detection in the financial domain .", "forward": false, "src_ids": "2022.acl-long.281_4679"}
+{"input": "measuring safety is done by using Method| context: the social impact of natural language processing and its applications has received increasing attention .", "entity": "measuring safety", "output": "safetykit", "neg_sample": ["measuring safety is done by using Method", "the social impact of natural language processing and its applications has received increasing attention ."], "relation": "used for", "id": "2022.acl-long.284", "year": 2022, "rel_sent": "SafetyKit : First Aid for Measuring Safety in Open - domain Conversational Systems.", "forward": false, "src_ids": "2022.acl-long.284_4680"}
+{"input": "safetykit is used for Task| context: the social impact of natural language processing and its applications has received increasing attention .", "entity": "safetykit", "output": "measuring safety", "neg_sample": ["safetykit is used for Task", "the social impact of natural language processing and its applications has received increasing attention ."], "relation": "used for", "id": "2022.acl-long.284", "year": 2022, "rel_sent": "SafetyKit : First Aid for Measuring Safety in Open - domain Conversational Systems.", "forward": true, "src_ids": "2022.acl-long.284_4681"}
+{"input": "nepal bhasa complement clauses is done by using Task| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "nepal bhasa complement clauses", "output": "shallow parsing", "neg_sample": ["nepal bhasa complement clauses is done by using Task", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "Shallow Parsing for Nepal Bhasa Complement Clauses.", "forward": false, "src_ids": "2022.computel-1.8_4682"}
+{"input": "endangered languages is done by using Task| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "endangered languages", "output": "shallow parsing", "neg_sample": ["endangered languages is done by using Task", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "The practice is easy to duplicate for training a shallow parser for other endangered languages in general .", "forward": false, "src_ids": "2022.computel-1.8_4683"}
+{"input": "shallow parsing is used for Material| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "shallow parsing", "output": "nepal bhasa complement clauses", "neg_sample": ["shallow parsing is used for Material", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "Shallow Parsing for Nepal Bhasa Complement Clauses.", "forward": true, "src_ids": "2022.computel-1.8_4684"}
+{"input": "raw data set is done by using OtherScientificTerm| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "raw data set", "output": "native speaker language consultants", "neg_sample": ["raw data set is done by using OtherScientificTerm", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "Native speaker language consultants were trained to annotate a minimally selected raw data set ( Suarez et al . ,2019 ) .", "forward": false, "src_ids": "2022.computel-1.8_4685"}
+{"input": "native speaker language consultants is used for Material| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "native speaker language consultants", "output": "raw data set", "neg_sample": ["native speaker language consultants is used for Material", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "Native speaker language consultants were trained to annotate a minimally selected raw data set ( Suarez et al . ,2019 ) .", "forward": true, "src_ids": "2022.computel-1.8_4686"}
+{"input": "training is done by using Method| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "training", "output": "statistical training algorithms", "neg_sample": ["training is done by using Method", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "We apply both statistical training algorithms and transfer learning in our training , including Naive Bayes , MaxEnt , and fine - tuning the pre - trained mBERT model ( Devlin et al . , 2018 ) .", "forward": false, "src_ids": "2022.computel-1.8_4687"}
+{"input": "transfer learning is used for Task| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "transfer learning", "output": "training", "neg_sample": ["transfer learning is used for Task", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "We apply both statistical training algorithms and transfer learning in our training , including Naive Bayes , MaxEnt , and fine - tuning the pre - trained mBERT model ( Devlin et al . , 2018 ) .", "forward": true, "src_ids": "2022.computel-1.8_4688"}
+{"input": "statistical training algorithms is used for Task| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "statistical training algorithms", "output": "training", "neg_sample": ["statistical training algorithms is used for Task", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "We apply both statistical training algorithms and transfer learning in our training , including Naive Bayes , MaxEnt , and fine - tuning the pre - trained mBERT model ( Devlin et al . , 2018 ) .", "forward": true, "src_ids": "2022.computel-1.8_4689"}
+{"input": "mbert model is done by using Method| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "mbert model", "output": "fine - tuning", "neg_sample": ["mbert model is done by using Method", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "We apply both statistical training algorithms and transfer learning in our training , including Naive Bayes , MaxEnt , and fine - tuning the pre - trained mBERT model ( Devlin et al . , 2018 ) .", "forward": false, "src_ids": "2022.computel-1.8_4690"}
+{"input": "fine - tuning is used for Method| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "fine - tuning", "output": "mbert model", "neg_sample": ["fine - tuning is used for Method", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "We apply both statistical training algorithms and transfer learning in our training , including Naive Bayes , MaxEnt , and fine - tuning the pre - trained mBERT model ( Devlin et al . , 2018 ) .", "forward": true, "src_ids": "2022.computel-1.8_4691"}
+{"input": "endangered languages is done by using Material| context: accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) .", "entity": "endangered languages", "output": "modeling resources", "neg_sample": ["endangered languages is done by using Material", "accelerating the process of data collection , annotation , and analysis is an urgent need for linguistic fieldwork and documentation of endangered languages ( bird , 2009 ) ."], "relation": "used for", "id": "2022.computel-1.8", "year": 2022, "rel_sent": "The modeling resources we used are largely available for many other endangered languages .", "forward": false, "src_ids": "2022.computel-1.8_4692"}
+{"input": "self - disclosure is done by using OtherScientificTerm| context: being able to reliably estimate self - disclosure - a key component of friendship and intimacy - from language is important for many psychology studies . we build single - task models on five self - disclosure corpora , but find that these models generalize poorly ; the within - domain accuracy of predicted message - level self - disclosure of the best - performing model ( mean pearson 's r=0.69 ) is much higher than the respective across data set accuracy ( mean pearson 's r=0.32 ) , due to both variations in the corpora ( e.g. , medical vs. general topics ) and labeling instructions ( target variables : self - disclosure , emotional disclosure , intimacy ) .", "entity": "self - disclosure", "output": "first person personal pronouns", "neg_sample": ["self - disclosure is done by using OtherScientificTerm", "being able to reliably estimate self - disclosure - a key component of friendship and intimacy - from language is important for many psychology studies .", "we build single - task models on five self - disclosure corpora , but find that these models generalize poorly ; the within - domain accuracy of predicted message - level self - disclosure of the best - performing model ( mean pearson 's r=0.69 ) is much higher than the respective across data set accuracy ( mean pearson 's r=0.32 ) , due to both variations in the corpora ( e.g.", ", medical vs. general topics ) and labeling instructions ( target variables : self - disclosure , emotional disclosure , intimacy ) ."], "relation": "used for", "id": "2022.findings-acl.83", "year": 2022, "rel_sent": "However , some lexical features , such as expression of negative emotions and use of first person personal pronouns such as ' I ' reliably predict self - disclosure across corpora .", "forward": false, "src_ids": "2022.findings-acl.83_4693"}
+{"input": "multilingual stereotypes dataset is done by using Material| context: warning : this paper contains explicit statements of offensive stereotypes which may be upsetting . much work on biases in natural language processing has addressed biases linked to the social and cultural experience of english speaking individuals in the united states .", "entity": "multilingual stereotypes dataset", "output": "us - centered crows - pairs dataset", "neg_sample": ["multilingual stereotypes dataset is done by using Material", "warning : this paper contains explicit statements of offensive stereotypes which may be upsetting .", "much work on biases in natural language processing has addressed biases linked to the social and cultural experience of english speaking individuals in the united states ."], "relation": "used for", "id": "2022.acl-long.583", "year": 2022, "rel_sent": "We build on the US - centered CrowS - pairs dataset to create a multilingual stereotypes dataset that allows for comparability across languages while also characterizing biases that are specific to each country and language .", "forward": false, "src_ids": "2022.acl-long.583_4694"}
+{"input": "us - centered crows - pairs dataset is used for Material| context: warning : this paper contains explicit statements of offensive stereotypes which may be upsetting . much work on biases in natural language processing has addressed biases linked to the social and cultural experience of english speaking individuals in the united states .", "entity": "us - centered crows - pairs dataset", "output": "multilingual stereotypes dataset", "neg_sample": ["us - centered crows - pairs dataset is used for Material", "warning : this paper contains explicit statements of offensive stereotypes which may be upsetting .", "much work on biases in natural language processing has addressed biases linked to the social and cultural experience of english speaking individuals in the united states ."], "relation": "used for", "id": "2022.acl-long.583", "year": 2022, "rel_sent": "We build on the US - centered CrowS - pairs dataset to create a multilingual stereotypes dataset that allows for comparability across languages while also characterizing biases that are specific to each country and language .", "forward": true, "src_ids": "2022.acl-long.583_4695"}
+{"input": "text style transfer is done by using Metric| context: text style transfer and paraphrasing of texts are actively growing areas of nlp , dozens of methods for solving these tasks have been recently introduced . in both tasks , the system is supposed to generate a text which should be semantically similar to the input text . therefore , these tasks are dependent on methods of measuring textual semantic similarity . however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text . according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches . the current problem is the lack of a thorough evaluation of the available measures .", "entity": "text style transfer", "output": "content preservation measures", "neg_sample": ["text style transfer is done by using Metric", "text style transfer and paraphrasing of texts are actively growing areas of nlp , dozens of methods for solving these tasks have been recently introduced .", "in both tasks , the system is supposed to generate a text which should be semantically similar to the input text .", "therefore , these tasks are dependent on methods of measuring textual semantic similarity .", "however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text .", "according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches .", "the current problem is the lack of a thorough evaluation of the available measures ."], "relation": "used for", "id": "2022.acl-srw.23", "year": 2022, "rel_sent": "A large - scale computational study of content preservation measures for text style transfer and paraphrase generation.", "forward": false, "src_ids": "2022.acl-srw.23_4696"}
+{"input": "paraphrase generation is done by using Metric| context: text style transfer and paraphrasing of texts are actively growing areas of nlp , dozens of methods for solving these tasks have been recently introduced . in both tasks , the system is supposed to generate a text which should be semantically similar to the input text . therefore , these tasks are dependent on methods of measuring textual semantic similarity . however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text . according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches . the current problem is the lack of a thorough evaluation of the available measures .", "entity": "paraphrase generation", "output": "content preservation measures", "neg_sample": ["paraphrase generation is done by using Metric", "text style transfer and paraphrasing of texts are actively growing areas of nlp , dozens of methods for solving these tasks have been recently introduced .", "in both tasks , the system is supposed to generate a text which should be semantically similar to the input text .", "therefore , these tasks are dependent on methods of measuring textual semantic similarity .", "however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text .", "according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches .", "the current problem is the lack of a thorough evaluation of the available measures ."], "relation": "used for", "id": "2022.acl-srw.23", "year": 2022, "rel_sent": "A large - scale computational study of content preservation measures for text style transfer and paraphrase generation.", "forward": false, "src_ids": "2022.acl-srw.23_4697"}
+{"input": "content preservation measures is used for Task| context: in both tasks , the system is supposed to generate a text which should be semantically similar to the input text . therefore , these tasks are dependent on methods of measuring textual semantic similarity . however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text . according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches . the current problem is the lack of a thorough evaluation of the available measures .", "entity": "content preservation measures", "output": "text style transfer", "neg_sample": ["content preservation measures is used for Task", "in both tasks , the system is supposed to generate a text which should be semantically similar to the input text .", "therefore , these tasks are dependent on methods of measuring textual semantic similarity .", "however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text .", "according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches .", "the current problem is the lack of a thorough evaluation of the available measures ."], "relation": "used for", "id": "2022.acl-srw.23", "year": 2022, "rel_sent": "A large - scale computational study of content preservation measures for text style transfer and paraphrase generation.", "forward": true, "src_ids": "2022.acl-srw.23_4698"}
+{"input": "content preservation measures is used for Task| context: text style transfer and paraphrasing of texts are actively growing areas of nlp , dozens of methods for solving these tasks have been recently introduced . in both tasks , the system is supposed to generate a text which should be semantically similar to the input text . therefore , these tasks are dependent on methods of measuring textual semantic similarity . however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text . according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches . the current problem is the lack of a thorough evaluation of the available measures .", "entity": "content preservation measures", "output": "paraphrase generation", "neg_sample": ["content preservation measures is used for Task", "text style transfer and paraphrasing of texts are actively growing areas of nlp , dozens of methods for solving these tasks have been recently introduced .", "in both tasks , the system is supposed to generate a text which should be semantically similar to the input text .", "therefore , these tasks are dependent on methods of measuring textual semantic similarity .", "however , it is still unclear which measures are the best to automatically evaluate content preservation between original and generated text .", "according to our observations , many researchers still use bleu - like measures , while there exist more advanced measures including neural - based that significantly outperform classic approaches .", "the current problem is the lack of a thorough evaluation of the available measures ."], "relation": "used for", "id": "2022.acl-srw.23", "year": 2022, "rel_sent": "A large - scale computational study of content preservation measures for text style transfer and paraphrase generation.", "forward": true, "src_ids": "2022.acl-srw.23_4699"}
+{"input": "neural machine translation is done by using Method| context: deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) . transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure .", "entity": "neural machine translation", "output": "deep decoder", "neg_sample": ["neural machine translation is done by using Method", "deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) .", "transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure ."], "relation": "used for", "id": "2022.findings-acl.39", "year": 2022, "rel_sent": "Inspired by this discovery , we then propose approaches to improving it , with respect to model structure and model training , to make the deep decoder practical in NMT .", "forward": false, "src_ids": "2022.findings-acl.39_4700"}
+{"input": "deep decoder is used for Task| context: transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure .", "entity": "deep decoder", "output": "neural machine translation", "neg_sample": ["deep decoder is used for Task", "transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure ."], "relation": "used for", "id": "2022.findings-acl.39", "year": 2022, "rel_sent": "Inspired by this discovery , we then propose approaches to improving it , with respect to model structure and model training , to make the deep decoder practical in NMT .", "forward": true, "src_ids": "2022.findings-acl.39_4701"}
+{"input": "deep - decoder training is done by using Method| context: deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) . transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure .", "entity": "deep - decoder training", "output": "cross - attention drop mechanism", "neg_sample": ["deep - decoder training is done by using Method", "deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) .", "transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure ."], "relation": "used for", "id": "2022.findings-acl.39", "year": 2022, "rel_sent": "Specifically , with respect to model structure , we propose a cross - attention drop mechanism to allow the decoder layers to perform their own different roles , to reduce the difficulty of deep - decoder learning .", "forward": false, "src_ids": "2022.findings-acl.39_4702"}
+{"input": "cross - attention drop mechanism is used for Method| context: deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) . transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure .", "entity": "cross - attention drop mechanism", "output": "deep - decoder training", "neg_sample": ["cross - attention drop mechanism is used for Method", "deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) .", "transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure ."], "relation": "used for", "id": "2022.findings-acl.39", "year": 2022, "rel_sent": "Specifically , with respect to model structure , we propose a cross - attention drop mechanism to allow the decoder layers to perform their own different roles , to reduce the difficulty of deep - decoder learning .", "forward": true, "src_ids": "2022.findings-acl.39_4703"}
+{"input": "collapse reducing training approach is used for Method| context: deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) . transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure .", "entity": "collapse reducing training approach", "output": "deep - decoder training", "neg_sample": ["collapse reducing training approach is used for Method", "deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) .", "transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure ."], "relation": "used for", "id": "2022.findings-acl.39", "year": 2022, "rel_sent": "For model training , we propose a collapse reducing training approach to improve the stability and effectiveness of deep - decoder training .", "forward": true, "src_ids": "2022.findings-acl.39_4704"}
+{"input": "deep - decoder training is done by using Method| context: deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) . transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure .", "entity": "deep - decoder training", "output": "collapse reducing training approach", "neg_sample": ["deep - decoder training is done by using Method", "deep learning has demonstrated performance advantages in a wide range of natural language processing tasks , including neural machine translation ( nmt ) .", "transformer nmt models are typically strengthened by deeper encoder layers , but deepening their decoder layers usually results in failure ."], "relation": "used for", "id": "2022.findings-acl.39", "year": 2022, "rel_sent": "For model training , we propose a collapse reducing training approach to improve the stability and effectiveness of deep - decoder training .", "forward": false, "src_ids": "2022.findings-acl.39_4705"}
+{"input": "ml model is used for Metric| context: although boguslav & cohen ( 2017 ) showed that this claim is falsified by many real - world ml systems , the claim has persisted .", "entity": "ml model", "output": "inter - annotator agreement ( iaa )", "neg_sample": ["ml model is used for Metric", "although boguslav & cohen ( 2017 ) showed that this claim is falsified by many real - world ml systems , the claim has persisted ."], "relation": "used for", "id": "2022.bionlp-1.26", "year": 2022, "rel_sent": "As a complement to this real - world evidence , we conducted a comprehensive set of simulations , and show that an ML model can beat IAA even if ( and especially if ) annotators are noisy and differ in their underlying classification functions , as long as the ML model is reasonably well - specified .", "forward": true, "src_ids": "2022.bionlp-1.26_4706"}
+{"input": "inter - annotator agreement ( iaa ) is done by using Method| context: it is commonly claimed that inter - annotator agreement ( iaa ) is the ceiling of machine learning ( ml ) performance , i.e. , that the agreement between an ml system 's predictions and an annotator can not be higher than the agreement between two annotators . although boguslav & cohen ( 2017 ) showed that this claim is falsified by many real - world ml systems , the claim has persisted .", "entity": "inter - annotator agreement ( iaa )", "output": "ml model", "neg_sample": ["inter - annotator agreement ( iaa ) is done by using Method", "it is commonly claimed that inter - annotator agreement ( iaa ) is the ceiling of machine learning ( ml ) performance , i.e.", ", that the agreement between an ml system 's predictions and an annotator can not be higher than the agreement between two annotators .", "although boguslav & cohen ( 2017 ) showed that this claim is falsified by many real - world ml systems , the claim has persisted ."], "relation": "used for", "id": "2022.bionlp-1.26", "year": 2022, "rel_sent": "As a complement to this real - world evidence , we conducted a comprehensive set of simulations , and show that an ML model can beat IAA even if ( and especially if ) annotators are noisy and differ in their underlying classification functions , as long as the ML model is reasonably well - specified .", "forward": false, "src_ids": "2022.bionlp-1.26_4707"}
+{"input": "structured sentiment analysis is done by using Method| context: the state - of - the - art model for structured sentiment analysis casts the task as a dependency parsing problem , which has some limitations : ( 1 ) the label proportions for span prediction and span relation prediction are imbalanced . ( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem . ( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized .", "entity": "structured sentiment analysis", "output": "token graph modeling", "neg_sample": ["structured sentiment analysis is done by using Method", "the state - of - the - art model for structured sentiment analysis casts the task as a dependency parsing problem , which has some limitations : ( 1 ) the label proportions for span prediction and span relation prediction are imbalanced .", "( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem .", "( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized ."], "relation": "used for", "id": "2022.acl-long.291", "year": 2022, "rel_sent": "Effective Token Graph Modeling using a Novel Labeling Strategy for Structured Sentiment Analysis.", "forward": false, "src_ids": "2022.acl-long.291_4708"}
+{"input": "labeling strategy is used for Task| context: ( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem . ( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized .", "entity": "labeling strategy", "output": "structured sentiment analysis", "neg_sample": ["labeling strategy is used for Task", "( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem .", "( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized ."], "relation": "used for", "id": "2022.acl-long.291", "year": 2022, "rel_sent": "Effective Token Graph Modeling using a Novel Labeling Strategy for Structured Sentiment Analysis.", "forward": true, "src_ids": "2022.acl-long.291_4709"}
+{"input": "token graph modeling is used for Task| context: ( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem . ( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized .", "entity": "token graph modeling", "output": "structured sentiment analysis", "neg_sample": ["token graph modeling is used for Task", "( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem .", "( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized ."], "relation": "used for", "id": "2022.acl-long.291", "year": 2022, "rel_sent": "Effective Token Graph Modeling using a Novel Labeling Strategy for Structured Sentiment Analysis.", "forward": true, "src_ids": "2022.acl-long.291_4710"}
+{"input": "token representations is done by using Method| context: the state - of - the - art model for structured sentiment analysis casts the task as a dependency parsing problem , which has some limitations : ( 1 ) the label proportions for span prediction and span relation prediction are imbalanced . ( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem . ( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized .", "entity": "token representations", "output": "graph attention networks", "neg_sample": ["token representations is done by using Method", "the state - of - the - art model for structured sentiment analysis casts the task as a dependency parsing problem , which has some limitations : ( 1 ) the label proportions for span prediction and span relation prediction are imbalanced .", "( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem .", "( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized ."], "relation": "used for", "id": "2022.acl-long.291", "year": 2022, "rel_sent": "Moreover , we also propose an effective model to well collaborate with our labeling strategy , which is equipped with the graph attention networks to iteratively refine token representations , and the adaptive multi - label classifier to dynamically predict multiple relations between token pairs .", "forward": false, "src_ids": "2022.acl-long.291_4711"}
+{"input": "graph attention networks is used for Method| context: the state - of - the - art model for structured sentiment analysis casts the task as a dependency parsing problem , which has some limitations : ( 1 ) the label proportions for span prediction and span relation prediction are imbalanced . ( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem . ( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized .", "entity": "graph attention networks", "output": "token representations", "neg_sample": ["graph attention networks is used for Method", "the state - of - the - art model for structured sentiment analysis casts the task as a dependency parsing problem , which has some limitations : ( 1 ) the label proportions for span prediction and span relation prediction are imbalanced .", "( 2 ) the span lengths of sentiment tuple components may be very large in this task , which will further exacerbates the imbalance problem .", "( 3 ) two nodes in a dependency graph can not have multiple arcs , therefore some overlapped sentiment tuples can not be recognized ."], "relation": "used for", "id": "2022.acl-long.291", "year": 2022, "rel_sent": "Moreover , we also propose an effective model to well collaborate with our labeling strategy , which is equipped with the graph attention networks to iteratively refine token representations , and the adaptive multi - label classifier to dynamically predict multiple relations between token pairs .", "forward": true, "src_ids": "2022.acl-long.291_4712"}
+{"input": "visual representations is done by using Method| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "visual representations", "output": "unified transformer model", "neg_sample": ["visual representations is done by using Method", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "We build a unified Transformer model to jointly learn visual representations , textual representations and semantic alignment between images and texts .", "forward": false, "src_ids": "2022.findings-acl.251_4713"}
+{"input": "textual representations is done by using Method| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "textual representations", "output": "unified transformer model", "neg_sample": ["textual representations is done by using Method", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "We build a unified Transformer model to jointly learn visual representations , textual representations and semantic alignment between images and texts .", "forward": false, "src_ids": "2022.findings-acl.251_4714"}
+{"input": "unified transformer model is used for Method| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "unified transformer model", "output": "visual representations", "neg_sample": ["unified transformer model is used for Method", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "We build a unified Transformer model to jointly learn visual representations , textual representations and semantic alignment between images and texts .", "forward": true, "src_ids": "2022.findings-acl.251_4715"}
+{"input": "textual and visual semantic alignment is done by using Method| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "textual and visual semantic alignment", "output": "grounded learning method", "neg_sample": ["textual and visual semantic alignment is done by using Method", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "The experiments show that our grounded learning method can improve textual and visual semantic alignment for improving performance on various cross - modal tasks .", "forward": false, "src_ids": "2022.findings-acl.251_4716"}
+{"input": "unaligned images and texts is done by using OtherScientificTerm| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "unaligned images and texts", "output": "sharing grounded space", "neg_sample": ["unaligned images and texts is done by using OtherScientificTerm", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "In particular , we propose to conduct grounded learning on both images and texts via a sharing grounded space , which helps bridge unaligned images and texts , and align the visual and textual semantic spaces on different types of corpora .", "forward": false, "src_ids": "2022.findings-acl.251_4717"}
+{"input": "sharing grounded space is used for Material| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "sharing grounded space", "output": "unaligned images and texts", "neg_sample": ["sharing grounded space is used for Material", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "In particular , we propose to conduct grounded learning on both images and texts via a sharing grounded space , which helps bridge unaligned images and texts , and align the visual and textual semantic spaces on different types of corpora .", "forward": true, "src_ids": "2022.findings-acl.251_4718"}
+{"input": "grounded learning method is used for Task| context: vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks . however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance .", "entity": "grounded learning method", "output": "textual and visual semantic alignment", "neg_sample": ["grounded learning method is used for Task", "vision - language pre - training ( vlp ) has achieved impressive performance on various cross - modal downstream tasks .", "however , most existing methods can only learn from aligned image - caption data and rely heavily on expensive regional features , which greatly limits their scalability and performance ."], "relation": "used for", "id": "2022.findings-acl.251", "year": 2022, "rel_sent": "The experiments show that our grounded learning method can improve textual and visual semantic alignment for improving performance on various cross - modal tasks .", "forward": true, "src_ids": "2022.findings-acl.251_4719"}
+{"input": "multilingual kb construction and completion is done by using Method| context: knowledge bases ( kbs ) contain plenty of structured world and commonsense knowledge . as such , they often complement distributional text - based information and facilitate various downstream tasks . since their manual construction is resource- and time - intensive , recent efforts have tried leveraging large pretrained language models ( plms ) to generate additional monolingual knowledge facts for kbs . however , such methods have not been attempted for building and enriching multilingual kbs . besides wider application , such multilingual kbs can provide richer combined knowledge than monolingual ( e.g. , english ) kbs . knowledge expressed in different languages may be complementary and unequally distributed : this implies that the knowledge available in high - resource languages can be transferred to low - resource ones . to achieve this , it is crucial to represent multilingual knowledge in a shared / unified space .", "entity": "multilingual kb construction and completion", "output": "prix - lm", "neg_sample": ["multilingual kb construction and completion is done by using Method", "knowledge bases ( kbs ) contain plenty of structured world and commonsense knowledge .", "as such , they often complement distributional text - based information and facilitate various downstream tasks .", "since their manual construction is resource- and time - intensive , recent efforts have tried leveraging large pretrained language models ( plms ) to generate additional monolingual knowledge facts for kbs .", "however , such methods have not been attempted for building and enriching multilingual kbs .", "besides wider application , such multilingual kbs can provide richer combined knowledge than monolingual ( e.g.", ", english ) kbs .", "knowledge expressed in different languages may be complementary and unequally distributed : this implies that the knowledge available in high - resource languages can be transferred to low - resource ones .", "to achieve this , it is crucial to represent multilingual knowledge in a shared / unified space ."], "relation": "used for", "id": "2022.acl-long.371", "year": 2022, "rel_sent": "To this end , we propose a unified representation model , Prix - LM , for multilingual KB construction and completion .", "forward": false, "src_ids": "2022.acl-long.371_4720"}
+{"input": "unified representation model is used for Task| context: knowledge bases ( kbs ) contain plenty of structured world and commonsense knowledge . as such , they often complement distributional text - based information and facilitate various downstream tasks . since their manual construction is resource- and time - intensive , recent efforts have tried leveraging large pretrained language models ( plms ) to generate additional monolingual knowledge facts for kbs . however , such methods have not been attempted for building and enriching multilingual kbs . besides wider application , such multilingual kbs can provide richer combined knowledge than monolingual ( e.g. , english ) kbs . knowledge expressed in different languages may be complementary and unequally distributed : this implies that the knowledge available in high - resource languages can be transferred to low - resource ones . to achieve this , it is crucial to represent multilingual knowledge in a shared / unified space .", "entity": "unified representation model", "output": "multilingual kb construction and completion", "neg_sample": ["unified representation model is used for Task", "knowledge bases ( kbs ) contain plenty of structured world and commonsense knowledge .", "as such , they often complement distributional text - based information and facilitate various downstream tasks .", "since their manual construction is resource- and time - intensive , recent efforts have tried leveraging large pretrained language models ( plms ) to generate additional monolingual knowledge facts for kbs .", "however , such methods have not been attempted for building and enriching multilingual kbs .", "besides wider application , such multilingual kbs can provide richer combined knowledge than monolingual ( e.g.", ", english ) kbs .", "knowledge expressed in different languages may be complementary and unequally distributed : this implies that the knowledge available in high - resource languages can be transferred to low - resource ones .", "to achieve this , it is crucial to represent multilingual knowledge in a shared / unified space ."], "relation": "used for", "id": "2022.acl-long.371", "year": 2022, "rel_sent": "To this end , we propose a unified representation model , Prix - LM , for multilingual KB construction and completion .", "forward": true, "src_ids": "2022.acl-long.371_4721"}
+{"input": "prix - lm is used for Task| context: knowledge bases ( kbs ) contain plenty of structured world and commonsense knowledge . as such , they often complement distributional text - based information and facilitate various downstream tasks . since their manual construction is resource- and time - intensive , recent efforts have tried leveraging large pretrained language models ( plms ) to generate additional monolingual knowledge facts for kbs . however , such methods have not been attempted for building and enriching multilingual kbs . besides wider application , such multilingual kbs can provide richer combined knowledge than monolingual ( e.g. , english ) kbs . knowledge expressed in different languages may be complementary and unequally distributed : this implies that the knowledge available in high - resource languages can be transferred to low - resource ones . to achieve this , it is crucial to represent multilingual knowledge in a shared / unified space .", "entity": "prix - lm", "output": "multilingual kb construction and completion", "neg_sample": ["prix - lm is used for Task", "knowledge bases ( kbs ) contain plenty of structured world and commonsense knowledge .", "as such , they often complement distributional text - based information and facilitate various downstream tasks .", "since their manual construction is resource- and time - intensive , recent efforts have tried leveraging large pretrained language models ( plms ) to generate additional monolingual knowledge facts for kbs .", "however , such methods have not been attempted for building and enriching multilingual kbs .", "besides wider application , such multilingual kbs can provide richer combined knowledge than monolingual ( e.g.", ", english ) kbs .", "knowledge expressed in different languages may be complementary and unequally distributed : this implies that the knowledge available in high - resource languages can be transferred to low - resource ones .", "to achieve this , it is crucial to represent multilingual knowledge in a shared / unified space ."], "relation": "used for", "id": "2022.acl-long.371", "year": 2022, "rel_sent": "To this end , we propose a unified representation model , Prix - LM , for multilingual KB construction and completion .", "forward": true, "src_ids": "2022.acl-long.371_4722"}
+{"input": "fieldwork annotation is done by using Method| context: there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently .", "entity": "fieldwork annotation", "output": "word - and - paradigm workflow", "neg_sample": ["fieldwork annotation is done by using Method", "there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently ."], "relation": "used for", "id": "2022.computel-1.20", "year": 2022, "rel_sent": "A Word - and - Paradigm Workflow for Fieldwork Annotation.", "forward": false, "src_ids": "2022.computel-1.20_4723"}
+{"input": "annotation process is done by using Method| context: there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently .", "entity": "annotation process", "output": "machine learning", "neg_sample": ["annotation process is done by using Method", "there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently ."], "relation": "used for", "id": "2022.computel-1.20", "year": 2022, "rel_sent": "Machine learning has the potential to greatly accelerate the annotation process and allow a human annotator tofocus on problematic cases , while the WP approach makes for an annotation system that is word - based and relational , removing the need to make decisions about feature labeling and segmentation early in the process and allowing speakers of the language of interest to participate more actively , since linguistic training is not necessary .", "forward": false, "src_ids": "2022.computel-1.20_4724"}
+{"input": "machine learning is used for Task| context: there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently .", "entity": "machine learning", "output": "annotation process", "neg_sample": ["machine learning is used for Task", "there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently ."], "relation": "used for", "id": "2022.computel-1.20", "year": 2022, "rel_sent": "Machine learning has the potential to greatly accelerate the annotation process and allow a human annotator tofocus on problematic cases , while the WP approach makes for an annotation system that is word - based and relational , removing the need to make decisions about feature labeling and segmentation early in the process and allowing speakers of the language of interest to participate more actively , since linguistic training is not necessary .", "forward": true, "src_ids": "2022.computel-1.20_4725"}
+{"input": "annotation system is done by using Method| context: there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently .", "entity": "annotation system", "output": "wp approach", "neg_sample": ["annotation system is done by using Method", "there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently ."], "relation": "used for", "id": "2022.computel-1.20", "year": 2022, "rel_sent": "Machine learning has the potential to greatly accelerate the annotation process and allow a human annotator tofocus on problematic cases , while the WP approach makes for an annotation system that is word - based and relational , removing the need to make decisions about feature labeling and segmentation early in the process and allowing speakers of the language of interest to participate more actively , since linguistic training is not necessary .", "forward": false, "src_ids": "2022.computel-1.20_4726"}
+{"input": "wp approach is used for Method| context: there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently .", "entity": "wp approach", "output": "annotation system", "neg_sample": ["wp approach is used for Method", "there are many challenges in morphological fieldwork annotation , it heavily relies on segmentation and feature labeling ( which have both practical and theoretical drawbacks ) , it 's time - intensive , and the annotator needs to be linguistically trained and may still annotate things inconsistently ."], "relation": "used for", "id": "2022.computel-1.20", "year": 2022, "rel_sent": "Machine learning has the potential to greatly accelerate the annotation process and allow a human annotator tofocus on problematic cases , while the WP approach makes for an annotation system that is word - based and relational , removing the need to make decisions about feature labeling and segmentation early in the process and allowing speakers of the language of interest to participate more actively , since linguistic training is not necessary .", "forward": true, "src_ids": "2022.computel-1.20_4727"}
+{"input": "automated claim verification is done by using Method| context: the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information . in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community . in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity .", "entity": "automated claim verification", "output": "semantics - aware approach", "neg_sample": ["automated claim verification is done by using Method", "the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information .", "in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community .", "in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity ."], "relation": "used for", "id": "2022.fever-1.5", "year": 2022, "rel_sent": "A Semantics - Aware Approach to Automated Claim Verification.", "forward": false, "src_ids": "2022.fever-1.5_4728"}
+{"input": "semantics - aware approach is used for Task| context: the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information . in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community .", "entity": "semantics - aware approach", "output": "automated claim verification", "neg_sample": ["semantics - aware approach is used for Task", "the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information .", "in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community ."], "relation": "used for", "id": "2022.fever-1.5", "year": 2022, "rel_sent": "A Semantics - Aware Approach to Automated Claim Verification.", "forward": true, "src_ids": "2022.fever-1.5_4729"}
+{"input": "claim verification is done by using Method| context: the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information . in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community . in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity .", "entity": "claim verification", "output": "bert model", "neg_sample": ["claim verification is done by using Method", "the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information .", "in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community .", "in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity ."], "relation": "used for", "id": "2022.fever-1.5", "year": 2022, "rel_sent": "In this work we empirically demonstrate that enriching a BERT model with explicit semantic information such as Semantic Role Labelling helps to improve results in claim verification as proposed by the FEVER benchmark .", "forward": false, "src_ids": "2022.fever-1.5_4730"}
+{"input": "bert model is used for Task| context: the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information . in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community .", "entity": "bert model", "output": "claim verification", "neg_sample": ["bert model is used for Task", "the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information .", "in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community ."], "relation": "used for", "id": "2022.fever-1.5", "year": 2022, "rel_sent": "In this work we empirically demonstrate that enriching a BERT model with explicit semantic information such as Semantic Role Labelling helps to improve results in claim verification as proposed by the FEVER benchmark .", "forward": true, "src_ids": "2022.fever-1.5_4731"}
+{"input": "complex cases is done by using Method| context: the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information . in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community . in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity .", "entity": "complex cases", "output": "semantically - enriched model", "neg_sample": ["complex cases is done by using Method", "the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information .", "in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community .", "in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity ."], "relation": "used for", "id": "2022.fever-1.5", "year": 2022, "rel_sent": "Furthermore , we perform a number of explainability tests that suggest that the semantically - enriched model is better at handling complex cases , such as those including passive forms or multiple propositions .", "forward": false, "src_ids": "2022.fever-1.5_4732"}
+{"input": "semantically - enriched model is used for OtherScientificTerm| context: the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information . in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community . in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity .", "entity": "semantically - enriched model", "output": "complex cases", "neg_sample": ["semantically - enriched model is used for OtherScientificTerm", "the influence of fake news in the perception of reality has become a mainstream topic in the last years due to the fast propagation of misleading information .", "in order to help in the fight against misinformation , automated solutions tofact - checking are being actively developed within the research community .", "in this context , the task of automated claim verification is defined as assessing the truthfulness of a claim by finding evidence about its veracity ."], "relation": "used for", "id": "2022.fever-1.5", "year": 2022, "rel_sent": "Furthermore , we perform a number of explainability tests that suggest that the semantically - enriched model is better at handling complex cases , such as those including passive forms or multiple propositions .", "forward": true, "src_ids": "2022.fever-1.5_4733"}
+{"input": "emergence of edge gemination / degemination is done by using Method| context: the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue . previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints . we argue that these edge gemination processes result from the same , independently established principles .", "entity": "emergence of edge gemination / degemination", "output": "coupled oscillator model of articulatory phonology", "neg_sample": ["emergence of edge gemination / degemination is done by using Method", "the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue .", "previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints .", "we argue that these edge gemination processes result from the same , independently established principles ."], "relation": "used for", "id": "2022.scil-1.2", "year": 2022, "rel_sent": "Thus , the coupled oscillator model of Articulatory Phonology , originally developed to model intergestural timing , predicts the emergence of edge gemination / degemination .", "forward": false, "src_ids": "2022.scil-1.2_4734"}
+{"input": "intergestural timing is done by using Method| context: the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue . previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints . we argue that these edge gemination processes result from the same , independently established principles .", "entity": "intergestural timing", "output": "coupled oscillator model of articulatory phonology", "neg_sample": ["intergestural timing is done by using Method", "the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue .", "previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints .", "we argue that these edge gemination processes result from the same , independently established principles ."], "relation": "used for", "id": "2022.scil-1.2", "year": 2022, "rel_sent": "Thus , the coupled oscillator model of Articulatory Phonology , originally developed to model intergestural timing , predicts the emergence of edge gemination / degemination .", "forward": false, "src_ids": "2022.scil-1.2_4735"}
+{"input": "coupled oscillator model of articulatory phonology is used for OtherScientificTerm| context: the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue . previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints . we argue that these edge gemination processes result from the same , independently established principles .", "entity": "coupled oscillator model of articulatory phonology", "output": "intergestural timing", "neg_sample": ["coupled oscillator model of articulatory phonology is used for OtherScientificTerm", "the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue .", "previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints .", "we argue that these edge gemination processes result from the same , independently established principles ."], "relation": "used for", "id": "2022.scil-1.2", "year": 2022, "rel_sent": "Thus , the coupled oscillator model of Articulatory Phonology , originally developed to model intergestural timing , predicts the emergence of edge gemination / degemination .", "forward": true, "src_ids": "2022.scil-1.2_4736"}
+{"input": "coupled oscillator model of articulatory phonology is used for Task| context: the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue . previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints . we argue that these edge gemination processes result from the same , independently established principles .", "entity": "coupled oscillator model of articulatory phonology", "output": "emergence of edge gemination / degemination", "neg_sample": ["coupled oscillator model of articulatory phonology is used for Task", "the phonological mechanisms responsible for the emergence of edge geminates in phonological processes like the italian raddoppiamento ( fono-)sintattico ( rs ) are an open issue .", "previous analyses of italian treat gemination of ( i ) word initial consonants , ( ii ) morpheme - final consonants , and ( iii ) word final consonants as separate processes brought about by dedicated rule / constraints .", "we argue that these edge gemination processes result from the same , independently established principles ."], "relation": "used for", "id": "2022.scil-1.2", "year": 2022, "rel_sent": "Thus , the coupled oscillator model of Articulatory Phonology , originally developed to model intergestural timing , predicts the emergence of edge gemination / degemination .", "forward": true, "src_ids": "2022.scil-1.2_4737"}
+{"input": "text retrieval is done by using Method| context: text retrieval has been widely - used in many online applications to help users find relevant information from a text collection . in this paper , we study a new attack scenario against text retrieval to evaluate its robustness to adversarial attacks under the black - box setting , in which attackers want their own texts to always get high relevance scores with different users ' input queries and thus be retrieved frequently and can receive large amounts of impressions for profits .", "entity": "text retrieval", "output": "text rewriting attack", "neg_sample": ["text retrieval is done by using Method", "text retrieval has been widely - used in many online applications to help users find relevant information from a text collection .", "in this paper , we study a new attack scenario against text retrieval to evaluate its robustness to adversarial attacks under the black - box setting , in which attackers want their own texts to always get high relevance scores with different users ' input queries and thus be retrieved frequently and can receive large amounts of impressions for profits ."], "relation": "used for", "id": "2022.repl4nlp-1.20", "year": 2022, "rel_sent": "TRAttack ' : ' Text Rewriting Attack Against Text Retrieval.", "forward": false, "src_ids": "2022.repl4nlp-1.20_4738"}
+{"input": "exploration is done by using OtherScientificTerm| context: text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces . in these games , the agent learns to explore the environment via natural language interactions with the game simulator . a fundamental challenge in tgs is the efficient exploration of the large action space when the agent has not yet acquired enough knowledge about the environment .", "entity": "exploration", "output": "commonsense inductive bias", "neg_sample": ["exploration is done by using OtherScientificTerm", "text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces .", "in these games , the agent learns to explore the environment via natural language interactions with the game simulator .", "a fundamental challenge in tgs is the efficient exploration of the large action space when the agent has not yet acquired enough knowledge about the environment ."], "relation": "used for", "id": "2022.acl-short.56", "year": 2022, "rel_sent": "Fire Burns , Sword Cuts : Commonsense Inductive Bias for Exploration in Text - based Games.", "forward": false, "src_ids": "2022.acl-short.56_4739"}
+{"input": "commonsense inductive bias is used for Task| context: text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces . in these games , the agent learns to explore the environment via natural language interactions with the game simulator .", "entity": "commonsense inductive bias", "output": "exploration", "neg_sample": ["commonsense inductive bias is used for Task", "text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces .", "in these games , the agent learns to explore the environment via natural language interactions with the game simulator ."], "relation": "used for", "id": "2022.acl-short.56", "year": 2022, "rel_sent": "Fire Burns , Sword Cuts : Commonsense Inductive Bias for Exploration in Text - based Games.", "forward": true, "src_ids": "2022.acl-short.56_4740"}
+{"input": "external commonsense knowledge is done by using Method| context: text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces . in these games , the agent learns to explore the environment via natural language interactions with the game simulator . a fundamental challenge in tgs is the efficient exploration of the large action space when the agent has not yet acquired enough knowledge about the environment .", "entity": "external commonsense knowledge", "output": "exploration technique", "neg_sample": ["external commonsense knowledge is done by using Method", "text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces .", "in these games , the agent learns to explore the environment via natural language interactions with the game simulator .", "a fundamental challenge in tgs is the efficient exploration of the large action space when the agent has not yet acquired enough knowledge about the environment ."], "relation": "used for", "id": "2022.acl-short.56", "year": 2022, "rel_sent": "We propose CommExpl , an exploration technique that injects external commonsense knowledge , via a pretrained language model ( LM ) , into the agent during training when the agent is the most uncertain about its next action .", "forward": false, "src_ids": "2022.acl-short.56_4741"}
+{"input": "exploration technique is used for OtherScientificTerm| context: text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces . in these games , the agent learns to explore the environment via natural language interactions with the game simulator . a fundamental challenge in tgs is the efficient exploration of the large action space when the agent has not yet acquired enough knowledge about the environment .", "entity": "exploration technique", "output": "external commonsense knowledge", "neg_sample": ["exploration technique is used for OtherScientificTerm", "text - based games ( tgs ) are exciting testbeds for developing deep reinforcement learning techniques due to their partially observed environments and large action spaces .", "in these games , the agent learns to explore the environment via natural language interactions with the game simulator .", "a fundamental challenge in tgs is the efficient exploration of the large action space when the agent has not yet acquired enough knowledge about the environment ."], "relation": "used for", "id": "2022.acl-short.56", "year": 2022, "rel_sent": "We propose CommExpl , an exploration technique that injects external commonsense knowledge , via a pretrained language model ( LM ) , into the agent during training when the agent is the most uncertain about its next action .", "forward": true, "src_ids": "2022.acl-short.56_4742"}
+{"input": "lm model is used for OtherScientificTerm| context: recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task .", "entity": "lm model", "output": "low - level structures", "neg_sample": ["lm model is used for OtherScientificTerm", "recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task ."], "relation": "used for", "id": "2022.acl-long.444", "year": 2022, "rel_sent": "Inspired by the natural reading process of human , we propose to regularize the parser with phrases extracted by an unsupervised phrase tagger to help the LM model quickly manage low - level structures .", "forward": true, "src_ids": "2022.acl-long.444_4743"}
+{"input": "low - level structures is done by using Method| context: recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task . despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context .", "entity": "low - level structures", "output": "lm model", "neg_sample": ["low - level structures is done by using Method", "recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task .", "despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context ."], "relation": "used for", "id": "2022.acl-long.444", "year": 2022, "rel_sent": "Inspired by the natural reading process of human , we propose to regularize the parser with phrases extracted by an unsupervised phrase tagger to help the LM model quickly manage low - level structures .", "forward": false, "src_ids": "2022.acl-long.444_4744"}
+{"input": "lm is done by using Method| context: recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task . despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context .", "entity": "lm", "output": "phrase - guided masking strategy", "neg_sample": ["lm is done by using Method", "recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task .", "despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context ."], "relation": "used for", "id": "2022.acl-long.444", "year": 2022, "rel_sent": "For a better understanding of high - level structures , we propose a phrase - guided masking strategy for LM to emphasize more on reconstructing non - phrase words .", "forward": false, "src_ids": "2022.acl-long.444_4745"}
+{"input": "identification of high - level structures is done by using Method| context: recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task . despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context .", "entity": "identification of high - level structures", "output": "phrase - guided masking strategy", "neg_sample": ["identification of high - level structures is done by using Method", "recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task .", "despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context ."], "relation": "used for", "id": "2022.acl-long.444", "year": 2022, "rel_sent": "We show that the initial phrase regularization serves as an effective bootstrap , and phrase - guided masking improves the identification of high - level structures .", "forward": false, "src_ids": "2022.acl-long.444_4746"}
+{"input": "phrase - guided masking strategy is used for Task| context: recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task . despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context .", "entity": "phrase - guided masking strategy", "output": "identification of high - level structures", "neg_sample": ["phrase - guided masking strategy is used for Task", "recent studies have achieved inspiring success in unsupervised grammar induction using masked language modeling ( mlm ) as the proxy task .", "despite their high accuracy in identifying low - level structures , prior arts tend to struggle in capturing high - level structures like clauses , since the mlm task usually only requires information from local context ."], "relation": "used for", "id": "2022.acl-long.444", "year": 2022, "rel_sent": "We show that the initial phrase regularization serves as an effective bootstrap , and phrase - guided masking improves the identification of high - level structures .", "forward": true, "src_ids": "2022.acl-long.444_4747"}
+{"input": "multilingual bert embedding space is done by using Method| context: several studies have explored various advantages of multilingual pre - trained models ( such as multilingual bert ) in capturing shared linguistic knowledge . however , less attention has been paid to their limitations .", "entity": "multilingual bert embedding space", "output": "isotropy analysis", "neg_sample": ["multilingual bert embedding space is done by using Method", "several studies have explored various advantages of multilingual pre - trained models ( such as multilingual bert ) in capturing shared linguistic knowledge .", "however , less attention has been paid to their limitations ."], "relation": "used for", "id": "2022.findings-acl.103", "year": 2022, "rel_sent": "An Isotropy Analysis in the Multilingual BERT Embedding Space.", "forward": false, "src_ids": "2022.findings-acl.103_4748"}
+{"input": "isotropy analysis is used for OtherScientificTerm| context: several studies have explored various advantages of multilingual pre - trained models ( such as multilingual bert ) in capturing shared linguistic knowledge . however , less attention has been paid to their limitations .", "entity": "isotropy analysis", "output": "multilingual bert embedding space", "neg_sample": ["isotropy analysis is used for OtherScientificTerm", "several studies have explored various advantages of multilingual pre - trained models ( such as multilingual bert ) in capturing shared linguistic knowledge .", "however , less attention has been paid to their limitations ."], "relation": "used for", "id": "2022.findings-acl.103", "year": 2022, "rel_sent": "An Isotropy Analysis in the Multilingual BERT Embedding Space.", "forward": true, "src_ids": "2022.findings-acl.103_4749"}
+{"input": "chinese popular song is done by using Method| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "chinese popular song", "output": "controllable lyric generation system", "neg_sample": ["chinese popular song is done by using Method", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "ChipSong : A Controllable Lyric Generation System for Chinese Popular Song.", "forward": false, "src_ids": "2022.in2writing-1.13_4750"}
+{"input": "controllable lyric generation system is used for Material| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "controllable lyric generation system", "output": "chinese popular song", "neg_sample": ["controllable lyric generation system is used for Material", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "ChipSong : A Controllable Lyric Generation System for Chinese Popular Song.", "forward": true, "src_ids": "2022.in2writing-1.13_4751"}
+{"input": "musical short - video display purpose is done by using OtherScientificTerm| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "musical short - video display purpose", "output": "controlled lyric paragraphs", "neg_sample": ["musical short - video display purpose is done by using OtherScientificTerm", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": false, "src_ids": "2022.in2writing-1.13_4752"}
+{"input": "controlled lyric paragraphs is used for Task| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "controlled lyric paragraphs", "output": "musical short - video display purpose", "neg_sample": ["controlled lyric paragraphs is used for Task", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": true, "src_ids": "2022.in2writing-1.13_4753"}
+{"input": "word - level length format control is done by using OtherScientificTerm| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "word - level length format control", "output": "guided attention mechanism", "neg_sample": ["word - level length format control is done by using OtherScientificTerm", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": false, "src_ids": "2022.in2writing-1.13_4754"}
+{"input": "guided attention mechanism is used for Task| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "guided attention mechanism", "output": "word - level length format control", "neg_sample": ["guided attention mechanism is used for Task", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": true, "src_ids": "2022.in2writing-1.13_4755"}
+{"input": "specific lyric contents generation is done by using Method| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "specific lyric contents generation", "output": "open - ended trigger word mechanism", "neg_sample": ["specific lyric contents generation is done by using Method", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": false, "src_ids": "2022.in2writing-1.13_4756"}
+{"input": "open - ended trigger word mechanism is used for Task| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "open - ended trigger word mechanism", "output": "specific lyric contents generation", "neg_sample": ["open - ended trigger word mechanism is used for Task", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": true, "src_ids": "2022.in2writing-1.13_4757"}
+{"input": "rhyme control is done by using Method| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "rhyme control", "output": "reverse order training and shielding decoding", "neg_sample": ["rhyme control is done by using Method", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": false, "src_ids": "2022.in2writing-1.13_4758"}
+{"input": "reverse order training and shielding decoding is used for Task| context: in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace . one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability . although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands .", "entity": "reverse order training and shielding decoding", "output": "rhyme control", "neg_sample": ["reverse order training and shielding decoding is used for Task", "in this work , we take a further step towards satisfying practical demands in chinese lyric generation from musical short - video creators , in respect of the challenges on songs ' format constraints , creating specific lyrics from open - ended inspiration inputs , and language rhyme grace .", "one representative detail in these demands is to control lyric format at word level , that is , for chinese songs , creators even expect fix - length words on certain positions in a lyric to match a special melody , while previous methods lack such ability .", "although recent lyric generation community has made gratifying progress , most methods are not comprehensive enough to simultaneously meet these demands ."], "relation": "used for", "id": "2022.in2writing-1.13", "year": 2022, "rel_sent": "As a result , we propose ChipSong , which is an assisted lyric generation system built based on a Transformer - based autoregressive language model architecture , and generates controlled lyric paragraphs fit for musical short - video display purpose , by designing 1 ) a novel Begin - Internal - End ( BIE ) word - granularity embedding sequence with its guided attention mechanism for word - level length format control , and an explicit symbol set for sentence - level length format control ; 2 ) an open - ended trigger word mechanism to guide specific lyric contents generation ; 3 ) a paradigm of reverse order training and shielding decoding for rhyme control .", "forward": true, "src_ids": "2022.in2writing-1.13_4759"}
+{"input": "continuous - space attention mechanism is used for OtherScientificTerm| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "continuous - space attention mechanism", "output": "long - term memory", "neg_sample": ["continuous - space attention mechanism is used for OtherScientificTerm", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.375", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision . In order to control where precision is more important , infty - former maintains ' sticky memories , ' being able to model arbitrarily long contexts while keeping the computation budget fixed . Experiments on a synthetic sorting task , language modeling , and document grounded dialogue generation demonstrate the infty - former 's ability to retain information from long sequences .", "forward": true, "src_ids": "2022.acl-long.375_4760"}
+{"input": "continuous - space attention mechanism is used for Method| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "continuous - space attention mechanism", "output": "infty - former", "neg_sample": ["continuous - space attention mechanism is used for Method", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.375", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision . In order to control where precision is more important , infty - former maintains ' sticky memories , ' being able to model arbitrarily long contexts while keeping the computation budget fixed . Experiments on a synthetic sorting task , language modeling , and document grounded dialogue generation demonstrate the infty - former 's ability to retain information from long sequences .", "forward": true, "src_ids": "2022.acl-long.375_4761"}
+{"input": "long - term memory is done by using Method| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "long - term memory", "output": "continuous - space attention mechanism", "neg_sample": ["long - term memory is done by using Method", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.375", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision . In order to control where precision is more important , infty - former maintains ' sticky memories , ' being able to model arbitrarily long contexts while keeping the computation budget fixed . Experiments on a synthetic sorting task , language modeling , and document grounded dialogue generation demonstrate the infty - former 's ability to retain information from long sequences .", "forward": false, "src_ids": "2022.acl-long.375_4762"}
+{"input": "infty - former is done by using Method| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "infty - former", "output": "continuous - space attention mechanism", "neg_sample": ["infty - former is done by using Method", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.375", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision . In order to control where precision is more important , infty - former maintains ' sticky memories , ' being able to model arbitrarily long contexts while keeping the computation budget fixed . Experiments on a synthetic sorting task , language modeling , and document grounded dialogue generation demonstrate the infty - former 's ability to retain information from long sequences .", "forward": false, "src_ids": "2022.acl-long.375_4763"}
+{"input": "multimodal distributions is done by using Method| context: non - autoregressive text to speech ( nar - tts ) models have attracted much attention from both academia and industry due to their fast generation speed . one limitation of nar - tts models is that they ignore the correlation in time and frequency domains while generating speech mel - spectrograms , and thus cause blurry and over - smoothed results .", "entity": "multimodal distributions", "output": "laplacian mixture loss", "neg_sample": ["multimodal distributions is done by using Method", "non - autoregressive text to speech ( nar - tts ) models have attracted much attention from both academia and industry due to their fast generation speed .", "one limitation of nar - tts models is that they ignore the correlation in time and frequency domains while generating speech mel - spectrograms , and thus cause blurry and over - smoothed results ."], "relation": "used for", "id": "2022.acl-long.564", "year": 2022, "rel_sent": "2 ) Among advanced modeling methods , Laplacian mixture loss performs well at modeling multimodal distributions and enjoys its simplicity , while GAN and Glow achieve the best voice quality while suffering from increased training or model complexity .", "forward": false, "src_ids": "2022.acl-long.564_4764"}
+{"input": "laplacian mixture loss is used for OtherScientificTerm| context: non - autoregressive text to speech ( nar - tts ) models have attracted much attention from both academia and industry due to their fast generation speed . one limitation of nar - tts models is that they ignore the correlation in time and frequency domains while generating speech mel - spectrograms , and thus cause blurry and over - smoothed results .", "entity": "laplacian mixture loss", "output": "multimodal distributions", "neg_sample": ["laplacian mixture loss is used for OtherScientificTerm", "non - autoregressive text to speech ( nar - tts ) models have attracted much attention from both academia and industry due to their fast generation speed .", "one limitation of nar - tts models is that they ignore the correlation in time and frequency domains while generating speech mel - spectrograms , and thus cause blurry and over - smoothed results ."], "relation": "used for", "id": "2022.acl-long.564", "year": 2022, "rel_sent": "2 ) Among advanced modeling methods , Laplacian mixture loss performs well at modeling multimodal distributions and enjoys its simplicity , while GAN and Glow achieve the best voice quality while suffering from increased training or model complexity .", "forward": true, "src_ids": "2022.acl-long.564_4765"}
+{"input": "dialogue generation is done by using Method| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "dialogue generation", "output": "prophetchat", "neg_sample": ["dialogue generation is done by using Method", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "ProphetChat : Enhancing Dialogue Generation with Simulation of Future Conversation.", "forward": false, "src_ids": "2022.acl-long.68_4766"}
+{"input": "prophetchat is used for Task| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "prophetchat", "output": "dialogue generation", "neg_sample": ["prophetchat is used for Task", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "ProphetChat : Enhancing Dialogue Generation with Simulation of Future Conversation.", "forward": true, "src_ids": "2022.acl-long.68_4767"}
+{"input": "future - to - response generator is used for OtherScientificTerm| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "entity": "future - to - response generator", "output": "informative response", "neg_sample": ["future - to - response generator is used for OtherScientificTerm", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "With the simulated futures , we then utilize the ensemble of a history - to - response generator and a future - to - response generator to jointly generate a more informative response .", "forward": true, "src_ids": "2022.acl-long.68_4768"}
+{"input": "response generation is done by using Method| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "response generation", "output": "dialogue generation framework", "neg_sample": ["response generation is done by using Method", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "Accordingly , we propose a novel dialogue generation framework named ProphetChat that utilizes the simulated dialogue futures in the inference phase to enhance response generation .", "forward": false, "src_ids": "2022.acl-long.68_4769"}
+{"input": "inference phase is done by using OtherScientificTerm| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "inference phase", "output": "dialogue future simulation", "neg_sample": ["inference phase is done by using OtherScientificTerm", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "Accordingly , we propose a novel dialogue generation framework named ProphetChat that utilizes the simulated dialogue futures in the inference phase to enhance response generation .", "forward": false, "src_ids": "2022.acl-long.68_4770"}
+{"input": "dialogue future simulation is used for Task| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "dialogue future simulation", "output": "inference phase", "neg_sample": ["dialogue future simulation is used for Task", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "Accordingly , we propose a novel dialogue generation framework named ProphetChat that utilizes the simulated dialogue futures in the inference phase to enhance response generation .", "forward": true, "src_ids": "2022.acl-long.68_4771"}
+{"input": "dialogue generation framework is used for Task| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "dialogue generation framework", "output": "response generation", "neg_sample": ["dialogue generation framework is used for Task", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "Accordingly , we propose a novel dialogue generation framework named ProphetChat that utilizes the simulated dialogue futures in the inference phase to enhance response generation .", "forward": true, "src_ids": "2022.acl-long.68_4772"}
+{"input": "informative response is done by using Method| context: typical generative dialogue models utilize the dialogue history to generate the response . however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy . intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e. , the dialogue future ) after receiving its response , it could possibly provide a more informative response .", "entity": "informative response", "output": "future - to - response generator", "neg_sample": ["informative response is done by using Method", "typical generative dialogue models utilize the dialogue history to generate the response .", "however , since one dialogue utterance can often be appropriately answered by multiple distinct responses , generating a desired response solely based on the historical information is not easy .", "intuitively , if the chatbot can foresee in advance what the user would talk about ( i.e.", ", the dialogue future ) after receiving its response , it could possibly provide a more informative response ."], "relation": "used for", "id": "2022.acl-long.68", "year": 2022, "rel_sent": "With the simulated futures , we then utilize the ensemble of a history - to - response generator and a future - to - response generator to jointly generate a more informative response .", "forward": false, "src_ids": "2022.acl-long.68_4773"}
+{"input": "identifying ( dis)similar functionalities of source code is done by using Method| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "identifying ( dis)similar functionalities of source code", "output": "self - supervised model", "neg_sample": ["identifying ( dis)similar functionalities of source code is done by using Method", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "We present DISCO ( DIS - similarity of COde ) , a novel self - supervised model focusing on identifying ( dis)similar functionalities of source code .", "forward": false, "src_ids": "2022.acl-long.436_4774"}
+{"input": "self - supervised model is used for Task| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "self - supervised model", "output": "identifying ( dis)similar functionalities of source code", "neg_sample": ["self - supervised model is used for Task", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "We present DISCO ( DIS - similarity of COde ) , a novel self - supervised model focusing on identifying ( dis)similar functionalities of source code .", "forward": true, "src_ids": "2022.acl-long.436_4775"}
+{"input": "synthetic code clones is done by using Method| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "synthetic code clones", "output": "structure - guided code transformation algorithms", "neg_sample": ["synthetic code clones is done by using Method", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "Rather , we design structure - guided code transformation algorithms to generate synthetic code clones and inject real - world security bugs , augmenting the collected datasets in a targeted way .", "forward": false, "src_ids": "2022.acl-long.436_4776"}
+{"input": "real - world security bugs is done by using Method| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "real - world security bugs", "output": "structure - guided code transformation algorithms", "neg_sample": ["real - world security bugs is done by using Method", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "Rather , we design structure - guided code transformation algorithms to generate synthetic code clones and inject real - world security bugs , augmenting the collected datasets in a targeted way .", "forward": false, "src_ids": "2022.acl-long.436_4777"}
+{"input": "structure - guided code transformation algorithms is used for OtherScientificTerm| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "structure - guided code transformation algorithms", "output": "synthetic code clones", "neg_sample": ["structure - guided code transformation algorithms is used for OtherScientificTerm", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "Rather , we design structure - guided code transformation algorithms to generate synthetic code clones and inject real - world security bugs , augmenting the collected datasets in a targeted way .", "forward": true, "src_ids": "2022.acl-long.436_4778"}
+{"input": "local tree - based context is done by using OtherScientificTerm| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "local tree - based context", "output": "cloze objective", "neg_sample": ["local tree - based context is done by using OtherScientificTerm", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "To better capture the structural features of source code , we propose a new cloze objective to encode the local tree - based context ( e.g. , parents or sibling nodes ) .", "forward": false, "src_ids": "2022.acl-long.436_4779"}
+{"input": "cloze objective is used for OtherScientificTerm| context: understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection .", "entity": "cloze objective", "output": "local tree - based context", "neg_sample": ["cloze objective is used for OtherScientificTerm", "understanding the functional ( dis)-similarity of source code is significant for code modeling tasks such as software vulnerability and code clone detection ."], "relation": "used for", "id": "2022.acl-long.436", "year": 2022, "rel_sent": "To better capture the structural features of source code , we propose a new cloze objective to encode the local tree - based context ( e.g. , parents or sibling nodes ) .", "forward": true, "src_ids": "2022.acl-long.436_4780"}
+{"input": "biomedical research articles is done by using Method| context: research papers reflect scientific advances . citations are widely used in research publications to support the new findings and show their benefits , while also regulating the information flow to make the contents clearer for the audience . a citation in a research article refers to the information 's source , but not the specific text span from that source article . in biomedical research articles , this task is challenging as the same chemical or biological component can be represented in multiple ways in different papers from various domains .", "entity": "biomedical research articles", "output": "deep learning - based citation linkage framework", "neg_sample": ["biomedical research articles is done by using Method", "research papers reflect scientific advances .", "citations are widely used in research publications to support the new findings and show their benefits , while also regulating the information flow to make the contents clearer for the audience .", "a citation in a research article refers to the information 's source , but not the specific text span from that source article .", "in biomedical research articles , this task is challenging as the same chemical or biological component can be represented in multiple ways in different papers from various domains ."], "relation": "used for", "id": "2022.bionlp-1.23", "year": 2022, "rel_sent": "BioCite : A Deep Learning - based Citation Linkage Framework for Biomedical Research Articles.", "forward": false, "src_ids": "2022.bionlp-1.23_4781"}
+{"input": "deep learning - based citation linkage framework is used for Material| context: research papers reflect scientific advances . citations are widely used in research publications to support the new findings and show their benefits , while also regulating the information flow to make the contents clearer for the audience . a citation in a research article refers to the information 's source , but not the specific text span from that source article .", "entity": "deep learning - based citation linkage framework", "output": "biomedical research articles", "neg_sample": ["deep learning - based citation linkage framework is used for Material", "research papers reflect scientific advances .", "citations are widely used in research publications to support the new findings and show their benefits , while also regulating the information flow to make the contents clearer for the audience .", "a citation in a research article refers to the information 's source , but not the specific text span from that source article ."], "relation": "used for", "id": "2022.bionlp-1.23", "year": 2022, "rel_sent": "BioCite : A Deep Learning - based Citation Linkage Framework for Biomedical Research Articles.", "forward": true, "src_ids": "2022.bionlp-1.23_4782"}
+{"input": "news recommendation is done by using Method| context: existing news recommendation methods usually learn news representations solely based on news titles . to sufficiently utilize other fields of news information such as category and entities , some methods treat each field as an additional feature and combine different feature vectors with attentive pooling . with the adoption of large pre - trained models like bert in news recommendation , the above way to incorporate multi - field information may encounter challenges : the shallow feature encoding to compress the category and entity information is not compatible with the deep bert encoding .", "entity": "news recommendation", "output": "multi - task learning", "neg_sample": ["news recommendation is done by using Method", "existing news recommendation methods usually learn news representations solely based on news titles .", "to sufficiently utilize other fields of news information such as category and entities , some methods treat each field as an additional feature and combine different feature vectors with attentive pooling .", "with the adoption of large pre - trained models like bert in news recommendation , the above way to incorporate multi - field information may encounter challenges : the shallow feature encoding to compress the category and entity information is not compatible with the deep bert encoding ."], "relation": "used for", "id": "2022.findings-acl.209", "year": 2022, "rel_sent": "MTRec : Multi - Task Learning over BERT for News Recommendation.", "forward": false, "src_ids": "2022.findings-acl.209_4783"}
+{"input": "multi - task learning is used for Task| context: to sufficiently utilize other fields of news information such as category and entities , some methods treat each field as an additional feature and combine different feature vectors with attentive pooling .", "entity": "multi - task learning", "output": "news recommendation", "neg_sample": ["multi - task learning is used for Task", "to sufficiently utilize other fields of news information such as category and entities , some methods treat each field as an additional feature and combine different feature vectors with attentive pooling ."], "relation": "used for", "id": "2022.findings-acl.209", "year": 2022, "rel_sent": "MTRec : Multi - Task Learning over BERT for News Recommendation.", "forward": true, "src_ids": "2022.findings-acl.209_4784"}
+{"input": "parsing knowledge is done by using Method| context: recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages . however , these advances assume access to high - quality machine translation systems and word alignment tools . we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e. , utterance - logical form pairs ) for new languages .", "entity": "parsing knowledge", "output": "multi - task encoder - decoder model", "neg_sample": ["parsing knowledge is done by using Method", "recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages .", "however , these advances assume access to high - quality machine translation systems and word alignment tools .", "we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e.", ", utterance - logical form pairs ) for new languages ."], "relation": "used for", "id": "2022.acl-long.285", "year": 2022, "rel_sent": "We propose a multi - task encoder - decoder model to transfer parsing knowledge to additional languages using only English - logical form paired data and in - domain natural language corpora in each new language .", "forward": false, "src_ids": "2022.acl-long.285_4785"}
+{"input": "multi - task encoder - decoder model is used for OtherScientificTerm| context: recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages . however , these advances assume access to high - quality machine translation systems and word alignment tools . we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e. , utterance - logical form pairs ) for new languages .", "entity": "multi - task encoder - decoder model", "output": "parsing knowledge", "neg_sample": ["multi - task encoder - decoder model is used for OtherScientificTerm", "recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages .", "however , these advances assume access to high - quality machine translation systems and word alignment tools .", "we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e.", ", utterance - logical form pairs ) for new languages ."], "relation": "used for", "id": "2022.acl-long.285", "year": 2022, "rel_sent": "We propose a multi - task encoder - decoder model to transfer parsing knowledge to additional languages using only English - logical form paired data and in - domain natural language corpora in each new language .", "forward": true, "src_ids": "2022.acl-long.285_4786"}
+{"input": "cross - lingual latent representation alignment is done by using OtherScientificTerm| context: recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages . however , these advances assume access to high - quality machine translation systems and word alignment tools . we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e. , utterance - logical form pairs ) for new languages .", "entity": "cross - lingual latent representation alignment", "output": "auxiliary objectives", "neg_sample": ["cross - lingual latent representation alignment is done by using OtherScientificTerm", "recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages .", "however , these advances assume access to high - quality machine translation systems and word alignment tools .", "we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e.", ", utterance - logical form pairs ) for new languages ."], "relation": "used for", "id": "2022.acl-long.285", "year": 2022, "rel_sent": "Our model encourages language - agnostic encodings by jointly optimizing for logical - form generation with auxiliary objectives designed for cross - lingual latent representation alignment .", "forward": false, "src_ids": "2022.acl-long.285_4787"}
+{"input": "auxiliary objectives is used for Task| context: recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages . however , these advances assume access to high - quality machine translation systems and word alignment tools . we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e. , utterance - logical form pairs ) for new languages .", "entity": "auxiliary objectives", "output": "cross - lingual latent representation alignment", "neg_sample": ["auxiliary objectives is used for Task", "recent work in cross - lingual semantic parsing has successfully applied machine translation to localize parsers to new languages .", "however , these advances assume access to high - quality machine translation systems and word alignment tools .", "we remove these assumptions and study cross - lingual semantic parsing as a zero - shot problem , without parallel data ( i.e.", ", utterance - logical form pairs ) for new languages ."], "relation": "used for", "id": "2022.acl-long.285", "year": 2022, "rel_sent": "Our model encourages language - agnostic encodings by jointly optimizing for logical - form generation with auxiliary objectives designed for cross - lingual latent representation alignment .", "forward": true, "src_ids": "2022.acl-long.285_4788"}
+{"input": "generation - based chatbot is done by using OtherScientificTerm| context: creating chatbots to behave like real people is important in terms of believability . errors in general chatbots and chatbots that follow a rough persona have been studied , but those in chatbots that behave like real people have not been thoroughly investigated .", "entity": "generation - based chatbot", "output": "user interactions", "neg_sample": ["generation - based chatbot is done by using OtherScientificTerm", "creating chatbots to behave like real people is important in terms of believability .", "errors in general chatbots and chatbots that follow a rough persona have been studied , but those in chatbots that behave like real people have not been thoroughly investigated ."], "relation": "used for", "id": "2022.acl-short.50", "year": 2022, "rel_sent": "We collected a large amount of user interactions of a generation - based chatbot trained from large - scale dialogue data of a specific character , i.e. , target person , and analyzed errors related to that person .", "forward": false, "src_ids": "2022.acl-short.50_4789"}
+{"input": "user interactions is used for Task| context: creating chatbots to behave like real people is important in terms of believability . errors in general chatbots and chatbots that follow a rough persona have been studied , but those in chatbots that behave like real people have not been thoroughly investigated .", "entity": "user interactions", "output": "generation - based chatbot", "neg_sample": ["user interactions is used for Task", "creating chatbots to behave like real people is important in terms of believability .", "errors in general chatbots and chatbots that follow a rough persona have been studied , but those in chatbots that behave like real people have not been thoroughly investigated ."], "relation": "used for", "id": "2022.acl-short.50", "year": 2022, "rel_sent": "We collected a large amount of user interactions of a generation - based chatbot trained from large - scale dialogue data of a specific character , i.e. , target person , and analyzed errors related to that person .", "forward": true, "src_ids": "2022.acl-short.50_4790"}
+{"input": "rare symptoms is done by using Method| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "rare symptoms", "output": "data augmentation", "neg_sample": ["rare symptoms is done by using Method", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "Data Augmentation for Rare Symptoms in Vaccine Side - Effect Detection.", "forward": false, "src_ids": "2022.bionlp-1.29_4791"}
+{"input": "vaccine side - effect detection is done by using Method| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "vaccine side - effect detection", "output": "data augmentation", "neg_sample": ["vaccine side - effect detection is done by using Method", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "Data Augmentation for Rare Symptoms in Vaccine Side - Effect Detection.", "forward": false, "src_ids": "2022.bionlp-1.29_4792"}
+{"input": "data augmentation is used for OtherScientificTerm| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "data augmentation", "output": "rare symptoms", "neg_sample": ["data augmentation is used for OtherScientificTerm", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "Data Augmentation for Rare Symptoms in Vaccine Side - Effect Detection.", "forward": true, "src_ids": "2022.bionlp-1.29_4793"}
+{"input": "data augmentation technique is used for OtherScientificTerm| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "data augmentation technique", "output": "rare symptoms", "neg_sample": ["data augmentation technique is used for OtherScientificTerm", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "We introduce a data augmentation technique to increase the number of training examples for rare symptoms .", "forward": true, "src_ids": "2022.bionlp-1.29_4794"}
+{"input": "data augmentation is used for Task| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "data augmentation", "output": "vaccine side - effect detection", "neg_sample": ["data augmentation is used for Task", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "Data Augmentation for Rare Symptoms in Vaccine Side - Effect Detection.", "forward": true, "src_ids": "2022.bionlp-1.29_4795"}
+{"input": "standardized names of symptoms is done by using Method| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "standardized names of symptoms", "output": "autoregressive model", "neg_sample": ["standardized names of symptoms is done by using Method", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "We tackle these challenges with an autoregressive model that generates standardized names of symptoms .", "forward": false, "src_ids": "2022.bionlp-1.29_4796"}
+{"input": "autoregressive model is used for OtherScientificTerm| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "autoregressive model", "output": "standardized names of symptoms", "neg_sample": ["autoregressive model is used for OtherScientificTerm", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "We tackle these challenges with an autoregressive model that generates standardized names of symptoms .", "forward": true, "src_ids": "2022.bionlp-1.29_4797"}
+{"input": "rare symptoms is done by using Method| context: we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines . our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type .", "entity": "rare symptoms", "output": "data augmentation technique", "neg_sample": ["rare symptoms is done by using Method", "we study the problem of entity detection and normalization applied to patient self - reports of symptoms that arise as side - effects of vaccines .", "our application domain presents unique challenges that render traditional classification methods ineffective : the number of entity types is large ; and many symptoms are rare , resulting in a long - tail distribution of training examples per entity type ."], "relation": "used for", "id": "2022.bionlp-1.29", "year": 2022, "rel_sent": "We introduce a data augmentation technique to increase the number of training examples for rare symptoms .", "forward": false, "src_ids": "2022.bionlp-1.29_4798"}
+{"input": "unsupervised domain adaptation of discourse dependency parsing is done by using Method| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "unsupervised domain adaptation of discourse dependency parsing", "output": "bootstrapping", "neg_sample": ["unsupervised domain adaptation of discourse dependency parsing is done by using Method", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "The experimental results show that bootstrapping is significantly and consistently effective for unsupervised domain adaptation of discourse dependency parsing , but the low coverage of accurately predicted pseudo labels is a bottleneck for further improvement .", "forward": false, "src_ids": "2022.tacl-1.8_4799"}
+{"input": "bert - based discourse dependency parsers is done by using Method| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "bert - based discourse dependency parsers", "output": "bootstrapping methods", "neg_sample": ["bert - based discourse dependency parsers is done by using Method", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "In this paper , we report and discuss the effectiveness and limitations of bootstrapping methods for adapting modern BERT - based discourse dependency parsers to out - of - domain text without relying on additional human supervision .", "forward": false, "src_ids": "2022.tacl-1.8_4800"}
+{"input": "bootstrapping methods is used for Method| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "bootstrapping methods", "output": "bert - based discourse dependency parsers", "neg_sample": ["bootstrapping methods is used for Method", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "In this paper , we report and discuss the effectiveness and limitations of bootstrapping methods for adapting modern BERT - based discourse dependency parsers to out - of - domain text without relying on additional human supervision .", "forward": true, "src_ids": "2022.tacl-1.8_4801"}
+{"input": "adaptation scenarios is done by using Method| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "adaptation scenarios", "output": "self - training", "neg_sample": ["adaptation scenarios is done by using Method", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "Specifically , we investigate self - training , co - training , tri - training , and asymmetric tri - training of graph - based and transition - based discourse dependency parsing models , as well as confidence measures and sample selection criteria in two adaptation scenarios : monologue adaptation between scientific disciplines and dialogue genre adaptation .", "forward": false, "src_ids": "2022.tacl-1.8_4802"}
+{"input": "adaptation scenarios is done by using Metric| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "adaptation scenarios", "output": "sample selection criteria", "neg_sample": ["adaptation scenarios is done by using Metric", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "Specifically , we investigate self - training , co - training , tri - training , and asymmetric tri - training of graph - based and transition - based discourse dependency parsing models , as well as confidence measures and sample selection criteria in two adaptation scenarios : monologue adaptation between scientific disciplines and dialogue genre adaptation .", "forward": false, "src_ids": "2022.tacl-1.8_4803"}
+{"input": "self - training is used for Generic| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "self - training", "output": "adaptation scenarios", "neg_sample": ["self - training is used for Generic", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "Specifically , we investigate self - training , co - training , tri - training , and asymmetric tri - training of graph - based and transition - based discourse dependency parsing models , as well as confidence measures and sample selection criteria in two adaptation scenarios : monologue adaptation between scientific disciplines and dialogue genre adaptation .", "forward": true, "src_ids": "2022.tacl-1.8_4804"}
+{"input": "sample selection criteria is used for Generic| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "sample selection criteria", "output": "adaptation scenarios", "neg_sample": ["sample selection criteria is used for Generic", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "Specifically , we investigate self - training , co - training , tri - training , and asymmetric tri - training of graph - based and transition - based discourse dependency parsing models , as well as confidence measures and sample selection criteria in two adaptation scenarios : monologue adaptation between scientific disciplines and dialogue genre adaptation .", "forward": true, "src_ids": "2022.tacl-1.8_4805"}
+{"input": "discourse dependency parsing is done by using Material| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "discourse dependency parsing", "output": "manually annotated resource", "neg_sample": ["discourse dependency parsing is done by using Material", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "We also release COVID-19 Discourse Dependency Treebank ( COVID19 - DTB ) , a new manually annotated resource for discourse dependency parsing of biomedical paper abstracts .", "forward": false, "src_ids": "2022.tacl-1.8_4806"}
+{"input": "manually annotated resource is used for Task| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "manually annotated resource", "output": "discourse dependency parsing", "neg_sample": ["manually annotated resource is used for Task", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "We also release COVID-19 Discourse Dependency Treebank ( COVID19 - DTB ) , a new manually annotated resource for discourse dependency parsing of biomedical paper abstracts .", "forward": true, "src_ids": "2022.tacl-1.8_4807"}
+{"input": "bootstrapping is used for Task| context: discourse parsing has been studied for decades . however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text .", "entity": "bootstrapping", "output": "unsupervised domain adaptation of discourse dependency parsing", "neg_sample": ["bootstrapping is used for Task", "discourse parsing has been studied for decades .", "however , it still remains challenging to utilize discourse parsing for real - world applications because the parsing accuracy degrades significantly on out - of - domain text ."], "relation": "used for", "id": "2022.tacl-1.8", "year": 2022, "rel_sent": "The experimental results show that bootstrapping is significantly and consistently effective for unsupervised domain adaptation of discourse dependency parsing , but the low coverage of accurately predicted pseudo labels is a bottleneck for further improvement .", "forward": true, "src_ids": "2022.tacl-1.8_4808"}
+{"input": "names is used for Material| context: before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness .", "entity": "names", "output": "lower - resourced languages", "neg_sample": ["names is used for Material", "before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness ."], "relation": "used for", "id": "2022.findings-acl.44", "year": 2022, "rel_sent": "We explore the contents of the names stored in Wikidata for a few lower - resourced languages and find that many of them are not in fact in the languages they claim to be , requiring non - trivial effort to correct .", "forward": true, "src_ids": "2022.findings-acl.44_4809"}
+{"input": "annotations is used for Material| context: before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness .", "entity": "annotations", "output": "lower - resourced languages", "neg_sample": ["annotations is used for Material", "before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness ."], "relation": "used for", "id": "2022.findings-acl.44", "year": 2022, "rel_sent": "We then discuss the importance of creating annotations for lower - resourced languages in a thoughtful and ethical way that includes the language speakers as part of the development process .", "forward": true, "src_ids": "2022.findings-acl.44_4810"}
+{"input": "lower - resourced languages is done by using OtherScientificTerm| context: in this position paper , we describe our perspective on how meaningful resources for lower - resourced languages should be developed in connection with the speakers of those languages . before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness .", "entity": "lower - resourced languages", "output": "names", "neg_sample": ["lower - resourced languages is done by using OtherScientificTerm", "in this position paper , we describe our perspective on how meaningful resources for lower - resourced languages should be developed in connection with the speakers of those languages .", "before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness ."], "relation": "used for", "id": "2022.findings-acl.44", "year": 2022, "rel_sent": "We explore the contents of the names stored in Wikidata for a few lower - resourced languages and find that many of them are not in fact in the languages they claim to be , requiring non - trivial effort to correct .", "forward": false, "src_ids": "2022.findings-acl.44_4811"}
+{"input": "lower - resourced languages is done by using OtherScientificTerm| context: in this position paper , we describe our perspective on how meaningful resources for lower - resourced languages should be developed in connection with the speakers of those languages . before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness .", "entity": "lower - resourced languages", "output": "annotations", "neg_sample": ["lower - resourced languages is done by using OtherScientificTerm", "in this position paper , we describe our perspective on how meaningful resources for lower - resourced languages should be developed in connection with the speakers of those languages .", "before advancing that position , we first examine two massively multilingual resources used in language technology development , identifying shortcomings that limit their usefulness ."], "relation": "used for", "id": "2022.findings-acl.44", "year": 2022, "rel_sent": "We then discuss the importance of creating annotations for lower - resourced languages in a thoughtful and ethical way that includes the language speakers as part of the development process .", "forward": false, "src_ids": "2022.findings-acl.44_4812"}
+{"input": "side information is done by using Material| context: recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks . however , existing legal event detection ( led ) datasets only concern incomprehensive event types and have limited annotated data , which restricts the development of led methods and their downstream applications .", "entity": "side information", "output": "legal events", "neg_sample": ["side information is done by using Material", "recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks .", "however , existing legal event detection ( led ) datasets only concern incomprehensive event types and have limited annotated data , which restricts the development of led methods and their downstream applications ."], "relation": "used for", "id": "2022.findings-acl.17", "year": 2022, "rel_sent": "Moreover , we simply utilize legal events as side information to promote downstream applications .", "forward": false, "src_ids": "2022.findings-acl.17_4813"}
+{"input": "downstream applications is done by using Material| context: recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks . however , existing legal event detection ( led ) datasets only concern incomprehensive event types and have limited annotated data , which restricts the development of led methods and their downstream applications .", "entity": "downstream applications", "output": "legal events", "neg_sample": ["downstream applications is done by using Material", "recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks .", "however , existing legal event detection ( led ) datasets only concern incomprehensive event types and have limited annotated data , which restricts the development of led methods and their downstream applications ."], "relation": "used for", "id": "2022.findings-acl.17", "year": 2022, "rel_sent": "Moreover , we simply utilize legal events as side information to promote downstream applications .", "forward": false, "src_ids": "2022.findings-acl.17_4814"}
+{"input": "legal events is used for OtherScientificTerm| context: recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks . however , existing legal event detection ( led ) datasets only concern incomprehensive event types and have limited annotated data , which restricts the development of led methods and their downstream applications .", "entity": "legal events", "output": "side information", "neg_sample": ["legal events is used for OtherScientificTerm", "recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks .", "however , existing legal event detection ( led ) datasets only concern incomprehensive event types and have limited annotated data , which restricts the development of led methods and their downstream applications ."], "relation": "used for", "id": "2022.findings-acl.17", "year": 2022, "rel_sent": "Moreover , we simply utilize legal events as side information to promote downstream applications .", "forward": true, "src_ids": "2022.findings-acl.17_4815"}
+{"input": "legal events is used for Task| context: recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks .", "entity": "legal events", "output": "downstream applications", "neg_sample": ["legal events is used for Task", "recognizing facts is the most fundamental step in making judgments , hence detecting events in the legal documents is important to legal case analysis tasks ."], "relation": "used for", "id": "2022.findings-acl.17", "year": 2022, "rel_sent": "Moreover , we simply utilize legal events as side information to promote downstream applications .", "forward": true, "src_ids": "2022.findings-acl.17_4816"}
+{"input": "paraphrase identification is done by using Method| context: paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings . while cross - encoders have achieved high performances across several benchmarks , bi - encoders such as sbert have been widely applied to sentence pair tasks . they exhibit substantially lower computation complexity and are better suited to symmetric tasks .", "entity": "paraphrase identification", "output": "predicate - argument based bi - encoder", "neg_sample": ["paraphrase identification is done by using Method", "paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings .", "while cross - encoders have achieved high performances across several benchmarks , bi - encoders such as sbert have been widely applied to sentence pair tasks .", "they exhibit substantially lower computation complexity and are better suited to symmetric tasks ."], "relation": "used for", "id": "2022.acl-long.382", "year": 2022, "rel_sent": "Predicate - Argument Based Bi - Encoder for Paraphrase Identification.", "forward": false, "src_ids": "2022.acl-long.382_4817"}
+{"input": "predicate - argument based bi - encoder is used for Task| context: while cross - encoders have achieved high performances across several benchmarks , bi - encoders such as sbert have been widely applied to sentence pair tasks . they exhibit substantially lower computation complexity and are better suited to symmetric tasks .", "entity": "predicate - argument based bi - encoder", "output": "paraphrase identification", "neg_sample": ["predicate - argument based bi - encoder is used for Task", "while cross - encoders have achieved high performances across several benchmarks , bi - encoders such as sbert have been widely applied to sentence pair tasks .", "they exhibit substantially lower computation complexity and are better suited to symmetric tasks ."], "relation": "used for", "id": "2022.acl-long.382", "year": 2022, "rel_sent": "Predicate - Argument Based Bi - Encoder for Paraphrase Identification.", "forward": true, "src_ids": "2022.acl-long.382_4818"}
+{"input": "paraphrase identification task is done by using Method| context: paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings . they exhibit substantially lower computation complexity and are better suited to symmetric tasks .", "entity": "paraphrase identification task", "output": "bi - encoder approach", "neg_sample": ["paraphrase identification task is done by using Method", "paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings .", "they exhibit substantially lower computation complexity and are better suited to symmetric tasks ."], "relation": "used for", "id": "2022.acl-long.382", "year": 2022, "rel_sent": "In this work , we adopt a bi - encoder approach to the paraphrase identification task , and investigate the impact of explicitly incorporating predicate - argument information into SBERT through weighted aggregation .", "forward": false, "src_ids": "2022.acl-long.382_4819"}
+{"input": "bi - encoder approach is used for Task| context: paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings . while cross - encoders have achieved high performances across several benchmarks , bi - encoders such as sbert have been widely applied to sentence pair tasks . they exhibit substantially lower computation complexity and are better suited to symmetric tasks .", "entity": "bi - encoder approach", "output": "paraphrase identification task", "neg_sample": ["bi - encoder approach is used for Task", "paraphrase identification involves identifying whether a pair of sentences express the same or similar meanings .", "while cross - encoders have achieved high performances across several benchmarks , bi - encoders such as sbert have been widely applied to sentence pair tasks .", "they exhibit substantially lower computation complexity and are better suited to symmetric tasks ."], "relation": "used for", "id": "2022.acl-long.382", "year": 2022, "rel_sent": "In this work , we adopt a bi - encoder approach to the paraphrase identification task , and investigate the impact of explicitly incorporating predicate - argument information into SBERT through weighted aggregation .", "forward": true, "src_ids": "2022.acl-long.382_4820"}
+{"input": "extractive summarization is done by using OtherScientificTerm| context: text summarization ( ts ) is an important nlp task . pre - trained language models ( plms ) have been used to improve the performance of ts . however , plms are limited by their need of labelled training data and by their attention mechanism , which often makes them unsuitable for use on long documents .", "entity": "extractive summarization", "output": "salient textual fragments", "neg_sample": ["extractive summarization is done by using OtherScientificTerm", "text summarization ( ts ) is an important nlp task .", "pre - trained language models ( plms ) have been used to improve the performance of ts .", "however , plms are limited by their need of labelled training data and by their attention mechanism , which often makes them unsuitable for use on long documents ."], "relation": "used for", "id": "2022.bionlp-1.22", "year": 2022, "rel_sent": "We evaluate the efficacy of generating and using salient textual fragments to guide extractive summarization on documents from the biomedical and general scientific domains .", "forward": false, "src_ids": "2022.bionlp-1.22_4821"}
+{"input": "salient textual fragments is used for Task| context: text summarization ( ts ) is an important nlp task . pre - trained language models ( plms ) have been used to improve the performance of ts . however , plms are limited by their need of labelled training data and by their attention mechanism , which often makes them unsuitable for use on long documents .", "entity": "salient textual fragments", "output": "extractive summarization", "neg_sample": ["salient textual fragments is used for Task", "text summarization ( ts ) is an important nlp task .", "pre - trained language models ( plms ) have been used to improve the performance of ts .", "however , plms are limited by their need of labelled training data and by their attention mechanism , which often makes them unsuitable for use on long documents ."], "relation": "used for", "id": "2022.bionlp-1.22", "year": 2022, "rel_sent": "We evaluate the efficacy of generating and using salient textual fragments to guide extractive summarization on documents from the biomedical and general scientific domains .", "forward": true, "src_ids": "2022.bionlp-1.22_4822"}
+{"input": "easy - to - read german is done by using Generic| context: low - literate users with intellectual or developmental disabilities ( idd ) and/or complex communication needs ( ccn ) require specific writing support .", "entity": "easy - to - read german", "output": "text - writing system", "neg_sample": ["easy - to - read german is done by using Generic", "low - literate users with intellectual or developmental disabilities ( idd ) and/or complex communication needs ( ccn ) require specific writing support ."], "relation": "used for", "id": "2022.in2writing-1.4", "year": 2022, "rel_sent": "A text - writing system for Easy - to - Read German evaluated with low - literate users with cognitive impairment.", "forward": false, "src_ids": "2022.in2writing-1.4_4823"}
+{"input": "text - writing system is used for Material| context: low - literate users with intellectual or developmental disabilities ( idd ) and/or complex communication needs ( ccn ) require specific writing support .", "entity": "text - writing system", "output": "easy - to - read german", "neg_sample": ["text - writing system is used for Material", "low - literate users with intellectual or developmental disabilities ( idd ) and/or complex communication needs ( ccn ) require specific writing support ."], "relation": "used for", "id": "2022.in2writing-1.4", "year": 2022, "rel_sent": "A text - writing system for Easy - to - Read German evaluated with low - literate users with cognitive impairment.", "forward": true, "src_ids": "2022.in2writing-1.4_4824"}
+{"input": "limitedink is used for Task| context: existing self - explaining models typically favor extracting the shortest possible rationales - snippets of an input text ' responsible for ' corresponding output - to explain the model prediction , with the assumption that shorter rationales are more intuitive to humans . however , this assumption has yet to be validated . is the shortest rationale indeed the most human - understandable ?", "entity": "limitedink", "output": "user study", "neg_sample": ["limitedink is used for Task", "existing self - explaining models typically favor extracting the shortest possible rationales - snippets of an input text ' responsible for ' corresponding output - to explain the model prediction , with the assumption that shorter rationales are more intuitive to humans .", "however , this assumption has yet to be validated .", "is the shortest rationale indeed the most human - understandable ?"], "relation": "used for", "id": "2022.acl-short.2", "year": 2022, "rel_sent": "We use LimitedInk to conduct a user study on the impact of rationale length , where we ask human judges to predict the sentiment label of documents based only on LimitedInk - generated rationales with different lengths .", "forward": true, "src_ids": "2022.acl-short.2_4825"}
+{"input": "limitedink is used for Method| context: however , this assumption has yet to be validated . is the shortest rationale indeed the most human - understandable ?", "entity": "limitedink", "output": "self - explaining models", "neg_sample": ["limitedink is used for Method", "however , this assumption has yet to be validated .", "is the shortest rationale indeed the most human - understandable ?"], "relation": "used for", "id": "2022.acl-short.2", "year": 2022, "rel_sent": "Compared to existing baselines , LimitedInk achieves compatible end - task performance and human - annotated rationale agreement , making it a suitable representation of the recent class of self - explaining models .", "forward": true, "src_ids": "2022.acl-short.2_4826"}
+{"input": "self - explaining models is done by using Method| context: existing self - explaining models typically favor extracting the shortest possible rationales - snippets of an input text ' responsible for ' corresponding output - to explain the model prediction , with the assumption that shorter rationales are more intuitive to humans . however , this assumption has yet to be validated . is the shortest rationale indeed the most human - understandable ?", "entity": "self - explaining models", "output": "limitedink", "neg_sample": ["self - explaining models is done by using Method", "existing self - explaining models typically favor extracting the shortest possible rationales - snippets of an input text ' responsible for ' corresponding output - to explain the model prediction , with the assumption that shorter rationales are more intuitive to humans .", "however , this assumption has yet to be validated .", "is the shortest rationale indeed the most human - understandable ?"], "relation": "used for", "id": "2022.acl-short.2", "year": 2022, "rel_sent": "Compared to existing baselines , LimitedInk achieves compatible end - task performance and human - annotated rationale agreement , making it a suitable representation of the recent class of self - explaining models .", "forward": false, "src_ids": "2022.acl-short.2_4827"}
+{"input": "user study is done by using Method| context: existing self - explaining models typically favor extracting the shortest possible rationales - snippets of an input text ' responsible for ' corresponding output - to explain the model prediction , with the assumption that shorter rationales are more intuitive to humans . however , this assumption has yet to be validated . is the shortest rationale indeed the most human - understandable ?", "entity": "user study", "output": "limitedink", "neg_sample": ["user study is done by using Method", "existing self - explaining models typically favor extracting the shortest possible rationales - snippets of an input text ' responsible for ' corresponding output - to explain the model prediction , with the assumption that shorter rationales are more intuitive to humans .", "however , this assumption has yet to be validated .", "is the shortest rationale indeed the most human - understandable ?"], "relation": "used for", "id": "2022.acl-short.2", "year": 2022, "rel_sent": "We use LimitedInk to conduct a user study on the impact of rationale length , where we ask human judges to predict the sentiment label of documents based only on LimitedInk - generated rationales with different lengths .", "forward": false, "src_ids": "2022.acl-short.2_4828"}
+{"input": "document - level argument extraction is done by using OtherScientificTerm| context: extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document . while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events .", "entity": "document - level argument extraction", "output": "dynamic global memory", "neg_sample": ["document - level argument extraction is done by using OtherScientificTerm", "extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document .", "while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events ."], "relation": "used for", "id": "2022.acl-long.361", "year": 2022, "rel_sent": "Dynamic Global Memory for Document - level Argument Extraction.", "forward": false, "src_ids": "2022.acl-long.361_4829"}
+{"input": "dynamic global memory is used for Task| context: extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document . while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events .", "entity": "dynamic global memory", "output": "document - level argument extraction", "neg_sample": ["dynamic global memory is used for Task", "extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document .", "while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events ."], "relation": "used for", "id": "2022.acl-long.361", "year": 2022, "rel_sent": "Dynamic Global Memory for Document - level Argument Extraction.", "forward": true, "src_ids": "2022.acl-long.361_4830"}
+{"input": "document - level event argument extraction is done by using Method| context: extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document . while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events .", "entity": "document - level event argument extraction", "output": "global neural generation - based framework", "neg_sample": ["document - level event argument extraction is done by using Method", "extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document .", "while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events ."], "relation": "used for", "id": "2022.acl-long.361", "year": 2022, "rel_sent": "To tackle this issue , we introduce a new global neural generation - based framework for document - level event argument extraction by constructing a document memory store to record the contextual event information and leveraging it to implicitly and explicitly help with decoding of arguments for later events .", "forward": false, "src_ids": "2022.acl-long.361_4831"}
+{"input": "global neural generation - based framework is used for Task| context: extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document . while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events .", "entity": "global neural generation - based framework", "output": "document - level event argument extraction", "neg_sample": ["global neural generation - based framework is used for Task", "extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document .", "while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events ."], "relation": "used for", "id": "2022.acl-long.361", "year": 2022, "rel_sent": "To tackle this issue , we introduce a new global neural generation - based framework for document - level event argument extraction by constructing a document memory store to record the contextual event information and leveraging it to implicitly and explicitly help with decoding of arguments for later events .", "forward": true, "src_ids": "2022.acl-long.361_4832"}
+{"input": "contextual event information is done by using OtherScientificTerm| context: extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document . while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events .", "entity": "contextual event information", "output": "document memory store", "neg_sample": ["contextual event information is done by using OtherScientificTerm", "extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document .", "while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events ."], "relation": "used for", "id": "2022.acl-long.361", "year": 2022, "rel_sent": "To tackle this issue , we introduce a new global neural generation - based framework for document - level event argument extraction by constructing a document memory store to record the contextual event information and leveraging it to implicitly and explicitly help with decoding of arguments for later events .", "forward": false, "src_ids": "2022.acl-long.361_4833"}
+{"input": "document memory store is used for OtherScientificTerm| context: extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document . while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events .", "entity": "document memory store", "output": "contextual event information", "neg_sample": ["document memory store is used for OtherScientificTerm", "extracting informative arguments of events from news articles is a challenging problem in information extraction , which requires a global contextual understanding of each document .", "while recent work on document - level extraction has gone beyond single - sentence and increased the cross - sentence inference capability of end - to - end models , they are still restricted by certain input sequence length constraints and usually ignore the global context between events ."], "relation": "used for", "id": "2022.acl-long.361", "year": 2022, "rel_sent": "To tackle this issue , we introduce a new global neural generation - based framework for document - level event argument extraction by constructing a document memory store to record the contextual event information and leveraging it to implicitly and explicitly help with decoding of arguments for later events .", "forward": true, "src_ids": "2022.acl-long.361_4834"}
+{"input": "cross - lingual transfer is done by using Method| context: pre - trained multilingual language models such as mbert and xlm - r have demonstrated great potential for zero - shot cross - lingual transfer to low web - resource languages ( lrl ) . however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls .", "entity": "cross - lingual transfer", "output": "overlap - based vocabulary generation", "neg_sample": ["cross - lingual transfer is done by using Method", "pre - trained multilingual language models such as mbert and xlm - r have demonstrated great potential for zero - shot cross - lingual transfer to low web - resource languages ( lrl ) .", "however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls ."], "relation": "used for", "id": "2022.acl-long.18", "year": 2022, "rel_sent": "Overlap - based Vocabulary Generation Improves Cross - lingual Transfer Among Related Languages.", "forward": false, "src_ids": "2022.acl-long.18_4835"}
+{"input": "overlap - based vocabulary generation is used for Task| context: however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls .", "entity": "overlap - based vocabulary generation", "output": "cross - lingual transfer", "neg_sample": ["overlap - based vocabulary generation is used for Task", "however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls ."], "relation": "used for", "id": "2022.acl-long.18", "year": 2022, "rel_sent": "Overlap - based Vocabulary Generation Improves Cross - lingual Transfer Among Related Languages.", "forward": true, "src_ids": "2022.acl-long.18_4836"}
+{"input": "representation of lrls is done by using Method| context: pre - trained multilingual language models such as mbert and xlm - r have demonstrated great potential for zero - shot cross - lingual transfer to low web - resource languages ( lrl ) . however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls .", "entity": "representation of lrls", "output": "overlap bpe", "neg_sample": ["representation of lrls is done by using Method", "pre - trained multilingual language models such as mbert and xlm - r have demonstrated great potential for zero - shot cross - lingual transfer to low web - resource languages ( lrl ) .", "however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls ."], "relation": "used for", "id": "2022.acl-long.18", "year": 2022, "rel_sent": "Through extensive experiments on multiple NLP tasks and datasets , we observe that OBPE generates a vocabulary that increases the representation of LRLs via tokens shared with HRLs .", "forward": false, "src_ids": "2022.acl-long.18_4837"}
+{"input": "overlap bpe is used for Task| context: pre - trained multilingual language models such as mbert and xlm - r have demonstrated great potential for zero - shot cross - lingual transfer to low web - resource languages ( lrl ) . however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls .", "entity": "overlap bpe", "output": "representation of lrls", "neg_sample": ["overlap bpe is used for Task", "pre - trained multilingual language models such as mbert and xlm - r have demonstrated great potential for zero - shot cross - lingual transfer to low web - resource languages ( lrl ) .", "however , due to limited model capacity , the large difference in the sizes of available monolingual corpora between high web - resource languages ( hrl ) and lrls does not provide enough scope of co - embedding the lrl with the hrl , thereby affecting the downstream task performance of lrls ."], "relation": "used for", "id": "2022.acl-long.18", "year": 2022, "rel_sent": "Through extensive experiments on multiple NLP tasks and datasets , we observe that OBPE generates a vocabulary that increases the representation of LRLs via tokens shared with HRLs .", "forward": true, "src_ids": "2022.acl-long.18_4838"}
+{"input": "conversational question answering is done by using Method| context: in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios . ( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing .", "entity": "conversational question answering", "output": "knowledge - aware fuzzy semantic parsing framework", "neg_sample": ["conversational question answering is done by using Method", "in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios .", "( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing ."], "relation": "used for", "id": "2022.acl-long.35", "year": 2022, "rel_sent": "KaFSP : Knowledge - Aware Fuzzy Semantic Parsing for Conversational Question Answering over a Large - Scale Knowledge Base.", "forward": false, "src_ids": "2022.acl-long.35_4839"}
+{"input": "knowledge - aware fuzzy semantic parsing framework is used for Task| context: ( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing .", "entity": "knowledge - aware fuzzy semantic parsing framework", "output": "conversational question answering", "neg_sample": ["knowledge - aware fuzzy semantic parsing framework is used for Task", "( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing ."], "relation": "used for", "id": "2022.acl-long.35", "year": 2022, "rel_sent": "KaFSP : Knowledge - Aware Fuzzy Semantic Parsing for Conversational Question Answering over a Large - Scale Knowledge Base.", "forward": true, "src_ids": "2022.acl-long.35_4840"}
+{"input": "grammar system is used for Task| context: ( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing .", "entity": "grammar system", "output": "uncertain reasoning", "neg_sample": ["grammar system is used for Task", "( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing ."], "relation": "used for", "id": "2022.acl-long.35", "year": 2022, "rel_sent": "It defines fuzzy comparison operations in the grammar system for uncertain reasoning based on the fuzzy set theory .", "forward": true, "src_ids": "2022.acl-long.35_4841"}
+{"input": "multi - label classification framework is used for OtherScientificTerm| context: in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios .", "entity": "multi - label classification framework", "output": "knowledge base information", "neg_sample": ["multi - label classification framework is used for OtherScientificTerm", "in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios ."], "relation": "used for", "id": "2022.acl-long.35", "year": 2022, "rel_sent": "Additionally , we propose a multi - label classification framework to not only capture correlations between entity types and relations but also detect knowledge base information relevant to the current utterance .", "forward": true, "src_ids": "2022.acl-long.35_4842"}
+{"input": "uncertain reasoning is done by using Method| context: in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios . ( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing .", "entity": "uncertain reasoning", "output": "grammar system", "neg_sample": ["uncertain reasoning is done by using Method", "in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios .", "( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing ."], "relation": "used for", "id": "2022.acl-long.35", "year": 2022, "rel_sent": "It defines fuzzy comparison operations in the grammar system for uncertain reasoning based on the fuzzy set theory .", "forward": false, "src_ids": "2022.acl-long.35_4843"}
+{"input": "knowledge base information is done by using Method| context: in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios . ( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing .", "entity": "knowledge base information", "output": "multi - label classification framework", "neg_sample": ["knowledge base information is done by using Method", "in this paper , we study two issues of semantic parsing approaches to conversational question answering over a large - scale knowledge base : ( 1 ) the actions defined in grammar are not sufficient to handle uncertain reasoning common in real - world scenarios .", "( 2 ) knowledge base information is not well exploited and incorporated into semantic parsing ."], "relation": "used for", "id": "2022.acl-long.35", "year": 2022, "rel_sent": "Additionally , we propose a multi - label classification framework to not only capture correlations between entity types and relations but also detect knowledge base information relevant to the current utterance .", "forward": false, "src_ids": "2022.acl-long.35_4844"}
+{"input": "multi - way aligned corpus is done by using Method| context: complete multi - lingual neural machine translation ( c - mnmt ) achieves superior performance against the conventional mnmt by constructing multi - way aligned corpus , i.e. , aligning bilingual training examples from different language pairs when either their source or target sides are identical . however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale .", "entity": "multi - way aligned corpus", "output": "extract and generate '", "neg_sample": ["multi - way aligned corpus is done by using Method", "complete multi - lingual neural machine translation ( c - mnmt ) achieves superior performance against the conventional mnmt by constructing multi - way aligned corpus , i.e.", ", aligning bilingual training examples from different language pairs when either their source or target sides are identical .", "however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale ."], "relation": "used for", "id": "2022.acl-long.560", "year": 2022, "rel_sent": "EAG : Extract and Generate Multi - way Aligned Corpus for Complete Multi - lingual Neural Machine Translation.", "forward": false, "src_ids": "2022.acl-long.560_4845"}
+{"input": "two - step pipeline is used for Method| context: complete multi - lingual neural machine translation ( c - mnmt ) achieves superior performance against the conventional mnmt by constructing multi - way aligned corpus , i.e. , aligning bilingual training examples from different language pairs when either their source or target sides are identical . however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale .", "entity": "two - step pipeline", "output": "extract and generate '", "neg_sample": ["two - step pipeline is used for Method", "complete multi - lingual neural machine translation ( c - mnmt ) achieves superior performance against the conventional mnmt by constructing multi - way aligned corpus , i.e.", ", aligning bilingual training examples from different language pairs when either their source or target sides are identical .", "however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale ."], "relation": "used for", "id": "2022.acl-long.560", "year": 2022, "rel_sent": "With this two - step pipeline , EAG can construct a large - scale and multi - way aligned corpus whose diversity is almost identical to the original bilingual corpus .", "forward": true, "src_ids": "2022.acl-long.560_4846"}
+{"input": "multi - way aligned corpus is used for Task| context: however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale .", "entity": "multi - way aligned corpus", "output": "multi - lingual neural machine translation", "neg_sample": ["multi - way aligned corpus is used for Task", "however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale ."], "relation": "used for", "id": "2022.acl-long.560", "year": 2022, "rel_sent": "EAG : Extract and Generate Multi - way Aligned Corpus for Complete Multi - lingual Neural Machine Translation.", "forward": true, "src_ids": "2022.acl-long.560_4847"}
+{"input": "extract and generate ' is done by using Generic| context: complete multi - lingual neural machine translation ( c - mnmt ) achieves superior performance against the conventional mnmt by constructing multi - way aligned corpus , i.e. , aligning bilingual training examples from different language pairs when either their source or target sides are identical . however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale .", "entity": "extract and generate '", "output": "two - step pipeline", "neg_sample": ["extract and generate ' is done by using Generic", "complete multi - lingual neural machine translation ( c - mnmt ) achieves superior performance against the conventional mnmt by constructing multi - way aligned corpus , i.e.", ", aligning bilingual training examples from different language pairs when either their source or target sides are identical .", "however , since exactly identical sentences from different language pairs are scarce , the power of the multi - way aligned corpus is limited by its scale ."], "relation": "used for", "id": "2022.acl-long.560", "year": 2022, "rel_sent": "With this two - step pipeline , EAG can construct a large - scale and multi - way aligned corpus whose diversity is almost identical to the original bilingual corpus .", "forward": false, "src_ids": "2022.acl-long.560_4848"}
+{"input": "text - to - sql parsing is done by using OtherScientificTerm| context: previous work has observed that leveraging lexico - logical alignments is very helpful to improve parsing performance . however , current attention - based approaches can only model such alignments at the token level and have unsatisfactory generalization capability .", "entity": "text - to - sql parsing", "output": "explicit lexico - logical alignments", "neg_sample": ["text - to - sql parsing is done by using OtherScientificTerm", "previous work has observed that leveraging lexico - logical alignments is very helpful to improve parsing performance .", "however , current attention - based approaches can only model such alignments at the token level and have unsatisfactory generalization capability ."], "relation": "used for", "id": "2022.acl-short.31", "year": 2022, "rel_sent": "Leveraging Explicit Lexico - logical Alignments in Text - to - SQL Parsing.", "forward": false, "src_ids": "2022.acl-short.31_4849"}
+{"input": "explicit lexico - logical alignments is used for Task| context: previous work has observed that leveraging lexico - logical alignments is very helpful to improve parsing performance . however , current attention - based approaches can only model such alignments at the token level and have unsatisfactory generalization capability .", "entity": "explicit lexico - logical alignments", "output": "text - to - sql parsing", "neg_sample": ["explicit lexico - logical alignments is used for Task", "previous work has observed that leveraging lexico - logical alignments is very helpful to improve parsing performance .", "however , current attention - based approaches can only model such alignments at the token level and have unsatisfactory generalization capability ."], "relation": "used for", "id": "2022.acl-short.31", "year": 2022, "rel_sent": "Leveraging Explicit Lexico - logical Alignments in Text - to - SQL Parsing.", "forward": true, "src_ids": "2022.acl-short.31_4850"}
+{"input": "data augmentation is used for Task| context: data augmentation strategies are increasingly important in nlp pipelines for low - resourced and endangered languages , and in neural morphological inflection , augmentation by so called data hallucination is a popular technique .", "entity": "data augmentation", "output": "computational morphology", "neg_sample": ["data augmentation is used for Task", "data augmentation strategies are increasingly important in nlp pipelines for low - resourced and endangered languages , and in neural morphological inflection , augmentation by so called data hallucination is a popular technique ."], "relation": "used for", "id": "2022.computel-1.5", "year": 2022, "rel_sent": "Our results indicate that further innovations in data augmentation for computational morphology are desirable .", "forward": true, "src_ids": "2022.computel-1.5_4851"}
+{"input": "factual knowledge is done by using OtherScientificTerm| context: large - scale pretrained language models are surprisingly good at recalling factual knowledge presented in the training corpus .", "entity": "factual knowledge", "output": "knowledge neurons", "neg_sample": ["factual knowledge is done by using OtherScientificTerm", "large - scale pretrained language models are surprisingly good at recalling factual knowledge presented in the training corpus ."], "relation": "used for", "id": "2022.acl-long.581", "year": 2022, "rel_sent": "In our case studies , we attempt to leverage knowledge neurons to edit ( such as update , and erase ) specific factual knowledge without fine - tuning .", "forward": false, "src_ids": "2022.acl-long.581_4852"}
+{"input": "data augmentation method is done by using Method| context: depression is a common mental disorder that severely affects the quality of life , and can lead to suicide . when diagnosed in time , mild , moderate , and even severe depression can be treated . this is why it is vital to detect signs of depression in time . one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected .", "entity": "data augmentation method", "output": "abstractive summarization techniques", "neg_sample": ["data augmentation method is done by using Method", "depression is a common mental disorder that severely affects the quality of life , and can lead to suicide .", "when diagnosed in time , mild , moderate , and even severe depression can be treated .", "this is why it is vital to detect signs of depression in time .", "one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected ."], "relation": "used for", "id": "2022.ltedi-1.41", "year": 2022, "rel_sent": "In particular , we use abstractive summarization techniques for data augmentation .", "forward": false, "src_ids": "2022.ltedi-1.41_4853"}
+{"input": "text classification is done by using Method| context: depression is a common mental disorder that severely affects the quality of life , and can lead to suicide . when diagnosed in time , mild , moderate , and even severe depression can be treated . this is why it is vital to detect signs of depression in time . one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected .", "entity": "text classification", "output": "data augmentation method", "neg_sample": ["text classification is done by using Method", "depression is a common mental disorder that severely affects the quality of life , and can lead to suicide .", "when diagnosed in time , mild , moderate , and even severe depression can be treated .", "this is why it is vital to detect signs of depression in time .", "one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected ."], "relation": "used for", "id": "2022.ltedi-1.41", "year": 2022, "rel_sent": "FilipN@LT - EDI - ACL2022 - Detecting signs of Depression from Social Media : Examining the use of summarization methods as data augmentation for text classification.", "forward": false, "src_ids": "2022.ltedi-1.41_4854"}
+{"input": "abstractive summarization techniques is used for Method| context: depression is a common mental disorder that severely affects the quality of life , and can lead to suicide . when diagnosed in time , mild , moderate , and even severe depression can be treated . this is why it is vital to detect signs of depression in time . one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected .", "entity": "abstractive summarization techniques", "output": "data augmentation method", "neg_sample": ["abstractive summarization techniques is used for Method", "depression is a common mental disorder that severely affects the quality of life , and can lead to suicide .", "when diagnosed in time , mild , moderate , and even severe depression can be treated .", "this is why it is vital to detect signs of depression in time .", "one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected ."], "relation": "used for", "id": "2022.ltedi-1.41", "year": 2022, "rel_sent": "In particular , we use abstractive summarization techniques for data augmentation .", "forward": true, "src_ids": "2022.ltedi-1.41_4855"}
+{"input": "data augmentation method is used for Task| context: depression is a common mental disorder that severely affects the quality of life , and can lead to suicide . when diagnosed in time , mild , moderate , and even severe depression can be treated . this is why it is vital to detect signs of depression in time . one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected .", "entity": "data augmentation method", "output": "text classification", "neg_sample": ["data augmentation method is used for Task", "depression is a common mental disorder that severely affects the quality of life , and can lead to suicide .", "when diagnosed in time , mild , moderate , and even severe depression can be treated .", "this is why it is vital to detect signs of depression in time .", "one drawback , however , is that when the dataset is imbalanced , the performance of these models may be negatively affected ."], "relation": "used for", "id": "2022.ltedi-1.41", "year": 2022, "rel_sent": "FilipN@LT - EDI - ACL2022 - Detecting signs of Depression from Social Media : Examining the use of summarization methods as data augmentation for text classification.", "forward": true, "src_ids": "2022.ltedi-1.41_4856"}
+{"input": "pre - trained language models is done by using Method| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "pre - trained language models", "output": "pluggable entity lookup table", "neg_sample": ["pre - trained language models is done by using Method", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "PELT can be compatibly plugged as inputs to infuse supplemental entity knowledge into PLMs .", "forward": false, "src_ids": "2022.acl-short.57_4857"}
+{"input": "entity knowledge is done by using Method| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "entity knowledge", "output": "pluggable entity lookup table", "neg_sample": ["entity knowledge is done by using Method", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "PELT can be compatibly plugged as inputs to infuse supplemental entity knowledge into PLMs .", "forward": false, "src_ids": "2022.acl-short.57_4858"}
+{"input": "domain adaptation scenario is done by using Method| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "domain adaptation scenario", "output": "pluggable entity lookup table", "neg_sample": ["domain adaptation scenario is done by using Method", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "Compared to previous knowledge - enhanced PLMs , PELT only requires 0.2%-5 % pre - computation with capability of acquiring knowledge from out - of - domain corpora for domain adaptation scenario .", "forward": false, "src_ids": "2022.acl-short.57_4859"}
+{"input": "entity knowledge is done by using Method| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "entity knowledge", "output": "pluggable entity lookup table", "neg_sample": ["entity knowledge is done by using Method", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "The experiments on knowledge - related tasks demonstrate that our method , PELT , can flexibly and effectively transfer entity knowledge from related corpora into PLMs with different architectures .", "forward": false, "src_ids": "2022.acl-short.57_4860"}
+{"input": "pluggable entity lookup table is used for Task| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "pluggable entity lookup table", "output": "domain adaptation scenario", "neg_sample": ["pluggable entity lookup table is used for Task", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "Compared to previous knowledge - enhanced PLMs , PELT only requires 0.2%-5 % pre - computation with capability of acquiring knowledge from out - of - domain corpora for domain adaptation scenario .", "forward": true, "src_ids": "2022.acl-short.57_4861"}
+{"input": "pre - trained language models is done by using OtherScientificTerm| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "pre - trained language models", "output": "entity knowledge", "neg_sample": ["pre - trained language models is done by using OtherScientificTerm", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "PELT can be compatibly plugged as inputs to infuse supplemental entity knowledge into PLMs .", "forward": false, "src_ids": "2022.acl-short.57_4862"}
+{"input": "pluggable entity lookup table is used for OtherScientificTerm| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "pluggable entity lookup table", "output": "entity knowledge", "neg_sample": ["pluggable entity lookup table is used for OtherScientificTerm", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "PELT can be compatibly plugged as inputs to infuse supplemental entity knowledge into PLMs .", "forward": true, "src_ids": "2022.acl-short.57_4863"}
+{"input": "pluggable entity lookup table is used for OtherScientificTerm| context: pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities .", "entity": "pluggable entity lookup table", "output": "entity knowledge", "neg_sample": ["pluggable entity lookup table is used for OtherScientificTerm", "pre - trained language models ( plms ) can not well recall rich factual knowledge of entities exhibited in large - scale corpora , especially those rare entities ."], "relation": "used for", "id": "2022.acl-short.57", "year": 2022, "rel_sent": "The experiments on knowledge - related tasks demonstrate that our method , PELT , can flexibly and effectively transfer entity knowledge from related corpora into PLMs with different architectures .", "forward": true, "src_ids": "2022.acl-short.57_4864"}
+{"input": "goal - directed natural language generation is done by using OtherScientificTerm| context: we consider the problem of generating natural language given a communicative goal and a world description . we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?", "entity": "goal - directed natural language generation", "output": "hybrid semantics", "neg_sample": ["goal - directed natural language generation is done by using OtherScientificTerm", "we consider the problem of generating natural language given a communicative goal and a world description .", "we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?"], "relation": "used for", "id": "2022.acl-long.136", "year": 2022, "rel_sent": "Hybrid Semantics for Goal - Directed Natural Language Generation.", "forward": false, "src_ids": "2022.acl-long.136_4865"}
+{"input": "hybrid semantics is used for Task| context: we consider the problem of generating natural language given a communicative goal and a world description . we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?", "entity": "hybrid semantics", "output": "goal - directed natural language generation", "neg_sample": ["hybrid semantics is used for Task", "we consider the problem of generating natural language given a communicative goal and a world description .", "we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?"], "relation": "used for", "id": "2022.acl-long.136", "year": 2022, "rel_sent": "Hybrid Semantics for Goal - Directed Natural Language Generation.", "forward": true, "src_ids": "2022.acl-long.136_4866"}
+{"input": "sentence generation is done by using Method| context: we consider the problem of generating natural language given a communicative goal and a world description . we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?", "entity": "sentence generation", "output": "s - struct", "neg_sample": ["sentence generation is done by using Method", "we consider the problem of generating natural language given a communicative goal and a world description .", "we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?"], "relation": "used for", "id": "2022.acl-long.136", "year": 2022, "rel_sent": "We build upon an existing goal - directed generation system , S - STRUCT , which models sentence generation as planning in a Markov decision process .", "forward": false, "src_ids": "2022.acl-long.136_4867"}
+{"input": "s - struct is used for Task| context: we consider the problem of generating natural language given a communicative goal and a world description . we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?", "entity": "s - struct", "output": "sentence generation", "neg_sample": ["s - struct is used for Task", "we consider the problem of generating natural language given a communicative goal and a world description .", "we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?"], "relation": "used for", "id": "2022.acl-long.136", "year": 2022, "rel_sent": "We build upon an existing goal - directed generation system , S - STRUCT , which models sentence generation as planning in a Markov decision process .", "forward": true, "src_ids": "2022.acl-long.136_4868"}
+{"input": "s - struct 's generation is done by using Method| context: we consider the problem of generating natural language given a communicative goal and a world description . we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?", "entity": "s - struct 's generation", "output": "hybrid method", "neg_sample": ["s - struct 's generation is done by using Method", "we consider the problem of generating natural language given a communicative goal and a world description .", "we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?"], "relation": "used for", "id": "2022.acl-long.136", "year": 2022, "rel_sent": "We find that our hybrid method allows S - STRUCT 's generation to scale significantly better in early phases of generation and that the hybrid can often generate sentences with the same quality as S - STRUCT in substantially less time .", "forward": false, "src_ids": "2022.acl-long.136_4869"}
+{"input": "hybrid method is used for Task| context: we consider the problem of generating natural language given a communicative goal and a world description . we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?", "entity": "hybrid method", "output": "s - struct 's generation", "neg_sample": ["hybrid method is used for Task", "we consider the problem of generating natural language given a communicative goal and a world description .", "we ask the question : is it possible to combine complementary meaning representations to scale a goal - directed nlg system without losing expressiveness ?"], "relation": "used for", "id": "2022.acl-long.136", "year": 2022, "rel_sent": "We find that our hybrid method allows S - STRUCT 's generation to scale significantly better in early phases of generation and that the hybrid can often generate sentences with the same quality as S - STRUCT in substantially less time .", "forward": true, "src_ids": "2022.acl-long.136_4870"}
+{"input": "pre - trained language models is used for Task| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "pre - trained language models", "output": "open relation modeling problem", "neg_sample": ["pre - trained language models is used for Task", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling , we incorporate reasoning paths in KGs and include a reasoning path selection mechanism .", "forward": true, "src_ids": "2022.findings-acl.26_4871"}
+{"input": "relational knowledge is used for Task| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "relational knowledge", "output": "open relation modeling problem", "neg_sample": ["relational knowledge is used for Task", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling , we incorporate reasoning paths in KGs and include a reasoning path selection mechanism .", "forward": true, "src_ids": "2022.findings-acl.26_4872"}
+{"input": "definition - like relation descriptions is done by using OtherScientificTerm| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "definition - like relation descriptions", "output": "machines", "neg_sample": ["definition - like relation descriptions is done by using OtherScientificTerm", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To solve this problem , we propose to teach machines to generate definition - like relation descriptions by letting them learn from defining entities .", "forward": false, "src_ids": "2022.findings-acl.26_4873"}
+{"input": "machines is used for OtherScientificTerm| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "machines", "output": "definition - like relation descriptions", "neg_sample": ["machines is used for OtherScientificTerm", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To solve this problem , we propose to teach machines to generate definition - like relation descriptions by letting them learn from defining entities .", "forward": true, "src_ids": "2022.findings-acl.26_4874"}
+{"input": "open relation modeling problem is done by using Method| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "open relation modeling problem", "output": "pre - trained language models", "neg_sample": ["open relation modeling problem is done by using Method", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling , we incorporate reasoning paths in KGs and include a reasoning path selection mechanism .", "forward": false, "src_ids": "2022.findings-acl.26_4875"}
+{"input": "relational knowledge is used for Method| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "relational knowledge", "output": "pre - trained language models", "neg_sample": ["relational knowledge is used for Method", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling , we incorporate reasoning paths in KGs and include a reasoning path selection mechanism .", "forward": true, "src_ids": "2022.findings-acl.26_4876"}
+{"input": "pre - trained language models is done by using OtherScientificTerm| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "pre - trained language models", "output": "relational knowledge", "neg_sample": ["pre - trained language models is done by using OtherScientificTerm", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling , we incorporate reasoning paths in KGs and include a reasoning path selection mechanism .", "forward": false, "src_ids": "2022.findings-acl.26_4877"}
+{"input": "open relation modeling problem is done by using OtherScientificTerm| context: relations between entities can be represented by different instances , e.g. , a sentence containing both entities or a fact in a knowledge graph ( kg ) . however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source .", "entity": "open relation modeling problem", "output": "relational knowledge", "neg_sample": ["open relation modeling problem is done by using OtherScientificTerm", "relations between entities can be represented by different instances , e.g.", ", a sentence containing both entities or a fact in a knowledge graph ( kg ) .", "however , these instances may not well capture the general relations between entities , may be difficult to understand by humans , even may not be found due to the incompleteness of the knowledge source ."], "relation": "used for", "id": "2022.findings-acl.26", "year": 2022, "rel_sent": "To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling , we incorporate reasoning paths in KGs and include a reasoning path selection mechanism .", "forward": false, "src_ids": "2022.findings-acl.26_4878"}
+{"input": "natural language processing is done by using OtherScientificTerm| context: meaning is context - dependent , but many properties of language ( should ) remain the same even if we transform the context . for example , sentiment or speaker properties should be the same in a translation and original of a text .", "entity": "natural language processing", "output": "language invariant properties", "neg_sample": ["natural language processing is done by using OtherScientificTerm", "meaning is context - dependent , but many properties of language ( should ) remain the same even if we transform the context .", "for example , sentiment or speaker properties should be the same in a translation and original of a text ."], "relation": "used for", "id": "2022.nlppower-1.9", "year": 2022, "rel_sent": "Language Invariant Properties in Natural Language Processing.", "forward": false, "src_ids": "2022.nlppower-1.9_4879"}
+{"input": "language invariant properties is used for Task| context: meaning is context - dependent , but many properties of language ( should ) remain the same even if we transform the context . for example , sentiment or speaker properties should be the same in a translation and original of a text .", "entity": "language invariant properties", "output": "natural language processing", "neg_sample": ["language invariant properties is used for Task", "meaning is context - dependent , but many properties of language ( should ) remain the same even if we transform the context .", "for example , sentiment or speaker properties should be the same in a translation and original of a text ."], "relation": "used for", "id": "2022.nlppower-1.9", "year": 2022, "rel_sent": "Language Invariant Properties in Natural Language Processing.", "forward": true, "src_ids": "2022.nlppower-1.9_4880"}
+{"input": "hope speech detection is done by using Method| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "hope speech detection", "output": "dual channel language modeling", "neg_sample": ["hope speech detection is done by using Method", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "The Best of both Worlds : Dual Channel Language modeling for Hope Speech Detection in low - resourced Kannada.", "forward": false, "src_ids": "2022.ltedi-1.14_4881"}
+{"input": "hope speech is done by using Method| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "hope speech", "output": "dual channel language modeling", "neg_sample": ["hope speech is done by using Method", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "As a result , we present DC - LM , a dual - channel language model that sees hope speech by using the English translations of the code - mixed dataset for additional training .", "forward": false, "src_ids": "2022.ltedi-1.14_4882"}
+{"input": "dual channel language modeling is used for Task| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "dual channel language modeling", "output": "hope speech detection", "neg_sample": ["dual channel language modeling is used for Task", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "The Best of both Worlds : Dual Channel Language modeling for Hope Speech Detection in low - resourced Kannada.", "forward": true, "src_ids": "2022.ltedi-1.14_4883"}
+{"input": "dual channel language modeling is used for OtherScientificTerm| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "entity": "dual channel language modeling", "output": "hope speech", "neg_sample": ["dual channel language modeling is used for OtherScientificTerm", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "As a result , we present DC - LM , a dual - channel language model that sees hope speech by using the English translations of the code - mixed dataset for additional training .", "forward": true, "src_ids": "2022.ltedi-1.14_4884"}
+{"input": "code - mixed dataset is done by using Material| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "code - mixed dataset", "output": "english translations", "neg_sample": ["code - mixed dataset is done by using Material", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "As a result , we present DC - LM , a dual - channel language model that sees hope speech by using the English translations of the code - mixed dataset for additional training .", "forward": false, "src_ids": "2022.ltedi-1.14_4885"}
+{"input": "english translations is used for Generic| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "english translations", "output": "code - mixed dataset", "neg_sample": ["english translations is used for Generic", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "As a result , we present DC - LM , a dual - channel language model that sees hope speech by using the English translations of the code - mixed dataset for additional training .", "forward": true, "src_ids": "2022.ltedi-1.14_4886"}
+{"input": "positive and supportive online content is done by using Method| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "positive and supportive online content", "output": "pragmatic approach", "neg_sample": ["positive and supportive online content is done by using Method", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "We aim to initiate research in Kannada while encouraging researchers to take a pragmatic approach to inspire positive and supportive online content .", "forward": false, "src_ids": "2022.ltedi-1.14_4887"}
+{"input": "pragmatic approach is used for Material| context: in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms . there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums . as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada .", "entity": "pragmatic approach", "output": "positive and supportive online content", "neg_sample": ["pragmatic approach is used for Material", "in recent years , various methods have been developed to control the spread of negativity by removing profane , aggressive , and offensive comments from social media platforms .", "there is , however , a scarcity of research focusing on embracing positivity and reinforcing supportive and reassuring content in online forums .", "as a result , we concentrate our research on developing systems to detect hope speech in code - mixed kannada ."], "relation": "used for", "id": "2022.ltedi-1.14", "year": 2022, "rel_sent": "We aim to initiate research in Kannada while encouraging researchers to take a pragmatic approach to inspire positive and supportive online content .", "forward": true, "src_ids": "2022.ltedi-1.14_4888"}
+{"input": "coarse - grained response selection is done by using Method| context: we study the problem of coarse - grained response selection in retrieval - based dialogue systems . the problem is equally important with fine - grained response selection , but is less explored in existing literature .", "entity": "coarse - grained response selection", "output": "contextual fine - to - coarse distillation", "neg_sample": ["coarse - grained response selection is done by using Method", "we study the problem of coarse - grained response selection in retrieval - based dialogue systems .", "the problem is equally important with fine - grained response selection , but is less explored in existing literature ."], "relation": "used for", "id": "2022.acl-long.334", "year": 2022, "rel_sent": "Contextual Fine - to - Coarse Distillation for Coarse - grained Response Selection in Open - Domain Conversations.", "forward": false, "src_ids": "2022.acl-long.334_4889"}
+{"input": "contextual fine - to - coarse distillation is used for Task| context: the problem is equally important with fine - grained response selection , but is less explored in existing literature .", "entity": "contextual fine - to - coarse distillation", "output": "coarse - grained response selection", "neg_sample": ["contextual fine - to - coarse distillation is used for Task", "the problem is equally important with fine - grained response selection , but is less explored in existing literature ."], "relation": "used for", "id": "2022.acl-long.334", "year": 2022, "rel_sent": "Contextual Fine - to - Coarse Distillation for Coarse - grained Response Selection in Open - Domain Conversations.", "forward": true, "src_ids": "2022.acl-long.334_4890"}
+{"input": "low - resource languages is done by using Method| context: cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones . recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources . however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e. sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs .", "entity": "low - resource languages", "output": "isomorphic cross - lingual embeddings", "neg_sample": ["low - resource languages is done by using Method", "cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones .", "recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources .", "however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e.", "sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs ."], "relation": "used for", "id": "2022.repl4nlp-1.14", "year": 2022, "rel_sent": "Isomorphic Cross - lingual Embeddings for Low - Resource Languages.", "forward": false, "src_ids": "2022.repl4nlp-1.14_4891"}
+{"input": "isomorphic cross - lingual embeddings is used for Material| context: cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones . recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources . however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e. sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs .", "entity": "isomorphic cross - lingual embeddings", "output": "low - resource languages", "neg_sample": ["isomorphic cross - lingual embeddings is used for Material", "cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones .", "recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources .", "however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e.", "sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs ."], "relation": "used for", "id": "2022.repl4nlp-1.14", "year": 2022, "rel_sent": "Isomorphic Cross - lingual Embeddings for Low - Resource Languages.", "forward": true, "src_ids": "2022.repl4nlp-1.14_4892"}
+{"input": "joint training methods is used for Method| context: recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources . however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e. sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs .", "entity": "joint training methods", "output": "cross - lingual word embeddings", "neg_sample": ["joint training methods is used for Method", "recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources .", "however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e.", "sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs ."], "relation": "used for", "id": "2022.repl4nlp-1.14", "year": 2022, "rel_sent": "Following this , we use joint training methods to develops CLWEs for the related language and the target embedding space .", "forward": true, "src_ids": "2022.repl4nlp-1.14_4893"}
+{"input": "offline methods is used for OtherScientificTerm| context: cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones . recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources . however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e. sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs .", "entity": "offline methods", "output": "isometry", "neg_sample": ["offline methods is used for OtherScientificTerm", "cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones .", "recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources .", "however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e.", "sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs ."], "relation": "used for", "id": "2022.repl4nlp-1.14", "year": 2022, "rel_sent": "In our work , we first pre - align the low - resource and related language embedding spaces using offline methods to mitigate the assumption of isometry .", "forward": true, "src_ids": "2022.repl4nlp-1.14_4894"}
+{"input": "isometry is done by using Method| context: cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones . recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources . however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e. sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs .", "entity": "isometry", "output": "offline methods", "neg_sample": ["isometry is done by using Method", "cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones .", "recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources .", "however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e.", "sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs ."], "relation": "used for", "id": "2022.repl4nlp-1.14", "year": 2022, "rel_sent": "In our work , we first pre - align the low - resource and related language embedding spaces using offline methods to mitigate the assumption of isometry .", "forward": false, "src_ids": "2022.repl4nlp-1.14_4895"}
+{"input": "cross - lingual word embeddings is done by using Method| context: cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones . recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources . however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e. sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs .", "entity": "cross - lingual word embeddings", "output": "joint training methods", "neg_sample": ["cross - lingual word embeddings is done by using Method", "cross - lingual word embeddings ( clwes ) are a key component to transfer linguistic information learnt from higher - resource settings into lower - resource ones .", "recent research in cross - lingual representation learning has focused on offline mapping approaches due to their simplicity , computational efficacy , and ability to work with minimal parallel resources .", "however , they crucially depend on the assumption of embedding spaces being approximately isomorphic i.e.", "sharing similar geometric structure , which does not hold in practice , leading to poorer performance on low - resource and distant language pairs ."], "relation": "used for", "id": "2022.repl4nlp-1.14", "year": 2022, "rel_sent": "Following this , we use joint training methods to develops CLWEs for the related language and the target embedding space .", "forward": false, "src_ids": "2022.repl4nlp-1.14_4896"}
+{"input": "semantic change is done by using Task| context: languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate .", "entity": "semantic change", "output": "causal analysis", "neg_sample": ["semantic change is done by using Task", "languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate ."], "relation": "used for", "id": "2022.acl-long.101", "year": 2022, "rel_sent": "Slangvolution : A Causal Analysis of Semantic Change and Frequency Dynamics in Slang.", "forward": false, "src_ids": "2022.acl-long.101_4897"}
+{"input": "frequency dynamics is done by using Task| context: languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate .", "entity": "frequency dynamics", "output": "causal analysis", "neg_sample": ["frequency dynamics is done by using Task", "languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate ."], "relation": "used for", "id": "2022.acl-long.101", "year": 2022, "rel_sent": "Slangvolution : A Causal Analysis of Semantic Change and Frequency Dynamics in Slang.", "forward": false, "src_ids": "2022.acl-long.101_4898"}
+{"input": "causal analysis is used for OtherScientificTerm| context: languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate .", "entity": "causal analysis", "output": "semantic change", "neg_sample": ["causal analysis is used for OtherScientificTerm", "languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate ."], "relation": "used for", "id": "2022.acl-long.101", "year": 2022, "rel_sent": "Slangvolution : A Causal Analysis of Semantic Change and Frequency Dynamics in Slang.", "forward": true, "src_ids": "2022.acl-long.101_4899"}
+{"input": "causal analysis is used for Task| context: languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate .", "entity": "causal analysis", "output": "frequency dynamics", "neg_sample": ["causal analysis is used for Task", "languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate ."], "relation": "used for", "id": "2022.acl-long.101", "year": 2022, "rel_sent": "Slangvolution : A Causal Analysis of Semantic Change and Frequency Dynamics in Slang.", "forward": true, "src_ids": "2022.acl-long.101_4900"}
+{"input": "language change is done by using OtherScientificTerm| context: languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate .", "entity": "language change", "output": "distributional factors", "neg_sample": ["language change is done by using OtherScientificTerm", "languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate ."], "relation": "used for", "id": "2022.acl-long.101", "year": 2022, "rel_sent": "In this work , we approach language evolution through the lens of causality in order to model not only how various distributional factors associate with language change , but how they causally affect it .", "forward": false, "src_ids": "2022.acl-long.101_4901"}
+{"input": "distributional factors is used for Task| context: languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate .", "entity": "distributional factors", "output": "language change", "neg_sample": ["distributional factors is used for Task", "languages are continuously undergoing changes , and the mechanisms that underlie these changes are still a matter of debate ."], "relation": "used for", "id": "2022.acl-long.101", "year": 2022, "rel_sent": "In this work , we approach language evolution through the lens of causality in order to model not only how various distributional factors associate with language change , but how they causally affect it .", "forward": true, "src_ids": "2022.acl-long.101_4902"}
+{"input": "re - ranking is done by using Method| context: sequence - to - sequence neural networks have recently achieved great success in abstractive summarization , especially through fine - tuning large pre - trained language models on the downstream dataset . these models are typically decoded with beam search to generate a unique summary . however , the search space is very large , and with the exposure bias , such decoding is not optimal .", "entity": "re - ranking", "output": "second - stage model", "neg_sample": ["re - ranking is done by using Method", "sequence - to - sequence neural networks have recently achieved great success in abstractive summarization , especially through fine - tuning large pre - trained language models on the downstream dataset .", "these models are typically decoded with beam search to generate a unique summary .", "however , the search space is very large , and with the exposure bias , such decoding is not optimal ."], "relation": "used for", "id": "2022.acl-long.309", "year": 2022, "rel_sent": "In this paper , we show that it is possible to directly train a second - stage model performing re - ranking on a set of summary candidates .", "forward": false, "src_ids": "2022.acl-long.309_4903"}
+{"input": "second - stage model is used for Task| context: sequence - to - sequence neural networks have recently achieved great success in abstractive summarization , especially through fine - tuning large pre - trained language models on the downstream dataset . these models are typically decoded with beam search to generate a unique summary . however , the search space is very large , and with the exposure bias , such decoding is not optimal .", "entity": "second - stage model", "output": "re - ranking", "neg_sample": ["second - stage model is used for Task", "sequence - to - sequence neural networks have recently achieved great success in abstractive summarization , especially through fine - tuning large pre - trained language models on the downstream dataset .", "these models are typically decoded with beam search to generate a unique summary .", "however , the search space is very large , and with the exposure bias , such decoding is not optimal ."], "relation": "used for", "id": "2022.acl-long.309", "year": 2022, "rel_sent": "In this paper , we show that it is possible to directly train a second - stage model performing re - ranking on a set of summary candidates .", "forward": true, "src_ids": "2022.acl-long.309_4904"}
+{"input": "question rewriting is used for Method| context: question rewriting ( qr ) is a subtask of conversational question answering ( cqa ) aiming to ease the challenges of understanding dependencies among dialogue history by reformulating questions in a self - contained form . despite seeming plausible , little evidence is available to justify qr as a mitigation method for cqa .", "entity": "question rewriting", "output": "reinforcement learning approach", "neg_sample": ["question rewriting is used for Method", "question rewriting ( qr ) is a subtask of conversational question answering ( cqa ) aiming to ease the challenges of understanding dependencies among dialogue history by reformulating questions in a self - contained form .", "despite seeming plausible , little evidence is available to justify qr as a mitigation method for cqa ."], "relation": "used for", "id": "2022.insights-1.13", "year": 2022, "rel_sent": "To verify the effectiveness of QR in CQA , we investigate a reinforcement learning approach that integrates QR and CQA tasks and does not require corresponding QR datasets for targeted CQA.We find , however , that the RL method is on par with the end - to - end baseline .", "forward": true, "src_ids": "2022.insights-1.13_4905"}
+{"input": "in - domain dataset is done by using Method| context: vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment . to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets . however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes .", "entity": "in - domain dataset", "output": "clip", "neg_sample": ["in - domain dataset is done by using Method", "vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment .", "to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets .", "however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes ."], "relation": "used for", "id": "2022.acl-long.332", "year": 2022, "rel_sent": "Our method fully utilizes the knowledge learned from CLIP to build an in - domain dataset by self - exploration without human labeling .", "forward": false, "src_ids": "2022.acl-long.332_4906"}
+{"input": "clip is used for Material| context: vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment . to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets . however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes .", "entity": "clip", "output": "in - domain dataset", "neg_sample": ["clip is used for Material", "vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment .", "to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets .", "however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes ."], "relation": "used for", "id": "2022.acl-long.332", "year": 2022, "rel_sent": "Our method fully utilizes the knowledge learned from CLIP to build an in - domain dataset by self - exploration without human labeling .", "forward": true, "src_ids": "2022.acl-long.332_4907"}
+{"input": "fast adaptation is done by using Method| context: vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment . to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets . however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes .", "entity": "fast adaptation", "output": "prompt tuning", "neg_sample": ["fast adaptation is done by using Method", "vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment .", "to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets .", "however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes ."], "relation": "used for", "id": "2022.acl-long.332", "year": 2022, "rel_sent": "Unlike the conventional approach of fine - tuning , we introduce prompt tuning to achieve fast adaptation for language embeddings , which substantially improves the learning efficiency by leveraging prior knowledge .", "forward": false, "src_ids": "2022.acl-long.332_4908"}
+{"input": "prompt tuning is used for Task| context: vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment . to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets . however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes .", "entity": "prompt tuning", "output": "fast adaptation", "neg_sample": ["prompt tuning is used for Task", "vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment .", "to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets .", "however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes ."], "relation": "used for", "id": "2022.acl-long.332", "year": 2022, "rel_sent": "Unlike the conventional approach of fine - tuning , we introduce prompt tuning to achieve fast adaptation for language embeddings , which substantially improves the learning efficiency by leveraging prior knowledge .", "forward": true, "src_ids": "2022.acl-long.332_4909"}
+{"input": "navigation model is done by using Method| context: vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment . to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets . however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes .", "entity": "navigation model", "output": "prompt - based environmental self - exploration", "neg_sample": ["navigation model is done by using Method", "vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment .", "to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets .", "however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes ."], "relation": "used for", "id": "2022.acl-long.332", "year": 2022, "rel_sent": "Both qualitative and quantitative results show that our ProbES significantly improves the generalization ability of the navigation model .", "forward": false, "src_ids": "2022.acl-long.332_4910"}
+{"input": "prompt - based environmental self - exploration is used for Method| context: vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment . to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets . however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes .", "entity": "prompt - based environmental self - exploration", "output": "navigation model", "neg_sample": ["prompt - based environmental self - exploration is used for Method", "vision - language navigation ( vln ) is a challenging task due to its large searching space in the environment .", "to address this problem , previous works have proposed some methods of fine - tuning a large model that pretrained on large - scale datasets .", "however , the conventional fine - tuning methods require extra human - labeled navigation data and lack self - exploration capabilities in environments , which hinders their generalization of unseen scenes ."], "relation": "used for", "id": "2022.acl-long.332", "year": 2022, "rel_sent": "Both qualitative and quantitative results show that our ProbES significantly improves the generalization ability of the navigation model .", "forward": true, "src_ids": "2022.acl-long.332_4911"}
+{"input": "sentential logical forms ( lfs ) is done by using Method| context: while corpora of child speech and child - directed speech ( cds ) have enabled major contributions to the study of child language acquisition , semantic annotation for such corpora is still scarce and lacks a uniform standard . we compile two cds corpora - in english and hebrew - with syntactic and semantic annotations .", "entity": "sentential logical forms ( lfs )", "output": "automatic method", "neg_sample": ["sentential logical forms ( lfs ) is done by using Method", "while corpora of child speech and child - directed speech ( cds ) have enabled major contributions to the study of child language acquisition , semantic annotation for such corpora is still scarce and lacks a uniform standard .", "we compile two cds corpora - in english and hebrew - with syntactic and semantic annotations ."], "relation": "used for", "id": "2022.scil-1.24", "year": 2022, "rel_sent": "To induce semantic forms , we develop an automatic method for transducing UD structures into sentential logical forms ( LFs ) , e.g.", "forward": false, "src_ids": "2022.scil-1.24_4912"}
+{"input": "ud structures is done by using Method| context: while corpora of child speech and child - directed speech ( cds ) have enabled major contributions to the study of child language acquisition , semantic annotation for such corpora is still scarce and lacks a uniform standard . we compile two cds corpora - in english and hebrew - with syntactic and semantic annotations .", "entity": "ud structures", "output": "automatic method", "neg_sample": ["ud structures is done by using Method", "while corpora of child speech and child - directed speech ( cds ) have enabled major contributions to the study of child language acquisition , semantic annotation for such corpora is still scarce and lacks a uniform standard .", "we compile two cds corpora - in english and hebrew - with syntactic and semantic annotations ."], "relation": "used for", "id": "2022.scil-1.24", "year": 2022, "rel_sent": "To induce semantic forms , we develop an automatic method for transducing UD structures into sentential logical forms ( LFs ) , e.g.", "forward": false, "src_ids": "2022.scil-1.24_4913"}
+{"input": "automatic method is used for OtherScientificTerm| context: while corpora of child speech and child - directed speech ( cds ) have enabled major contributions to the study of child language acquisition , semantic annotation for such corpora is still scarce and lacks a uniform standard . we compile two cds corpora - in english and hebrew - with syntactic and semantic annotations .", "entity": "automatic method", "output": "ud structures", "neg_sample": ["automatic method is used for OtherScientificTerm", "while corpora of child speech and child - directed speech ( cds ) have enabled major contributions to the study of child language acquisition , semantic annotation for such corpora is still scarce and lacks a uniform standard .", "we compile two cds corpora - in english and hebrew - with syntactic and semantic annotations ."], "relation": "used for", "id": "2022.scil-1.24", "year": 2022, "rel_sent": "To induce semantic forms , we develop an automatic method for transducing UD structures into sentential logical forms ( LFs ) , e.g.", "forward": true, "src_ids": "2022.scil-1.24_4914"}
+{"input": "mixup strategy is used for Method| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy . while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks .", "entity": "mixup strategy", "output": "pre - trained language models", "neg_sample": ["mixup strategy is used for Method", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": true, "src_ids": "2022.acl-long.368_4915"}
+{"input": "model calibration is done by using Method| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "entity": "model calibration", "output": "mixup", "neg_sample": ["model calibration is done by using Method", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": false, "src_ids": "2022.acl-long.368_4916"}
+{"input": "nlu tasks is done by using Method| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "entity": "nlu tasks", "output": "mixup", "neg_sample": ["nlu tasks is done by using Method", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": false, "src_ids": "2022.acl-long.368_4917"}
+{"input": "nlu tasks is done by using Task| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "entity": "nlu tasks", "output": "model calibration", "neg_sample": ["nlu tasks is done by using Task", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": false, "src_ids": "2022.acl-long.368_4918"}
+{"input": "mixup is used for Task| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "entity": "mixup", "output": "model calibration", "neg_sample": ["mixup is used for Task", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": true, "src_ids": "2022.acl-long.368_4919"}
+{"input": "mixup strategy is used for Task| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "entity": "mixup strategy", "output": "model calibration", "neg_sample": ["mixup strategy is used for Task", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": true, "src_ids": "2022.acl-long.368_4920"}
+{"input": "mixup is used for Task| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy . while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks .", "entity": "mixup", "output": "nlu tasks", "neg_sample": ["mixup is used for Task", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": true, "src_ids": "2022.acl-long.368_4921"}
+{"input": "model calibration is used for Task| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy . while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks .", "entity": "model calibration", "output": "nlu tasks", "neg_sample": ["model calibration is used for Task", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": true, "src_ids": "2022.acl-long.368_4922"}
+{"input": "model calibration is done by using Method| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy . while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks .", "entity": "model calibration", "output": "mixup strategy", "neg_sample": ["model calibration is done by using Method", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": false, "src_ids": "2022.acl-long.368_4923"}
+{"input": "pre - trained language models is done by using Method| context: a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy . while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks .", "entity": "pre - trained language models", "output": "mixup strategy", "neg_sample": ["pre - trained language models is done by using Method", "a well - calibrated neural model produces confidence ( probability outputs ) closely approximated by the expected accuracy .", "while prior studies have shown that mixup training as a data augmentation technique can improve model calibration on image classification tasks , little is known about using mixup for model calibration on natural language understanding ( nlu ) tasks ."], "relation": "used for", "id": "2022.acl-long.368", "year": 2022, "rel_sent": "In this paper , we explore mixup for model calibration on several NLU tasks and propose a novel mixup strategy for pre - trained language models that improves model calibration further .", "forward": false, "src_ids": "2022.acl-long.368_4924"}
+{"input": "detecting harmful memes is done by using Method| context: identifying good and evil through representations of victimhood , heroism , and villainy ( i.e. , role labeling of entities ) has recently caught the research community 's interest . because of the growing popularity of memes , the amount of offensive information published on the internet is expanding at an alarming rate . it generated a larger need to address this issue and analyze the memes for content moderation .", "entity": "detecting harmful memes", "output": "semantic role labelling approach", "neg_sample": ["detecting harmful memes is done by using Method", "identifying good and evil through representations of victimhood , heroism , and villainy ( i.e.", ", role labeling of entities ) has recently caught the research community 's interest .", "because of the growing popularity of memes , the amount of offensive information published on the internet is expanding at an alarming rate .", "it generated a larger need to address this issue and analyze the memes for content moderation ."], "relation": "used for", "id": "2022.constraint-1.3", "year": 2022, "rel_sent": "Are you a hero or a villain ? A semantic role labelling approach for detecting harmful memes ..", "forward": false, "src_ids": "2022.constraint-1.3_4925"}
+{"input": "semantic role labelling approach is used for Task| context: identifying good and evil through representations of victimhood , heroism , and villainy ( i.e. , role labeling of entities ) has recently caught the research community 's interest . because of the growing popularity of memes , the amount of offensive information published on the internet is expanding at an alarming rate . it generated a larger need to address this issue and analyze the memes for content moderation .", "entity": "semantic role labelling approach", "output": "detecting harmful memes", "neg_sample": ["semantic role labelling approach is used for Task", "identifying good and evil through representations of victimhood , heroism , and villainy ( i.e.", ", role labeling of entities ) has recently caught the research community 's interest .", "because of the growing popularity of memes , the amount of offensive information published on the internet is expanding at an alarming rate .", "it generated a larger need to address this issue and analyze the memes for content moderation ."], "relation": "used for", "id": "2022.constraint-1.3", "year": 2022, "rel_sent": "Are you a hero or a villain ? A semantic role labelling approach for detecting harmful memes ..", "forward": true, "src_ids": "2022.constraint-1.3_4926"}
+{"input": "multilingual document understanding is done by using Method| context: multimodal pre - training with text , layout , and image has achieved sota performance for visually rich document understanding tasks recently , which demonstrates the great potential for joint learning across different modalities . however , the existed research work has focused only on the english domain while neglecting the importance of multilingual generalization .", "entity": "multilingual document understanding", "output": "multimodal pre - trained model", "neg_sample": ["multilingual document understanding is done by using Method", "multimodal pre - training with text , layout , and image has achieved sota performance for visually rich document understanding tasks recently , which demonstrates the great potential for joint learning across different modalities .", "however , the existed research work has focused only on the english domain while neglecting the importance of multilingual generalization ."], "relation": "used for", "id": "2022.findings-acl.253", "year": 2022, "rel_sent": "Meanwhile , we present LayoutXLM , a multimodal pre - trained model for multilingual document understanding , which aims to bridge the language barriers for visually rich document understanding .", "forward": false, "src_ids": "2022.findings-acl.253_4927"}
+{"input": "multimodal pre - trained model is used for Task| context: multimodal pre - training with text , layout , and image has achieved sota performance for visually rich document understanding tasks recently , which demonstrates the great potential for joint learning across different modalities . however , the existed research work has focused only on the english domain while neglecting the importance of multilingual generalization .", "entity": "multimodal pre - trained model", "output": "multilingual document understanding", "neg_sample": ["multimodal pre - trained model is used for Task", "multimodal pre - training with text , layout , and image has achieved sota performance for visually rich document understanding tasks recently , which demonstrates the great potential for joint learning across different modalities .", "however , the existed research work has focused only on the english domain while neglecting the importance of multilingual generalization ."], "relation": "used for", "id": "2022.findings-acl.253", "year": 2022, "rel_sent": "Meanwhile , we present LayoutXLM , a multimodal pre - trained model for multilingual document understanding , which aims to bridge the language barriers for visually rich document understanding .", "forward": true, "src_ids": "2022.findings-acl.253_4928"}
+{"input": "homophobia and transphobia detection is done by using Method| context: the dataset consists of youtube comments , and it has been released for the shared task on homophobia / transphobia detection in social media comments .", "entity": "homophobia and transphobia detection", "output": "ensemble modeling", "neg_sample": ["homophobia and transphobia detection is done by using Method", "the dataset consists of youtube comments , and it has been released for the shared task on homophobia / transphobia detection in social media comments ."], "relation": "used for", "id": "2022.ltedi-1.37", "year": 2022, "rel_sent": "Nozza@LT - EDI - ACL2022 : Ensemble Modeling for Homophobia and Transphobia Detection.", "forward": false, "src_ids": "2022.ltedi-1.37_4929"}
+{"input": "temporal verbal aspect is done by using Method| context: aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time .", "entity": "temporal verbal aspect", "output": "transformers", "neg_sample": ["temporal verbal aspect is done by using Method", "aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time ."], "relation": "used for", "id": "2022.cmcl-1.10", "year": 2022, "rel_sent": "About Time : Do Transformers Learn Temporal Verbal Aspect ?.", "forward": false, "src_ids": "2022.cmcl-1.10_4930"}
+{"input": "transformers is used for OtherScientificTerm| context: aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time .", "entity": "transformers", "output": "temporal verbal aspect", "neg_sample": ["transformers is used for OtherScientificTerm", "aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time ."], "relation": "used for", "id": "2022.cmcl-1.10", "year": 2022, "rel_sent": "About Time : Do Transformers Learn Temporal Verbal Aspect ?.", "forward": true, "src_ids": "2022.cmcl-1.10_4931"}
+{"input": "aspectual features is done by using Method| context: aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time .", "entity": "aspectual features", "output": "transformer models", "neg_sample": ["aspectual features is done by using Method", "aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time ."], "relation": "used for", "id": "2022.cmcl-1.10", "year": 2022, "rel_sent": "In this paper , we explore whether different transformer models are capable of identifying aspectual features .", "forward": false, "src_ids": "2022.cmcl-1.10_4932"}
+{"input": "transformer models is used for OtherScientificTerm| context: aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time .", "entity": "transformer models", "output": "aspectual features", "neg_sample": ["transformer models is used for OtherScientificTerm", "aspect is a linguistic concept that describes how an action , event , or state of a verb phrase is situated in time ."], "relation": "used for", "id": "2022.cmcl-1.10", "year": 2022, "rel_sent": "In this paper , we explore whether different transformer models are capable of identifying aspectual features .", "forward": true, "src_ids": "2022.cmcl-1.10_4933"}
+{"input": "cs speech is done by using Method| context: code switching ( cs ) refers to the phenomenon of interchangeably using words and phrases from different languages . cs can pose significant accuracy challenges to nlp , due to the often monolingual nature of the underlying systems .", "entity": "cs speech", "output": "st architectures", "neg_sample": ["cs speech is done by using Method", "code switching ( cs ) refers to the phenomenon of interchangeably using words and phrases from different languages .", "cs can pose significant accuracy challenges to nlp , due to the often monolingual nature of the underlying systems ."], "relation": "used for", "id": "2022.findings-acl.113", "year": 2022, "rel_sent": "We show that our ST architectures , and especially our bidirectional end - to - end architecture , perform well on CS speech , even when no CS training data is used .", "forward": false, "src_ids": "2022.findings-acl.113_4934"}
+{"input": "code switching is used for Task| context: code switching ( cs ) refers to the phenomenon of interchangeably using words and phrases from different languages . cs can pose significant accuracy challenges to nlp , due to the often monolingual nature of the underlying systems .", "entity": "code switching", "output": "speech translation ( st )", "neg_sample": ["code switching is used for Task", "code switching ( cs ) refers to the phenomenon of interchangeably using words and phrases from different languages .", "cs can pose significant accuracy challenges to nlp , due to the often monolingual nature of the underlying systems ."], "relation": "used for", "id": "2022.findings-acl.113", "year": 2022, "rel_sent": "In this work , we focus on CS in the context of English / Spanish conversations for the task of speech translation ( ST ) , generating and evaluating both transcript and translation .", "forward": true, "src_ids": "2022.findings-acl.113_4935"}
+{"input": "st architectures is used for Material| context: code switching ( cs ) refers to the phenomenon of interchangeably using words and phrases from different languages . cs can pose significant accuracy challenges to nlp , due to the often monolingual nature of the underlying systems .", "entity": "st architectures", "output": "cs speech", "neg_sample": ["st architectures is used for Material", "code switching ( cs ) refers to the phenomenon of interchangeably using words and phrases from different languages .", "cs can pose significant accuracy challenges to nlp , due to the often monolingual nature of the underlying systems ."], "relation": "used for", "id": "2022.findings-acl.113", "year": 2022, "rel_sent": "We show that our ST architectures , and especially our bidirectional end - to - end architecture , perform well on CS speech , even when no CS training data is used .", "forward": true, "src_ids": "2022.findings-acl.113_4936"}
+{"input": "imagination of unseen counterfactual is done by using Method| context: neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning . however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. ' what the annualized rate of return would be if the revenue in 2020 was doubled ' . the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models .", "entity": "imagination of unseen counterfactual", "output": "learning to imagine ( l2i ) module", "neg_sample": ["imagination of unseen counterfactual is done by using Method", "neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning .", "however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. '", "what the annualized rate of return would be if the revenue in 2020 was doubled ' .", "the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models ."], "relation": "used for", "id": "2022.acl-long.5", "year": 2022, "rel_sent": "In this work , we devise a Learning to Imagine ( L2I ) module , which can be seamlessly incorporated into NDR models to perform the imagination of unseen counterfactual .", "forward": false, "src_ids": "2022.acl-long.5_4937"}
+{"input": "ndr models is used for Task| context: neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning . however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. ' what the annualized rate of return would be if the revenue in 2020 was doubled ' . the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models .", "entity": "ndr models", "output": "imagination of unseen counterfactual", "neg_sample": ["ndr models is used for Task", "neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning .", "however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. '", "what the annualized rate of return would be if the revenue in 2020 was doubled ' .", "the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models ."], "relation": "used for", "id": "2022.acl-long.5", "year": 2022, "rel_sent": "In this work , we devise a Learning to Imagine ( L2I ) module , which can be seamlessly incorporated into NDR models to perform the imagination of unseen counterfactual .", "forward": true, "src_ids": "2022.acl-long.5_4938"}
+{"input": "learning to imagine ( l2i ) module is used for Task| context: neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning . however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. ' what the annualized rate of return would be if the revenue in 2020 was doubled ' . the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models .", "entity": "learning to imagine ( l2i ) module", "output": "imagination of unseen counterfactual", "neg_sample": ["learning to imagine ( l2i ) module is used for Task", "neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning .", "however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. '", "what the annualized rate of return would be if the revenue in 2020 was doubled ' .", "the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models ."], "relation": "used for", "id": "2022.acl-long.5", "year": 2022, "rel_sent": "In this work , we devise a Learning to Imagine ( L2I ) module , which can be seamlessly incorporated into NDR models to perform the imagination of unseen counterfactual .", "forward": true, "src_ids": "2022.acl-long.5_4939"}
+{"input": "l2i is used for Method| context: neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning . however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. ' what the annualized rate of return would be if the revenue in 2020 was doubled ' . the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models .", "entity": "l2i", "output": "tat - qa", "neg_sample": ["l2i is used for Method", "neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning .", "however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. '", "what the annualized rate of return would be if the revenue in 2020 was doubled ' .", "the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models ."], "relation": "used for", "id": "2022.acl-long.5", "year": 2022, "rel_sent": "We apply the proposed L2I to TAGOP , the state - of - the - art solution on TAT - QA , validating the rationality and effectiveness of our approach .", "forward": true, "src_ids": "2022.acl-long.5_4940"}
+{"input": "tagop is done by using Method| context: neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning . however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. ' what the annualized rate of return would be if the revenue in 2020 was doubled ' . the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models .", "entity": "tagop", "output": "l2i", "neg_sample": ["tagop is done by using Method", "neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning .", "however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. '", "what the annualized rate of return would be if the revenue in 2020 was doubled ' .", "the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models ."], "relation": "used for", "id": "2022.acl-long.5", "year": 2022, "rel_sent": "We apply the proposed L2I to TAGOP , the state - of - the - art solution on TAT - QA , validating the rationality and effectiveness of our approach .", "forward": false, "src_ids": "2022.acl-long.5_4941"}
+{"input": "tat - qa is done by using Method| context: neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning . however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. ' what the annualized rate of return would be if the revenue in 2020 was doubled ' . the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models .", "entity": "tat - qa", "output": "l2i", "neg_sample": ["tat - qa is done by using Method", "neural discrete reasoning ( ndr ) has shown remarkable progress in combining deep models with discrete reasoning .", "however , we find that existing ndr solution suffers from large performance drop on hypothetical questions , e.g. '", "what the annualized rate of return would be if the revenue in 2020 was doubled ' .", "the key to hypothetical question answering ( hqa ) is counterfactual thinking , which is a natural ability of human reasoning but difficult for deep models ."], "relation": "used for", "id": "2022.acl-long.5", "year": 2022, "rel_sent": "We apply the proposed L2I to TAGOP , the state - of - the - art solution on TAT - QA , validating the rationality and effectiveness of our approach .", "forward": false, "src_ids": "2022.acl-long.5_4942"}
+{"input": "long - term open - domain conversation is done by using OtherScientificTerm| context: despite recent improvements in open - domain dialogue models , state of the art models are trained and evaluated on short conversations with little context . in contrast , the long - term conversation setting has hardly been studied .", "entity": "long - term open - domain conversation", "output": "goldfish memory", "neg_sample": ["long - term open - domain conversation is done by using OtherScientificTerm", "despite recent improvements in open - domain dialogue models , state of the art models are trained and evaluated on short conversations with little context .", "in contrast , the long - term conversation setting has hardly been studied ."], "relation": "used for", "id": "2022.acl-long.356", "year": 2022, "rel_sent": "Beyond Goldfish Memory : Long - Term Open - Domain Conversation.", "forward": false, "src_ids": "2022.acl-long.356_4943"}
+{"input": "goldfish memory is used for Task| context: despite recent improvements in open - domain dialogue models , state of the art models are trained and evaluated on short conversations with little context . in contrast , the long - term conversation setting has hardly been studied .", "entity": "goldfish memory", "output": "long - term open - domain conversation", "neg_sample": ["goldfish memory is used for Task", "despite recent improvements in open - domain dialogue models , state of the art models are trained and evaluated on short conversations with little context .", "in contrast , the long - term conversation setting has hardly been studied ."], "relation": "used for", "id": "2022.acl-long.356", "year": 2022, "rel_sent": "Beyond Goldfish Memory : Long - Term Open - Domain Conversation.", "forward": true, "src_ids": "2022.acl-long.356_4944"}
+{"input": "uniformly distributed variance is done by using Method| context: the recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution . several studies have suggested that contextualized word embedding models do not isotropically project tokens into vector space . however , current methods designed to measure isotropy , such as average random cosine similarity and the partition score , have not been thoroughly analyzed and are not appropriate for measuring isotropy .", "entity": "uniformly distributed variance", "output": "isoscore", "neg_sample": ["uniformly distributed variance is done by using Method", "the recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution .", "several studies have suggested that contextualized word embedding models do not isotropically project tokens into vector space .", "however , current methods designed to measure isotropy , such as average random cosine similarity and the partition score , have not been thoroughly analyzed and are not appropriate for measuring isotropy ."], "relation": "used for", "id": "2022.findings-acl.262", "year": 2022, "rel_sent": "Using rigorously designed tests , we demonstrate that IsoScore is the only tool available in the literature that accurately measures how uniformly distributed variance is across dimensions in vector space .", "forward": false, "src_ids": "2022.findings-acl.262_4945"}
+{"input": "isoscore is used for OtherScientificTerm| context: the recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution . several studies have suggested that contextualized word embedding models do not isotropically project tokens into vector space . however , current methods designed to measure isotropy , such as average random cosine similarity and the partition score , have not been thoroughly analyzed and are not appropriate for measuring isotropy .", "entity": "isoscore", "output": "uniformly distributed variance", "neg_sample": ["isoscore is used for OtherScientificTerm", "the recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution .", "several studies have suggested that contextualized word embedding models do not isotropically project tokens into vector space .", "however , current methods designed to measure isotropy , such as average random cosine similarity and the partition score , have not been thoroughly analyzed and are not appropriate for measuring isotropy ."], "relation": "used for", "id": "2022.findings-acl.262", "year": 2022, "rel_sent": "Using rigorously designed tests , we demonstrate that IsoScore is the only tool available in the literature that accurately measures how uniformly distributed variance is across dimensions in vector space .", "forward": true, "src_ids": "2022.findings-acl.262_4946"}
+{"input": "prompt is done by using Method| context: natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples . we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time . particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time .", "entity": "prompt", "output": "pada", "neg_sample": ["prompt is done by using Method", "natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples .", "we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time .", "particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time ."], "relation": "used for", "id": "2022.tacl-1.24", "year": 2022, "rel_sent": "PADA is trained to generate a prompt that is a token sequence of unrestricted length , consisting of Domain Related Features ( DRFs ) that characterize each of the source domains .", "forward": false, "src_ids": "2022.tacl-1.24_4947"}
+{"input": "on - the - fly adaptation is done by using Method| context: natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples . we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time . particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time .", "entity": "on - the - fly adaptation", "output": "example - based prompt learning", "neg_sample": ["on - the - fly adaptation is done by using Method", "natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples .", "we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time .", "particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time ."], "relation": "used for", "id": "2022.tacl-1.24", "year": 2022, "rel_sent": "PADA : Example - based Prompt Learning for on - the - fly Adaptation to Unseen Domains.", "forward": false, "src_ids": "2022.tacl-1.24_4948"}
+{"input": "example - based prompt learning is used for Task| context: natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples . we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time . particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time .", "entity": "example - based prompt learning", "output": "on - the - fly adaptation", "neg_sample": ["example - based prompt learning is used for Task", "natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples .", "we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time .", "particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time ."], "relation": "used for", "id": "2022.tacl-1.24", "year": 2022, "rel_sent": "PADA : Example - based Prompt Learning for on - the - fly Adaptation to Unseen Domains.", "forward": true, "src_ids": "2022.tacl-1.24_4949"}
+{"input": "pada is used for OtherScientificTerm| context: natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples . we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time . particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time .", "entity": "pada", "output": "prompt", "neg_sample": ["pada is used for OtherScientificTerm", "natural language processing algorithms have made incredible progress , but they still struggle when applied to out - of - distribution examples .", "we address a challenging and underexplored version of this domain adaptation problem , where an algorithm is trained on several source domains , and then applied to examples from unseen domains that are unknown at training time .", "particularly , no examples , labeled or unlabeled , or any other knowledge about the target domain are available to the algorithm at training time ."], "relation": "used for", "id": "2022.tacl-1.24", "year": 2022, "rel_sent": "PADA is trained to generate a prompt that is a token sequence of unrestricted length , consisting of Domain Related Features ( DRFs ) that characterize each of the source domains .", "forward": true, "src_ids": "2022.tacl-1.24_4950"}
+{"input": "equality is done by using Task| context: designing a robust model that helps in predicting the target class with higher accuracy is a challenging task in machine learning , especially when the distribution of the class labels is highly imbalanced .", "entity": "equality", "output": "hope speech detection", "neg_sample": ["equality is done by using Task", "designing a robust model that helps in predicting the target class with higher accuracy is a challenging task in machine learning , especially when the distribution of the class labels is highly imbalanced ."], "relation": "used for", "id": "2022.ltedi-1.13", "year": 2022, "rel_sent": "The English language dataset used in this research was developed by collecting YouTube comments and is part of the task ' ACL-2022 : Hope Speech Detection for Equality , Diversity , and Inclusion ' .", "forward": false, "src_ids": "2022.ltedi-1.13_4951"}
+{"input": "hope speech detection is used for OtherScientificTerm| context: designing a robust model that helps in predicting the target class with higher accuracy is a challenging task in machine learning , especially when the distribution of the class labels is highly imbalanced .", "entity": "hope speech detection", "output": "equality", "neg_sample": ["hope speech detection is used for OtherScientificTerm", "designing a robust model that helps in predicting the target class with higher accuracy is a challenging task in machine learning , especially when the distribution of the class labels is highly imbalanced ."], "relation": "used for", "id": "2022.ltedi-1.13", "year": 2022, "rel_sent": "The English language dataset used in this research was developed by collecting YouTube comments and is part of the task ' ACL-2022 : Hope Speech Detection for Equality , Diversity , and Inclusion ' .", "forward": true, "src_ids": "2022.ltedi-1.13_4952"}
+{"input": "sentence- and clause - final rts is done by using OtherScientificTerm| context: numerous analyses of reading time ( rt ) data have been undertaken in the effort to learn more about the internal processes that occur during reading comprehension . however , data measured on words at the end of a sentence - or even clause - is often omitted due to the confounding factors introduced by so - called ' wrap - up effects , ' which manifests as a skewed distribution of rts for these words . consequently , the understanding of the cognitive processes that might be involved in these effects is limited .", "entity": "sentence- and clause - final rts", "output": "information distribution", "neg_sample": ["sentence- and clause - final rts is done by using OtherScientificTerm", "numerous analyses of reading time ( rt ) data have been undertaken in the effort to learn more about the internal processes that occur during reading comprehension .", "however , data measured on words at the end of a sentence - or even clause - is often omitted due to the confounding factors introduced by so - called ' wrap - up effects , ' which manifests as a skewed distribution of rts for these words .", "consequently , the understanding of the cognitive processes that might be involved in these effects is limited ."], "relation": "used for", "id": "2022.acl-short.3", "year": 2022, "rel_sent": "We find that the information distribution of prior context is often predictive of sentence- and clause - final RTs ( while not of sentence - medial RTs ) , which lends support to several prior hypotheses about the processes involved in wrap - up effects .", "forward": false, "src_ids": "2022.acl-short.3_4953"}
+{"input": "information distribution is used for OtherScientificTerm| context: numerous analyses of reading time ( rt ) data have been undertaken in the effort to learn more about the internal processes that occur during reading comprehension . however , data measured on words at the end of a sentence - or even clause - is often omitted due to the confounding factors introduced by so - called ' wrap - up effects , ' which manifests as a skewed distribution of rts for these words . consequently , the understanding of the cognitive processes that might be involved in these effects is limited .", "entity": "information distribution", "output": "sentence- and clause - final rts", "neg_sample": ["information distribution is used for OtherScientificTerm", "numerous analyses of reading time ( rt ) data have been undertaken in the effort to learn more about the internal processes that occur during reading comprehension .", "however , data measured on words at the end of a sentence - or even clause - is often omitted due to the confounding factors introduced by so - called ' wrap - up effects , ' which manifests as a skewed distribution of rts for these words .", "consequently , the understanding of the cognitive processes that might be involved in these effects is limited ."], "relation": "used for", "id": "2022.acl-short.3", "year": 2022, "rel_sent": "We find that the information distribution of prior context is often predictive of sentence- and clause - final RTs ( while not of sentence - medial RTs ) , which lends support to several prior hypotheses about the processes involved in wrap - up effects .", "forward": true, "src_ids": "2022.acl-short.3_4954"}
+{"input": "long document classification tasks is done by using Method| context: research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not . building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy . however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp . furthermore , the existing methods can not utilize a large size of unlabeled dataset tofurther improve the model interpretability .", "entity": "long document classification tasks", "output": "variational contextual consistency sentence masking", "neg_sample": ["long document classification tasks is done by using Method", "research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not .", "building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy .", "however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp .", "furthermore , the existing methods can not utilize a large size of unlabeled dataset tofurther improve the model interpretability ."], "relation": "used for", "id": "2022.findings-acl.305", "year": 2022, "rel_sent": "Results of our experiments on RRP along with European Convention of Human Rights ( ECHR ) datasets demonstrate that VCCSM is able to improve the model interpretability for the long document classification tasks using the area over the perturbation curve and post - hoc accuracy as evaluation metrics .", "forward": false, "src_ids": "2022.findings-acl.305_4955"}
+{"input": "variational contextual consistency sentence masking is used for Task| context: research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not . building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy . however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp . furthermore , the existing methods can not utilize a large size of unlabeled dataset tofurther improve the model interpretability .", "entity": "variational contextual consistency sentence masking", "output": "long document classification tasks", "neg_sample": ["variational contextual consistency sentence masking is used for Task", "research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not .", "building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy .", "however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp .", "furthermore , the existing methods can not utilize a large size of unlabeled dataset tofurther improve the model interpretability ."], "relation": "used for", "id": "2022.findings-acl.305", "year": 2022, "rel_sent": "Results of our experiments on RRP along with European Convention of Human Rights ( ECHR ) datasets demonstrate that VCCSM is able to improve the model interpretability for the long document classification tasks using the area over the perturbation curve and post - hoc accuracy as evaluation metrics .", "forward": true, "src_ids": "2022.findings-acl.305_4956"}
+{"input": "low - resource ner is done by using Method| context: recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates . similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence . however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence .", "entity": "low - resource ner", "output": "demonstration - based learning method", "neg_sample": ["low - resource ner is done by using Method", "recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates .", "similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence .", "however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence ."], "relation": "used for", "id": "2022.acl-long.192", "year": 2022, "rel_sent": "Good Examples Make A Faster Learner : Simple Demonstration - based Learning for Low - resource NER.", "forward": false, "src_ids": "2022.acl-long.192_4957"}
+{"input": "named entity recognition is done by using Method| context: recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates . similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence . however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence .", "entity": "named entity recognition", "output": "demonstration - based learning method", "neg_sample": ["named entity recognition is done by using Method", "recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates .", "similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence .", "however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence ."], "relation": "used for", "id": "2022.acl-long.192", "year": 2022, "rel_sent": "Here we present a simple demonstration - based learning method for NER , which lets the input be prefaced by task demonstrations for in - context learning .", "forward": false, "src_ids": "2022.acl-long.192_4958"}
+{"input": "demonstration - based learning method is used for Task| context: recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates . similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence . however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence .", "entity": "demonstration - based learning method", "output": "low - resource ner", "neg_sample": ["demonstration - based learning method is used for Task", "recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates .", "similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence .", "however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence ."], "relation": "used for", "id": "2022.acl-long.192", "year": 2022, "rel_sent": "Good Examples Make A Faster Learner : Simple Demonstration - based Learning for Low - resource NER.", "forward": true, "src_ids": "2022.acl-long.192_4959"}
+{"input": "demonstration - based learning method is used for Task| context: however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence .", "entity": "demonstration - based learning method", "output": "named entity recognition", "neg_sample": ["demonstration - based learning method is used for Task", "however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence ."], "relation": "used for", "id": "2022.acl-long.192", "year": 2022, "rel_sent": "Here we present a simple demonstration - based learning method for NER , which lets the input be prefaced by task demonstrations for in - context learning .", "forward": true, "src_ids": "2022.acl-long.192_4960"}
+{"input": "in - context learning is done by using OtherScientificTerm| context: recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates . similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence . however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence .", "entity": "in - context learning", "output": "task demonstrations", "neg_sample": ["in - context learning is done by using OtherScientificTerm", "recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates .", "similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence .", "however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence ."], "relation": "used for", "id": "2022.acl-long.192", "year": 2022, "rel_sent": "Here we present a simple demonstration - based learning method for NER , which lets the input be prefaced by task demonstrations for in - context learning .", "forward": false, "src_ids": "2022.acl-long.192_4961"}
+{"input": "task demonstrations is used for Method| context: recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates . similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence . however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence .", "entity": "task demonstrations", "output": "in - context learning", "neg_sample": ["task demonstrations is used for Method", "recent advances in prompt - based learning have shown strong results on few - shot text classification by using cloze - style templates .", "similar attempts have been made on named entity recognition ( ner ) which manually design templates to predict entity types for every text span in a sentence .", "however , such methods may suffer from error propagation induced by entity span detection , high cost due to enumeration of all possible text spans , and omission of inter - dependencies among token labels in a sentence ."], "relation": "used for", "id": "2022.acl-long.192", "year": 2022, "rel_sent": "Here we present a simple demonstration - based learning method for NER , which lets the input be prefaced by task demonstrations for in - context learning .", "forward": true, "src_ids": "2022.acl-long.192_4962"}
+{"input": "downstream tasks is done by using Method| context: language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks . however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "downstream tasks", "output": "linkbert", "neg_sample": ["downstream tasks is done by using Method", "language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks .", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "We show that LinkBERT outperforms BERT on various downstream tasks across two domains : the general domain ( pretrained on Wikipedia with hyperlinks ) and biomedical domain ( pretrained on PubMed with citation links ) .", "forward": false, "src_ids": "2022.acl-long.551_4963"}
+{"input": "multi - hop reasoning is done by using Method| context: language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks . however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "multi - hop reasoning", "output": "linkbert", "neg_sample": ["multi - hop reasoning is done by using Method", "language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks .", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "LinkBERT is especially effective for multi - hop reasoning and few - shot QA ( +5 % absolute improvement on HotpotQA and TriviaQA ) , and our biomedical LinkBERT sets new states of the art on various BioNLP tasks ( +7 % on BioASQ and USMLE ) .", "forward": false, "src_ids": "2022.acl-long.551_4964"}
+{"input": "few - shot qa is done by using Method| context: language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks . however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "few - shot qa", "output": "linkbert", "neg_sample": ["few - shot qa is done by using Method", "language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks .", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "LinkBERT is especially effective for multi - hop reasoning and few - shot QA ( +5 % absolute improvement on HotpotQA and TriviaQA ) , and our biomedical LinkBERT sets new states of the art on various BioNLP tasks ( +7 % on BioASQ and USMLE ) .", "forward": false, "src_ids": "2022.acl-long.551_4965"}
+{"input": "bert is used for Task| context: however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "bert", "output": "downstream tasks", "neg_sample": ["bert is used for Task", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "We show that LinkBERT outperforms BERT on various downstream tasks across two domains : the general domain ( pretrained on Wikipedia with hyperlinks ) and biomedical domain ( pretrained on PubMed with citation links ) .", "forward": true, "src_ids": "2022.acl-long.551_4966"}
+{"input": "linkbert is used for Task| context: however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "linkbert", "output": "downstream tasks", "neg_sample": ["linkbert is used for Task", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "We show that LinkBERT outperforms BERT on various downstream tasks across two domains : the general domain ( pretrained on Wikipedia with hyperlinks ) and biomedical domain ( pretrained on PubMed with citation links ) .", "forward": true, "src_ids": "2022.acl-long.551_4967"}
+{"input": "downstream tasks is done by using Method| context: language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks .", "entity": "downstream tasks", "output": "bert", "neg_sample": ["downstream tasks is done by using Method", "language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "We show that LinkBERT outperforms BERT on various downstream tasks across two domains : the general domain ( pretrained on Wikipedia with hyperlinks ) and biomedical domain ( pretrained on PubMed with citation links ) .", "forward": false, "src_ids": "2022.acl-long.551_4968"}
+{"input": "linkbert is used for Method| context: language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks . however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "linkbert", "output": "multi - hop reasoning", "neg_sample": ["linkbert is used for Method", "language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks .", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "LinkBERT is especially effective for multi - hop reasoning and few - shot QA ( +5 % absolute improvement on HotpotQA and TriviaQA ) , and our biomedical LinkBERT sets new states of the art on various BioNLP tasks ( +7 % on BioASQ and USMLE ) .", "forward": true, "src_ids": "2022.acl-long.551_4969"}
+{"input": "linkbert is used for Task| context: language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks . however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents .", "entity": "linkbert", "output": "few - shot qa", "neg_sample": ["linkbert is used for Task", "language model ( lm ) pretraining captures various knowledge from text corpora , helping downstream tasks .", "however , existing methods such as bert model a single document , and do not capture dependencies or knowledge that span across documents ."], "relation": "used for", "id": "2022.acl-long.551", "year": 2022, "rel_sent": "LinkBERT is especially effective for multi - hop reasoning and few - shot QA ( +5 % absolute improvement on HotpotQA and TriviaQA ) , and our biomedical LinkBERT sets new states of the art on various BioNLP tasks ( +7 % on BioASQ and USMLE ) .", "forward": true, "src_ids": "2022.acl-long.551_4970"}
+{"input": "downstream tasks is done by using OtherScientificTerm| context: pretrained language models have served as important backbones for natural language processing . recently , in - domain pretraining has been shown to benefit various domain - specific downstream tasks . in the biomedical domain , natural language generation ( nlg ) tasks are of critical importance , while understudied . approaching natural language understanding ( nlu ) tasks as nlg achieves satisfying performance in the general domain through constrained language generation or language prompting . we emphasize the lack of in - domain generative language models and the unsystematic generative downstream benchmarks in the biomedical domain , hindering the development of the research community . in this work , we introduce the generative language model biobart that adapts bart to the biomedical domain .", "entity": "downstream tasks", "output": "sentence permutation", "neg_sample": ["downstream tasks is done by using OtherScientificTerm", "pretrained language models have served as important backbones for natural language processing .", "recently , in - domain pretraining has been shown to benefit various domain - specific downstream tasks .", "in the biomedical domain , natural language generation ( nlg ) tasks are of critical importance , while understudied .", "approaching natural language understanding ( nlu ) tasks as nlg achieves satisfying performance in the general domain through constrained language generation or language prompting .", "we emphasize the lack of in - domain generative language models and the unsystematic generative downstream benchmarks in the biomedical domain , hindering the development of the research community .", "in this work , we introduce the generative language model biobart that adapts bart to the biomedical domain ."], "relation": "used for", "id": "2022.bionlp-1.9", "year": 2022, "rel_sent": "Furthermore , we conduct ablation studies on the pretraining tasks for BioBART and find that sentence permutation has negative effects on downstream tasks .", "forward": false, "src_ids": "2022.bionlp-1.9_4971"}
+{"input": "sentence permutation is used for Task| context: pretrained language models have served as important backbones for natural language processing . in the biomedical domain , natural language generation ( nlg ) tasks are of critical importance , while understudied . approaching natural language understanding ( nlu ) tasks as nlg achieves satisfying performance in the general domain through constrained language generation or language prompting . we emphasize the lack of in - domain generative language models and the unsystematic generative downstream benchmarks in the biomedical domain , hindering the development of the research community . in this work , we introduce the generative language model biobart that adapts bart to the biomedical domain .", "entity": "sentence permutation", "output": "downstream tasks", "neg_sample": ["sentence permutation is used for Task", "pretrained language models have served as important backbones for natural language processing .", "in the biomedical domain , natural language generation ( nlg ) tasks are of critical importance , while understudied .", "approaching natural language understanding ( nlu ) tasks as nlg achieves satisfying performance in the general domain through constrained language generation or language prompting .", "we emphasize the lack of in - domain generative language models and the unsystematic generative downstream benchmarks in the biomedical domain , hindering the development of the research community .", "in this work , we introduce the generative language model biobart that adapts bart to the biomedical domain ."], "relation": "used for", "id": "2022.bionlp-1.9", "year": 2022, "rel_sent": "Furthermore , we conduct ablation studies on the pretraining tasks for BioBART and find that sentence permutation has negative effects on downstream tasks .", "forward": true, "src_ids": "2022.bionlp-1.9_4972"}
+{"input": "creole languages is done by using Method| context: we find that standard transfer methods do not facilitate ancestry transfer .", "entity": "creole languages", "output": "language models", "neg_sample": ["creole languages is done by using Method", "we find that standard transfer methods do not facilitate ancestry transfer ."], "relation": "used for", "id": "2022.insights-1.9", "year": 2022, "rel_sent": "We aim to learn language models for Creole languages for which large volumes of data are not readily available , and therefore explore the potential transfer from ancestor languages ( the ' Ancestry Transfer Hypothesis ' ) .", "forward": false, "src_ids": "2022.insights-1.9_4973"}
+{"input": "compression phase is used for Method| context: we find that standard transfer methods do not facilitate ancestry transfer .", "entity": "compression phase", "output": "language models", "neg_sample": ["compression phase is used for Method", "we find that standard transfer methods do not facilitate ancestry transfer ."], "relation": "used for", "id": "2022.insights-1.9", "year": 2022, "rel_sent": "We explore if this compression phase can lead to practically useful language models ( the ' Ancestry Bottleneck Hypothesis ' ) , but alsofalsify this .", "forward": true, "src_ids": "2022.insights-1.9_4974"}
+{"input": "language models is used for Material| context: we find that standard transfer methods do not facilitate ancestry transfer .", "entity": "language models", "output": "creole languages", "neg_sample": ["language models is used for Material", "we find that standard transfer methods do not facilitate ancestry transfer ."], "relation": "used for", "id": "2022.insights-1.9", "year": 2022, "rel_sent": "We aim to learn language models for Creole languages for which large volumes of data are not readily available , and therefore explore the potential transfer from ancestor languages ( the ' Ancestry Transfer Hypothesis ' ) .", "forward": true, "src_ids": "2022.insights-1.9_4975"}
+{"input": "language models is done by using Generic| context: we find that standard transfer methods do not facilitate ancestry transfer . surprisingly , different from other non - creole languages , a very distinct two - phase pattern emerges for creoles : as our training losses plateau , and language models begin to overfit on their source languages , perplexity on the creoles drop .", "entity": "language models", "output": "compression phase", "neg_sample": ["language models is done by using Generic", "we find that standard transfer methods do not facilitate ancestry transfer .", "surprisingly , different from other non - creole languages , a very distinct two - phase pattern emerges for creoles : as our training losses plateau , and language models begin to overfit on their source languages , perplexity on the creoles drop ."], "relation": "used for", "id": "2022.insights-1.9", "year": 2022, "rel_sent": "We explore if this compression phase can lead to practically useful language models ( the ' Ancestry Bottleneck Hypothesis ' ) , but alsofalsify this .", "forward": false, "src_ids": "2022.insights-1.9_4976"}
+{"input": "syntactic representation is used for OtherScientificTerm| context: we present an incremental syntactic representation that consists of assigning a single discrete label to each word in a sentence , where the label is predicted using strictly incremental processing of a prefix of the sentence , and the sequence of labels for a sentence fully determines a parse tree .", "entity": "syntactic representation", "output": "syntactic choices", "neg_sample": ["syntactic representation is used for OtherScientificTerm", "we present an incremental syntactic representation that consists of assigning a single discrete label to each word in a sentence , where the label is predicted using strictly incremental processing of a prefix of the sentence , and the sequence of labels for a sentence fully determines a parse tree ."], "relation": "used for", "id": "2022.acl-long.220", "year": 2022, "rel_sent": "Our goal is to induce a syntactic representation that commits to syntactic choices only as they are incrementally revealed by the input , in contrast with standard representations that must make output choices such as attachments speculatively and later throw out conflicting analyses .", "forward": true, "src_ids": "2022.acl-long.220_4977"}
+{"input": "visual commonsense tasks is done by using Task| context: we study event understanding as a critical step towards visual commonsense tasks . meanwhile , we argue that current object - based event understanding is purely likelihood - based , leading to incorrect event prediction , due to biased correlation between events and objects .", "entity": "visual commonsense tasks", "output": "debiasing event understanding", "neg_sample": ["visual commonsense tasks is done by using Task", "we study event understanding as a critical step towards visual commonsense tasks .", "meanwhile , we argue that current object - based event understanding is purely likelihood - based , leading to incorrect event prediction , due to biased correlation between events and objects ."], "relation": "used for", "id": "2022.findings-acl.351", "year": 2022, "rel_sent": "Debiasing Event Understanding for Visual Commonsense Tasks.", "forward": false, "src_ids": "2022.findings-acl.351_4978"}
+{"input": "debiasing event understanding is used for Task| context: meanwhile , we argue that current object - based event understanding is purely likelihood - based , leading to incorrect event prediction , due to biased correlation between events and objects .", "entity": "debiasing event understanding", "output": "visual commonsense tasks", "neg_sample": ["debiasing event understanding is used for Task", "meanwhile , we argue that current object - based event understanding is purely likelihood - based , leading to incorrect event prediction , due to biased correlation between events and objects ."], "relation": "used for", "id": "2022.findings-acl.351", "year": 2022, "rel_sent": "Debiasing Event Understanding for Visual Commonsense Tasks.", "forward": true, "src_ids": "2022.findings-acl.351_4979"}
+{"input": "iterative fact reasoning is done by using Method| context: current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts . however , it neglects the n - ary facts , which contain more than two entities . in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e. , answering n - ary facts questions upon n - ary kgs . nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa .", "entity": "iterative fact reasoning", "output": "factree", "neg_sample": ["iterative fact reasoning is done by using Method", "current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts .", "however , it neglects the n - ary facts , which contain more than two entities .", "in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e.", ", answering n - ary facts questions upon n - ary kgs .", "nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa ."], "relation": "used for", "id": "2022.findings-acl.352", "year": 2022, "rel_sent": "FacTree transforms the question into a fact tree and performs iterative fact reasoning on the fact tree to infer the correct answer .", "forward": false, "src_ids": "2022.findings-acl.352_4980"}
+{"input": "factree is used for Method| context: current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts . however , it neglects the n - ary facts , which contain more than two entities . in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e. , answering n - ary facts questions upon n - ary kgs . nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa .", "entity": "factree", "output": "iterative fact reasoning", "neg_sample": ["factree is used for Method", "current question answering over knowledge graphs ( kgqa ) task mainly focuses on performing answer reasoning upon kgs with binary facts .", "however , it neglects the n - ary facts , which contain more than two entities .", "in this work , we highlight a more challenging but under - explored task : n - ary kgqa , i.e.", ", answering n - ary facts questions upon n - ary kgs .", "nevertheless , the multi - hop reasoning framework popular in binary kgqa task is not directly applicable on n - ary kgqa ."], "relation": "used for", "id": "2022.findings-acl.352", "year": 2022, "rel_sent": "FacTree transforms the question into a fact tree and performs iterative fact reasoning on the fact tree to infer the correct answer .", "forward": true, "src_ids": "2022.findings-acl.352_4981"}
+{"input": "multi - domain specialization is done by using Method| context: recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) . these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "multi - domain specialization", "output": "light - weight adapter - based specialization", "neg_sample": ["multi - domain specialization is done by using Method", "recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) .", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.401", "year": 2022, "rel_sent": "Moreover , we show that the light - weight adapter - based specialization ( 1 ) performs comparably to full fine - tuning in single domain setups and ( 2 ) is particularly suitable for multi - domain specialization , where besides advantageous computational footprint , it can offer better TOD performance .", "forward": false, "src_ids": "2022.findings-acl.401_4982"}
+{"input": "light - weight adapter - based specialization is used for Task| context: recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) . these approaches , however , exploit general dialogic corpora ( e.g. , reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains .", "entity": "light - weight adapter - based specialization", "output": "multi - domain specialization", "neg_sample": ["light - weight adapter - based specialization is used for Task", "recent work has shown that self - supervised dialog - specific pretraining on large conversational datasets yields substantial gains over traditional language modeling ( lm ) pretraining in downstream task - oriented dialog ( tod ) .", "these approaches , however , exploit general dialogic corpora ( e.g.", ", reddit ) and thus presumably fail to reliably embed domain - specific knowledge useful for concrete downstream tod domains ."], "relation": "used for", "id": "2022.findings-acl.401", "year": 2022, "rel_sent": "Moreover , we show that the light - weight adapter - based specialization ( 1 ) performs comparably to full fine - tuning in single domain setups and ( 2 ) is particularly suitable for multi - domain specialization , where besides advantageous computational footprint , it can offer better TOD performance .", "forward": true, "src_ids": "2022.findings-acl.401_4983"}
+{"input": "generating augmented data is done by using Method| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "generating augmented data", "output": "anomaly detector", "neg_sample": ["generating augmented data is done by using Method", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.409", "year": 2022, "rel_sent": "We also investigate two applications of the anomaly detector : ( 1 ) In data augmentation , we employ the anomaly detector to force generating augmented data that are distinguished as non - natural , which brings larger gains to the accuracy of PrLMs .", "forward": false, "src_ids": "2022.findings-acl.409_4984"}
+{"input": "anomaly detector is used for Task| context: recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest . latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust . however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality .", "entity": "anomaly detector", "output": "generating augmented data", "neg_sample": ["anomaly detector is used for Task", "recently , the problem of robustness of pre - trained language models ( prlms ) has received increasing research interest .", "latest studies on adversarial attacks achieve high attack success rates against prlms , claiming that prlms are not robust .", "however , we find that the adversarial samples that prlms fail are mostly non - natural and do not appear in reality ."], "relation": "used for", "id": "2022.findings-acl.409", "year": 2022, "rel_sent": "We also investigate two applications of the anomaly detector : ( 1 ) In data augmentation , we employ the anomaly detector to force generating augmented data that are distinguished as non - natural , which brings larger gains to the accuracy of PrLMs .", "forward": true, "src_ids": "2022.findings-acl.409_4985"}
+{"input": "qa systems is done by using Material| context: most research on question answering focuses on the pre - deployment stage ; i.e. , building an accurate model for deployment .", "entity": "qa systems", "output": "feedback data", "neg_sample": ["qa systems is done by using Material", "most research on question answering focuses on the pre - deployment stage ; i.e.", ", building an accurate model for deployment ."], "relation": "used for", "id": "2022.findings-acl.429", "year": 2022, "rel_sent": "We show that feedback data not only improves the accuracy of the deployed QA system but also other stronger non - deployed systems .", "forward": false, "src_ids": "2022.findings-acl.429_4986"}
+{"input": "feedback data is used for Method| context: most research on question answering focuses on the pre - deployment stage ; i.e. , building an accurate model for deployment .", "entity": "feedback data", "output": "qa systems", "neg_sample": ["feedback data is used for Method", "most research on question answering focuses on the pre - deployment stage ; i.e.", ", building an accurate model for deployment ."], "relation": "used for", "id": "2022.findings-acl.429", "year": 2022, "rel_sent": "We show that feedback data not only improves the accuracy of the deployed QA system but also other stronger non - deployed systems .", "forward": true, "src_ids": "2022.findings-acl.429_4987"}
+{"input": "morphological segmentation datasets is used for Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "morphological segmentation datasets", "output": "raramuri", "neg_sample": ["morphological segmentation datasets is used for Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.446", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri -- Spanish .", "forward": true, "src_ids": "2022.findings-acl.446_4988"}
+{"input": "raramuri is done by using Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "raramuri", "output": "morphological segmentation datasets", "neg_sample": ["raramuri is done by using Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.446", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri -- Spanish .", "forward": false, "src_ids": "2022.findings-acl.446_4989"}
+{"input": "shipibo - konibo is done by using Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "shipibo - konibo", "output": "morphological segmentation datasets", "neg_sample": ["shipibo - konibo is done by using Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.446", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri -- Spanish .", "forward": false, "src_ids": "2022.findings-acl.446_4990"}
+{"input": "raramuri -- spanish is done by using Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "raramuri -- spanish", "output": "parallel corpus", "neg_sample": ["raramuri -- spanish is done by using Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.446", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri -- Spanish .", "forward": false, "src_ids": "2022.findings-acl.446_4991"}
+{"input": "parallel corpus is used for Material| context: morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation .", "entity": "parallel corpus", "output": "raramuri -- spanish", "neg_sample": ["parallel corpus is used for Material", "morphologically - rich polysynthetic languages present a challenge for nlp systems due to data sparsity , and a common strategy to handle this issue is to apply subword segmentation ."], "relation": "used for", "id": "2022.findings-acl.446", "year": 2022, "rel_sent": "Finally , we contribute two new morphological segmentation datasets for Raramuri and Shipibo - Konibo , and a parallel corpus for Raramuri -- Spanish .", "forward": true, "src_ids": "2022.findings-acl.446_4992"}
+{"input": "overfitting terms is done by using Method| context: natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability . e.g. , neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance . most mitigation techniques use lists of identity terms or samples from the target domain during training . however , this approach requires a - priori knowledge and introduces further bias if important terms are neglected .", "entity": "overfitting terms", "output": "entropy - based attention regularization", "neg_sample": ["overfitting terms is done by using Method", "natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability .", "e.g.", ", neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance .", "most mitigation techniques use lists of identity terms or samples from the target domain during training .", "however , this approach requires a - priori knowledge and introduces further bias if important terms are neglected ."], "relation": "used for", "id": "2022.findings-acl.506", "year": 2022, "rel_sent": "EAR also reveals overfitting terms , i.e. , terms most likely to induce bias , to help identify their effect on the model , task , and predictions .", "forward": false, "src_ids": "2022.findings-acl.506_4993"}
+{"input": "entropy - based attention regularization is used for OtherScientificTerm| context: natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability . e.g. , neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance . most mitigation techniques use lists of identity terms or samples from the target domain during training . however , this approach requires a - priori knowledge and introduces further bias if important terms are neglected .", "entity": "entropy - based attention regularization", "output": "overfitting terms", "neg_sample": ["entropy - based attention regularization is used for OtherScientificTerm", "natural language processing ( nlp ) models risk overfitting to specific terms in the training data , thereby reducing their performance , fairness , and generalizability .", "e.g.", ", neural hate speech detection models are strongly influenced by identity terms like gay , or women , resulting in false positives , severe unintended bias , and lower performance .", "most mitigation techniques use lists of identity terms or samples from the target domain during training .", "however , this approach requires a - priori knowledge and introduces further bias if important terms are neglected ."], "relation": "used for", "id": "2022.findings-acl.506", "year": 2022, "rel_sent": "EAR also reveals overfitting terms , i.e. , terms most likely to induce bias , to help identify their effect on the model , task , and predictions .", "forward": true, "src_ids": "2022.findings-acl.506_4994"}
+{"input": "bert is used for OtherScientificTerm| context: chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts . csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings . many recent works use bert - based language models to directly correct each character of the input sentence . however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters .", "entity": "bert", "output": "visual and phonological features", "neg_sample": ["bert is used for OtherScientificTerm", "chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts .", "csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings .", "many recent works use bert - based language models to directly correct each character of the input sentence .", "however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters ."], "relation": "used for", "id": "2022.findings-acl.563", "year": 2022, "rel_sent": "In this work , we propose a novel general detector - corrector multi - task framework where the corrector uses BERT to capture the visual and phonological features from each character in the raw sentence and uses a late fusion strategy to fuse the hidden states of the corrector with that of the detector to minimize the negative impact from the misspelled characters .", "forward": true, "src_ids": "2022.findings-acl.563_4995"}
+{"input": "visual and phonological features is done by using Method| context: chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts . csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings . however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters .", "entity": "visual and phonological features", "output": "bert", "neg_sample": ["visual and phonological features is done by using Method", "chinese spelling correction ( csc ) is a task to detect and correct misspelled characters in chinese texts .", "csc is challenging since many chinese characters are visually or phonologically similar but with quite different semantic meanings .", "however , these methods can be sub - optimal since they correct every character of the sentence only by the context which is easily negatively affected by the misspelled characters ."], "relation": "used for", "id": "2022.findings-acl.563", "year": 2022, "rel_sent": "In this work , we propose a novel general detector - corrector multi - task framework where the corrector uses BERT to capture the visual and phonological features from each character in the raw sentence and uses a late fusion strategy to fuse the hidden states of the corrector with that of the detector to minimize the negative impact from the misspelled characters .", "forward": false, "src_ids": "2022.findings-acl.563_4996"}
+{"input": "extractive text summarization is done by using Metric| context: transformer - based language models usually treat texts as linear sequences . however , most texts also have an inherent hierarchical structure , i.e. , parts of a text can be identified using their position in this hierarchy . in addition , section titles usually indicate the common topic of their respective sentences .", "entity": "extractive text summarization", "output": "sota rouges", "neg_sample": ["extractive text summarization is done by using Metric", "transformer - based language models usually treat texts as linear sequences .", "however , most texts also have an inherent hierarchical structure , i.e.", ", parts of a text can be identified using their position in this hierarchy .", "in addition , section titles usually indicate the common topic of their respective sentences ."], "relation": "used for", "id": "2022.findings-acl.574", "year": 2022, "rel_sent": "We propose a novel approach to formulate , extract , encode and inject hierarchical structure information explicitly into an extractive summarization model based on a pre - trained , encoder - only Transformer language model ( HiStruct+ model ) , which improves SOTA ROUGEs for extractive summarization on PubMed and arXiv substantially .", "forward": false, "src_ids": "2022.findings-acl.574_4997"}
+{"input": "sota rouges is used for Task| context: transformer - based language models usually treat texts as linear sequences . however , most texts also have an inherent hierarchical structure , i.e. , parts of a text can be identified using their position in this hierarchy . in addition , section titles usually indicate the common topic of their respective sentences .", "entity": "sota rouges", "output": "extractive text summarization", "neg_sample": ["sota rouges is used for Task", "transformer - based language models usually treat texts as linear sequences .", "however , most texts also have an inherent hierarchical structure , i.e.", ", parts of a text can be identified using their position in this hierarchy .", "in addition , section titles usually indicate the common topic of their respective sentences ."], "relation": "used for", "id": "2022.findings-acl.574", "year": 2022, "rel_sent": "We propose a novel approach to formulate , extract , encode and inject hierarchical structure information explicitly into an extractive summarization model based on a pre - trained , encoder - only Transformer language model ( HiStruct+ model ) , which improves SOTA ROUGEs for extractive summarization on PubMed and arXiv substantially .", "forward": true, "src_ids": "2022.findings-acl.574_4998"}
+{"input": "edit action is done by using Method| context: current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps .", "entity": "edit action", "output": "multi - aspect scoring functions", "neg_sample": ["edit action is done by using Method", "current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps ."], "relation": "used for", "id": "2022.findings-acl.609", "year": 2022, "rel_sent": "Then it introduces four multi - aspect scoring functions to select edit action to further reduce search difficulty .", "forward": false, "src_ids": "2022.findings-acl.609_4999"}
+{"input": "search difficulty is done by using Method| context: current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps .", "entity": "search difficulty", "output": "multi - aspect scoring functions", "neg_sample": ["search difficulty is done by using Method", "current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps ."], "relation": "used for", "id": "2022.findings-acl.609", "year": 2022, "rel_sent": "Then it introduces four multi - aspect scoring functions to select edit action to further reduce search difficulty .", "forward": false, "src_ids": "2022.findings-acl.609_5000"}
+{"input": "multi - aspect scoring functions is used for OtherScientificTerm| context: current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps .", "entity": "multi - aspect scoring functions", "output": "edit action", "neg_sample": ["multi - aspect scoring functions is used for OtherScientificTerm", "current state - of - the - art methods stochastically sample edit positions and actions , which may cause unnecessary search steps ."], "relation": "used for", "id": "2022.findings-acl.609", "year": 2022, "rel_sent": "Then it introduces four multi - aspect scoring functions to select edit action to further reduce search difficulty .", "forward": true, "src_ids": "2022.findings-acl.609_5001"}
+{"input": "winning lottery ticket is used for Method| context: the lottery ticket hypothesis suggests that for any over - parameterized model , a small subnetwork exists to achieve competitive performance compared to the backbone architecture .", "entity": "winning lottery ticket", "output": "pre - trained language models", "neg_sample": ["winning lottery ticket is used for Method", "the lottery ticket hypothesis suggests that for any over - parameterized model , a small subnetwork exists to achieve competitive performance compared to the backbone architecture ."], "relation": "used for", "id": "2022.findings-acl.654", "year": 2022, "rel_sent": "In this paper , we study whether there is a winning lottery ticket for pre - trained language models , which allow the practitioners to fine - tune the parameters in the ticket but achieve good downstream performance .", "forward": true, "src_ids": "2022.findings-acl.654_5002"}
+{"input": "pre - trained language models is done by using OtherScientificTerm| context: the lottery ticket hypothesis suggests that for any over - parameterized model , a small subnetwork exists to achieve competitive performance compared to the backbone architecture .", "entity": "pre - trained language models", "output": "winning lottery ticket", "neg_sample": ["pre - trained language models is done by using OtherScientificTerm", "the lottery ticket hypothesis suggests that for any over - parameterized model , a small subnetwork exists to achieve competitive performance compared to the backbone architecture ."], "relation": "used for", "id": "2022.findings-acl.654", "year": 2022, "rel_sent": "In this paper , we study whether there is a winning lottery ticket for pre - trained language models , which allow the practitioners to fine - tune the parameters in the ticket but achieve good downstream performance .", "forward": false, "src_ids": "2022.findings-acl.654_5003"}
+{"input": "distributed robust optimization setting is done by using OtherScientificTerm| context: analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms . while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored .", "entity": "distributed robust optimization setting", "output": "set linguistic transformations", "neg_sample": ["distributed robust optimization setting is done by using OtherScientificTerm", "analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms .", "while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored ."], "relation": "used for", "id": "2022.findings-acl.701", "year": 2022, "rel_sent": "In this paper , we present SDRO , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference .", "forward": false, "src_ids": "2022.findings-acl.701_5004"}
+{"input": "set linguistic transformations is used for Task| context: analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms . while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored .", "entity": "set linguistic transformations", "output": "distributed robust optimization setting", "neg_sample": ["set linguistic transformations is used for Task", "analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms .", "while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored ."], "relation": "used for", "id": "2022.findings-acl.701", "year": 2022, "rel_sent": "In this paper , we present SDRO , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference .", "forward": true, "src_ids": "2022.findings-acl.701_5005"}
+{"input": "inference is done by using Method| context: analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms . while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored .", "entity": "inference", "output": "ensembling technique", "neg_sample": ["inference is done by using Method", "analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms .", "while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored ."], "relation": "used for", "id": "2022.findings-acl.701", "year": 2022, "rel_sent": "In this paper , we present SDRO , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference .", "forward": false, "src_ids": "2022.findings-acl.701_5006"}
+{"input": "ensembling technique is used for Task| context: analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms . while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored .", "entity": "ensembling technique", "output": "inference", "neg_sample": ["ensembling technique is used for Task", "analysis of vision - and - language models has revealed their brittleness under linguistic phenomena such as paraphrasing , negation , textual entailment , and word substitutions with synonyms or antonyms .", "while data augmentation techniques have been designed to mitigate against these failure modes , methods that can integrate this knowledge into the training pipeline remain under - explored ."], "relation": "used for", "id": "2022.findings-acl.701", "year": 2022, "rel_sent": "In this paper , we present SDRO , a model - agnostic method that utilizes a set linguistic transformations in a distributed robust optimization setting , along with an ensembling technique to leverage these transformations during inference .", "forward": true, "src_ids": "2022.findings-acl.701_5007"}
+{"input": "embedding space is done by using Method| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "embedding space", "output": "maml - enhanced prototypical networks", "neg_sample": ["embedding space is done by using Method", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.729", "year": 2022, "rel_sent": "For few - shot entity typing , we propose MAML - ProtoNet , i.e. , MAML - enhanced prototypical networks to find a good embedding space that can better distinguish text span representations from different entity classes .", "forward": false, "src_ids": "2022.findings-acl.729_5008"}
+{"input": "text span representations is done by using OtherScientificTerm| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "text span representations", "output": "embedding space", "neg_sample": ["text span representations is done by using OtherScientificTerm", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.729", "year": 2022, "rel_sent": "For few - shot entity typing , we propose MAML - ProtoNet , i.e. , MAML - enhanced prototypical networks to find a good embedding space that can better distinguish text span representations from different entity classes .", "forward": false, "src_ids": "2022.findings-acl.729_5009"}
+{"input": "maml - enhanced prototypical networks is used for OtherScientificTerm| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "maml - enhanced prototypical networks", "output": "embedding space", "neg_sample": ["maml - enhanced prototypical networks is used for OtherScientificTerm", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.729", "year": 2022, "rel_sent": "For few - shot entity typing , we propose MAML - ProtoNet , i.e. , MAML - enhanced prototypical networks to find a good embedding space that can better distinguish text span representations from different entity classes .", "forward": true, "src_ids": "2022.findings-acl.729_5010"}
+{"input": "embedding space is used for Method| context: few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples .", "entity": "embedding space", "output": "text span representations", "neg_sample": ["embedding space is used for Method", "few - shot named entity recognition ( ner ) systems aim at recognizing novel - class named entities based on only a few labeled examples ."], "relation": "used for", "id": "2022.findings-acl.729", "year": 2022, "rel_sent": "For few - shot entity typing , we propose MAML - ProtoNet , i.e. , MAML - enhanced prototypical networks to find a good embedding space that can better distinguish text span representations from different entity classes .", "forward": true, "src_ids": "2022.findings-acl.729_5011"}
+{"input": "dialogue augmentation is done by using OtherScientificTerm| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "dialogue augmentation", "output": "belief state annotations", "neg_sample": ["dialogue augmentation is done by using OtherScientificTerm", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.754", "year": 2022, "rel_sent": "We conclude that exploiting belief state annotations enhances dialogue augmentation and results in improved models in n - shot training scenarios .", "forward": false, "src_ids": "2022.findings-acl.754_5012"}
+{"input": "belief state annotations is used for Task| context: augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure .", "entity": "belief state annotations", "output": "dialogue augmentation", "neg_sample": ["belief state annotations is used for Task", "augmentation of task - oriented dialogues has followed standard methods used for plain - text such as back - translation , word - level manipulation , and paraphrasing despite its richly annotated structure ."], "relation": "used for", "id": "2022.findings-acl.754", "year": 2022, "rel_sent": "We conclude that exploiting belief state annotations enhances dialogue augmentation and results in improved models in n - shot training scenarios .", "forward": true, "src_ids": "2022.findings-acl.754_5013"}
+{"input": "symmetric classification tasks is done by using Method| context: specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores .", "entity": "symmetric classification tasks", "output": "pre - trained models", "neg_sample": ["symmetric classification tasks is done by using Method", "specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores ."], "relation": "used for", "id": "2022.findings-acl.842", "year": 2022, "rel_sent": "Striking a Balance : Alleviating Inconsistency in Pre - trained Models for Symmetric Classification Tasks.", "forward": false, "src_ids": "2022.findings-acl.842_5014"}
+{"input": "symmetric classification is done by using Method| context: while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models . specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores .", "entity": "symmetric classification", "output": "consistency loss function", "neg_sample": ["symmetric classification is done by using Method", "while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models .", "specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores ."], "relation": "used for", "id": "2022.findings-acl.842", "year": 2022, "rel_sent": "We highlight this model shortcoming and apply a consistency loss function to alleviate inconsistency in symmetric classification .", "forward": false, "src_ids": "2022.findings-acl.842_5015"}
+{"input": "consistency loss function is used for Task| context: while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models . specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores .", "entity": "consistency loss function", "output": "symmetric classification", "neg_sample": ["consistency loss function is used for Task", "while fine - tuning pre - trained models for downstream classification is the conventional paradigm in nlp , often task - specific nuances may not get captured in the resultant models .", "specifically , for tasks that take two inputs and require the output to be invariant of the order of the inputs , inconsistency is often observed in the predicted labels or confidence scores ."], "relation": "used for", "id": "2022.findings-acl.842", "year": 2022, "rel_sent": "We highlight this model shortcoming and apply a consistency loss function to alleviate inconsistency in symmetric classification .", "forward": true, "src_ids": "2022.findings-acl.842_5016"}
+{"input": "under - documented languages is done by using OtherScientificTerm| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific . with this paper , we make the case that igt data can be leveraged successfully provided that target language expertise is available .", "entity": "under - documented languages", "output": "linguistic expertise", "neg_sample": ["under - documented languages is done by using OtherScientificTerm", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "with this paper , we make the case that igt data can be leveraged successfully provided that target language expertise is available ."], "relation": "used for", "id": "2022.findings-acl.967", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in { NLP } for Under - Documented Languages.", "forward": false, "src_ids": "2022.findings-acl.967_5017"}
+{"input": "linguistic expertise is used for Material| context: recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers . technologically underserved languages are left behind because they lack such resources . hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts . igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific . with this paper , we make the case that igt data can be leveraged successfully provided that target language expertise is available .", "entity": "linguistic expertise", "output": "under - documented languages", "neg_sample": ["linguistic expertise is used for Material", "recent progress in nlp is driven by pretrained models leveraging massive datasets and has predominantly benefited the world 's political and economic superpowers .", "technologically underserved languages are left behind because they lack such resources .", "hundreds of underserved languages , nevertheless , have available data sources in the form of interlinear glossed text ( igt ) from language documentation efforts .", "igt remains underutilized in nlp work , perhaps because its annotations are only semi - structured and often language - specific .", "with this paper , we make the case that igt data can be leveraged successfully provided that target language expertise is available ."], "relation": "used for", "id": "2022.findings-acl.967", "year": 2022, "rel_sent": "Dim Wihl Gat Tun : The Case for Linguistic Expertise in { NLP } for Under - Documented Languages.", "forward": true, "src_ids": "2022.findings-acl.967_5018"}
+{"input": "continual ner is done by using Method| context: traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types . however , in many real - world scenarios , new entity types are incrementally involved . to investigate this problem , continual learning is introduced for ner . however , the existing method depends on the relevance between tasks and is prone to inter - type confusion .", "entity": "continual ner", "output": "learn - and - review ( l&r )", "neg_sample": ["continual ner is done by using Method", "traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types .", "however , in many real - world scenarios , new entity types are incrementally involved .", "to investigate this problem , continual learning is introduced for ner .", "however , the existing method depends on the relevance between tasks and is prone to inter - type confusion ."], "relation": "used for", "id": "2022.findings-acl.1013", "year": 2022, "rel_sent": "In this paper , we propose a novel two - stage framework Learn - and - Review ( L&R ) for continual NER under the type - incremental setting to alleviate the above issues .", "forward": false, "src_ids": "2022.findings-acl.1013_5019"}
+{"input": "learn - and - review ( l&r ) is used for Task| context: traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types . however , in many real - world scenarios , new entity types are incrementally involved . to investigate this problem , continual learning is introduced for ner . however , the existing method depends on the relevance between tasks and is prone to inter - type confusion .", "entity": "learn - and - review ( l&r )", "output": "continual ner", "neg_sample": ["learn - and - review ( l&r ) is used for Task", "traditional methods for named entity recognition ( ner ) classify mentions into a fixed set of pre - defined entity types .", "however , in many real - world scenarios , new entity types are incrementally involved .", "to investigate this problem , continual learning is introduced for ner .", "however , the existing method depends on the relevance between tasks and is prone to inter - type confusion ."], "relation": "used for", "id": "2022.findings-acl.1013", "year": 2022, "rel_sent": "In this paper , we propose a novel two - stage framework Learn - and - Review ( L&R ) for continual NER under the type - incremental setting to alleviate the above issues .", "forward": true, "src_ids": "2022.findings-acl.1013_5020"}
+{"input": "morphologically complex languages is done by using Method| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "morphologically complex languages", "output": "phoneme recognition setup", "neg_sample": ["morphologically complex languages is done by using Method", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.1014", "year": 2022, "rel_sent": "To automate data preparation , training and evaluation steps , we also developed a phoneme recognition setup which handles morphologically complex languages and writing systems for which no pronunciation dictionary exists .", "forward": false, "src_ids": "2022.findings-acl.1014_5021"}
+{"input": "phoneme recognition setup is used for Material| context: transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers . in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data . however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training .", "entity": "phoneme recognition setup", "output": "morphologically complex languages", "neg_sample": ["phoneme recognition setup is used for Material", "transcription is often reported as the bottleneck in endangered language documentation , requiring large efforts from scarce speakers and transcribers .", "in general , automatic speech recognition ( asr ) can be accurate enough to accelerate transcription only if trained on large amounts of transcribed data .", "however , when a single speaker is involved , several studies have reported encouraging results for phonetic transcription even with small amounts of training ."], "relation": "used for", "id": "2022.findings-acl.1014", "year": 2022, "rel_sent": "To automate data preparation , training and evaluation steps , we also developed a phoneme recognition setup which handles morphologically complex languages and writing systems for which no pronunciation dictionary exists .", "forward": true, "src_ids": "2022.findings-acl.1014_5022"}
+{"input": "ontological relations is used for OtherScientificTerm| context: many tasks in text - based computational social science ( css ) involve the classification of political statements into categories based on a domain - specific codebook . in order to be useful for css analysis , these categories must be fine - grained . the typically skewed distribution of fine - grained categories , however , results in a challenging classification problem on the nlp side .", "entity": "ontological relations", "output": "prior knowledge", "neg_sample": ["ontological relations is used for OtherScientificTerm", "many tasks in text - based computational social science ( css ) involve the classification of political statements into categories based on a domain - specific codebook .", "in order to be useful for css analysis , these categories must be fine - grained .", "the typically skewed distribution of fine - grained categories , however , results in a challenging classification problem on the nlp side ."], "relation": "used for", "id": "2022.findings-acl.1049", "year": 2022, "rel_sent": "We use these ontological relations as prior knowledge to establish additional constraints on the learned model , thus improving performance overall and in particular for infrequent categories .", "forward": true, "src_ids": "2022.findings-acl.1049_5023"}
+{"input": "prior knowledge is done by using OtherScientificTerm| context: many tasks in text - based computational social science ( css ) involve the classification of political statements into categories based on a domain - specific codebook . in order to be useful for css analysis , these categories must be fine - grained . the typically skewed distribution of fine - grained categories , however , results in a challenging classification problem on the nlp side .", "entity": "prior knowledge", "output": "ontological relations", "neg_sample": ["prior knowledge is done by using OtherScientificTerm", "many tasks in text - based computational social science ( css ) involve the classification of political statements into categories based on a domain - specific codebook .", "in order to be useful for css analysis , these categories must be fine - grained .", "the typically skewed distribution of fine - grained categories , however , results in a challenging classification problem on the nlp side ."], "relation": "used for", "id": "2022.findings-acl.1049", "year": 2022, "rel_sent": "We use these ontological relations as prior knowledge to establish additional constraints on the learned model , thus improving performance overall and in particular for infrequent categories .", "forward": false, "src_ids": "2022.findings-acl.1049_5024"}
+{"input": "multimodal transformer - based models is done by using Method| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "multimodal transformer - based models", "output": "adapter - based approaches", "neg_sample": ["multimodal transformer - based models is done by using Method", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.1113", "year": 2022, "rel_sent": "We further propose new adapter - based approaches to adapt multimodal transformer - based models to become multilingual , and --- vice versa --- multilingual models to become multimodal .", "forward": false, "src_ids": "2022.findings-acl.1113_5025"}
+{"input": "adapter - based approaches is used for Method| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "adapter - based approaches", "output": "multimodal transformer - based models", "neg_sample": ["adapter - based approaches is used for Method", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.1113", "year": 2022, "rel_sent": "We further propose new adapter - based approaches to adapt multimodal transformer - based models to become multilingual , and --- vice versa --- multilingual models to become multimodal .", "forward": true, "src_ids": "2022.findings-acl.1113_5026"}
+{"input": "latent multilingual multimodal misalignment is done by using Task| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "latent multilingual multimodal misalignment", "output": "cross - lingual transfer", "neg_sample": ["latent multilingual multimodal misalignment is done by using Task", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.1113", "year": 2022, "rel_sent": "Our results suggest that simple cross - lingual transfer of multimodal models yields latent multilingual multimodal misalignment , calling for more sophisticated methods for vision and multilingual language modeling .", "forward": false, "src_ids": "2022.findings-acl.1113_5027"}
+{"input": "multimodal models is done by using Task| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "multimodal models", "output": "cross - lingual transfer", "neg_sample": ["multimodal models is done by using Task", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.1113", "year": 2022, "rel_sent": "Our results suggest that simple cross - lingual transfer of multimodal models yields latent multilingual multimodal misalignment , calling for more sophisticated methods for vision and multilingual language modeling .", "forward": false, "src_ids": "2022.findings-acl.1113_5028"}
+{"input": "cross - lingual transfer is used for Method| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "cross - lingual transfer", "output": "multimodal models", "neg_sample": ["cross - lingual transfer is used for Method", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.1113", "year": 2022, "rel_sent": "Our results suggest that simple cross - lingual transfer of multimodal models yields latent multilingual multimodal misalignment , calling for more sophisticated methods for vision and multilingual language modeling .", "forward": true, "src_ids": "2022.findings-acl.1113_5029"}
+{"input": "cross - lingual transfer is used for OtherScientificTerm| context: recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts .", "entity": "cross - lingual transfer", "output": "latent multilingual multimodal misalignment", "neg_sample": ["cross - lingual transfer is used for OtherScientificTerm", "recent advances in multimodal vision and language modeling have predominantly focused on the english language , mostly due to the lack of multilingual multimodal datasets to steer modeling efforts ."], "relation": "used for", "id": "2022.findings-acl.1113", "year": 2022, "rel_sent": "Our results suggest that simple cross - lingual transfer of multimodal models yields latent multilingual multimodal misalignment , calling for more sophisticated methods for vision and multilingual language modeling .", "forward": true, "src_ids": "2022.findings-acl.1113_5030"}
+{"input": "event connections is done by using OtherScientificTerm| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "event connections", "output": "structured variable", "neg_sample": ["event connections is done by using OtherScientificTerm", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.1158", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process .", "forward": false, "src_ids": "2022.findings-acl.1158_5031"}
+{"input": "event connections is done by using Method| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "event connections", "output": "bert", "neg_sample": ["event connections is done by using Method", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.1158", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process .", "forward": false, "src_ids": "2022.findings-acl.1158_5032"}
+{"input": "structured variable is used for OtherScientificTerm| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "structured variable", "output": "event connections", "neg_sample": ["structured variable is used for OtherScientificTerm", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.1158", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process .", "forward": true, "src_ids": "2022.findings-acl.1158_5033"}
+{"input": "bert is used for OtherScientificTerm| context: predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events . previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation . however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance .", "entity": "bert", "output": "event connections", "neg_sample": ["bert is used for OtherScientificTerm", "predicting the subsequent event for an existing event context is an important but challenging task , as it requires understanding the underlying relationship between events .", "previous methods propose to retrieve relational features from event graph to enhance the modeling of event correlation .", "however , the sparsity of event graph may restrict the acquisition of relevant graph information , and hence influence the model performance ."], "relation": "used for", "id": "2022.findings-acl.1158", "year": 2022, "rel_sent": "To this end , we incorporate an additional structured variable into BERT to learn to predict the event connections in the training process .", "forward": true, "src_ids": "2022.findings-acl.1158_5034"}
+{"input": "training is done by using OtherScientificTerm| context: identifying the relation between two sentences requires datasets with pairwise annotations . in many cases , these datasets contain instances that are annotated multiple times as part of different pairs . they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations .", "entity": "training", "output": "structural dataset information", "neg_sample": ["training is done by using OtherScientificTerm", "identifying the relation between two sentences requires datasets with pairwise annotations .", "in many cases , these datasets contain instances that are annotated multiple times as part of different pairs .", "they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations ."], "relation": "used for", "id": "2022.findings-acl.1353", "year": 2022, "rel_sent": "This paper investigates how this kind of structural dataset information can be exploited during training .", "forward": false, "src_ids": "2022.findings-acl.1353_5035"}
+{"input": "structural dataset information is used for Task| context: identifying the relation between two sentences requires datasets with pairwise annotations . in many cases , these datasets contain instances that are annotated multiple times as part of different pairs . they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations .", "entity": "structural dataset information", "output": "training", "neg_sample": ["structural dataset information is used for Task", "identifying the relation between two sentences requires datasets with pairwise annotations .", "in many cases , these datasets contain instances that are annotated multiple times as part of different pairs .", "they constitute a structure that contains additional helpful information about the inter - relatedness of the text instances based on the annotations ."], "relation": "used for", "id": "2022.findings-acl.1353", "year": 2022, "rel_sent": "This paper investigates how this kind of structural dataset information can be exploited during training .", "forward": true, "src_ids": "2022.findings-acl.1353_5036"}
+{"input": "glyph similarity measurement is used for OtherScientificTerm| context: modern chinese characters evolved from 3,000 years ago . up to now , tens of thousands of glyphs of ancient characters have been discovered , which must be deciphered by experts to interpret unearthed documents . experts usually need to compare each ancient character to be examined with similar known ones in whole historical periods . however , it is inevitably limited by human memory and experience , which often cost a lot of time but associations are limited to a small scope .", "entity": "glyph similarity measurement", "output": "similar glyph pairs", "neg_sample": ["glyph similarity measurement is used for OtherScientificTerm", "modern chinese characters evolved from 3,000 years ago .", "up to now , tens of thousands of glyphs of ancient characters have been discovered , which must be deciphered by experts to interpret unearthed documents .", "experts usually need to compare each ancient character to be examined with similar known ones in whole historical periods .", "however , it is inevitably limited by human memory and experience , which often cost a lot of time but associations are limited to a small scope ."], "relation": "used for", "id": "2022.findings-acl.1376", "year": 2022, "rel_sent": "In addition , powered by the knowledge of radical systems in ZiNet , this paper introduces glyph similarity measurement between ancient Chinese characters , which could capture similar glyph pairs that are potentially related in origins or semantics .", "forward": true, "src_ids": "2022.findings-acl.1376_5037"}
+{"input": "similar glyph pairs is done by using Method| context: modern chinese characters evolved from 3,000 years ago . up to now , tens of thousands of glyphs of ancient characters have been discovered , which must be deciphered by experts to interpret unearthed documents . experts usually need to compare each ancient character to be examined with similar known ones in whole historical periods . however , it is inevitably limited by human memory and experience , which often cost a lot of time but associations are limited to a small scope .", "entity": "similar glyph pairs", "output": "glyph similarity measurement", "neg_sample": ["similar glyph pairs is done by using Method", "modern chinese characters evolved from 3,000 years ago .", "up to now , tens of thousands of glyphs of ancient characters have been discovered , which must be deciphered by experts to interpret unearthed documents .", "experts usually need to compare each ancient character to be examined with similar known ones in whole historical periods .", "however , it is inevitably limited by human memory and experience , which often cost a lot of time but associations are limited to a small scope ."], "relation": "used for", "id": "2022.findings-acl.1376", "year": 2022, "rel_sent": "In addition , powered by the knowledge of radical systems in ZiNet , this paper introduces glyph similarity measurement between ancient Chinese characters , which could capture similar glyph pairs that are potentially related in origins or semantics .", "forward": false, "src_ids": "2022.findings-acl.1376_5038"}
+{"input": "nlp is done by using Method| context: abstract meaning representation ( amr ) is a semantic representation for nlp / nlu .", "entity": "nlp", "output": "data augmentation", "neg_sample": ["nlp is done by using Method", "abstract meaning representation ( amr ) is a semantic representation for nlp / nlu ."], "relation": "used for", "id": "2022.findings-acl.1381", "year": 2022, "rel_sent": "In this paper , we propose to use it for data augmentation in NLP .", "forward": false, "src_ids": "2022.findings-acl.1381_5039"}
+{"input": "nested named entity recognition is done by using OtherScientificTerm| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem .", "entity": "nested named entity recognition", "output": "heterogeneous factors", "neg_sample": ["nested named entity recognition is done by using OtherScientificTerm", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Fusing Heterogeneous Factors with Triaffine Mechanism for Nested Named Entity Recognition.", "forward": false, "src_ids": "2022.findings-acl.1427_5040"}
+{"input": "nested named entity recognition is done by using Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "entity": "nested named entity recognition", "output": "triaffine mechanism", "neg_sample": ["nested named entity recognition is done by using Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Fusing Heterogeneous Factors with Triaffine Mechanism for Nested Named Entity Recognition.", "forward": false, "src_ids": "2022.findings-acl.1427_5041"}
+{"input": "classification is done by using Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem .", "entity": "classification", "output": "span representations", "neg_sample": ["classification is done by using Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": false, "src_ids": "2022.findings-acl.1427_5042"}
+{"input": "triaffine attention is used for Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem .", "entity": "triaffine attention", "output": "span representations", "neg_sample": ["triaffine attention is used for Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine attention uses boundaries and labels as queries and uses inside tokens and related spans as keys and values for span representations .", "forward": true, "src_ids": "2022.findings-acl.1427_5043"}
+{"input": "triaffine scoring is used for Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem .", "entity": "triaffine scoring", "output": "span representations", "neg_sample": ["triaffine scoring is used for Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": true, "src_ids": "2022.findings-acl.1427_5044"}
+{"input": "triaffine scoring is used for Task| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "entity": "triaffine scoring", "output": "classification", "neg_sample": ["triaffine scoring is used for Task", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": true, "src_ids": "2022.findings-acl.1427_5045"}
+{"input": "span representations is used for Task| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "entity": "span representations", "output": "classification", "neg_sample": ["span representations is used for Task", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": true, "src_ids": "2022.findings-acl.1427_5046"}
+{"input": "triaffine scoring is used for OtherScientificTerm| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem .", "entity": "triaffine scoring", "output": "boundaries", "neg_sample": ["triaffine scoring is used for OtherScientificTerm", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": true, "src_ids": "2022.findings-acl.1427_5047"}
+{"input": "span representations is done by using Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "entity": "span representations", "output": "triaffine attention", "neg_sample": ["span representations is done by using Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine attention uses boundaries and labels as queries and uses inside tokens and related spans as keys and values for span representations .", "forward": false, "src_ids": "2022.findings-acl.1427_5048"}
+{"input": "classification is done by using Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "entity": "classification", "output": "triaffine scoring", "neg_sample": ["classification is done by using Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": false, "src_ids": "2022.findings-acl.1427_5049"}
+{"input": "boundaries is done by using Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "entity": "boundaries", "output": "triaffine scoring", "neg_sample": ["boundaries is done by using Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": false, "src_ids": "2022.findings-acl.1427_5050"}
+{"input": "span representations is done by using Method| context: nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework . a natural solution is to treat the task as a span classification problem . to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition .", "entity": "span representations", "output": "triaffine scoring", "neg_sample": ["span representations is done by using Method", "nested entities are observed in many domains due to their compositionality , which can not be easily recognized by the widely - used sequence labeling framework .", "a natural solution is to treat the task as a span classification problem .", "to learn better span representation and increase classification performance , it is crucial to effectively integrate heterogeneous factors including inside tokens , boundaries , labels , and related spans which could be contributing to nested entities recognition ."], "relation": "used for", "id": "2022.findings-acl.1427", "year": 2022, "rel_sent": "Triaffine scoring interacts with boundaries and span representations for classification .", "forward": false, "src_ids": "2022.findings-acl.1427_5051"}
+{"input": "grammatical error correction is done by using Method| context: in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference . previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks . first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections . second , they ignore the interdependence between different types of corrections .", "entity": "grammatical error correction", "output": "type - driven multi - turn corrections approach", "neg_sample": ["grammatical error correction is done by using Method", "in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference .", "previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks .", "first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections .", "second , they ignore the interdependence between different types of corrections ."], "relation": "used for", "id": "2022.findings-acl.1437", "year": 2022, "rel_sent": "In this paper , we propose a Type - Driven Multi - Turn Corrections approach for GEC .", "forward": false, "src_ids": "2022.findings-acl.1437_5052"}
+{"input": "type - driven multi - turn corrections approach is used for Task| context: in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference . previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks . first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections . second , they ignore the interdependence between different types of corrections .", "entity": "type - driven multi - turn corrections approach", "output": "grammatical error correction", "neg_sample": ["type - driven multi - turn corrections approach is used for Task", "in this aspect , dominant models are trained by one - iteration learning while performing multiple iterations of corrections during inference .", "previous studies mainly focus on the data augmentation approach to combat the exposure bias , which suffers from two drawbacks .", "first , they simply mix additionally - constructed training instances and original ones to train models , which fails to help models be explicitly aware of the procedure of gradual corrections .", "second , they ignore the interdependence between different types of corrections ."], "relation": "used for", "id": "2022.findings-acl.1437", "year": 2022, "rel_sent": "In this paper , we propose a Type - Driven Multi - Turn Corrections approach for GEC .", "forward": true, "src_ids": "2022.findings-acl.1437_5053"}
+{"input": "compositional word replacement is used for OtherScientificTerm| context: we present two simple modifications for word - level perturbation : word replacement considering length ( wr - l ) and compositional word replacement ( cwr ) .", "entity": "compositional word replacement", "output": "compositional candidates", "neg_sample": ["compositional word replacement is used for OtherScientificTerm", "we present two simple modifications for word - level perturbation : word replacement considering length ( wr - l ) and compositional word replacement ( cwr ) ."], "relation": "used for", "id": "2022.findings-acl.1467", "year": 2022, "rel_sent": "CWR considers the compositional candidates by restricting the source of sampling to related words that appear in subword regularization .", "forward": true, "src_ids": "2022.findings-acl.1467_5054"}
+{"input": "wr - l is used for Task| context: we present two simple modifications for word - level perturbation : word replacement considering length ( wr - l ) and compositional word replacement ( cwr ) .", "entity": "wr - l", "output": "text classification", "neg_sample": ["wr - l is used for Task", "we present two simple modifications for word - level perturbation : word replacement considering length ( wr - l ) and compositional word replacement ( cwr ) ."], "relation": "used for", "id": "2022.findings-acl.1467", "year": 2022, "rel_sent": "Experimental results showed that the combination of WR - L and CWR improved the performance of text classification and machine translation .", "forward": true, "src_ids": "2022.findings-acl.1467_5055"}
+{"input": "compositional word replacement is used for Task| context: we present two simple modifications for word - level perturbation : word replacement considering length ( wr - l ) and compositional word replacement ( cwr ) .", "entity": "compositional word replacement", "output": "text classification", "neg_sample": ["compositional word replacement is used for Task", "we present two simple modifications for word - level perturbation : word replacement considering length ( wr - l ) and compositional word replacement ( cwr ) ."], "relation": "used for", "id": "2022.findings-acl.1467", "year": 2022, "rel_sent": "Experimental results showed that the combination of WR - L and CWR improved the performance of text classification and machine translation .", "forward": true, "src_ids": "2022.findings-acl.1467_5056"}
+{"input": "unsupervised pos tagging task is done by using OtherScientificTerm| context: in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks . but , in the unsupervised pos tagging task , works utilizing plms are few and fail to achieve state - of - the - art ( sota ) performance . the recent sota performance is yielded by a guassian hmm variant proposed by he et al . ( 2018 ) . however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms .", "entity": "unsupervised pos tagging task", "output": "hand - crafted features", "neg_sample": ["unsupervised pos tagging task is done by using OtherScientificTerm", "in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks .", "but , in the unsupervised pos tagging task , works utilizing plms are few and fail to achieve state - of - the - art ( sota ) performance .", "the recent sota performance is yielded by a guassian hmm variant proposed by he et al .", "( 2018 ) .", "however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms ."], "relation": "used for", "id": "2022.findings-acl.1475", "year": 2022, "rel_sent": "Bridging Pre - trained Language Models and Hand - crafted Features for Unsupervised { POS } Tagging.", "forward": false, "src_ids": "2022.findings-acl.1475_5057"}
+{"input": "hand - crafted features is used for Task| context: in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks . the recent sota performance is yielded by a guassian hmm variant proposed by he et al . ( 2018 ) . however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms .", "entity": "hand - crafted features", "output": "unsupervised pos tagging task", "neg_sample": ["hand - crafted features is used for Task", "in recent years , large - scale pre - trained language models ( plms ) have made extraordinary progress in most nlp tasks .", "the recent sota performance is yielded by a guassian hmm variant proposed by he et al .", "( 2018 ) .", "however , as a generative model , hmm makes very strong independence assumptions , making it very challenging to incorporate contexualized word representations from plms ."], "relation": "used for", "id": "2022.findings-acl.1475", "year": 2022, "rel_sent": "Bridging Pre - trained Language Models and Hand - crafted Features for Unsupervised { POS } Tagging.", "forward": true, "src_ids": "2022.findings-acl.1475_5058"}
+{"input": "contrastive summaries is done by using Task| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? ''", "entity": "contrastive summaries", "output": "comparative opinion summarization task", "neg_sample": ["contrastive summaries is done by using Task", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? ''"], "relation": "used for", "id": "2022.findings-acl.1493", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews .", "forward": false, "src_ids": "2022.findings-acl.1493_5059"}
+{"input": "comparative opinion summarization task is used for OtherScientificTerm| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? ''", "entity": "comparative opinion summarization task", "output": "contrastive summaries", "neg_sample": ["comparative opinion summarization task is used for OtherScientificTerm", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? ''"], "relation": "used for", "id": "2022.findings-acl.1493", "year": 2022, "rel_sent": "In this paper , we propose the comparative opinion summarization task , which aims at generating two contrastive summaries and one common summary from two different candidate sets of reviews .", "forward": true, "src_ids": "2022.findings-acl.1493_5060"}
+{"input": "contrastive and common summaries is done by using Method| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? ''", "entity": "contrastive and common summaries", "output": "base summarization models", "neg_sample": ["contrastive and common summaries is done by using Method", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? ''"], "relation": "used for", "id": "2022.findings-acl.1493", "year": 2022, "rel_sent": "We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries .", "forward": false, "src_ids": "2022.findings-acl.1493_5061"}
+{"input": "base summarization models is used for OtherScientificTerm| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? ''", "entity": "base summarization models", "output": "contrastive and common summaries", "neg_sample": ["base summarization models is used for OtherScientificTerm", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? ''"], "relation": "used for", "id": "2022.findings-acl.1493", "year": 2022, "rel_sent": "We develop a comparative summarization framework CoCoSum , which consists of two base summarization models that jointly generate contrastive and common summaries .", "forward": true, "src_ids": "2022.findings-acl.1493_5062"}
+{"input": "cocosum is used for OtherScientificTerm| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? ''", "entity": "cocosum", "output": "contrastive and common summaries", "neg_sample": ["cocosum is used for OtherScientificTerm", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? ''"], "relation": "used for", "id": "2022.findings-acl.1493", "year": 2022, "rel_sent": "Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": true, "src_ids": "2022.findings-acl.1493_5063"}
+{"input": "contrastive and common summaries is done by using Method| context: opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews . while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices . thus , the user may still struggle with the question ' which one should i pick ? ''", "entity": "contrastive and common summaries", "output": "cocosum", "neg_sample": ["contrastive and common summaries is done by using Method", "opinion summarization focuses on generating summaries that reflect popular subjective information expressed in multiple online reviews .", "while generated summaries offer general and concise information about a particular hotel or product , the information may be insufficient to help the user compare multiple different choices .", "thus , the user may still struggle with the question ' which one should i pick ? ''"], "relation": "used for", "id": "2022.findings-acl.1493", "year": 2022, "rel_sent": "Experimental results on a newly created benchmark CoCoTrip show that CoCoSum can produce higher - quality contrastive and common summaries than state - of - the - art opinion summarization models .", "forward": false, "src_ids": "2022.findings-acl.1493_5064"}
+{"input": "umls candidate mentions is done by using Method| context: we study cross - lingual umls named entity linking , where mentions in a given source language are mapped to umls concepts , most of which are labeled in english .", "entity": "umls candidate mentions", "output": "per - document pipeline", "neg_sample": ["umls candidate mentions is done by using Method", "we study cross - lingual umls named entity linking , where mentions in a given source language are mapped to umls concepts , most of which are labeled in english ."], "relation": "used for", "id": "2022.findings-acl.1523", "year": 2022, "rel_sent": "Our cross - lingual framework includes an offline unsupervised construction of a translated UMLS dictionary and a per - document pipeline which identifies UMLS candidate mentions and uses a fine - tuned pretrained transformer language model to filter candidates according to context .", "forward": false, "src_ids": "2022.findings-acl.1523_5065"}
+{"input": "per - document pipeline is used for OtherScientificTerm| context: we study cross - lingual umls named entity linking , where mentions in a given source language are mapped to umls concepts , most of which are labeled in english .", "entity": "per - document pipeline", "output": "umls candidate mentions", "neg_sample": ["per - document pipeline is used for OtherScientificTerm", "we study cross - lingual umls named entity linking , where mentions in a given source language are mapped to umls concepts , most of which are labeled in english ."], "relation": "used for", "id": "2022.findings-acl.1523", "year": 2022, "rel_sent": "Our cross - lingual framework includes an offline unsupervised construction of a translated UMLS dictionary and a per - document pipeline which identifies UMLS candidate mentions and uses a fine - tuned pretrained transformer language model to filter candidates according to context .", "forward": true, "src_ids": "2022.findings-acl.1523_5066"}
+{"input": "' coarse '' plot skeleton is done by using Method| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "' coarse '' plot skeleton", "output": "event transition planner", "neg_sample": ["' coarse '' plot skeleton is done by using Method", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.1544", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse '' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": false, "src_ids": "2022.findings-acl.1544_5067"}
+{"input": "text generator is done by using Method| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "text generator", "output": "event transition planner", "neg_sample": ["text generator is done by using Method", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.1544", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse '' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": false, "src_ids": "2022.findings-acl.1544_5068"}
+{"input": "event transition planner is used for OtherScientificTerm| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "event transition planner", "output": "' coarse '' plot skeleton", "neg_sample": ["event transition planner is used for OtherScientificTerm", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.1544", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse '' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": true, "src_ids": "2022.findings-acl.1544_5069"}
+{"input": "skeleton is done by using Method| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "skeleton", "output": "text generator", "neg_sample": ["skeleton is done by using Method", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.1544", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse '' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": false, "src_ids": "2022.findings-acl.1544_5070"}
+{"input": "event transition planner is used for Method| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "event transition planner", "output": "text generator", "neg_sample": ["event transition planner is used for Method", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.1544", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse '' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": true, "src_ids": "2022.findings-acl.1544_5071"}
+{"input": "text generator is used for OtherScientificTerm| context: open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context . the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays . despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events .", "entity": "text generator", "output": "skeleton", "neg_sample": ["text generator is used for OtherScientificTerm", "open - ended text generation tasks , such as dialogue generation and story completion , require models to generate a coherent continuation given limited preceding context .", "the open - ended nature of these tasks brings new challenges to the neural auto - regressive text generators nowadays .", "despite these neural models are good at producing human - like text , it is difficult for them to arrange causalities and relations between given facts and possible ensuing events ."], "relation": "used for", "id": "2022.findings-acl.1544", "year": 2022, "rel_sent": "Our approach can be understood as a specially - trained coarse - to - fine algorithm , where an event transition planner provides a ' coarse '' plot skeleton and a text generator in the second stage refines the skeleton .", "forward": true, "src_ids": "2022.findings-acl.1544_5072"}
+{"input": "anaphoric phenomena is done by using Method| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "anaphoric phenomena", "output": "anaphora annotation framework", "neg_sample": ["anaphoric phenomena is done by using Method", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.1578", "year": 2022, "rel_sent": "To fill this gap , we investigate the textual properties of two types of procedural text , recipes and chemical patents , and generalize an anaphora annotation framework developed for the chemical domain for modeling anaphoric phenomena in recipes .", "forward": false, "src_ids": "2022.findings-acl.1578_5073"}
+{"input": "chemical domain is done by using Method| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "chemical domain", "output": "anaphora annotation framework", "neg_sample": ["chemical domain is done by using Method", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.1578", "year": 2022, "rel_sent": "To fill this gap , we investigate the textual properties of two types of procedural text , recipes and chemical patents , and generalize an anaphora annotation framework developed for the chemical domain for modeling anaphoric phenomena in recipes .", "forward": false, "src_ids": "2022.findings-acl.1578_5074"}
+{"input": "anaphora annotation framework is used for Material| context: procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp .", "entity": "anaphora annotation framework", "output": "chemical domain", "neg_sample": ["anaphora annotation framework is used for Material", "procedural text contains rich anaphoric phenomena , yet has not received much attention in nlp ."], "relation": "used for", "id": "2022.findings-acl.1578", "year": 2022, "rel_sent": "To fill this gap , we investigate the textual properties of two types of procedural text , recipes and chemical patents , and generalize an anaphora annotation framework developed for the chemical domain for modeling anaphoric phenomena in recipes .", "forward": true, "src_ids": "2022.findings-acl.1578_5075"}
+{"input": "hybrid method is used for Task| context: entity linking ( el ) is the task of linking entity mentions in a document to referent entities in a knowledge base ( kb ) . many previous studies focus on wikipedia - derived kbs . there is little work on el over wikidata , even though it is the most extensive crowdsourced kb . the scale of wikidata can open up many new real - world applications , but its massive number of entities also makes el challenging .", "entity": "hybrid method", "output": "candidate retrieval", "neg_sample": ["hybrid method is used for Task", "entity linking ( el ) is the task of linking entity mentions in a document to referent entities in a knowledge base ( kb ) .", "many previous studies focus on wikipedia - derived kbs .", "there is little work on el over wikidata , even though it is the most extensive crowdsourced kb .", "the scale of wikidata can open up many new real - world applications , but its massive number of entities also makes el challenging ."], "relation": "used for", "id": "2022.findings-acl.1676", "year": 2022, "rel_sent": "Our approach complements the traditional approach of using a Wikipedia anchor - text dictionary , enabling us to further design a highly effective hybrid method for candidate retrieval .", "forward": true, "src_ids": "2022.findings-acl.1676_5076"}
+{"input": "candidate retrieval is done by using Method| context: entity linking ( el ) is the task of linking entity mentions in a document to referent entities in a knowledge base ( kb ) . many previous studies focus on wikipedia - derived kbs . there is little work on el over wikidata , even though it is the most extensive crowdsourced kb . the scale of wikidata can open up many new real - world applications , but its massive number of entities also makes el challenging .", "entity": "candidate retrieval", "output": "hybrid method", "neg_sample": ["candidate retrieval is done by using Method", "entity linking ( el ) is the task of linking entity mentions in a document to referent entities in a knowledge base ( kb ) .", "many previous studies focus on wikipedia - derived kbs .", "there is little work on el over wikidata , even though it is the most extensive crowdsourced kb .", "the scale of wikidata can open up many new real - world applications , but its massive number of entities also makes el challenging ."], "relation": "used for", "id": "2022.findings-acl.1676", "year": 2022, "rel_sent": "Our approach complements the traditional approach of using a Wikipedia anchor - text dictionary , enabling us to further design a highly effective hybrid method for candidate retrieval .", "forward": false, "src_ids": "2022.findings-acl.1676_5077"}
+{"input": "neural models is done by using OtherScientificTerm| context: recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words .", "entity": "neural models", "output": "order - altering perturbations", "neg_sample": ["neural models is done by using OtherScientificTerm", "recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words ."], "relation": "used for", "id": "2022.findings-acl.1680", "year": 2022, "rel_sent": "In this paper , we investigate this phenomenon by developing order - altering perturbations on the order of words , subwords , and characters to analyze their effect on neural models ' performance on language understanding tasks .", "forward": false, "src_ids": "2022.findings-acl.1680_5078"}
+{"input": "order of words is done by using OtherScientificTerm| context: recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words .", "entity": "order of words", "output": "order - altering perturbations", "neg_sample": ["order of words is done by using OtherScientificTerm", "recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words ."], "relation": "used for", "id": "2022.findings-acl.1680", "year": 2022, "rel_sent": "In this paper , we investigate this phenomenon by developing order - altering perturbations on the order of words , subwords , and characters to analyze their effect on neural models ' performance on language understanding tasks .", "forward": false, "src_ids": "2022.findings-acl.1680_5079"}
+{"input": "global ordering is done by using OtherScientificTerm| context: recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words .", "entity": "global ordering", "output": "perturbation functions", "neg_sample": ["global ordering is done by using OtherScientificTerm", "recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words ."], "relation": "used for", "id": "2022.findings-acl.1680", "year": 2022, "rel_sent": "We experiment with measuring the impact of perturbations to the local neighborhood of characters and global position of characters in the perturbed texts and observe that perturbation functions found in prior literature only affect the global ordering while the local ordering remains relatively unperturbed .", "forward": false, "src_ids": "2022.findings-acl.1680_5080"}
+{"input": "perturbation functions is used for OtherScientificTerm| context: recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words .", "entity": "perturbation functions", "output": "global ordering", "neg_sample": ["perturbation functions is used for OtherScientificTerm", "recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words ."], "relation": "used for", "id": "2022.findings-acl.1680", "year": 2022, "rel_sent": "We experiment with measuring the impact of perturbations to the local neighborhood of characters and global position of characters in the perturbed texts and observe that perturbation functions found in prior literature only affect the global ordering while the local ordering remains relatively unperturbed .", "forward": true, "src_ids": "2022.findings-acl.1680_5081"}
+{"input": "understanding is done by using OtherScientificTerm| context: recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words .", "entity": "understanding", "output": "local structure of text", "neg_sample": ["understanding is done by using OtherScientificTerm", "recent research analyzing the sensitivity of natural language understanding models to word - order perturbations has shown that neural models are surprisingly insensitive to the order of words ."], "relation": "used for", "id": "2022.findings-acl.1680", "year": 2022, "rel_sent": "We empirically show that neural models , invariant of their inductive biases , pretraining scheme , or the choice of tokenization , mostly rely on the local structure of text to build understanding and make limited use of the global structure .", "forward": false, "src_ids": "2022.findings-acl.1680_5082"}
+{"input": "sign language translation is done by using Method| context: this paper attacks the challenging problem of sign language translation ( slt ) , which involves not only visual and textual understanding but also additional prior knowledge learning ( i.e. performing style , syntax ) . however , the majority of existing methods with vanilla encoder - decoder structures fail to sufficiently explore all of them .", "entity": "sign language translation", "output": "memory enriched transformer", "neg_sample": ["sign language translation is done by using Method", "this paper attacks the challenging problem of sign language translation ( slt ) , which involves not only visual and textual understanding but also additional prior knowledge learning ( i.e.", "performing style , syntax ) .", "however , the majority of existing methods with vanilla encoder - decoder structures fail to sufficiently explore all of them ."], "relation": "used for", "id": "2022.findings-acl.1688", "year": 2022, "rel_sent": "Prior Knowledge and Memory Enriched Transformer for Sign Language Translation.", "forward": false, "src_ids": "2022.findings-acl.1688_5083"}
+{"input": "memory enriched transformer is used for Task| context: performing style , syntax ) . however , the majority of existing methods with vanilla encoder - decoder structures fail to sufficiently explore all of them .", "entity": "memory enriched transformer", "output": "sign language translation", "neg_sample": ["memory enriched transformer is used for Task", "performing style , syntax ) .", "however , the majority of existing methods with vanilla encoder - decoder structures fail to sufficiently explore all of them ."], "relation": "used for", "id": "2022.findings-acl.1688", "year": 2022, "rel_sent": "Prior Knowledge and Memory Enriched Transformer for Sign Language Translation.", "forward": true, "src_ids": "2022.findings-acl.1688_5084"}
+{"input": "summary templates is used for Task| context: annotating task - oriented dialogues is notorious for the expensive and difficult data collection process . few - shot dialogue state tracking ( dst ) is a realistic solution to this problem .", "entity": "summary templates", "output": "training", "neg_sample": ["summary templates is used for Task", "annotating task - oriented dialogues is notorious for the expensive and difficult data collection process .", "few - shot dialogue state tracking ( dst ) is a realistic solution to this problem ."], "relation": "used for", "id": "2022.findings-acl.1711", "year": 2022, "rel_sent": "Finally , based on our analysis , we discover that the naturalness of the summary templates plays a key role for successful training .", "forward": true, "src_ids": "2022.findings-acl.1711_5085"}
+{"input": "training is done by using OtherScientificTerm| context: annotating task - oriented dialogues is notorious for the expensive and difficult data collection process . few - shot dialogue state tracking ( dst ) is a realistic solution to this problem .", "entity": "training", "output": "summary templates", "neg_sample": ["training is done by using OtherScientificTerm", "annotating task - oriented dialogues is notorious for the expensive and difficult data collection process .", "few - shot dialogue state tracking ( dst ) is a realistic solution to this problem ."], "relation": "used for", "id": "2022.findings-acl.1711", "year": 2022, "rel_sent": "Finally , based on our analysis , we discover that the naturalness of the summary templates plays a key role for successful training .", "forward": false, "src_ids": "2022.findings-acl.1711_5086"}
+{"input": "sentence - level explanations is done by using Method| context: research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not . building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy . however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp . furthermore , the existing methods can not utilize a large size of unlabeled dataset to further improve the model interpretability .", "entity": "sentence - level explanations", "output": "interpretable neural model", "neg_sample": ["sentence - level explanations is done by using Method", "research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not .", "building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy .", "however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp .", "furthermore , the existing methods can not utilize a large size of unlabeled dataset to further improve the model interpretability ."], "relation": "used for", "id": "2022.findings-acl.1727", "year": 2022, "rel_sent": "To address these limitations , we aim to build an interpretable neural model which can provide sentence - level explanations and apply weakly supervised approach to further leverage the large corpus of unlabeled datasets to boost the interpretability in addition to improving prediction performance as existing works have done .", "forward": false, "src_ids": "2022.findings-acl.1727_5087"}
+{"input": "interpretable neural model is used for OtherScientificTerm| context: research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not . building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy . however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp . furthermore , the existing methods can not utilize a large size of unlabeled dataset to further improve the model interpretability .", "entity": "interpretable neural model", "output": "sentence - level explanations", "neg_sample": ["interpretable neural model is used for OtherScientificTerm", "research replication prediction ( rrp ) is the task of predicting whether a published research result can be replicated or not .", "building an interpretable neural text classifier for rrp promotes the understanding of why a research paper is predicted as replicable or non - replicable and therefore makes its real - world application more reliable and trustworthy .", "however , the prior works on model interpretation mainly focused on improving the model interpretability at the word / phrase level , which are insufficient especially for long research papers in rrp .", "furthermore , the existing methods can not utilize a large size of unlabeled dataset to further improve the model interpretability ."], "relation": "used for", "id": "2022.findings-acl.1727", "year": 2022, "rel_sent": "To address these limitations , we aim to build an interpretable neural model which can provide sentence - level explanations and apply weakly supervised approach to further leverage the large corpus of unlabeled datasets to boost the interpretability in addition to improving prediction performance as existing works have done .", "forward": true, "src_ids": "2022.findings-acl.1727_5088"}
+{"input": "precise dialogue states is done by using Method| context: previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set . approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations . furthermore , previously proposed dialogue state representations are ambiguous and lack the precision necessary for building an effective agent .", "entity": "precise dialogue states", "output": "sample - efficient methodology", "neg_sample": ["precise dialogue states is done by using Method", "previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set .", "approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations .", "furthermore , previously proposed dialogue state representations are ambiguous and lack the precision necessary for building an effective agent ."], "relation": "used for", "id": "2022.findings-acl.1814", "year": 2022, "rel_sent": "This paper proposes a new dialogue representation and a sample - efficient methodology that can predict precise dialogue states in WOZ conversations .", "forward": false, "src_ids": "2022.findings-acl.1814_5089"}
+{"input": "sample - efficient methodology is used for OtherScientificTerm| context: previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set . approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations . furthermore , previously proposed dialogue state representations are ambiguous and lack the precision necessary for building an effective agent .", "entity": "sample - efficient methodology", "output": "precise dialogue states", "neg_sample": ["sample - efficient methodology is used for OtherScientificTerm", "previous attempts to build effective semantic parsers for wizard - of - oz ( woz ) conversations suffer from the difficulty in acquiring a high - quality , manually annotated training set .", "approaches based only on dialogue synthesis are insufficient , as dialogues generated from state - machine based models are poor approximations of real - life conversations .", "furthermore , previously proposed dialogue state representations are ambiguous and lack the precision necessary for building an effective agent ."], "relation": "used for", "id": "2022.findings-acl.1814", "year": 2022, "rel_sent": "This paper proposes a new dialogue representation and a sample - efficient methodology that can predict precise dialogue states in WOZ conversations .", "forward": true, "src_ids": "2022.findings-acl.1814_5090"}
+{"input": "mention representations is done by using OtherScientificTerm| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "mention representations", "output": "sibling mentions", "neg_sample": ["mention representations is done by using OtherScientificTerm", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.611", "year": 2022, "rel_sent": "To this end , we propose to exploit sibling mentions for enhancing the mention representations .", "forward": false, "src_ids": "2022.acl-long.611_5091"}
+{"input": "sibling mentions is used for Method| context: in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance .", "entity": "sibling mentions", "output": "mention representations", "neg_sample": ["sibling mentions is used for Method", "in this paper , we firstly empirically find that existing models struggle to handle hard mentions due to their insufficient contexts , which consequently limits their overall typing performance ."], "relation": "used for", "id": "2022.acl-long.611", "year": 2022, "rel_sent": "To this end , we propose to exploit sibling mentions for enhancing the mention representations .", "forward": true, "src_ids": "2022.acl-long.611_5092"}
+{"input": "pretraining is used for Task| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "pretraining", "output": "sign language recognition", "neg_sample": ["pretraining is used for Task", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.615", "year": 2022, "rel_sent": "Fourth , we compare different pretraining strategies and for the first time establish that pretraining is effective for sign language recognition by demonstrating ( a ) improved fine - tuning performance especially in low - resource settings , and ( b ) high crosslingual transfer from Indian - SL to few other sign languages .", "forward": true, "src_ids": "2022.acl-long.615_5093"}
+{"input": "sign language recognition is done by using Method| context: ai technologies for natural languages have made tremendous progress recently . however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences .", "entity": "sign language recognition", "output": "pretraining", "neg_sample": ["sign language recognition is done by using Method", "ai technologies for natural languages have made tremendous progress recently .", "however , commensurate progress has not been made on sign languages , in particular , in recognizing signs as individual words or as complete sentences ."], "relation": "used for", "id": "2022.acl-long.615", "year": 2022, "rel_sent": "Fourth , we compare different pretraining strategies and for the first time establish that pretraining is effective for sign language recognition by demonstrating ( a ) improved fine - tuning performance especially in low - resource settings , and ( b ) high crosslingual transfer from Indian - SL to few other sign languages .", "forward": false, "src_ids": "2022.acl-long.615_5094"}
+{"input": "pretrained language models is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "pretrained language models", "output": "bert2{bert }", "neg_sample": ["pretrained language models is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "bert2{BERT } : Towards Reusable Pretrained Language Models.", "forward": false, "src_ids": "2022.acl-long.618_5095"}
+{"input": "bert2{bert } is used for Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "bert2{bert }", "output": "pretrained language models", "neg_sample": ["bert2{bert } is used for Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "bert2{BERT } : Towards Reusable Pretrained Language Models.", "forward": true, "src_ids": "2022.acl-long.618_5096"}
+{"input": "pre - training bert_base is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "pre - training bert_base", "output": "bert2bert", "neg_sample": ["pre - training bert_base is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "In particular , bert2BERT saves about 45 % and 47 % computational cost of pre - training BERT_BASE and GPT_BASE by reusing the models of almost their half sizes .", "forward": false, "src_ids": "2022.acl-long.618_5097"}
+{"input": "gpt_base is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "gpt_base", "output": "bert2bert", "neg_sample": ["gpt_base is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "In particular , bert2BERT saves about 45 % and 47 % computational cost of pre - training BERT_BASE and GPT_BASE by reusing the models of almost their half sizes .", "forward": false, "src_ids": "2022.acl-long.618_5098"}
+{"input": "two - stage learning method is used for Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "entity": "two - stage learning method", "output": "pre - training", "neg_sample": ["two - stage learning method is used for Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "In addition , a two - stage learning method is proposed to further accelerate the pre - training .", "forward": true, "src_ids": "2022.acl-long.618_5099"}
+{"input": "pre - training is done by using Method| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "pre - training", "output": "two - stage learning method", "neg_sample": ["pre - training is done by using Method", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "In addition , a two - stage learning method is proposed to further accelerate the pre - training .", "forward": false, "src_ids": "2022.acl-long.618_5100"}
+{"input": "bert2bert is used for Task| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "bert2bert", "output": "pre - training bert_base", "neg_sample": ["bert2bert is used for Task", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "In particular , bert2BERT saves about 45 % and 47 % computational cost of pre - training BERT_BASE and GPT_BASE by reusing the models of almost their half sizes .", "forward": true, "src_ids": "2022.acl-long.618_5101"}
+{"input": "bert2bert is used for OtherScientificTerm| context: in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models . however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful .", "entity": "bert2bert", "output": "gpt_base", "neg_sample": ["bert2bert is used for OtherScientificTerm", "in recent years , researchers tend to pre - train ever - larger language models to explore the upper limit of deep models .", "however , large language model pre - training costs intensive computational resources , and most of the models are trained from scratch without reusing the existing pre - trained models , which is wasteful ."], "relation": "used for", "id": "2022.acl-long.618", "year": 2022, "rel_sent": "In particular , bert2BERT saves about 45 % and 47 % computational cost of pre - training BERT_BASE and GPT_BASE by reusing the models of almost their half sizes .", "forward": true, "src_ids": "2022.acl-long.618_5102"}
+{"input": "pretraining and downstream tasks is done by using Method| context: as an important task in sentiment analysis , multimodal aspect - based sentiment analysis ( mabsa ) has attracted increasing attention in recent years . however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodal alignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grained aspects , opinions , and their alignments across modalities .", "entity": "pretraining and downstream tasks", "output": "unified multimodal encoder - decoder architecture", "neg_sample": ["pretraining and downstream tasks is done by using Method", "as an important task in sentiment analysis , multimodal aspect - based sentiment analysis ( mabsa ) has attracted increasing attention in recent years .", "however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodal alignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grained aspects , opinions , and their alignments across modalities ."], "relation": "used for", "id": "2022.acl-long.620", "year": 2022, "rel_sent": "To tackle these limitations , we propose a task - specific Vision - Language Pre - training framework for MABSA ( VLP - MABSA ) , which is a unified multimodal encoder - decoder architecture for all the pretraining and downstream tasks .", "forward": false, "src_ids": "2022.acl-long.620_5103"}
+{"input": "unified multimodal encoder - decoder architecture is used for Task| context: as an important task in sentiment analysis , multimodal aspect - based sentiment analysis ( mabsa ) has attracted increasing attention in recent years . however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodal alignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grained aspects , opinions , and their alignments across modalities .", "entity": "unified multimodal encoder - decoder architecture", "output": "pretraining and downstream tasks", "neg_sample": ["unified multimodal encoder - decoder architecture is used for Task", "as an important task in sentiment analysis , multimodal aspect - based sentiment analysis ( mabsa ) has attracted increasing attention in recent years .", "however , previous approaches either ( i ) use separately pre - trained visual and textual models , which ignore the crossmodal alignment or ( ii ) use vision - language models pre - trained with general pre - training tasks , which are inadequate to identify fine - grained aspects , opinions , and their alignments across modalities ."], "relation": "used for", "id": "2022.acl-long.620", "year": 2022, "rel_sent": "To tackle these limitations , we propose a task - specific Vision - Language Pre - training framework for MABSA ( VLP - MABSA ) , which is a unified multimodal encoder - decoder architecture for all the pretraining and downstream tasks .", "forward": true, "src_ids": "2022.acl-long.620_5104"}
+{"input": "computational framework is used for OtherScientificTerm| context: in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "computational framework", "output": "hedges", "neg_sample": ["computational framework is used for OtherScientificTerm", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": true, "src_ids": "2022.acl-long.621_5105"}
+{"input": "features is used for OtherScientificTerm| context: in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "features", "output": "hedges", "neg_sample": ["features is used for OtherScientificTerm", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "We employ a model explainability tool to explore the features that characterize hedges in peer - tutoring conversations , and we identify some novel features , and the benefits of a such a hybrid model approach .", "forward": true, "src_ids": "2022.acl-long.621_5106"}
+{"input": "learning is done by using Method| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "learning", "output": "tutoring agent", "neg_sample": ["learning is done by using Method", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.621_5107"}
+{"input": "rapport is done by using Method| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "rapport", "output": "tutoring agent", "neg_sample": ["rapport is done by using Method", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.621_5108"}
+{"input": "tutoring agent is used for Task| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "tutoring agent", "output": "learning", "neg_sample": ["tutoring agent is used for Task", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": true, "src_ids": "2022.acl-long.621_5109"}
+{"input": "computational framework is done by using Material| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "computational framework", "output": "multimodal peer - tutoring dataset", "neg_sample": ["computational framework is done by using Material", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.621_5110"}
+{"input": "hedges is done by using Method| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "hedges", "output": "computational framework", "neg_sample": ["hedges is done by using Method", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": false, "src_ids": "2022.acl-long.621_5111"}
+{"input": "multimodal peer - tutoring dataset is used for Method| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "multimodal peer - tutoring dataset", "output": "computational framework", "neg_sample": ["multimodal peer - tutoring dataset is used for Method", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "Pursuing the objective of building a tutoring agent that manages rapport with teenagers in order to improve learning , we used a multimodal peer - tutoring dataset to construct a computational framework for identifying hedges .", "forward": true, "src_ids": "2022.acl-long.621_5112"}
+{"input": "hedges is done by using OtherScientificTerm| context: hedges have an important role in the management of rapport . in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback .", "entity": "hedges", "output": "features", "neg_sample": ["hedges is done by using OtherScientificTerm", "hedges have an important role in the management of rapport .", "in peer - tutoring , they are notably used by tutors in dyads experiencing low rapport to tone down the impact of instructions and negative feedback ."], "relation": "used for", "id": "2022.acl-long.621", "year": 2022, "rel_sent": "We employ a model explainability tool to explore the features that characterize hedges in peer - tutoring conversations , and we identify some novel features , and the benefits of a such a hybrid model approach .", "forward": false, "src_ids": "2022.acl-long.621_5113"}
+{"input": "redundant nodes is done by using Method| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "redundant nodes", "output": "cluster - based pruning solution", "neg_sample": ["redundant nodes is done by using Method", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.622", "year": 2022, "rel_sent": "We then suggest a cluster - based pruning solution to filter out 10%~40 % redundant nodes in large datastores while retaining translation quality .", "forward": false, "src_ids": "2022.acl-long.622_5114"}
+{"input": "cluster - based pruning solution is used for OtherScientificTerm| context: k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) . previous studies ( khandelwal et al . , 2021 ; zheng et al . , 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data . in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores .", "entity": "cluster - based pruning solution", "output": "redundant nodes", "neg_sample": ["cluster - based pruning solution is used for OtherScientificTerm", "k - nearest - neighbor machine translation ( knn - mt ) has been recently proposed as a non - parametric solution for domain adaptation in neural machine translation ( nmt ) .", "previous studies ( khandelwal et al .", ", 2021 ; zheng et al .", ", 2021 ) have already demonstrated that non - parametric nmt is even superior to models fine - tuned on out - of - domain data .", "in spite of this success , knn retrieval is at the expense of high latency , in particular for large datastores ."], "relation": "used for", "id": "2022.acl-long.622", "year": 2022, "rel_sent": "We then suggest a cluster - based pruning solution to filter out 10%~40 % redundant nodes in large datastores while retaining translation quality .", "forward": true, "src_ids": "2022.acl-long.622_5115"}
+{"input": "backbone plms is done by using Method| context: for fget , a key challenge is the low - resource problem --- the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "backbone plms", "output": "cross - lingual contrastive learning framework", "neg_sample": ["backbone plms is done by using Method", "for fget , a key challenge is the low - resource problem --- the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.640", "year": 2022, "rel_sent": "Furthermore , we introduce entity - pair - oriented heuristic rules as well as machine translation to obtain cross - lingual distantly - supervised data , and apply cross - lingual contrastive learning on the distantly - supervised data to enhance the backbone PLMs .", "forward": false, "src_ids": "2022.acl-long.640_5116"}
+{"input": "cross - lingual contrastive learning framework is used for Method| context: for fget , a key challenge is the low - resource problem --- the complex entity type hierarchy makes it difficult to manually label data . especially for those languages other than english , human - labeled data is extremely scarce .", "entity": "cross - lingual contrastive learning framework", "output": "backbone plms", "neg_sample": ["cross - lingual contrastive learning framework is used for Method", "for fget , a key challenge is the low - resource problem --- the complex entity type hierarchy makes it difficult to manually label data .", "especially for those languages other than english , human - labeled data is extremely scarce ."], "relation": "used for", "id": "2022.acl-long.640", "year": 2022, "rel_sent": "Furthermore , we introduce entity - pair - oriented heuristic rules as well as machine translation to obtain cross - lingual distantly - supervised data , and apply cross - lingual contrastive learning on the distantly - supervised data to enhance the backbone PLMs .", "forward": true, "src_ids": "2022.acl-long.640_5117"}
+{"input": "spatial knowledge is done by using Method| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "spatial knowledge", "output": "image synthesis models", "neg_sample": ["spatial knowledge is done by using Method", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.680", "year": 2022, "rel_sent": "We probe PLMs and models with visual signals , including vision - language pretrained models and image synthesis models , on this benchmark , and find that image synthesis models are more capable of learning accurate and consistent spatial knowledge than other models .", "forward": false, "src_ids": "2022.acl-long.680_5118"}
+{"input": "image synthesis models is used for OtherScientificTerm| context: spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge . although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning .", "entity": "image synthesis models", "output": "spatial knowledge", "neg_sample": ["image synthesis models is used for OtherScientificTerm", "spatial commonsense , the knowledge about spatial position and relationship between objects ( like the relative size of a lion and a girl , and the position of a boy relative to a bicycle when cycling ) , is an important part of commonsense knowledge .", "although pretrained language models ( plms ) succeed in many nlp tasks , they are shown to be ineffective in spatial commonsense reasoning ."], "relation": "used for", "id": "2022.acl-long.680", "year": 2022, "rel_sent": "We probe PLMs and models with visual signals , including vision - language pretrained models and image synthesis models , on this benchmark , and find that image synthesis models are more capable of learning accurate and consistent spatial knowledge than other models .", "forward": true, "src_ids": "2022.acl-long.680_5119"}
+{"input": "machine reading comprehension ( mrc ) framework is used for Method| context: higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level .", "entity": "machine reading comprehension ( mrc ) framework", "output": "span linking module", "neg_sample": ["machine reading comprehension ( mrc ) framework is used for Method", "higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level ."], "relation": "used for", "id": "2022.acl-long.694", "year": 2022, "rel_sent": "We use the machine reading comprehension ( MRC ) framework as the backbone to formalize the span linking module , where one span is used as query to extract the text span / subtree it should be linked to .", "forward": true, "src_ids": "2022.acl-long.694_5120"}
+{"input": "span linking module is done by using Method| context: higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level .", "entity": "span linking module", "output": "machine reading comprehension ( mrc ) framework", "neg_sample": ["span linking module is done by using Method", "higher - order methods for dependency parsing can partially but not fully address the issue that edges in dependency trees should be constructed at the text span / subtree level rather than word level ."], "relation": "used for", "id": "2022.acl-long.694", "year": 2022, "rel_sent": "We use the machine reading comprehension ( MRC ) framework as the backbone to formalize the span linking module , where one span is used as query to extract the text span / subtree it should be linked to .", "forward": false, "src_ids": "2022.acl-long.694_5121"}
+{"input": "cross - lingual spoken language understanding is done by using Method| context: due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort . however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "cross - lingual spoken language understanding", "output": "global -- local contrastive learning framework", "neg_sample": ["cross - lingual spoken language understanding is done by using Method", "due to high data demands of current methods , attention to zero - shot cross - lingual spoken language understanding ( slu ) has grown , as such approaches greatly reduce human annotation effort .", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.756", "year": 2022, "rel_sent": "GL - CLeF : A Global -- Local Contrastive Learning Framework for Cross - lingual Spoken Language Understanding.", "forward": false, "src_ids": "2022.acl-long.756_5122"}
+{"input": "global -- local contrastive learning framework is used for Task| context: however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages .", "entity": "global -- local contrastive learning framework", "output": "cross - lingual spoken language understanding", "neg_sample": ["global -- local contrastive learning framework is used for Task", "however , existing models solely rely on shared parameters , which can only perform implicit alignment across languages ."], "relation": "used for", "id": "2022.acl-long.756", "year": 2022, "rel_sent": "GL - CLeF : A Global -- Local Contrastive Learning Framework for Cross - lingual Spoken Language Understanding.", "forward": true, "src_ids": "2022.acl-long.756_5123"}
+{"input": "meta - framework is done by using Method| context: news events are often associated with quantities ( e.g. , the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events .", "entity": "meta - framework", "output": "meta - framework", "neg_sample": ["meta - framework is done by using Method", "news events are often associated with quantities ( e.g.", ", the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events ."], "relation": "used for", "id": "2022.acl-long.779", "year": 2022, "rel_sent": "We demonstrate the meta - framework in three domains --- the COVID-19 pandemic , Black Lives Matter protests , and 2020 California wildfires --- to show that the formalism is general and extensible , the crowdsourcing pipeline facilitates fast and high - quality data annotation , and the baseline system can handle spatiotemporal quantity extraction well enough to be practically useful .", "forward": false, "src_ids": "2022.acl-long.779_5124"}
+{"input": "meta - framework is used for Method| context: news events are often associated with quantities ( e.g. , the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events .", "entity": "meta - framework", "output": "meta - framework", "neg_sample": ["meta - framework is used for Method", "news events are often associated with quantities ( e.g.", ", the number of covid-19 patients or the number of arrests in a protest ) , and it is often important to extract their type , time , and location from unstructured text in order to analyze these quantity events ."], "relation": "used for", "id": "2022.acl-long.779", "year": 2022, "rel_sent": "We demonstrate the meta - framework in three domains --- the COVID-19 pandemic , Black Lives Matter protests , and 2020 California wildfires --- to show that the formalism is general and extensible , the crowdsourcing pipeline facilitates fast and high - quality data annotation , and the baseline system can handle spatiotemporal quantity extraction well enough to be practically useful .", "forward": true, "src_ids": "2022.acl-long.779_5125"}
+{"input": "vision - language models is done by using Method| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "vision - language models", "output": "low - resource prompt - based learning", "neg_sample": ["vision - language models is done by using Method", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "A Good Prompt Is Worth Millions of Parameters : Low - resource Prompt - based Learning for Vision - Language Models.", "forward": false, "src_ids": "2022.acl-long.783_5126"}
+{"input": "low - resource prompt - based learning is used for Method| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "low - resource prompt - based learning", "output": "vision - language models", "neg_sample": ["low - resource prompt - based learning is used for Method", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "A Good Prompt Is Worth Millions of Parameters : Low - resource Prompt - based Learning for Vision - Language Models.", "forward": true, "src_ids": "2022.acl-long.783_5127"}
+{"input": "captioning is done by using Method| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "captioning", "output": "prefix language modeling ( prefixlm )", "neg_sample": ["captioning is done by using Method", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "In our analysis , we observe that ( 1 ) prompts significantly affect zero - shot performance but marginally affect few - shot performance , ( 2 ) models with noisy prompts learn as quickly as hand - crafted prompts given larger training data , and ( 3 ) MaskedLM helps VQA tasks while PrefixLM boosts captioning performance .", "forward": false, "src_ids": "2022.acl-long.783_5128"}
+{"input": "vqa tasks is done by using Method| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "vqa tasks", "output": "masked language modeling ( maskedlm )", "neg_sample": ["vqa tasks is done by using Method", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "In our analysis , we observe that ( 1 ) prompts significantly affect zero - shot performance but marginally affect few - shot performance , ( 2 ) models with noisy prompts learn as quickly as hand - crafted prompts given larger training data , and ( 3 ) MaskedLM helps VQA tasks while PrefixLM boosts captioning performance .", "forward": false, "src_ids": "2022.acl-long.783_5129"}
+{"input": "few - shot tasks is done by using OtherScientificTerm| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "few - shot tasks", "output": "diverse prompts", "neg_sample": ["few - shot tasks is done by using OtherScientificTerm", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "Furthermore , we analyze the effect of diverse prompts for few - shot tasks .", "forward": false, "src_ids": "2022.acl-long.783_5130"}
+{"input": "diverse prompts is used for Task| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "diverse prompts", "output": "few - shot tasks", "neg_sample": ["diverse prompts is used for Task", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "Furthermore , we analyze the effect of diverse prompts for few - shot tasks .", "forward": true, "src_ids": "2022.acl-long.783_5131"}
+{"input": "masked language modeling ( maskedlm ) is used for Task| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "masked language modeling ( maskedlm )", "output": "vqa tasks", "neg_sample": ["masked language modeling ( maskedlm ) is used for Task", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "In our analysis , we observe that ( 1 ) prompts significantly affect zero - shot performance but marginally affect few - shot performance , ( 2 ) models with noisy prompts learn as quickly as hand - crafted prompts given larger training data , and ( 3 ) MaskedLM helps VQA tasks while PrefixLM boosts captioning performance .", "forward": true, "src_ids": "2022.acl-long.783_5132"}
+{"input": "prefix language modeling ( prefixlm ) is used for Task| context: large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning . however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed .", "entity": "prefix language modeling ( prefixlm )", "output": "captioning", "neg_sample": ["prefix language modeling ( prefixlm ) is used for Task", "large pre - trained vision - language ( vl ) models can learn a new task with a handful of examples and generalize to a new task without fine - tuning .", "however , these vl models are hard to deploy for real - world applications due to their impractically huge sizes and slow inference speed ."], "relation": "used for", "id": "2022.acl-long.783", "year": 2022, "rel_sent": "In our analysis , we observe that ( 1 ) prompts significantly affect zero - shot performance but marginally affect few - shot performance , ( 2 ) models with noisy prompts learn as quickly as hand - crafted prompts given larger training data , and ( 3 ) MaskedLM helps VQA tasks while PrefixLM boosts captioning performance .", "forward": true, "src_ids": "2022.acl-long.783_5133"}
+{"input": "joint commonsense and fact - view link prediction is done by using OtherScientificTerm| context: knowledge graphs store a large number of factual triples while they are still incomplete , inevitably . the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance .", "entity": "joint commonsense and fact - view link prediction", "output": "commonsense", "neg_sample": ["joint commonsense and fact - view link prediction is done by using OtherScientificTerm", "knowledge graphs store a large number of factual triples while they are still incomplete , inevitably .", "the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance ."], "relation": "used for", "id": "2022.acl-long.810", "year": 2022, "rel_sent": "The generated commonsense augments effective self - supervision to facilitate both high - quality negative sampling ( NS ) and joint commonsense and fact - view link prediction .", "forward": false, "src_ids": "2022.acl-long.810_5134"}
+{"input": "high - quality negative sampling ( ns ) is done by using OtherScientificTerm| context: knowledge graphs store a large number of factual triples while they are still incomplete , inevitably . the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance .", "entity": "high - quality negative sampling ( ns )", "output": "commonsense", "neg_sample": ["high - quality negative sampling ( ns ) is done by using OtherScientificTerm", "knowledge graphs store a large number of factual triples while they are still incomplete , inevitably .", "the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance ."], "relation": "used for", "id": "2022.acl-long.810", "year": 2022, "rel_sent": "The generated commonsense augments effective self - supervision to facilitate both high - quality negative sampling ( NS ) and joint commonsense and fact - view link prediction .", "forward": false, "src_ids": "2022.acl-long.810_5135"}
+{"input": "high - quality negative sampling ( ns ) is done by using Method| context: knowledge graphs store a large number of factual triples while they are still incomplete , inevitably . the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge . the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance .", "entity": "high - quality negative sampling ( ns )", "output": "self - supervision", "neg_sample": ["high - quality negative sampling ( ns ) is done by using Method", "knowledge graphs store a large number of factual triples while they are still incomplete , inevitably .", "the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge .", "the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance ."], "relation": "used for", "id": "2022.acl-long.810", "year": 2022, "rel_sent": "The generated commonsense augments effective self - supervision to facilitate both high - quality negative sampling ( NS ) and joint commonsense and fact - view link prediction .", "forward": false, "src_ids": "2022.acl-long.810_5136"}
+{"input": "joint commonsense and fact - view link prediction is done by using Method| context: knowledge graphs store a large number of factual triples while they are still incomplete , inevitably . the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge . the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance .", "entity": "joint commonsense and fact - view link prediction", "output": "self - supervision", "neg_sample": ["joint commonsense and fact - view link prediction is done by using Method", "knowledge graphs store a large number of factual triples while they are still incomplete , inevitably .", "the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge .", "the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance ."], "relation": "used for", "id": "2022.acl-long.810", "year": 2022, "rel_sent": "The generated commonsense augments effective self - supervision to facilitate both high - quality negative sampling ( NS ) and joint commonsense and fact - view link prediction .", "forward": false, "src_ids": "2022.acl-long.810_5137"}
+{"input": "self - supervision is used for Task| context: knowledge graphs store a large number of factual triples while they are still incomplete , inevitably . the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge . the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance .", "entity": "self - supervision", "output": "high - quality negative sampling ( ns )", "neg_sample": ["self - supervision is used for Task", "knowledge graphs store a large number of factual triples while they are still incomplete , inevitably .", "the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge .", "the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance ."], "relation": "used for", "id": "2022.acl-long.810", "year": 2022, "rel_sent": "The generated commonsense augments effective self - supervision to facilitate both high - quality negative sampling ( NS ) and joint commonsense and fact - view link prediction .", "forward": true, "src_ids": "2022.acl-long.810_5138"}
+{"input": "commonsense is used for Task| context: knowledge graphs store a large number of factual triples while they are still incomplete , inevitably . the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge . the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance .", "entity": "commonsense", "output": "high - quality negative sampling ( ns )", "neg_sample": ["commonsense is used for Task", "knowledge graphs store a large number of factual triples while they are still incomplete , inevitably .", "the previous knowledge graph completion ( kgc ) models predict missing links between entities merely relying on fact - view data , ignoring the valuable commonsense knowledge .", "the previous knowledge graph embedding ( kge ) techniques suffer from invalid negative sampling and the uncertainty of fact - view link prediction , limiting kgc 's performance ."], "relation": "used for", "id": "2022.acl-long.810", "year": 2022, "rel_sent": "The generated commonsense augments effective self - supervision to facilitate both high - quality negative sampling ( NS ) and joint commonsense and fact - view link prediction .", "forward": true, "src_ids": "2022.acl-long.810_5139"}
+{"input": "clip is used for Method| context: we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions .", "entity": "clip", "output": "fine - grained semantic representations of sentences", "neg_sample": ["clip is used for Method", "we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions ."], "relation": "used for", "id": "2022.acl-long.854", "year": 2022, "rel_sent": "CLIP also forms fine - grained semantic representations of sentences , and obtains Spearman 's rho = .73 on the SemEval-2017 Semantic Textual Similarity Benchmark with no fine - tuning , compared to no greater than rho = .45 in any layer of GPT-2 .", "forward": true, "src_ids": "2022.acl-long.854_5140"}
+{"input": "visual semantic pretraining is used for Task| context: we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions .", "entity": "visual semantic pretraining", "output": "ordering visual representations", "neg_sample": ["visual semantic pretraining is used for Task", "we examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized english language representations formed by gpt-2 and clip , a zero - shot multimodal image classifier which adapts the gpt-2 architecture to encode image captions ."], "relation": "used for", "id": "2022.acl-long.854", "year": 2022, "rel_sent": "Our results indicate that high anisotropy is not an inevitable consequence of contextualization , and that visual semantic pretraining is beneficial not only for ordering visual representations , but also for encoding useful semantic representations of language , both on the word level and the sentence level .", "forward": true, "src_ids": "2022.acl-long.854_5141"}
+{"input": "full puzzle solutions is done by using Method| context: we present the berkeley crossword solver , a state - of - the - art approach for automatically solving crossword puzzles .", "entity": "full puzzle solutions", "output": "loopy belief propagation", "neg_sample": ["full puzzle solutions is done by using Method", "we present the berkeley crossword solver , a state - of - the - art approach for automatically solving crossword puzzles ."], "relation": "used for", "id": "2022.acl-long.857", "year": 2022, "rel_sent": "Our system works by generating answer candidates for each crossword clue using neural question answering models and then combines loopy belief propagation with local search to find full puzzle solutions .", "forward": false, "src_ids": "2022.acl-long.857_5142"}
+{"input": "local search is used for OtherScientificTerm| context: we present the berkeley crossword solver , a state - of - the - art approach for automatically solving crossword puzzles .", "entity": "local search", "output": "full puzzle solutions", "neg_sample": ["local search is used for OtherScientificTerm", "we present the berkeley crossword solver , a state - of - the - art approach for automatically solving crossword puzzles ."], "relation": "used for", "id": "2022.acl-long.857", "year": 2022, "rel_sent": "Our system works by generating answer candidates for each crossword clue using neural question answering models and then combines loopy belief propagation with local search to find full puzzle solutions .", "forward": true, "src_ids": "2022.acl-long.857_5143"}
+{"input": "loopy belief propagation is used for OtherScientificTerm| context: we present the berkeley crossword solver , a state - of - the - art approach for automatically solving crossword puzzles .", "entity": "loopy belief propagation", "output": "full puzzle solutions", "neg_sample": ["loopy belief propagation is used for OtherScientificTerm", "we present the berkeley crossword solver , a state - of - the - art approach for automatically solving crossword puzzles ."], "relation": "used for", "id": "2022.acl-long.857", "year": 2022, "rel_sent": "Our system works by generating answer candidates for each crossword clue using neural question answering models and then combines loopy belief propagation with local search to find full puzzle solutions .", "forward": true, "src_ids": "2022.acl-long.857_5144"}
+{"input": "machine - generated toxicity is done by using Method| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate . such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language .", "entity": "machine - generated toxicity", "output": "toxigen", "neg_sample": ["machine - generated toxicity is done by using Method", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language ."], "relation": "used for", "id": "2022.acl-long.888", "year": 2022, "rel_sent": "We also demonstrate that ToxiGen can be used to fight machine - generated toxicity as finetuning improves the classifier significantly on our evaluation subset .", "forward": false, "src_ids": "2022.acl-long.888_5145"}
+{"input": "toxigen is used for OtherScientificTerm| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate . such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language .", "entity": "toxigen", "output": "machine - generated toxicity", "neg_sample": ["toxigen is used for OtherScientificTerm", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language ."], "relation": "used for", "id": "2022.acl-long.888", "year": 2022, "rel_sent": "We also demonstrate that ToxiGen can be used to fight machine - generated toxicity as finetuning improves the classifier significantly on our evaluation subset .", "forward": true, "src_ids": "2022.acl-long.888_5146"}
+{"input": "classifier is done by using Method| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate . such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language .", "entity": "classifier", "output": "finetuning", "neg_sample": ["classifier is done by using Method", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language ."], "relation": "used for", "id": "2022.acl-long.888", "year": 2022, "rel_sent": "We also demonstrate that ToxiGen can be used to fight machine - generated toxicity as finetuning improves the classifier significantly on our evaluation subset .", "forward": false, "src_ids": "2022.acl-long.888_5147"}
+{"input": "finetuning is used for Method| context: toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate . such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language .", "entity": "finetuning", "output": "classifier", "neg_sample": ["finetuning is used for Method", "toxic language detection systems often falsely flag text that contains minority group mentions as toxic , as those groups are often the targets of online hate .", "such over - reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language ."], "relation": "used for", "id": "2022.acl-long.888", "year": 2022, "rel_sent": "We also demonstrate that ToxiGen can be used to fight machine - generated toxicity as finetuning improves the classifier significantly on our evaluation subset .", "forward": true, "src_ids": "2022.acl-long.888_5148"}
+{"input": "summarization system is done by using OtherScientificTerm| context: state - of - the - art abstractive summarization systems often generate hallucinations ; i.e. , content that is not directly inferable from the source text . despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations . including these factual hallucinations in a summary can be beneficial because they provide useful background information .", "entity": "summarization system", "output": "reward signal", "neg_sample": ["summarization system is done by using OtherScientificTerm", "state - of - the - art abstractive summarization systems often generate hallucinations ; i.e.", ", content that is not directly inferable from the source text .", "despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations .", "including these factual hallucinations in a summary can be beneficial because they provide useful background information ."], "relation": "used for", "id": "2022.acl-long.893", "year": 2022, "rel_sent": "Furthermore , we use our method as a reward signal to train a summarization system using an off - line reinforcement learning ( RL ) algorithm that can significantly improve the factuality of generated summaries while maintaining the level of abstractiveness .", "forward": false, "src_ids": "2022.acl-long.893_5149"}
+{"input": "reward signal is used for Method| context: despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations . including these factual hallucinations in a summary can be beneficial because they provide useful background information .", "entity": "reward signal", "output": "summarization system", "neg_sample": ["reward signal is used for Method", "despite being assumed to be incorrect , we find that much hallucinated content is actually consistent with world knowledge , which we call factual hallucinations .", "including these factual hallucinations in a summary can be beneficial because they provide useful background information ."], "relation": "used for", "id": "2022.acl-long.893", "year": 2022, "rel_sent": "Furthermore , we use our method as a reward signal to train a summarization system using an off - line reinforcement learning ( RL ) algorithm that can significantly improve the factuality of generated summaries while maintaining the level of abstractiveness .", "forward": true, "src_ids": "2022.acl-long.893_5150"}
+{"input": "image retrieval from contextual descriptions is used for Task| context: the ability to integrate context , including perceptual and temporal cues , plays a pivotal role in grounding the meaning of a linguistic utterance . because of this , descriptions tend to be complex in terms of syntax and discourse and require drawing pragmatic inferences . images are sourced from both static pictures and video frames .", "entity": "image retrieval from contextual descriptions", "output": "grounded language understanding", "neg_sample": ["image retrieval from contextual descriptions is used for Task", "the ability to integrate context , including perceptual and temporal cues , plays a pivotal role in grounding the meaning of a linguistic utterance .", "because of this , descriptions tend to be complex in terms of syntax and discourse and require drawing pragmatic inferences .", "images are sourced from both static pictures and video frames ."], "relation": "used for", "id": "2022.acl-long.901", "year": 2022, "rel_sent": "Our hope is that ImageCoDE will foster progress in grounded language understanding by encouraging models to focus on fine - grained visual differences .", "forward": true, "src_ids": "2022.acl-long.901_5151"}
+{"input": "grounded language understanding is done by using Material| context: the ability to integrate context , including perceptual and temporal cues , plays a pivotal role in grounding the meaning of a linguistic utterance . because of this , descriptions tend to be complex in terms of syntax and discourse and require drawing pragmatic inferences . images are sourced from both static pictures and video frames .", "entity": "grounded language understanding", "output": "image retrieval from contextual descriptions", "neg_sample": ["grounded language understanding is done by using Material", "the ability to integrate context , including perceptual and temporal cues , plays a pivotal role in grounding the meaning of a linguistic utterance .", "because of this , descriptions tend to be complex in terms of syntax and discourse and require drawing pragmatic inferences .", "images are sourced from both static pictures and video frames ."], "relation": "used for", "id": "2022.acl-long.901", "year": 2022, "rel_sent": "Our hope is that ImageCoDE will foster progress in grounded language understanding by encouraging models to focus on fine - grained visual differences .", "forward": false, "src_ids": "2022.acl-long.901_5152"}
+{"input": "languagemodels is done by using OtherScientificTerm| context: state - of - the - art nlp systems represent inputs with word embeddings , but these are brittle when faced with out - of - vocabulary ( oov ) words .", "entity": "languagemodels", "output": "love", "neg_sample": ["languagemodels is done by using OtherScientificTerm", "state - of - the - art nlp systems represent inputs with word embeddings , but these are brittle when faced with out - of - vocabulary ( oov ) words ."], "relation": "used for", "id": "2022.acl-long.912", "year": 2022, "rel_sent": "Imputing Out - of - Vocabulary Embeddings with LOVE Makes LanguageModels Robust with Little Cost.", "forward": false, "src_ids": "2022.acl-long.912_5153"}
+{"input": "love is used for Method| context: state - of - the - art nlp systems represent inputs with word embeddings , but these are brittle when faced with out - of - vocabulary ( oov ) words .", "entity": "love", "output": "languagemodels", "neg_sample": ["love is used for Method", "state - of - the - art nlp systems represent inputs with word embeddings , but these are brittle when faced with out - of - vocabulary ( oov ) words ."], "relation": "used for", "id": "2022.acl-long.912", "year": 2022, "rel_sent": "Imputing Out - of - Vocabulary Embeddings with LOVE Makes LanguageModels Robust with Little Cost.", "forward": true, "src_ids": "2022.acl-long.912_5154"}
+{"input": "downstream disparities is done by using Method| context: a few large , homogenous , pre - trained models undergird many machine learning systems --- and often , these models contain harmful stereotypes learned from the internet . we investigate the bias transfer hypothesis : the theory that social biases ( such as stereotypes ) internalized by large language models during pre - training transfer into harmful task - specific behavior after fine - tuning .", "entity": "downstream disparities", "output": "regression analysis", "neg_sample": ["downstream disparities is done by using Method", "a few large , homogenous , pre - trained models undergird many machine learning systems --- and often , these models contain harmful stereotypes learned from the internet .", "we investigate the bias transfer hypothesis : the theory that social biases ( such as stereotypes ) internalized by large language models during pre - training transfer into harmful task - specific behavior after fine - tuning ."], "relation": "used for", "id": "2022.acl-long.919", "year": 2022, "rel_sent": "Regression analysis suggests that downstream disparities are better explained by biases in the fine - tuning dataset .", "forward": false, "src_ids": "2022.acl-long.919_5155"}
+{"input": "regression analysis is used for OtherScientificTerm| context: a few large , homogenous , pre - trained models undergird many machine learning systems --- and often , these models contain harmful stereotypes learned from the internet . we investigate the bias transfer hypothesis : the theory that social biases ( such as stereotypes ) internalized by large language models during pre - training transfer into harmful task - specific behavior after fine - tuning .", "entity": "regression analysis", "output": "downstream disparities", "neg_sample": ["regression analysis is used for OtherScientificTerm", "a few large , homogenous , pre - trained models undergird many machine learning systems --- and often , these models contain harmful stereotypes learned from the internet .", "we investigate the bias transfer hypothesis : the theory that social biases ( such as stereotypes ) internalized by large language models during pre - training transfer into harmful task - specific behavior after fine - tuning ."], "relation": "used for", "id": "2022.acl-long.919", "year": 2022, "rel_sent": "Regression analysis suggests that downstream disparities are better explained by biases in the fine - tuning dataset .", "forward": true, "src_ids": "2022.acl-long.919_5156"}
+{"input": "non - literal translation is done by using Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "non - literal translation", "output": "transformer", "neg_sample": ["non - literal translation is done by using Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.929", "year": 2022, "rel_sent": "When Transformer emits a non - literal translation - i.e.", "forward": false, "src_ids": "2022.acl-long.929_5157"}
+{"input": "transformer is used for Method| context: unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) . nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations .", "entity": "transformer", "output": "non - literal translation", "neg_sample": ["transformer is used for Method", "unlike literal expressions , idioms ' meanings do not directly follow from their parts , posing a challenge for neural machine translation ( nmt ) .", "nmt models are often unable to translate idioms accurately and over - generate compositional , literal translations ."], "relation": "used for", "id": "2022.acl-long.929", "year": 2022, "rel_sent": "When Transformer emits a non - literal translation - i.e.", "forward": true, "src_ids": "2022.acl-long.929_5158"}
+{"input": "knowledge transfer is done by using Method| context: continual learning is essential for real - world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks . existing work on continual sequence generation either always reuses existing parameters to learn new tasks , which is vulnerable to catastrophic forgetting on dissimilar tasks , or blindly adds new parameters for every new task , which could prevent knowledge sharing between similar tasks .", "entity": "knowledge transfer", "output": "pseudo experience replay", "neg_sample": ["knowledge transfer is done by using Method", "continual learning is essential for real - world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks .", "existing work on continual sequence generation either always reuses existing parameters to learn new tasks , which is vulnerable to catastrophic forgetting on dissimilar tasks , or blindly adds new parameters for every new task , which could prevent knowledge sharing between similar tasks ."], "relation": "used for", "id": "2022.acl-long.938", "year": 2022, "rel_sent": "We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules .", "forward": false, "src_ids": "2022.acl-long.938_5159"}
+{"input": "pseudo experience replay is used for Task| context: continual learning is essential for real - world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks . existing work on continual sequence generation either always reuses existing parameters to learn new tasks , which is vulnerable to catastrophic forgetting on dissimilar tasks , or blindly adds new parameters for every new task , which could prevent knowledge sharing between similar tasks .", "entity": "pseudo experience replay", "output": "knowledge transfer", "neg_sample": ["pseudo experience replay is used for Task", "continual learning is essential for real - world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks .", "existing work on continual sequence generation either always reuses existing parameters to learn new tasks , which is vulnerable to catastrophic forgetting on dissimilar tasks , or blindly adds new parameters for every new task , which could prevent knowledge sharing between similar tasks ."], "relation": "used for", "id": "2022.acl-long.938", "year": 2022, "rel_sent": "We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules .", "forward": true, "src_ids": "2022.acl-long.938_5160"}
+{"input": "positive reframing is done by using OtherScientificTerm| context: with a sentiment reversal comes also a reversal in meaning . our insistence on meaning preservation makes positive reframing a challenging and semantically rich task .", "entity": "positive reframing", "output": "structured annotations", "neg_sample": ["positive reframing is done by using OtherScientificTerm", "with a sentiment reversal comes also a reversal in meaning .", "our insistence on meaning preservation makes positive reframing a challenging and semantically rich task ."], "relation": "used for", "id": "2022.acl-long.948", "year": 2022, "rel_sent": "To facilitate rapid progress , we introduce a large - scale benchmark , Positive Psychology Frames , with 8,349 sentence pairs and 12,755 structured annotations to explain positive reframing in terms of six theoretically - motivated reframing strategies .", "forward": false, "src_ids": "2022.acl-long.948_5161"}
+{"input": "structured annotations is used for Task| context: with a sentiment reversal comes also a reversal in meaning .", "entity": "structured annotations", "output": "positive reframing", "neg_sample": ["structured annotations is used for Task", "with a sentiment reversal comes also a reversal in meaning ."], "relation": "used for", "id": "2022.acl-long.948", "year": 2022, "rel_sent": "To facilitate rapid progress , we introduce a large - scale benchmark , Positive Psychology Frames , with 8,349 sentence pairs and 12,755 structured annotations to explain positive reframing in terms of six theoretically - motivated reframing strategies .", "forward": true, "src_ids": "2022.acl-long.948_5162"}
+{"input": "moral ambiguities is done by using Material| context: conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system . moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously .", "entity": "moral ambiguities", "output": "moral integrity corpus", "neg_sample": ["moral ambiguities is done by using Material", "conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system .", "moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously ."], "relation": "used for", "id": "2022.acl-long.958", "year": 2022, "rel_sent": "In this work , we introduce a new resource , not to authoritatively resolve moral ambiguities , but instead to facilitate systematic understanding of the intuitions , values and moral judgments reflected in the utterances of dialogue systems .", "forward": false, "src_ids": "2022.acl-long.958_5163"}
+{"input": "moral integrity corpus is used for OtherScientificTerm| context: conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system . moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously .", "entity": "moral integrity corpus", "output": "moral ambiguities", "neg_sample": ["moral integrity corpus is used for OtherScientificTerm", "conversational agents have come increasingly closer to human competence in open - domain dialogue settings ; however , such models can reflect insensitive , hurtful , or entirely incoherent viewpoints that erode a user 's trust in the moral integrity of the system .", "moral deviations are difficult to mitigate because moral judgments are not universal , and there may be multiple competing judgments that apply to a situation simultaneously ."], "relation": "used for", "id": "2022.acl-long.958", "year": 2022, "rel_sent": "In this work , we introduce a new resource , not to authoritatively resolve moral ambiguities , but instead to facilitate systematic understanding of the intuitions , values and moral judgments reflected in the utterances of dialogue systems .", "forward": true, "src_ids": "2022.acl-long.958_5164"}
+{"input": "predicting code - switching points is done by using Task| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "predicting code - switching points", "output": "speaker - driven task", "neg_sample": ["predicting code - switching points is done by using Task", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.981", "year": 2022, "rel_sent": "For the speaker - driven task of predicting code - switching points in English -- Spanish bilingual dialogues , we show that adding sociolinguistically - grounded speaker features as prepended prompts significantly improves accuracy .", "forward": false, "src_ids": "2022.acl-long.981_5165"}
+{"input": "speaker - driven task is used for Task| context: natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task .", "entity": "speaker - driven task", "output": "predicting code - switching points", "neg_sample": ["speaker - driven task is used for Task", "natural language processing ( nlp ) models trained on people - generated data can be unreliable because , without any constraints , they can learn from spurious correlations that are not relevant to the task ."], "relation": "used for", "id": "2022.acl-long.981", "year": 2022, "rel_sent": "For the speaker - driven task of predicting code - switching points in English -- Spanish bilingual dialogues , we show that adding sociolinguistically - grounded speaker features as prepended prompts significantly improves accuracy .", "forward": true, "src_ids": "2022.acl-long.981_5166"}
+{"input": "tweets is used for OtherScientificTerm| context: it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "tweets", "output": "bragging", "neg_sample": ["tweets is used for OtherScientificTerm", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.1008", "year": 2022, "rel_sent": "To facilitate this , we introduce a new publicly available data set of tweets annotated for bragging and their types .", "forward": true, "src_ids": "2022.acl-long.1008_5167"}
+{"input": "bragging is done by using Material| context: bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself . it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "bragging", "output": "tweets", "neg_sample": ["bragging is done by using Material", "bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself .", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.1008", "year": 2022, "rel_sent": "To facilitate this , we introduce a new publicly available data set of tweets annotated for bragging and their types .", "forward": false, "src_ids": "2022.acl-long.1008_5168"}
+{"input": "binary bragging classification is done by using OtherScientificTerm| context: bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself . it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "binary bragging classification", "output": "linguistic information", "neg_sample": ["binary bragging classification is done by using OtherScientificTerm", "bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself .", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.1008", "year": 2022, "rel_sent": "We empirically evaluate different transformer - based models injected with linguistic information in ( a ) binary bragging classification , i.e. , if tweets contain bragging statements or not ; and ( b ) multi - class bragging type prediction including not bragging .", "forward": false, "src_ids": "2022.acl-long.1008_5169"}
+{"input": "linguistic information is used for Task| context: bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself . it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly .", "entity": "linguistic information", "output": "binary bragging classification", "neg_sample": ["linguistic information is used for Task", "bragging is a speech act employed with the goal of constructing a favorable self - image through positive statements about oneself .", "it is widespread in daily communication and especially popular in social media , where users aim to build a positive image of their persona directly or indirectly ."], "relation": "used for", "id": "2022.acl-long.1008", "year": 2022, "rel_sent": "We empirically evaluate different transformer - based models injected with linguistic information in ( a ) binary bragging classification , i.e. , if tweets contain bragging statements or not ; and ( b ) multi - class bragging type prediction including not bragging .", "forward": true, "src_ids": "2022.acl-long.1008_5170"}
+{"input": "cross - target stance detection is done by using OtherScientificTerm| context: research in stance detection has so far focused on models which leverage purely textual input .", "entity": "cross - target stance detection", "output": "multiple input signals", "neg_sample": ["cross - target stance detection is done by using OtherScientificTerm", "research in stance detection has so far focused on models which leverage purely textual input ."], "relation": "used for", "id": "2022.acl-long.1026", "year": 2022, "rel_sent": "We show experimentally and through detailed result analysis that our stance detection system benefits from financial information , and achieves state - of - the - art results on the wt -- wt dataset : this demonstrates that the combination of multiple input signals is effective for cross - target stance detection , and opens interesting research directions for future work .", "forward": false, "src_ids": "2022.acl-long.1026_5171"}
+{"input": "multiple input signals is used for Task| context: research in stance detection has so far focused on models which leverage purely textual input .", "entity": "multiple input signals", "output": "cross - target stance detection", "neg_sample": ["multiple input signals is used for Task", "research in stance detection has so far focused on models which leverage purely textual input ."], "relation": "used for", "id": "2022.acl-long.1026", "year": 2022, "rel_sent": "We show experimentally and through detailed result analysis that our stance detection system benefits from financial information , and achieves state - of - the - art results on the wt -- wt dataset : this demonstrates that the combination of multiple input signals is effective for cross - target stance detection , and opens interesting research directions for future work .", "forward": true, "src_ids": "2022.acl-long.1026_5172"}
+{"input": "cross - lingual phrase retriever is used for Method| context: current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level .", "entity": "cross - lingual phrase retriever", "output": "phrase representations", "neg_sample": ["cross - lingual phrase retriever is used for Method", "current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level ."], "relation": "used for", "id": "2022.acl-long.1045", "year": 2022, "rel_sent": "In this paper , we propose XPR , a cross - lingual phrase retriever that extracts phrase representations from unlabeled example sentences .", "forward": true, "src_ids": "2022.acl-long.1045_5173"}
+{"input": "phrase representations is done by using Method| context: current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level . however , how to learn phrase representations for cross - lingual phrase retrieval is still an open problem .", "entity": "phrase representations", "output": "cross - lingual phrase retriever", "neg_sample": ["phrase representations is done by using Method", "current methods typically achieve cross - lingual retrieval by learning language - agnostic text representations in word or sentence level .", "however , how to learn phrase representations for cross - lingual phrase retrieval is still an open problem ."], "relation": "used for", "id": "2022.acl-long.1045", "year": 2022, "rel_sent": "In this paper , we propose XPR , a cross - lingual phrase retriever that extracts phrase representations from unlabeled example sentences .", "forward": false, "src_ids": "2022.acl-long.1045_5174"}
+{"input": "pseudo - response selection is done by using OtherScientificTerm| context: data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) . such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs .", "entity": "pseudo - response selection", "output": "fine - tuned bleurt", "neg_sample": ["pseudo - response selection is done by using OtherScientificTerm", "data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) .", "such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs ."], "relation": "used for", "id": "2022.acl-long.1046", "year": 2022, "rel_sent": "To further improve the model 's performance , we propose an approach based on self - training using fine - tuned BLEURT for pseudo - response selection .", "forward": false, "src_ids": "2022.acl-long.1046_5175"}
+{"input": "fine - tuned bleurt is used for Task| context: data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) . such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs .", "entity": "fine - tuned bleurt", "output": "pseudo - response selection", "neg_sample": ["fine - tuned bleurt is used for Task", "data - to - text generation focuses on generating fluent natural language responses from structured meaning representations ( mrs ) .", "such representations are compositional and it is costly to collect responses for all possible combinations of atomic meaning schemata , thereby necessitating few - shot generalization to novel mrs ."], "relation": "used for", "id": "2022.acl-long.1046", "year": 2022, "rel_sent": "To further improve the model 's performance , we propose an approach based on self - training using fine - tuned BLEURT for pseudo - response selection .", "forward": true, "src_ids": "2022.acl-long.1046_5176"}
+{"input": "edu - level pre - trained edu representation is done by using Method| context: pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing . current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e. element discourse unit ( edu ) .", "entity": "edu - level pre - trained edu representation", "output": "light bi - gram edu modification", "neg_sample": ["edu - level pre - trained edu representation is done by using Method", "pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing .", "current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e.", "element discourse unit ( edu ) ."], "relation": "used for", "id": "2022.acl-long.1057", "year": 2022, "rel_sent": "We take a state - of - the - art transition - based neural parser as baseline , and adopt it with a light bi - gram EDU modification to effectively explore the EDU - level pre - trained EDU representation .", "forward": false, "src_ids": "2022.acl-long.1057_5177"}
+{"input": "light bi - gram edu modification is used for Method| context: pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing . current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e. element discourse unit ( edu ) .", "entity": "light bi - gram edu modification", "output": "edu - level pre - trained edu representation", "neg_sample": ["light bi - gram edu modification is used for Method", "pre - trained language models ( plms ) have shown great potentials in natural language processing ( nlp ) including rhetorical structure theory ( rst ) discourse parsing .", "current plms are obtained by sentence - level pre - training , which is different from the basic processing unit , i.e.", "element discourse unit ( edu ) ."], "relation": "used for", "id": "2022.acl-long.1057", "year": 2022, "rel_sent": "We take a state - of - the - art transition - based neural parser as baseline , and adopt it with a light bi - gram EDU modification to effectively explore the EDU - level pre - trained EDU representation .", "forward": true, "src_ids": "2022.acl-long.1057_5178"}
+{"input": "multimodal event sequencing is done by using OtherScientificTerm| context: the ability to sequence unordered events is evidence of comprehension and reasoning about real world tasks / procedures . it is essential for applications such as task planning and multi - source instruction summarization . it often requires thorough understanding of temporal common sense and multimodal information , since these procedures are often conveyed by a combination of texts and images . while humans are capable of reasoning about and sequencing unordered procedural instructions , the extent to which the current machine learning methods possess such capability is still an open question .", "entity": "multimodal event sequencing", "output": "machines", "neg_sample": ["multimodal event sequencing is done by using OtherScientificTerm", "the ability to sequence unordered events is evidence of comprehension and reasoning about real world tasks / procedures .", "it is essential for applications such as task planning and multi - source instruction summarization .", "it often requires thorough understanding of temporal common sense and multimodal information , since these procedures are often conveyed by a combination of texts and images .", "while humans are capable of reasoning about and sequencing unordered procedural instructions , the extent to which the current machine learning methods possess such capability is still an open question ."], "relation": "used for", "id": "2022.acl-long.1120", "year": 2022, "rel_sent": "To improve machines ' performance on multimodal event sequencing , we propose sequence - aware pretraining techniques exploiting the sequential alignment properties of both texts and images , resulting in > 5 % improvements on perfect match ratio .", "forward": false, "src_ids": "2022.acl-long.1120_5179"}
+{"input": "machines is used for Task| context: the ability to sequence unordered events is evidence of comprehension and reasoning about real world tasks / procedures . it is essential for applications such as task planning and multi - source instruction summarization . it often requires thorough understanding of temporal common sense and multimodal information , since these procedures are often conveyed by a combination of texts and images . while humans are capable of reasoning about and sequencing unordered procedural instructions , the extent to which the current machine learning methods possess such capability is still an open question .", "entity": "machines", "output": "multimodal event sequencing", "neg_sample": ["machines is used for Task", "the ability to sequence unordered events is evidence of comprehension and reasoning about real world tasks / procedures .", "it is essential for applications such as task planning and multi - source instruction summarization .", "it often requires thorough understanding of temporal common sense and multimodal information , since these procedures are often conveyed by a combination of texts and images .", "while humans are capable of reasoning about and sequencing unordered procedural instructions , the extent to which the current machine learning methods possess such capability is still an open question ."], "relation": "used for", "id": "2022.acl-long.1120", "year": 2022, "rel_sent": "To improve machines ' performance on multimodal event sequencing , we propose sequence - aware pretraining techniques exploiting the sequential alignment properties of both texts and images , resulting in > 5 % improvements on perfect match ratio .", "forward": true, "src_ids": "2022.acl-long.1120_5180"}
+{"input": "training mechanism is used for Task| context: finding the causes of skewed performance is crucial for the robustness of an ed model , but to date there has been little exploration of this problem .", "entity": "training mechanism", "output": "event detection", "neg_sample": ["training mechanism is used for Task", "finding the causes of skewed performance is crucial for the robustness of an ed model , but to date there has been little exploration of this problem ."], "relation": "used for", "id": "2022.acl-long.1124", "year": 2022, "rel_sent": "On this foundation , we develop a new training mechanism for ED , which can distinguish between trigger - dependent and context - dependent types and achieve promising performance on two benchmarks .", "forward": true, "src_ids": "2022.acl-long.1124_5181"}
+{"input": "event detection is done by using Method| context: event detection ( ed ) is a critical subtask of event extraction that seeks to identify event triggers of certain types in texts . despite significant advances in ed , existing methods typically follow a ' one model fits all types '' approach , which sees no differences between event types and often results in a quite skewed performance . finding the causes of skewed performance is crucial for the robustness of an ed model , but to date there has been little exploration of this problem .", "entity": "event detection", "output": "training mechanism", "neg_sample": ["event detection is done by using Method", "event detection ( ed ) is a critical subtask of event extraction that seeks to identify event triggers of certain types in texts .", "despite significant advances in ed , existing methods typically follow a ' one model fits all types '' approach , which sees no differences between event types and often results in a quite skewed performance .", "finding the causes of skewed performance is crucial for the robustness of an ed model , but to date there has been little exploration of this problem ."], "relation": "used for", "id": "2022.acl-long.1124", "year": 2022, "rel_sent": "On this foundation , we develop a new training mechanism for ED , which can distinguish between trigger - dependent and context - dependent types and achieve promising performance on two benchmarks .", "forward": false, "src_ids": "2022.acl-long.1124_5182"}
+{"input": "ood detection is done by using OtherScientificTerm| context: the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems . then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples .", "entity": "ood detection", "output": "discriminative semantic features", "neg_sample": ["ood detection is done by using OtherScientificTerm", "the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems .", "then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples ."], "relation": "used for", "id": "2022.acl-long.1247", "year": 2022, "rel_sent": "Our approach utilizes k - nearest neighbors ( KNN ) of IND intents to learn discriminative semantic features that are more conducive to OOD detection .", "forward": false, "src_ids": "2022.acl-long.1247_5183"}
+{"input": "discriminative semantic features is used for Task| context: the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems . previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features . then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples .", "entity": "discriminative semantic features", "output": "ood detection", "neg_sample": ["discriminative semantic features is used for Task", "the out - of - domain ( ood ) intent classification is a basic and challenging task for dialogue systems .", "previous methods commonly restrict the region ( in feature space ) of in - domain ( ind ) intent features to be compact or simply - connected implicitly , which assumes no ood intents reside , to learn discriminative semantic features .", "then the distribution of the ind intent features is often assumed to obey a hypothetical distribution ( gaussian mostly ) and samples outside this distribution are regarded as ood samples ."], "relation": "used for", "id": "2022.acl-long.1247", "year": 2022, "rel_sent": "Our approach utilizes k - nearest neighbors ( KNN ) of IND intents to learn discriminative semantic features that are more conducive to OOD detection .", "forward": true, "src_ids": "2022.acl-long.1247_5184"}
+{"input": "generative template - based event extraction is done by using Method| context: we consider event extraction in a generative manner with template - based conditional generation . although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information .", "entity": "generative template - based event extraction", "output": "dynamic prefix - tuning", "neg_sample": ["generative template - based event extraction is done by using Method", "we consider event extraction in a generative manner with template - based conditional generation .", "although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information ."], "relation": "used for", "id": "2022.acl-long.1266", "year": 2022, "rel_sent": "Dynamic Prefix - Tuning for Generative Template - based Event Extraction.", "forward": false, "src_ids": "2022.acl-long.1266_5185"}
+{"input": "dynamic prefix - tuning is used for Task| context: we consider event extraction in a generative manner with template - based conditional generation . although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information .", "entity": "dynamic prefix - tuning", "output": "generative template - based event extraction", "neg_sample": ["dynamic prefix - tuning is used for Task", "we consider event extraction in a generative manner with template - based conditional generation .", "although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information ."], "relation": "used for", "id": "2022.acl-long.1266", "year": 2022, "rel_sent": "Dynamic Prefix - Tuning for Generative Template - based Event Extraction.", "forward": true, "src_ids": "2022.acl-long.1266_5186"}
+{"input": "context - specific prefix is done by using OtherScientificTerm| context: we consider event extraction in a generative manner with template - based conditional generation . although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information .", "entity": "context - specific prefix", "output": "context information", "neg_sample": ["context - specific prefix is done by using OtherScientificTerm", "we consider event extraction in a generative manner with template - based conditional generation .", "although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information ."], "relation": "used for", "id": "2022.acl-long.1266", "year": 2022, "rel_sent": "In this paper , we propose a generative template - based event extraction method with dynamic prefix ( GTEE - DynPref ) by integrating context information with type - specific prefixes to learn a context - specific prefix for each context .", "forward": false, "src_ids": "2022.acl-long.1266_5187"}
+{"input": "context information is used for OtherScientificTerm| context: we consider event extraction in a generative manner with template - based conditional generation . although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information .", "entity": "context information", "output": "context - specific prefix", "neg_sample": ["context information is used for OtherScientificTerm", "we consider event extraction in a generative manner with template - based conditional generation .", "although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information ."], "relation": "used for", "id": "2022.acl-long.1266", "year": 2022, "rel_sent": "In this paper , we propose a generative template - based event extraction method with dynamic prefix ( GTEE - DynPref ) by integrating context information with type - specific prefixes to learn a context - specific prefix for each context .", "forward": true, "src_ids": "2022.acl-long.1266_5188"}
+{"input": "type - specific prefixes is used for OtherScientificTerm| context: we consider event extraction in a generative manner with template - based conditional generation . although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information .", "entity": "type - specific prefixes", "output": "context - specific prefix", "neg_sample": ["type - specific prefixes is used for OtherScientificTerm", "we consider event extraction in a generative manner with template - based conditional generation .", "although there is a rising trend of casting the task of event extraction as a sequence generation problem with prompts , these generation - based methods have two significant challenges , including using suboptimal prompts and static event type information ."], "relation": "used for", "id": "2022.acl-long.1266", "year": 2022, "rel_sent": "In this paper , we propose a generative template - based event extraction method with dynamic prefix ( GTEE - DynPref ) by integrating context information with type - specific prefixes to learn a context - specific prefix for each context .", "forward": true, "src_ids": "2022.acl-long.1266_5189"}
+{"input": "human reading behavior is done by using OtherScientificTerm| context: there is a growing interest in the combined use of nlp and machine learning methods to predict gaze patterns during naturalistic reading . while promising results have been obtained through the use of transformer - based language models , little work has been undertaken to relate the performance of such models to general text characteristics .", "entity": "human reading behavior", "output": "features", "neg_sample": ["human reading behavior is done by using OtherScientificTerm", "there is a growing interest in the combined use of nlp and machine learning methods to predict gaze patterns during naturalistic reading .", "while promising results have been obtained through the use of transformer - based language models , little work has been undertaken to relate the performance of such models to general text characteristics ."], "relation": "used for", "id": "2022.acl-long.1282", "year": 2022, "rel_sent": "In all experiments , we test effects of a broad spectrum of features for predicting human reading behavior that fall into five categories ( syntactic complexity , lexical richness , register - based multiword combinations , readability and psycholinguistic word properties ) .", "forward": false, "src_ids": "2022.acl-long.1282_5190"}
+{"input": "features is used for OtherScientificTerm| context: there is a growing interest in the combined use of nlp and machine learning methods to predict gaze patterns during naturalistic reading . while promising results have been obtained through the use of transformer - based language models , little work has been undertaken to relate the performance of such models to general text characteristics .", "entity": "features", "output": "human reading behavior", "neg_sample": ["features is used for OtherScientificTerm", "there is a growing interest in the combined use of nlp and machine learning methods to predict gaze patterns during naturalistic reading .", "while promising results have been obtained through the use of transformer - based language models , little work has been undertaken to relate the performance of such models to general text characteristics ."], "relation": "used for", "id": "2022.acl-long.1282", "year": 2022, "rel_sent": "In all experiments , we test effects of a broad spectrum of features for predicting human reading behavior that fall into five categories ( syntactic complexity , lexical richness , register - based multiword combinations , readability and psycholinguistic word properties ) .", "forward": true, "src_ids": "2022.acl-long.1282_5191"}
+{"input": "two - tier bert architecture is used for OtherScientificTerm| context: pre - trained language models such as bert have been successful at tackling many natural language processing tasks . however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g. , byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages . despite the success of bert , most of its evaluations have been conducted on high - resource languages , obscuring its applicability on low - resource languages .", "entity": "two - tier bert architecture", "output": "morphological compositionality", "neg_sample": ["two - tier bert architecture is used for OtherScientificTerm", "pre - trained language models such as bert have been successful at tackling many natural language processing tasks .", "however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g.", ", byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages .", "despite the success of bert , most of its evaluations have been conducted on high - resource languages , obscuring its applicability on low - resource languages ."], "relation": "used for", "id": "2022.acl-long.1296", "year": 2022, "rel_sent": "We address these challenges by proposing a simple yet effective two - tier BERT architecture that leverages a morphological analyzer and explicitly represents morphological compositionality .", "forward": true, "src_ids": "2022.acl-long.1296_5192"}
+{"input": "morphological compositionality is done by using Method| context: pre - trained language models such as bert have been successful at tackling many natural language processing tasks . however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g. , byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages . even given a morphological analyzer , naive sequencing of morphemes into a standard bert architecture is inefficient at capturing morphological compositionality and expressing word - relative syntactic regularities . despite the success of bert , most of its evaluations have been conducted on high - resource languages , obscuring its applicability on low - resource languages .", "entity": "morphological compositionality", "output": "two - tier bert architecture", "neg_sample": ["morphological compositionality is done by using Method", "pre - trained language models such as bert have been successful at tackling many natural language processing tasks .", "however , the unsupervised sub - word tokenization methods commonly used in these models ( e.g.", ", byte - pair encoding - bpe ) are sub - optimal at handling morphologically rich languages .", "even given a morphological analyzer , naive sequencing of morphemes into a standard bert architecture is inefficient at capturing morphological compositionality and expressing word - relative syntactic regularities .", "despite the success of bert , most of its evaluations have been conducted on high - resource languages , obscuring its applicability on low - resource languages ."], "relation": "used for", "id": "2022.acl-long.1296", "year": 2022, "rel_sent": "We address these challenges by proposing a simple yet effective two - tier BERT architecture that leverages a morphological analyzer and explicitly represents morphological compositionality .", "forward": false, "src_ids": "2022.acl-long.1296_5193"}
+{"input": "continuous - space attention mechanism is used for OtherScientificTerm| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "continuous - space attention mechanism", "output": "long - term memory", "neg_sample": ["continuous - space attention mechanism is used for OtherScientificTerm", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.1311", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision .", "forward": true, "src_ids": "2022.acl-long.1311_5194"}
+{"input": "continuous - space attention mechanism is used for Method| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "continuous - space attention mechanism", "output": "infty - former", "neg_sample": ["continuous - space attention mechanism is used for Method", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.1311", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision .", "forward": true, "src_ids": "2022.acl-long.1311_5195"}
+{"input": "long - term memory is done by using Method| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "long - term memory", "output": "continuous - space attention mechanism", "neg_sample": ["long - term memory is done by using Method", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.1311", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision .", "forward": false, "src_ids": "2022.acl-long.1311_5196"}
+{"input": "infty - former is done by using Method| context: transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length . while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information .", "entity": "infty - former", "output": "continuous - space attention mechanism", "neg_sample": ["infty - former is done by using Method", "transformers are unable to model long - term memories effectively , since the amount of computation they need to perform grows with the context length .", "while variations of efficient transformers have been proposed , they all have a finite memory capacity and are forced to drop old information ."], "relation": "used for", "id": "2022.acl-long.1311", "year": 2022, "rel_sent": "By making use of a continuous - space attention mechanism to attend over the long - term memory , the infty - former 's attention complexity becomes independent of the context length , trading off memory length with precision .", "forward": false, "src_ids": "2022.acl-long.1311_5197"}
+{"input": "fully hyperbolic framework is used for Method| context: hyperbolic neural networks have shown great potential for modeling complex data . this hybrid method greatly limits the modeling ability of networks .", "entity": "fully hyperbolic framework", "output": "hyperbolic networks", "neg_sample": ["fully hyperbolic framework is used for Method", "hyperbolic neural networks have shown great potential for modeling complex data .", "this hybrid method greatly limits the modeling ability of networks ."], "relation": "used for", "id": "2022.acl-long.1354", "year": 2022, "rel_sent": "In this paper , we propose a fully hyperbolic framework to build hyperbolic networks based on the Lorentz model by adapting the Lorentz transformations ( including boost and rotation ) to formalize essential operations of neural networks .", "forward": true, "src_ids": "2022.acl-long.1354_5198"}
+{"input": "hyperbolic networks is done by using Method| context: hyperbolic neural networks have shown great potential for modeling complex data . however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model . this hybrid method greatly limits the modeling ability of networks .", "entity": "hyperbolic networks", "output": "fully hyperbolic framework", "neg_sample": ["hyperbolic networks is done by using Method", "hyperbolic neural networks have shown great potential for modeling complex data .", "however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model .", "this hybrid method greatly limits the modeling ability of networks ."], "relation": "used for", "id": "2022.acl-long.1354", "year": 2022, "rel_sent": "In this paper , we propose a fully hyperbolic framework to build hyperbolic networks based on the Lorentz model by adapting the Lorentz transformations ( including boost and rotation ) to formalize essential operations of neural networks .", "forward": false, "src_ids": "2022.acl-long.1354_5199"}
+{"input": "neural networks is done by using OtherScientificTerm| context: hyperbolic neural networks have shown great potential for modeling complex data . however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model . this hybrid method greatly limits the modeling ability of networks .", "entity": "neural networks", "output": "lorentz transformations", "neg_sample": ["neural networks is done by using OtherScientificTerm", "hyperbolic neural networks have shown great potential for modeling complex data .", "however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model .", "this hybrid method greatly limits the modeling ability of networks ."], "relation": "used for", "id": "2022.acl-long.1354", "year": 2022, "rel_sent": "In this paper , we propose a fully hyperbolic framework to build hyperbolic networks based on the Lorentz model by adapting the Lorentz transformations ( including boost and rotation ) to formalize essential operations of neural networks .", "forward": false, "src_ids": "2022.acl-long.1354_5200"}
+{"input": "lorentz transformations is used for Method| context: however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model . this hybrid method greatly limits the modeling ability of networks .", "entity": "lorentz transformations", "output": "neural networks", "neg_sample": ["lorentz transformations is used for Method", "however , existing hyperbolic networks are not completely hyperbolic , as they encode features in the hyperbolic space yet formalize most of their operations in the tangent space ( a euclidean subspace ) at the origin of the hyperbolic model .", "this hybrid method greatly limits the modeling ability of networks ."], "relation": "used for", "id": "2022.acl-long.1354", "year": 2022, "rel_sent": "In this paper , we propose a fully hyperbolic framework to build hyperbolic networks based on the Lorentz model by adapting the Lorentz transformations ( including boost and rotation ) to formalize essential operations of neural networks .", "forward": true, "src_ids": "2022.acl-long.1354_5201"}
+{"input": "emotion recognition is done by using Method| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "emotion recognition", "output": "multimodal dialogue - aware interaction framework", "neg_sample": ["emotion recognition is done by using Method", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.1358", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": false, "src_ids": "2022.acl-long.1358_5202"}
+{"input": "multimodal dialogue - aware interaction framework is used for Task| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "multimodal dialogue - aware interaction framework", "output": "emotion recognition", "neg_sample": ["multimodal dialogue - aware interaction framework is used for Task", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.1358", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": true, "src_ids": "2022.acl-long.1358_5203"}
+{"input": "mdi is used for Task| context: the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus . the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity .", "entity": "mdi", "output": "emotion recognition", "neg_sample": ["mdi is used for Task", "the emotional state of a speaker can be influenced by many different factors in dialogues , such as dialogue scene , dialogue topic , and interlocutor stimulus .", "the currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity ."], "relation": "used for", "id": "2022.acl-long.1358", "year": 2022, "rel_sent": "We also propose a general Multimodal Dialogue - aware Interaction framework , MDI , to model the dialogue context for emotion recognition , which achieves comparable performance to the state - of - the - art methods on the M^3ED .", "forward": true, "src_ids": "2022.acl-long.1358_5204"}
+{"input": "consultation note generation is done by using Metric| context: in recent years , machine learning models have rapidly become better at generating clinical consultation notes ; yet , there is little work on how to properly evaluate the generated consultation notes to understand the impact they may have on both the clinician using them and the patient 's clinical safety .", "entity": "consultation note generation", "output": "automatic metrics", "neg_sample": ["consultation note generation is done by using Metric", "in recent years , machine learning models have rapidly become better at generating clinical consultation notes ; yet , there is little work on how to properly evaluate the generated consultation notes to understand the impact they may have on both the clinician using them and the patient 's clinical safety ."], "relation": "used for", "id": "2022.acl-long.1365", "year": 2022, "rel_sent": "Human Evaluation and Correlation with Automatic Metrics in Consultation Note Generation.", "forward": false, "src_ids": "2022.acl-long.1365_5205"}
+{"input": "automatic metrics is used for Task| context: in recent years , machine learning models have rapidly become better at generating clinical consultation notes ; yet , there is little work on how to properly evaluate the generated consultation notes to understand the impact they may have on both the clinician using them and the patient 's clinical safety .", "entity": "automatic metrics", "output": "consultation note generation", "neg_sample": ["automatic metrics is used for Task", "in recent years , machine learning models have rapidly become better at generating clinical consultation notes ; yet , there is little work on how to properly evaluate the generated consultation notes to understand the impact they may have on both the clinician using them and the patient 's clinical safety ."], "relation": "used for", "id": "2022.acl-long.1365", "year": 2022, "rel_sent": "Human Evaluation and Correlation with Automatic Metrics in Consultation Note Generation.", "forward": true, "src_ids": "2022.acl-long.1365_5206"}
+{"input": "zero - shot label matching ability is done by using OtherScientificTerm| context: few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities .", "entity": "zero - shot label matching ability", "output": "triplet paraphrase", "neg_sample": ["zero - shot label matching ability is done by using OtherScientificTerm", "few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities ."], "relation": "used for", "id": "2022.acl-long.1369", "year": 2022, "rel_sent": "To fill in the gap between zero - shot and few - shot RE , we propose the triplet - paraphrase meta - training , which leverages triplet paraphrase to pre - train zero - shot label matching ability and uses meta - learning paradigm to learn few - shot instance summarizing ability .", "forward": false, "src_ids": "2022.acl-long.1369_5207"}
+{"input": "triplet paraphrase is used for OtherScientificTerm| context: few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities .", "entity": "triplet paraphrase", "output": "zero - shot label matching ability", "neg_sample": ["triplet paraphrase is used for OtherScientificTerm", "few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities ."], "relation": "used for", "id": "2022.acl-long.1369", "year": 2022, "rel_sent": "To fill in the gap between zero - shot and few - shot RE , we propose the triplet - paraphrase meta - training , which leverages triplet paraphrase to pre - train zero - shot label matching ability and uses meta - learning paradigm to learn few - shot instance summarizing ability .", "forward": true, "src_ids": "2022.acl-long.1369_5208"}
+{"input": "few - shot instance summarizing ability is done by using Method| context: few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities .", "entity": "few - shot instance summarizing ability", "output": "meta - learning paradigm", "neg_sample": ["few - shot instance summarizing ability is done by using Method", "few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities ."], "relation": "used for", "id": "2022.acl-long.1369", "year": 2022, "rel_sent": "To fill in the gap between zero - shot and few - shot RE , we propose the triplet - paraphrase meta - training , which leverages triplet paraphrase to pre - train zero - shot label matching ability and uses meta - learning paradigm to learn few - shot instance summarizing ability .", "forward": false, "src_ids": "2022.acl-long.1369_5209"}
+{"input": "meta - learning paradigm is used for OtherScientificTerm| context: few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities .", "entity": "meta - learning paradigm", "output": "few - shot instance summarizing ability", "neg_sample": ["meta - learning paradigm is used for OtherScientificTerm", "few - shot and zero - shot re are two representative low - shot re tasks , which seem to be with similar target but require totally different underlying abilities ."], "relation": "used for", "id": "2022.acl-long.1369", "year": 2022, "rel_sent": "To fill in the gap between zero - shot and few - shot RE , we propose the triplet - paraphrase meta - training , which leverages triplet paraphrase to pre - train zero - shot label matching ability and uses meta - learning paradigm to learn few - shot instance summarizing ability .", "forward": true, "src_ids": "2022.acl-long.1369_5210"}
+{"input": "decoding algorithm via third - order tensor isomorphism is done by using Generic| context: for a long time , most researchers have regarded ea as a pure graph representation learning task and focused on improving graph encoders while paying little attention to the decoding process .", "entity": "decoding algorithm via third - order tensor isomorphism", "output": "equations", "neg_sample": ["decoding algorithm via third - order tensor isomorphism is done by using Generic", "for a long time , most researchers have regarded ea as a pure graph representation learning task and focused on improving graph encoders while paying little attention to the decoding process ."], "relation": "used for", "id": "2022.acl-long.1402", "year": 2022, "rel_sent": "By combining these equations , DATTI could effectively utilize the adjacency and inner correlation isomorphisms of KGs to enhance the decoding process of EA .", "forward": false, "src_ids": "2022.acl-long.1402_5211"}
+{"input": "decoding process is done by using Method| context: for a long time , most researchers have regarded ea as a pure graph representation learning task and focused on improving graph encoders while paying little attention to the decoding process .", "entity": "decoding process", "output": "decoding algorithm via third - order tensor isomorphism", "neg_sample": ["decoding process is done by using Method", "for a long time , most researchers have regarded ea as a pure graph representation learning task and focused on improving graph encoders while paying little attention to the decoding process ."], "relation": "used for", "id": "2022.acl-long.1402", "year": 2022, "rel_sent": "By combining these equations , DATTI could effectively utilize the adjacency and inner correlation isomorphisms of KGs to enhance the decoding process of EA .", "forward": false, "src_ids": "2022.acl-long.1402_5212"}
+{"input": "equations is used for Method| context: for a long time , most researchers have regarded ea as a pure graph representation learning task and focused on improving graph encoders while paying little attention to the decoding process .", "entity": "equations", "output": "decoding algorithm via third - order tensor isomorphism", "neg_sample": ["equations is used for Method", "for a long time , most researchers have regarded ea as a pure graph representation learning task and focused on improving graph encoders while paying little attention to the decoding process ."], "relation": "used for", "id": "2022.acl-long.1402", "year": 2022, "rel_sent": "By combining these equations , DATTI could effectively utilize the adjacency and inner correlation isomorphisms of KGs to enhance the decoding process of EA .", "forward": true, "src_ids": "2022.acl-long.1402_5213"}
+{"input": "graph self - supervised training is used for Method| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "entity": "graph self - supervised training", "output": "pre - trained language models", "neg_sample": ["graph self - supervised training is used for Method", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "To this end , we investigate graph self - supervised training to improve the structure awareness of PLMs over AMR graphs .", "forward": true, "src_ids": "2022.acl-long.1426_5214"}
+{"input": "pre - trained language models is done by using Method| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "pre - trained language models", "output": "graph self - supervised training", "neg_sample": ["pre - trained language models is done by using Method", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "To this end , we investigate graph self - supervised training to improve the structure awareness of PLMs over AMR graphs .", "forward": false, "src_ids": "2022.acl-long.1426_5215"}
+{"input": "structure awareness is done by using Method| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "structure awareness", "output": "graph self - supervised training", "neg_sample": ["structure awareness is done by using Method", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "To this end , we investigate graph self - supervised training to improve the structure awareness of PLMs over AMR graphs .", "forward": false, "src_ids": "2022.acl-long.1426_5216"}
+{"input": "graph self - supervised training is used for OtherScientificTerm| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "graph self - supervised training", "output": "structure awareness", "neg_sample": ["graph self - supervised training is used for OtherScientificTerm", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "To this end , we investigate graph self - supervised training to improve the structure awareness of PLMs over AMR graphs .", "forward": true, "src_ids": "2022.acl-long.1426_5217"}
+{"input": "graph - to - graph pre - training is done by using Method| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "graph - to - graph pre - training", "output": "graph auto - encoding strategies", "neg_sample": ["graph - to - graph pre - training is done by using Method", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "In particular , we introduce two graph auto - encoding strategies for graph - to - graph pre - training and four tasks to integrate text and graph information during pre - training .", "forward": false, "src_ids": "2022.acl-long.1426_5218"}
+{"input": "graph auto - encoding strategies is used for Task| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "graph auto - encoding strategies", "output": "graph - to - graph pre - training", "neg_sample": ["graph auto - encoding strategies is used for Task", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "In particular , we introduce two graph auto - encoding strategies for graph - to - graph pre - training and four tasks to integrate text and graph information during pre - training .", "forward": true, "src_ids": "2022.acl-long.1426_5219"}
+{"input": "pre - training is done by using OtherScientificTerm| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "pre - training", "output": "text and graph information", "neg_sample": ["pre - training is done by using OtherScientificTerm", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "In particular , we introduce two graph auto - encoding strategies for graph - to - graph pre - training and four tasks to integrate text and graph information during pre - training .", "forward": false, "src_ids": "2022.acl-long.1426_5220"}
+{"input": "text and graph information is used for Method| context: abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure . recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively . however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge .", "entity": "text and graph information", "output": "pre - training", "neg_sample": ["text and graph information is used for Method", "abstract meaning representation ( amr ) highlights the core semantic information of text in a graph structure .", "recently , pre - trained language models ( plms ) have advanced tasks of amr parsing and amr - to - text generation , respectively .", "however , plms are typically pre - trained on textual data , thus are sub - optimal for modeling structural knowledge ."], "relation": "used for", "id": "2022.acl-long.1426", "year": 2022, "rel_sent": "In particular , we introduce two graph auto - encoding strategies for graph - to - graph pre - training and four tasks to integrate text and graph information during pre - training .", "forward": true, "src_ids": "2022.acl-long.1426_5221"}
+{"input": "eisner - satta algorithm is used for Task| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "eisner - satta algorithm", "output": "partial marginalization", "neg_sample": ["eisner - satta algorithm is used for Task", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.1452", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently .", "forward": true, "src_ids": "2022.acl-long.1452_5222"}
+{"input": "partial marginalization is done by using Method| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "partial marginalization", "output": "eisner - satta algorithm", "neg_sample": ["partial marginalization is done by using Method", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.1452", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently .", "forward": false, "src_ids": "2022.acl-long.1452_5223"}
+{"input": "inference is done by using Method| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "inference", "output": "eisner - satta algorithm", "neg_sample": ["inference is done by using Method", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.1452", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently .", "forward": false, "src_ids": "2022.acl-long.1452_5224"}
+{"input": "eisner - satta algorithm is used for Task| context: nested named entity recognition ( ner ) has been receiving increasing attention . recently , fu et al . ( 2020 ) adapt a span - based constituency parser to tackle nested ner . they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization . however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing .", "entity": "eisner - satta algorithm", "output": "inference", "neg_sample": ["eisner - satta algorithm is used for Task", "nested named entity recognition ( ner ) has been receiving increasing attention .", "recently , fu et al .", "( 2020 ) adapt a span - based constituency parser to tackle nested ner .", "they treat nested entities as partially - observed constituency trees and propose the masked inside algorithm for partial marginalization .", "however , their method can not leverage entity heads , which have been shown useful in entity mention detection and entity typing ."], "relation": "used for", "id": "2022.acl-long.1452", "year": 2022, "rel_sent": "We leverage the Eisner - Satta algorithm to perform partial marginalization and inference efficiently .", "forward": true, "src_ids": "2022.acl-long.1452_5225"}
+{"input": "dependency parsing is done by using Task| context: compositionality--- the ability to combine familiar units like words into novel phrases and sentences--- has been the focus of intense interest in artificial intelligence in recent years . ( 2020 ) introduced compositional freebase queries ( cfq ) . this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases .", "entity": "dependency parsing", "output": "compositional generalization", "neg_sample": ["dependency parsing is done by using Task", "compositionality--- the ability to combine familiar units like words into novel phrases and sentences--- has been the focus of intense interest in artificial intelligence in recent years .", "( 2020 ) introduced compositional freebase queries ( cfq ) .", "this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases ."], "relation": "used for", "id": "2022.acl-long.1497", "year": 2022, "rel_sent": "Compositional Generalization in Dependency Parsing.", "forward": false, "src_ids": "2022.acl-long.1497_5226"}
+{"input": "compositional generalization is used for Task| context: compositionality--- the ability to combine familiar units like words into novel phrases and sentences--- has been the focus of intense interest in artificial intelligence in recent years . to test compositional generalization in semantic parsing , keysers et al . ( 2020 ) introduced compositional freebase queries ( cfq ) . this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases .", "entity": "compositional generalization", "output": "dependency parsing", "neg_sample": ["compositional generalization is used for Task", "compositionality--- the ability to combine familiar units like words into novel phrases and sentences--- has been the focus of intense interest in artificial intelligence in recent years .", "to test compositional generalization in semantic parsing , keysers et al .", "( 2020 ) introduced compositional freebase queries ( cfq ) .", "this dataset maximizes the similarity between the test and train distributions over primitive units , like words , while maximizing the compound divergence : the dissimilarity between test and train distributions over larger structures , like phrases ."], "relation": "used for", "id": "2022.acl-long.1497", "year": 2022, "rel_sent": "Compositional Generalization in Dependency Parsing.", "forward": true, "src_ids": "2022.acl-long.1497_5227"}
+{"input": "numerical reasoning is done by using OtherScientificTerm| context: numerical reasoning over hybrid data containing both textual and tabular content ( e.g. , financial reports ) has recently attracted much attention in the nlp community . however , existing question answering ( qa ) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi - step numerical reasoning across multiple hierarchical tables .", "entity": "numerical reasoning", "output": "fine - grained annotations of reasoning processes", "neg_sample": ["numerical reasoning is done by using OtherScientificTerm", "numerical reasoning over hybrid data containing both textual and tabular content ( e.g.", ", financial reports ) has recently attracted much attention in the nlp community .", "however , existing question answering ( qa ) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi - step numerical reasoning across multiple hierarchical tables ."], "relation": "used for", "id": "2022.acl-long.1506", "year": 2022, "rel_sent": "MultiHiertt is built from a wealth of financial reports and has the following unique characteristics : 1 ) each document contain multiple tables and longer unstructured texts ; 2 ) most of tables contained are hierarchical ; 3 ) the reasoning process required for each question is more complex and challenging than existing benchmarks ; and 4 ) fine - grained annotations of reasoning processes and supporting facts are provided to reveal complex numerical reasoning .", "forward": false, "src_ids": "2022.acl-long.1506_5228"}
+{"input": "compressed intermediate document representations is done by using Method| context: bert based ranking models have achieved superior performance on various information retrieval tasks . however , the large number of parameters and complex self - attention operations come at a significant latency overhead . to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency . nonetheless , having solved the immediate latency issue , these methods now introduce storage costs and network fetching latency , which limit their adoption in real - life production systems .", "entity": "compressed intermediate document representations", "output": "succinct document representation ( sdr ) scheme", "neg_sample": ["compressed intermediate document representations is done by using Method", "bert based ranking models have achieved superior performance on various information retrieval tasks .", "however , the large number of parameters and complex self - attention operations come at a significant latency overhead .", "to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency .", "nonetheless , having solved the immediate latency issue , these methods now introduce storage costs and network fetching latency , which limit their adoption in real - life production systems ."], "relation": "used for", "id": "2022.acl-long.1511", "year": 2022, "rel_sent": "In this work , we propose the Succinct Document Representation ( SDR ) scheme that computes highly compressed intermediate document representations , mitigating the storage / network issue .", "forward": false, "src_ids": "2022.acl-long.1511_5229"}
+{"input": "succinct document representation ( sdr ) scheme is used for Method| context: bert based ranking models have achieved superior performance on various information retrieval tasks . however , the large number of parameters and complex self - attention operations come at a significant latency overhead . to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency . nonetheless , having solved the immediate latency issue , these methods now introduce storage costs and network fetching latency , which limit their adoption in real - life production systems .", "entity": "succinct document representation ( sdr ) scheme", "output": "compressed intermediate document representations", "neg_sample": ["succinct document representation ( sdr ) scheme is used for Method", "bert based ranking models have achieved superior performance on various information retrieval tasks .", "however , the large number of parameters and complex self - attention operations come at a significant latency overhead .", "to remedy this , recent works propose late - interaction architectures , which allow pre - computation of intermediate document representations , thus reducing latency .", "nonetheless , having solved the immediate latency issue , these methods now introduce storage costs and network fetching latency , which limit their adoption in real - life production systems ."], "relation": "used for", "id": "2022.acl-long.1511", "year": 2022, "rel_sent": "In this work , we propose the Succinct Document Representation ( SDR ) scheme that computes highly compressed intermediate document representations , mitigating the storage / network issue .", "forward": true, "src_ids": "2022.acl-long.1511_5230"}
+{"input": "healthcare applications is done by using Task| context: although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent . as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area .", "entity": "healthcare applications", "output": "task - oriented dialogue systems", "neg_sample": ["healthcare applications is done by using Task", "although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent .", "as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area ."], "relation": "used for", "id": "2022.acl-long.1513", "year": 2022, "rel_sent": "To fill this gap , we investigated an initial pool of 4070 papers from well - known computer science , natural language processing , and artificial intelligence venues , identifying 70 papers discussing the system - level implementation of task - oriented dialogue systems for healthcare applications .", "forward": false, "src_ids": "2022.acl-long.1513_5231"}
+{"input": "task - oriented dialogue systems is used for Task| context: task - oriented dialogue systems are increasingly prevalent in healthcare settings , and have been characterized by a diverse range of architectures and objectives . although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent . as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area .", "entity": "task - oriented dialogue systems", "output": "healthcare applications", "neg_sample": ["task - oriented dialogue systems is used for Task", "task - oriented dialogue systems are increasingly prevalent in healthcare settings , and have been characterized by a diverse range of architectures and objectives .", "although these systems have been surveyed in the medical community from a non - technical perspective , a systematic review from a rigorous computational perspective has to date remained noticeably absent .", "as a result , many important implementation details of healthcare - oriented dialogue systems remain limited or underspecified , slowing the pace of innovation in this area ."], "relation": "used for", "id": "2022.acl-long.1513", "year": 2022, "rel_sent": "To fill this gap , we investigated an initial pool of 4070 papers from well - known computer science , natural language processing , and artificial intelligence venues , identifying 70 papers discussing the system - level implementation of task - oriented dialogue systems for healthcare applications .", "forward": true, "src_ids": "2022.acl-long.1513_5232"}
+{"input": "phonological information is done by using Method| context: character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue .", "entity": "phonological information", "output": "probing classifiers", "neg_sample": ["phonological information is done by using Method", "character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue ."], "relation": "used for", "id": "2022.acl-long.1551", "year": 2022, "rel_sent": "We further develop a set of probing classifiers to intrinsically evaluate what phonological information is encoded in character embeddings .", "forward": false, "src_ids": "2022.acl-long.1551_5233"}
+{"input": "probing classifiers is used for OtherScientificTerm| context: character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue .", "entity": "probing classifiers", "output": "phonological information", "neg_sample": ["probing classifiers is used for OtherScientificTerm", "character - level information is included in many nlp models , but evaluating the information encoded in character representations is an open issue ."], "relation": "used for", "id": "2022.acl-long.1551", "year": 2022, "rel_sent": "We further develop a set of probing classifiers to intrinsically evaluate what phonological information is encoded in character embeddings .", "forward": true, "src_ids": "2022.acl-long.1551_5234"}
+{"input": "text - to - speech model is done by using Method| context: while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data .", "entity": "text - to - speech model", "output": "language - agnostic meta - learning", "neg_sample": ["text - to - speech model is done by using Method", "while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data ."], "relation": "used for", "id": "2022.acl-long.1553", "year": 2022, "rel_sent": "In conjunction with language agnostic meta learning , this enables us to fine - tune a high - quality text - to - speech model on just 30 minutes of data in a previously unseen language spoken by a previously unseen speaker .", "forward": false, "src_ids": "2022.acl-long.1553_5235"}
+{"input": "language - agnostic meta - learning is used for Method| context: while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data .", "entity": "language - agnostic meta - learning", "output": "text - to - speech model", "neg_sample": ["language - agnostic meta - learning is used for Method", "while neural text - to - speech systems perform remarkably well in high - resource scenarios , they can not be applied to the majority of the over 6,000 spoken languages in the world due to a lack of appropriate training data ."], "relation": "used for", "id": "2022.acl-long.1553", "year": 2022, "rel_sent": "In conjunction with language agnostic meta learning , this enables us to fine - tune a high - quality text - to - speech model on just 30 minutes of data in a previously unseen language spoken by a previously unseen speaker .", "forward": true, "src_ids": "2022.acl-long.1553_5236"}
+{"input": "subword segmentation is done by using OtherScientificTerm| context: recent studies have shown that language models pretrained and/or fine - tuned on randomly permuted sentences exhibit competitive performance on glue , putting into question the importance of word order information . somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text .", "entity": "subword segmentation", "output": "shuffling", "neg_sample": ["subword segmentation is done by using OtherScientificTerm", "recent studies have shown that language models pretrained and/or fine - tuned on randomly permuted sentences exhibit competitive performance on glue , putting into question the importance of word order information .", "somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text ."], "relation": "used for", "id": "2022.acl-long.1561", "year": 2022, "rel_sent": "We show this is in part due to a subtlety in how shuffling is implemented in previous work -- before rather than after subword segmentation .", "forward": false, "src_ids": "2022.acl-long.1561_5237"}
+{"input": "shuffling is used for Task| context: recent studies have shown that language models pretrained and/or fine - tuned on randomly permuted sentences exhibit competitive performance on glue , putting into question the importance of word order information . somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text .", "entity": "shuffling", "output": "subword segmentation", "neg_sample": ["shuffling is used for Task", "recent studies have shown that language models pretrained and/or fine - tuned on randomly permuted sentences exhibit competitive performance on glue , putting into question the importance of word order information .", "somewhat counter - intuitively , some of these studies also report that position embeddings appear to be crucial for models ' good performance with shuffled text ."], "relation": "used for", "id": "2022.acl-long.1561", "year": 2022, "rel_sent": "We show this is in part due to a subtlety in how shuffling is implemented in previous work -- before rather than after subword segmentation .", "forward": true, "src_ids": "2022.acl-long.1561_5238"}
+{"input": "spoilers is done by using Method| context: clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary .", "entity": "spoilers", "output": "question answering model", "neg_sample": ["spoilers is done by using Method", "clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary ."], "relation": "used for", "id": "2022.acl-long.1592", "year": 2022, "rel_sent": "A large - scale evaluation and error analysis on a new corpus of 5,000 ~ manually spoiled clickbait posts --- the Webis Clickbait Spoiling Corpus~2022 - --shows that our spoiler type classifier achieves an accuracy of~80 % , while the question answering model DeBERTa - large outperforms all others in generating spoilers for both types .", "forward": false, "src_ids": "2022.acl-long.1592_5239"}
+{"input": "question answering model is used for OtherScientificTerm| context: clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary .", "entity": "question answering model", "output": "spoilers", "neg_sample": ["question answering model is used for OtherScientificTerm", "clickbait links to a web page and advertises its contents by arousing curiosity instead of providing an informative summary ."], "relation": "used for", "id": "2022.acl-long.1592", "year": 2022, "rel_sent": "A large - scale evaluation and error analysis on a new corpus of 5,000 ~ manually spoiled clickbait posts --- the Webis Clickbait Spoiling Corpus~2022 - --shows that our spoiler type classifier achieves an accuracy of~80 % , while the question answering model DeBERTa - large outperforms all others in generating spoilers for both types .", "forward": true, "src_ids": "2022.acl-long.1592_5240"}
+{"input": "language learners is done by using Metric| context: in addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing .", "entity": "language learners", "output": "interpretability", "neg_sample": ["language learners is done by using Metric", "in addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing ."], "relation": "used for", "id": "2022.acl-long.1623", "year": 2022, "rel_sent": "Interpretability for Language Learners Using Example - Based Grammatical Error Correction.", "forward": false, "src_ids": "2022.acl-long.1623_5241"}
+{"input": "eb - gec is used for OtherScientificTerm| context: grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning . however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored . a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections . in addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing .", "entity": "eb - gec", "output": "language learners", "neg_sample": ["eb - gec is used for OtherScientificTerm", "grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning .", "however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored .", "a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections .", "in addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing ."], "relation": "used for", "id": "2022.acl-long.1623", "year": 2022, "rel_sent": "Experiments demonstrate that the examples presented by EB - GEC help language learners decide to accept or refuse suggestions from the GEC output .", "forward": true, "src_ids": "2022.acl-long.1623_5242"}
+{"input": "language learners is done by using Method| context: grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning . however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored . a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections . in addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing .", "entity": "language learners", "output": "eb - gec", "neg_sample": ["language learners is done by using Method", "grammatical error correction ( gec ) should not focus only on high accuracy of corrections but also on interpretability for language learning .", "however , existing neural - based gec models mainly aim at improving accuracy , and their interpretability has not been explored .", "a promising approach for improving interpretability is an example - based method , which uses similar retrieved examples to generate corrections .", "in addition , examples are beneficial in language learning , helping learners understand the basis of grammatically incorrect / correct texts and improve their confidence in writing ."], "relation": "used for", "id": "2022.acl-long.1623", "year": 2022, "rel_sent": "Experiments demonstrate that the examples presented by EB - GEC help language learners decide to accept or refuse suggestions from the GEC output .", "forward": false, "src_ids": "2022.acl-long.1623_5243"}
+{"input": "encoder is used for Task| context: we investigate what kind of structural knowledge learned in neural network encoders is transferable to processing natural language .", "entity": "encoder", "output": "downstream tasks", "neg_sample": ["encoder is used for Task", "we investigate what kind of structural knowledge learned in neural network encoders is transferable to processing natural language ."], "relation": "used for", "id": "2022.acl-long.1654", "year": 2022, "rel_sent": "We design artificial languages with structural properties that mimic natural language , pretrain encoders on the data , and see how much performance the encoder exhibits on downstream tasks in natural language .", "forward": true, "src_ids": "2022.acl-long.1654_5244"}
+{"input": "neural network encoders is used for Material| context: we investigate what kind of structural knowledge learned in neural network encoders is transferable to processing natural language .", "entity": "neural network encoders", "output": "human languages", "neg_sample": ["neural network encoders is used for Material", "we investigate what kind of structural knowledge learned in neural network encoders is transferable to processing natural language ."], "relation": "used for", "id": "2022.acl-long.1654", "year": 2022, "rel_sent": "Our results provide insights into how neural network encoders process human languages and the source of cross - lingual transferability of recent multilingual language models .", "forward": true, "src_ids": "2022.acl-long.1654_5245"}
+{"input": "multilingual pretrained language models is done by using Method| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "multilingual pretrained language models", "output": "entity representations", "neg_sample": ["multilingual pretrained language models is done by using Method", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "m{LUKE } : The Power of Entity Representations in Multilingual Pretrained Language Models.", "forward": false, "src_ids": "2022.acl-long.1655_5246"}
+{"input": "downstream cross - lingual tasks is done by using Method| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "downstream cross - lingual tasks", "output": "entity representations", "neg_sample": ["downstream cross - lingual tasks is done by using Method", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "In this study , we explore the effectiveness of leveraging entity representations for downstream cross - lingual tasks .", "forward": false, "src_ids": "2022.acl-long.1655_5247"}
+{"input": "language - agnostic features is done by using Method| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "language - agnostic features", "output": "entity representations", "neg_sample": ["language - agnostic features is done by using Method", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "We also analyze the model and the key insight is that incorporating entity representations into the input allows us to extract more language - agnostic features .", "forward": false, "src_ids": "2022.acl-long.1655_5248"}
+{"input": "entity representations is used for Method| context: however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "entity representations", "output": "multilingual pretrained language models", "neg_sample": ["entity representations is used for Method", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "m{LUKE } : The Power of Entity Representations in Multilingual Pretrained Language Models.", "forward": true, "src_ids": "2022.acl-long.1655_5249"}
+{"input": "entity representations is used for Task| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "entity representations", "output": "downstream cross - lingual tasks", "neg_sample": ["entity representations is used for Task", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "In this study , we explore the effectiveness of leveraging entity representations for downstream cross - lingual tasks .", "forward": true, "src_ids": "2022.acl-long.1655_5250"}
+{"input": "entity representations is used for OtherScientificTerm| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "entity representations", "output": "language - agnostic features", "neg_sample": ["entity representations is used for OtherScientificTerm", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "We also analyze the model and the key insight is that incorporating entity representations into the input allows us to extract more language - agnostic features .", "forward": true, "src_ids": "2022.acl-long.1655_5251"}
+{"input": "correct factual knowledge is done by using OtherScientificTerm| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "correct factual knowledge", "output": "entity - based prompt", "neg_sample": ["correct factual knowledge is done by using OtherScientificTerm", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "We show that entity - based prompt elicits correct factual knowledge more likely than using only word representations .", "forward": false, "src_ids": "2022.acl-long.1655_5252"}
+{"input": "entity - based prompt is used for OtherScientificTerm| context: recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities . however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks .", "entity": "entity - based prompt", "output": "correct factual knowledge", "neg_sample": ["entity - based prompt is used for OtherScientificTerm", "recent studies have shown that multilingual pretrained language models can be effectively improved with cross - lingual alignment information from wikipedia entities .", "however , existing methods only exploit entity information in pretraining and do not explicitly use entities in downstream tasks ."], "relation": "used for", "id": "2022.acl-long.1655", "year": 2022, "rel_sent": "We show that entity - based prompt elicits correct factual knowledge more likely than using only word representations .", "forward": true, "src_ids": "2022.acl-long.1655_5253"}
+{"input": "language models is done by using Method| context: the allure of superhuman - level capabilities has led to considerable interest in language models like gpt-3 and t5 , wherein the research has , by and large , revolved around new model architectures , training tasks , and loss objectives , along with substantial engineering efforts to scale up model capacity and dataset size . comparatively little work has been done to improve the generalization of these models through better optimization .", "entity": "language models", "output": "sharpness - aware minimization ( sam )", "neg_sample": ["language models is done by using Method", "the allure of superhuman - level capabilities has led to considerable interest in language models like gpt-3 and t5 , wherein the research has , by and large , revolved around new model architectures , training tasks , and loss objectives , along with substantial engineering efforts to scale up model capacity and dataset size .", "comparatively little work has been done to improve the generalization of these models through better optimization ."], "relation": "used for", "id": "2022.acl-long.1658", "year": 2022, "rel_sent": "In this work , we show that Sharpness - Aware Minimization ( SAM ) , a recently proposed optimization procedure that encourages convergence to flatter minima , can substantially improve the generalization of language models without much computational overhead .", "forward": false, "src_ids": "2022.acl-long.1658_5254"}
+{"input": "sharpness - aware minimization ( sam ) is used for Method| context: comparatively little work has been done to improve the generalization of these models through better optimization .", "entity": "sharpness - aware minimization ( sam )", "output": "language models", "neg_sample": ["sharpness - aware minimization ( sam ) is used for Method", "comparatively little work has been done to improve the generalization of these models through better optimization ."], "relation": "used for", "id": "2022.acl-long.1658", "year": 2022, "rel_sent": "In this work , we show that Sharpness - Aware Minimization ( SAM ) , a recently proposed optimization procedure that encourages convergence to flatter minima , can substantially improve the generalization of language models without much computational overhead .", "forward": true, "src_ids": "2022.acl-long.1658_5255"}
+{"input": "deobfuscation is done by using Task| context: recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution . to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches . however , existing authorship obfuscation approaches do not consider the adversarial threat model .", "entity": "deobfuscation", "output": "adversarially trained authorship attributors", "neg_sample": ["deobfuscation is done by using Task", "recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution .", "to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches .", "however , existing authorship obfuscation approaches do not consider the adversarial threat model ."], "relation": "used for", "id": "2022.acl-long.1664", "year": 2022, "rel_sent": "To fill this gap , we investigate the problem of adversarial authorship attribution for deobfuscation .", "forward": false, "src_ids": "2022.acl-long.1664_5256"}
+{"input": "adversarially trained authorship attributors is used for Task| context: recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution . to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches . however , existing authorship obfuscation approaches do not consider the adversarial threat model . specifically , they are not evaluated against adversarially trained authorship attributors that are aware of potential obfuscation .", "entity": "adversarially trained authorship attributors", "output": "deobfuscation", "neg_sample": ["adversarially trained authorship attributors is used for Task", "recent advances in natural language processing have enabled powerful privacy - invasive authorship attribution .", "to counter authorship attribution , researchers have proposed a variety of rule - based and learning - based text obfuscation approaches .", "however , existing authorship obfuscation approaches do not consider the adversarial threat model .", "specifically , they are not evaluated against adversarially trained authorship attributors that are aware of potential obfuscation ."], "relation": "used for", "id": "2022.acl-long.1664", "year": 2022, "rel_sent": "To fill this gap , we investigate the problem of adversarial authorship attribution for deobfuscation .", "forward": true, "src_ids": "2022.acl-long.1664_5257"}
+{"input": "image information is used for Task| context: given a natural language navigation instruction , a visual agent interacts with a graph - based environment equipped with panorama images and tries to follow the described route . most prior work has been conducted in indoor scenarios where best results were obtained for navigation on routes that are similar to the training routes , with sharp drops in performance when testing on unseen environments .", "entity": "image information", "output": "vision and language navigation ( vln )", "neg_sample": ["image information is used for Task", "given a natural language navigation instruction , a visual agent interacts with a graph - based environment equipped with panorama images and tries to follow the described route .", "most prior work has been conducted in indoor scenarios where best results were obtained for navigation on routes that are similar to the training routes , with sharp drops in performance when testing on unseen environments ."], "relation": "used for", "id": "2022.acl-long.1684", "year": 2022, "rel_sent": "We focus on VLN in outdoor scenarios and find that in contrast to indoor VLN , most of the gain in outdoor VLN on unseen data is due to features like junction type embedding or heading delta that are specific to the respective environment graph , while image information plays a very minor role in generalizing VLN to unseen outdoor areas .", "forward": true, "src_ids": "2022.acl-long.1684_5258"}
+{"input": "vision and language navigation ( vln ) is done by using OtherScientificTerm| context: vision and language navigation ( vln ) is a challenging visually - grounded language understanding task . given a natural language navigation instruction , a visual agent interacts with a graph - based environment equipped with panorama images and tries to follow the described route . most prior work has been conducted in indoor scenarios where best results were obtained for navigation on routes that are similar to the training routes , with sharp drops in performance when testing on unseen environments .", "entity": "vision and language navigation ( vln )", "output": "image information", "neg_sample": ["vision and language navigation ( vln ) is done by using OtherScientificTerm", "vision and language navigation ( vln ) is a challenging visually - grounded language understanding task .", "given a natural language navigation instruction , a visual agent interacts with a graph - based environment equipped with panorama images and tries to follow the described route .", "most prior work has been conducted in indoor scenarios where best results were obtained for navigation on routes that are similar to the training routes , with sharp drops in performance when testing on unseen environments ."], "relation": "used for", "id": "2022.acl-long.1684", "year": 2022, "rel_sent": "We focus on VLN in outdoor scenarios and find that in contrast to indoor VLN , most of the gain in outdoor VLN on unseen data is due to features like junction type embedding or heading delta that are specific to the respective environment graph , while image information plays a very minor role in generalizing VLN to unseen outdoor areas .", "forward": false, "src_ids": "2022.acl-long.1684_5259"}
+{"input": "refined objective function is used for OtherScientificTerm| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "entity": "refined objective function", "output": "spurious programs", "neg_sample": ["refined objective function is used for OtherScientificTerm", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally ."], "relation": "used for", "id": "2022.acl-long.1707", "year": 2022, "rel_sent": "Moreover , we design a refined objective function with lexical features and violation punishments to further avoid spurious programs .", "forward": true, "src_ids": "2022.acl-long.1707_5260"}
+{"input": "spurious programs is done by using Method| context: table fact verification aims to check the correctness of textual statements based on given semi - structured data . most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally . however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs .", "entity": "spurious programs", "output": "refined objective function", "neg_sample": ["spurious programs is done by using Method", "table fact verification aims to check the correctness of textual statements based on given semi - structured data .", "most existing methods are devoted to better comprehending logical operations and tables , but they hardly study generating latent programs from statements , with which we can not only retrieve evidences efficiently but also explain reasons behind verifications naturally .", "however , it is challenging to get correct programs with existing weakly supervised semantic parsers due to the huge search space with lots of spurious programs ."], "relation": "used for", "id": "2022.acl-long.1707", "year": 2022, "rel_sent": "Moreover , we design a refined objective function with lexical features and violation punishments to further avoid spurious programs .", "forward": false, "src_ids": "2022.acl-long.1707_5261"}
+{"input": "topical classification tasks is done by using Task| context: in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce . in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance .", "entity": "topical classification tasks", "output": "classification phase", "neg_sample": ["topical classification tasks is done by using Task", "in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce .", "in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance ."], "relation": "used for", "id": "2022.acl-long.1710", "year": 2022, "rel_sent": "We test this hypothesis on various data sets , and show that this additional classification phase can significantly improve performance , mainly for topical classification tasks , when the number of labeled instances available for fine - tuning is only a couple of dozen to a few hundred .", "forward": false, "src_ids": "2022.acl-long.1710_5262"}
+{"input": "classification phase is used for Task| context: in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce . in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance .", "entity": "classification phase", "output": "topical classification tasks", "neg_sample": ["classification phase is used for Task", "in real - world scenarios , a text classification task often begins with a cold start , when labeled data is scarce .", "in such cases , the common practice of fine - tuning pre - trained models , such as bert , for a target classification task , is prone to produce poor performance ."], "relation": "used for", "id": "2022.acl-long.1710", "year": 2022, "rel_sent": "We test this hypothesis on various data sets , and show that this additional classification phase can significantly improve performance , mainly for topical classification tasks , when the number of labeled instances available for fine - tuning is only a couple of dozen to a few hundred .", "forward": true, "src_ids": "2022.acl-long.1710_5263"}
+{"input": "machine translation is done by using Task| context: although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with . hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer .", "entity": "machine translation", "output": "length generalization", "neg_sample": ["machine translation is done by using Task", "although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with .", "hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer ."], "relation": "used for", "id": "2022.acl-long.1712", "year": 2022, "rel_sent": "Third , when transformers need to focus on a single position , as for FIRST , we find that they can fail to generalize to longer strings ; we offer a simple remedy to this problem that also improves length generalization in machine translation .", "forward": false, "src_ids": "2022.acl-long.1712_5264"}
+{"input": "length generalization is used for Task| context: although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with . hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer .", "entity": "length generalization", "output": "machine translation", "neg_sample": ["length generalization is used for Task", "although transformers are remarkably effective for many tasks , there are some surprisingly easy - looking regular languages that they struggle with .", "hahn shows that for languages where acceptance depends on a single input symbol , a transformer 's classification decisions get closer and closer to random guessing ( that is , a cross - entropy of 1 ) as input strings get longer and longer ."], "relation": "used for", "id": "2022.acl-long.1712", "year": 2022, "rel_sent": "Third , when transformers need to focus on a single position , as for FIRST , we find that they can fail to generalize to longer strings ; we offer a simple remedy to this problem that also improves length generalization in machine translation .", "forward": true, "src_ids": "2022.acl-long.1712_5265"}
+{"input": "docogen algorithm is done by using Material| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "docogen algorithm", "output": "text example", "neg_sample": ["docogen algorithm is done by using Material", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.1725", "year": 2022, "rel_sent": "Given an input text example , our DoCoGen algorithm generates a domain - counterfactual textual example ( D - con ) - that is similar to the original in all aspects , including the task label , but its domain is changed to a desired one .", "forward": false, "src_ids": "2022.acl-long.1725_5266"}
+{"input": "text example is used for Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "text example", "output": "docogen algorithm", "neg_sample": ["text example is used for Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.1725", "year": 2022, "rel_sent": "Given an input text example , our DoCoGen algorithm generates a domain - counterfactual textual example ( D - con ) - that is similar to the original in all aspects , including the task label , but its domain is changed to a desired one .", "forward": true, "src_ids": "2022.acl-long.1725_5267"}
+{"input": "da setups is done by using Method| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "da setups", "output": "multi - label intent classifier", "neg_sample": ["da setups is done by using Method", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.1725", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": false, "src_ids": "2022.acl-long.1725_5268"}
+{"input": "multi - label intent classifier is used for Material| context: natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples .", "entity": "multi - label intent classifier", "output": "da setups", "neg_sample": ["multi - label intent classifier is used for Material", "natural language processing ( nlp ) algorithms have become very successful , but they still struggle when applied to out - of - distribution examples ."], "relation": "used for", "id": "2022.acl-long.1725", "year": 2022, "rel_sent": "We use the D - cons generated by DoCoGen to augment a sentiment classifier and a multi - label intent classifier in 20 and 78 DA setups , respectively , where source - domain labeled data is scarce .", "forward": true, "src_ids": "2022.acl-long.1725_5269"}
+{"input": "hop patterns is done by using OtherScientificTerm| context: semantic dependencies in srl are modeled as a distribution over semantic dependency labels conditioned on a predicate and an argument word .", "entity": "hop patterns", "output": "semantic label distributions", "neg_sample": ["hop patterns is done by using OtherScientificTerm", "semantic dependencies in srl are modeled as a distribution over semantic dependency labels conditioned on a predicate and an argument word ."], "relation": "used for", "id": "2022.acl-long.1756", "year": 2022, "rel_sent": "We target the variation of semantic label distributions using a mixture model , separately estimating semantic label distributions for different hop patterns and probabilistically clustering hop patterns with similar semantic label distributions .", "forward": false, "src_ids": "2022.acl-long.1756_5270"}
+{"input": "semantic label distributions is used for OtherScientificTerm| context: semantic dependencies in srl are modeled as a distribution over semantic dependency labels conditioned on a predicate and an argument word .", "entity": "semantic label distributions", "output": "hop patterns", "neg_sample": ["semantic label distributions is used for OtherScientificTerm", "semantic dependencies in srl are modeled as a distribution over semantic dependency labels conditioned on a predicate and an argument word ."], "relation": "used for", "id": "2022.acl-long.1756", "year": 2022, "rel_sent": "We target the variation of semantic label distributions using a mixture model , separately estimating semantic label distributions for different hop patterns and probabilistically clustering hop patterns with similar semantic label distributions .", "forward": true, "src_ids": "2022.acl-long.1756_5271"}
+{"input": "zero - shot nlp tasks is done by using Method| context: long - range semantic coherence remains a challenge in automatic language generation and understanding . we demonstrate that large language models have insufficiently learned the effect of distant words on next - token prediction .", "entity": "zero - shot nlp tasks", "output": "coherence boosting", "neg_sample": ["zero - shot nlp tasks is done by using Method", "long - range semantic coherence remains a challenge in automatic language generation and understanding .", "we demonstrate that large language models have insufficiently learned the effect of distant words on next - token prediction ."], "relation": "used for", "id": "2022.acl-long.1798", "year": 2022, "rel_sent": "It is also found that coherence boosting with state - of - the - art models for various zero - shot NLP tasks yields performance gains with no additional training .", "forward": false, "src_ids": "2022.acl-long.1798_5272"}
+{"input": "coherence boosting is used for Task| context: long - range semantic coherence remains a challenge in automatic language generation and understanding . we demonstrate that large language models have insufficiently learned the effect of distant words on next - token prediction .", "entity": "coherence boosting", "output": "zero - shot nlp tasks", "neg_sample": ["coherence boosting is used for Task", "long - range semantic coherence remains a challenge in automatic language generation and understanding .", "we demonstrate that large language models have insufficiently learned the effect of distant words on next - token prediction ."], "relation": "used for", "id": "2022.acl-long.1798", "year": 2022, "rel_sent": "It is also found that coherence boosting with state - of - the - art models for various zero - shot NLP tasks yields performance gains with no additional training .", "forward": true, "src_ids": "2022.acl-long.1798_5273"}
+{"input": "text classification tasks is done by using Task| context: most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks . little attention has been paid to ue in natural language processing .", "entity": "text classification tasks", "output": "misclassification detection", "neg_sample": ["text classification tasks is done by using Task", "most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks .", "little attention has been paid to ue in natural language processing ."], "relation": "used for", "id": "2022.acl-long.1801", "year": 2022, "rel_sent": "To fill this gap , we perform a vast empirical investigation of state - of - the - art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications , one of which approaches or even outperforms computationally intensive methods .", "forward": false, "src_ids": "2022.acl-long.1801_5274"}
+{"input": "transformer models is done by using Method| context: uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc . most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks . little attention has been paid to ue in natural language processing .", "entity": "transformer models", "output": "ue methods", "neg_sample": ["transformer models is done by using Method", "uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc .", "most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks .", "little attention has been paid to ue in natural language processing ."], "relation": "used for", "id": "2022.acl-long.1801", "year": 2022, "rel_sent": "To fill this gap , we perform a vast empirical investigation of state - of - the - art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications , one of which approaches or even outperforms computationally intensive methods .", "forward": false, "src_ids": "2022.acl-long.1801_5275"}
+{"input": "ue methods is used for Method| context: uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc . most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks . little attention has been paid to ue in natural language processing .", "entity": "ue methods", "output": "transformer models", "neg_sample": ["ue methods is used for Method", "uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc .", "most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks .", "little attention has been paid to ue in natural language processing ."], "relation": "used for", "id": "2022.acl-long.1801", "year": 2022, "rel_sent": "To fill this gap , we perform a vast empirical investigation of state - of - the - art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications , one of which approaches or even outperforms computationally intensive methods .", "forward": true, "src_ids": "2022.acl-long.1801_5276"}
+{"input": "misclassification detection is used for Task| context: uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc . most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks . little attention has been paid to ue in natural language processing .", "entity": "misclassification detection", "output": "text classification tasks", "neg_sample": ["misclassification detection is used for Task", "uncertainty estimation ( ue ) of model predictions is a crucial step for a variety of tasks such as active learning , misclassification detection , adversarial attack detection , out - of - distribution detection , etc .", "most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks .", "little attention has been paid to ue in natural language processing ."], "relation": "used for", "id": "2022.acl-long.1801", "year": 2022, "rel_sent": "To fill this gap , we perform a vast empirical investigation of state - of - the - art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications , one of which approaches or even outperforms computationally intensive methods .", "forward": true, "src_ids": "2022.acl-long.1801_5277"}
+{"input": "ai tasks is done by using OtherScientificTerm| context: several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized . at issue here are not just individual systems and datasets , but also the ai tasks themselves .", "entity": "ai tasks", "output": "ethics sheets", "neg_sample": ["ai tasks is done by using OtherScientificTerm", "several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized .", "at issue here are not just individual systems and datasets , but also the ai tasks themselves ."], "relation": "used for", "id": "2022.acl-long.1822", "year": 2022, "rel_sent": "I will present a new form of such an effort , Ethics Sheets for AI Tasks , dedicated to fleshing out the assumptions and ethical considerations hidden in how a task is commonly framed and in the choices we make regarding the data , method , and evaluation .", "forward": false, "src_ids": "2022.acl-long.1822_5278"}
+{"input": "ethics sheets is used for Task| context: several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized .", "entity": "ethics sheets", "output": "ai tasks", "neg_sample": ["ethics sheets is used for Task", "several high - profile events , such as the mass testing of emotion recognition systems on vulnerable sub - populations and using question answering systems to make moral judgments , have highlighted how technology will often lead to more adverse outcomes for those that are already marginalized ."], "relation": "used for", "id": "2022.acl-long.1822", "year": 2022, "rel_sent": "I will present a new form of such an effort , Ethics Sheets for AI Tasks , dedicated to fleshing out the assumptions and ethical considerations hidden in how a task is commonly framed and in the choices we make regarding the data , method , and evaluation .", "forward": true, "src_ids": "2022.acl-long.1822_5279"}
+{"input": "parallel text generation is done by using Method| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "parallel text generation", "output": "latent - glat : glancing at latent variables", "neg_sample": ["parallel text generation is done by using Method", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "latent - GLAT : Glancing at Latent Variables for Parallel Text Generation.", "forward": false, "src_ids": "2022.acl-long.1825_5280"}
+{"input": "latent - glat : glancing at latent variables is used for Task| context: although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "latent - glat : glancing at latent variables", "output": "parallel text generation", "neg_sample": ["latent - glat : glancing at latent variables is used for Task", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "latent - GLAT : Glancing at Latent Variables for Parallel Text Generation.", "forward": true, "src_ids": "2022.acl-long.1825_5281"}
+{"input": "multi - modality problem is done by using Method| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "multi - modality problem", "output": "latent - glat", "neg_sample": ["multi - modality problem is done by using Method", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "In this paper , we propose latent - GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": false, "src_ids": "2022.acl-long.1825_5282"}
+{"input": "word categorical information is done by using OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "word categorical information", "output": "discrete latent variables", "neg_sample": ["word categorical information is done by using OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "In this paper , we propose latent - GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": false, "src_ids": "2022.acl-long.1825_5283"}
+{"input": "discrete latent variables is used for OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "discrete latent variables", "output": "word categorical information", "neg_sample": ["discrete latent variables is used for OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "In this paper , we propose latent - GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": true, "src_ids": "2022.acl-long.1825_5284"}
+{"input": "curriculum learning technique is used for OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "curriculum learning technique", "output": "multi - modality problem", "neg_sample": ["curriculum learning technique is used for OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "In this paper , we propose latent - GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": true, "src_ids": "2022.acl-long.1825_5285"}
+{"input": "latent - glat is used for OtherScientificTerm| context: recently , parallel text generation has received widespread attention due to its success in generation efficiency . although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications .", "entity": "latent - glat", "output": "multi - modality problem", "neg_sample": ["latent - glat is used for OtherScientificTerm", "recently , parallel text generation has received widespread attention due to its success in generation efficiency .", "although many advanced techniques are proposed to improve its generation quality , they still need the help of an autoregressive model for training to overcome the one - to - many multi - modal phenomenon in the dataset , limiting their applications ."], "relation": "used for", "id": "2022.acl-long.1825", "year": 2022, "rel_sent": "In this paper , we propose latent - GLAT , which employs the discrete latent variables to capture word categorical information and invoke an advanced curriculum learning technique , alleviating the multi - modality problem .", "forward": true, "src_ids": "2022.acl-long.1825_5286"}
+{"input": "data augmentation is done by using OtherScientificTerm| context: wpd measures the degree of structural alteration , while ld measures the difference in vocabulary used .", "entity": "data augmentation", "output": "paraphrases", "neg_sample": ["data augmentation is done by using OtherScientificTerm", "wpd measures the degree of structural alteration , while ld measures the difference in vocabulary used ."], "relation": "used for", "id": "2022.acl-long.1863", "year": 2022, "rel_sent": "Lastly , we apply our metrics to filter the output of a paraphrase generation model and show how it can be used to generate specific forms of paraphrases for data augmentation or robustness testing of NLP models .", "forward": false, "src_ids": "2022.acl-long.1863_5287"}
+{"input": "paraphrases is used for Method| context: wpd measures the degree of structural alteration , while ld measures the difference in vocabulary used .", "entity": "paraphrases", "output": "data augmentation", "neg_sample": ["paraphrases is used for Method", "wpd measures the degree of structural alteration , while ld measures the difference in vocabulary used ."], "relation": "used for", "id": "2022.acl-long.1863", "year": 2022, "rel_sent": "Lastly , we apply our metrics to filter the output of a paraphrase generation model and show how it can be used to generate specific forms of paraphrases for data augmentation or robustness testing of NLP models .", "forward": true, "src_ids": "2022.acl-long.1863_5288"}
+{"input": "snare is done by using Method| context: natural language applied to natural 2d images describes a fundamentally 3d world .", "entity": "snare", "output": "voxel - informed language grounder", "neg_sample": ["snare is done by using Method", "natural language applied to natural 2d images describes a fundamentally 3d world ."], "relation": "used for", "id": "2022.acl-short.206", "year": 2022, "rel_sent": "We show that VLG significantly improves grounding accuracy on SNARE , an object reference game task .", "forward": false, "src_ids": "2022.acl-short.206_5289"}
+{"input": "voxel - informed language grounder is used for Task| context: natural language applied to natural 2d images describes a fundamentally 3d world .", "entity": "voxel - informed language grounder", "output": "snare", "neg_sample": ["voxel - informed language grounder is used for Task", "natural language applied to natural 2d images describes a fundamentally 3d world ."], "relation": "used for", "id": "2022.acl-short.206", "year": 2022, "rel_sent": "We show that VLG significantly improves grounding accuracy on SNARE , an object reference game task .", "forward": true, "src_ids": "2022.acl-short.206_5290"}
+{"input": "deep prompt tuning is used for Task| context: prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training . we also find that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality .", "entity": "deep prompt tuning", "output": "nlu", "neg_sample": ["deep prompt tuning is used for Task", "prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training .", "we also find that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality ."], "relation": "used for", "id": "2022.acl-short.263", "year": 2022, "rel_sent": "Our method P - Tuning v2 is an implementation of Deep Prompt Tuning   optimized and adapted for NLU .", "forward": true, "src_ids": "2022.acl-short.263_5291"}
+{"input": "nlu is done by using Method| context: prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training . however , in the context of nlu , prior work reveals that prompt tuning does not perform well for normal - sized pretrained models . we also find that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality .", "entity": "nlu", "output": "deep prompt tuning", "neg_sample": ["nlu is done by using Method", "prompt tuning , which only tunes continuous prompts with a frozen language model , substantially reduces per - task storage and memory usage at training .", "however , in the context of nlu , prior work reveals that prompt tuning does not perform well for normal - sized pretrained models .", "we also find that existing methods of prompt tuning can not handle hard sequence labeling tasks , indicating a lack of universality ."], "relation": "used for", "id": "2022.acl-short.263", "year": 2022, "rel_sent": "Our method P - Tuning v2 is an implementation of Deep Prompt Tuning   optimized and adapted for NLU .", "forward": false, "src_ids": "2022.acl-short.263_5292"}
+{"input": "prompt tuning is used for Task| context: prompt tuning has recently emerged as an effective method for adapting pre - trained language models to a number of language understanding and generation tasks .", "entity": "prompt tuning", "output": "semantic parsing", "neg_sample": ["prompt tuning is used for Task", "prompt tuning has recently emerged as an effective method for adapting pre - trained language models to a number of language understanding and generation tasks ."], "relation": "used for", "id": "2022.acl-short.415", "year": 2022, "rel_sent": "In this paper , we investigate prompt tuning for semantic parsing - the task of mapping natural language utterances onto formal meaning representations .", "forward": true, "src_ids": "2022.acl-short.415_5293"}
+{"input": "semantic distinction is done by using Method| context: specifically , none of the existing few - shot approaches ( including the in - context learning of gpt-3 ) can attain a performance that is meaningfully different from the random baseline .", "entity": "semantic distinction", "output": "prompt - based techniques", "neg_sample": ["semantic distinction is done by using Method", "specifically , none of the existing few - shot approaches ( including the in - context learning of gpt-3 ) can attain a performance that is meaningfully different from the random baseline ."], "relation": "used for", "id": "2022.acl-short.717", "year": 2022, "rel_sent": "Our simple adaptation shows that the failure of existing prompt - based techniques in semantic distinction is due to their improper configuration , rather than lack of relevant knowledge in the representations .", "forward": false, "src_ids": "2022.acl-short.717_5294"}
+{"input": "debiased test set is done by using Material| context: we present a debiased dataset for the person - centric visual grounding ( pcvg ) task first proposed by cui et al . given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image . naturally , models trained on these biased data lead to over - estimation of performance on the benchmark .", "entity": "debiased test set", "output": "biased ( and larger ) training set", "neg_sample": ["debiased test set is done by using Material", "we present a debiased dataset for the person - centric visual grounding ( pcvg ) task first proposed by cui et al .", "given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image .", "naturally , models trained on these biased data lead to over - estimation of performance on the benchmark ."], "relation": "used for", "id": "2022.acl-short.768", "year": 2022, "rel_sent": "We also demonstrate the same benchmark model trained on our debiased training set outperforms that trained on the original biased ( and larger ) training set on our debiased test set .", "forward": false, "src_ids": "2022.acl-short.768_5295"}
+{"input": "biased ( and larger ) training set is used for Material| context: we present a debiased dataset for the person - centric visual grounding ( pcvg ) task first proposed by cui et al . given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image . naturally , models trained on these biased data lead to over - estimation of performance on the benchmark .", "entity": "biased ( and larger ) training set", "output": "debiased test set", "neg_sample": ["biased ( and larger ) training set is used for Material", "we present a debiased dataset for the person - centric visual grounding ( pcvg ) task first proposed by cui et al .", "given an image and a caption , pcvg requires pairing up a person 's name mentioned in a caption with a bounding box that points to the person in the image .", "naturally , models trained on these biased data lead to over - estimation of performance on the benchmark ."], "relation": "used for", "id": "2022.acl-short.768", "year": 2022, "rel_sent": "We also demonstrate the same benchmark model trained on our debiased training set outperforms that trained on the original biased ( and larger ) training set on our debiased test set .", "forward": true, "src_ids": "2022.acl-short.768_5296"}
+{"input": "language understanding is done by using Method| context: transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks .", "entity": "language understanding", "output": "cross - modal systems", "neg_sample": ["language understanding is done by using Method", "transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks ."], "relation": "used for", "id": "2022.acl-short.1108", "year": 2022, "rel_sent": "{ XDBERT } : { D}istilling Visual Information to { BERT } from Cross - Modal Systems to Improve Language Understanding.", "forward": false, "src_ids": "2022.acl-short.1108_5297"}
+{"input": "language - heavy characteristics is done by using OtherScientificTerm| context: transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks .", "entity": "language - heavy characteristics", "output": "learning objective", "neg_sample": ["language - heavy characteristics is done by using OtherScientificTerm", "transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks ."], "relation": "used for", "id": "2022.acl-short.1108", "year": 2022, "rel_sent": "Our framework is inspired by cross - modal encoders ' success in visual - language tasks while we alter the learning objective to cater to the language - heavy characteristics of NLU .", "forward": false, "src_ids": "2022.acl-short.1108_5298"}
+{"input": "learning objective is used for OtherScientificTerm| context: transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks .", "entity": "learning objective", "output": "language - heavy characteristics", "neg_sample": ["learning objective is used for OtherScientificTerm", "transformer - based models are widely used in natural language understanding ( nlu ) tasks , and multimodal transformers have been effective in visual - language tasks ."], "relation": "used for", "id": "2022.acl-short.1108", "year": 2022, "rel_sent": "Our framework is inspired by cross - modal encoders ' success in visual - language tasks while we alter the learning objective to cater to the language - heavy characteristics of NLU .", "forward": true, "src_ids": "2022.acl-short.1108_5299"}
+{"input": "low - resource scenarios is done by using Method| context: the emergence of multilingual pre - trained language models makes it possible to adapt to target languages with only few labeled examples . however , vanilla fine - tuning tends to achieve degenerated and unstable results , owing to the language interference among different languages , and parameter overload under the few - sample transfer learning scenarios .", "entity": "low - resource scenarios", "output": "language sub - networks", "neg_sample": ["low - resource scenarios is done by using Method", "the emergence of multilingual pre - trained language models makes it possible to adapt to target languages with only few labeled examples .", "however , vanilla fine - tuning tends to achieve degenerated and unstable results , owing to the language interference among different languages , and parameter overload under the few - sample transfer learning scenarios ."], "relation": "used for", "id": "2022.acl-short.1240", "year": 2022, "rel_sent": "In this way , the language sub - networks lower the scale of trainable parameters , and hence better suit the low - resource scenarios .", "forward": false, "src_ids": "2022.acl-short.1240_5300"}
+{"input": "language sub - networks is used for OtherScientificTerm| context: the emergence of multilingual pre - trained language models makes it possible to adapt to target languages with only few labeled examples . however , vanilla fine - tuning tends to achieve degenerated and unstable results , owing to the language interference among different languages , and parameter overload under the few - sample transfer learning scenarios .", "entity": "language sub - networks", "output": "low - resource scenarios", "neg_sample": ["language sub - networks is used for OtherScientificTerm", "the emergence of multilingual pre - trained language models makes it possible to adapt to target languages with only few labeled examples .", "however , vanilla fine - tuning tends to achieve degenerated and unstable results , owing to the language interference among different languages , and parameter overload under the few - sample transfer learning scenarios ."], "relation": "used for", "id": "2022.acl-short.1240", "year": 2022, "rel_sent": "In this way , the language sub - networks lower the scale of trainable parameters , and hence better suit the low - resource scenarios .", "forward": true, "src_ids": "2022.acl-short.1240_5301"}
+{"input": "goal - oriented document - grounded dialogue is done by using Method| context: existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation . however , such pipeline methods would unavoidably suffer from the error propagation issue .", "entity": "goal - oriented document - grounded dialogue", "output": "unified generative framework", "neg_sample": ["goal - oriented document - grounded dialogue is done by using Method", "existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation .", "however , such pipeline methods would unavoidably suffer from the error propagation issue ."], "relation": "used for", "id": "2022.acl-short.1374", "year": 2022, "rel_sent": "Uni{GDD } : A Unified Generative Framework for Goal - Oriented Document - Grounded Dialogue.", "forward": false, "src_ids": "2022.acl-short.1374_5302"}
+{"input": "unified generative framework is used for Task| context: existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation . however , such pipeline methods would unavoidably suffer from the error propagation issue .", "entity": "unified generative framework", "output": "goal - oriented document - grounded dialogue", "neg_sample": ["unified generative framework is used for Task", "existing studies tackle this problem by decomposing it into two sub - tasks : knowledge identification and response generation .", "however , such pipeline methods would unavoidably suffer from the error propagation issue ."], "relation": "used for", "id": "2022.acl-short.1374", "year": 2022, "rel_sent": "Uni{GDD } : A Unified Generative Framework for Goal - Oriented Document - Grounded Dialogue.", "forward": true, "src_ids": "2022.acl-short.1374_5303"}
+{"input": "hyperbolic space is used for Metric| context: interpolation - based regularisation methods such as mixup , which generate virtual training samples , have proven to be effective for various tasks and modalities .", "entity": "hyperbolic space", "output": "similarity measures", "neg_sample": ["hyperbolic space is used for Metric", "interpolation - based regularisation methods such as mixup , which generate virtual training samples , have proven to be effective for various tasks and modalities ."], "relation": "used for", "id": "2022.acl-short.1378", "year": 2022, "rel_sent": "DMix leverages the hyperbolic space as a similarity measure among input samples for a richer encoded representation .", "forward": true, "src_ids": "2022.acl-short.1378_5304"}
+{"input": "similarity measures is done by using OtherScientificTerm| context: interpolation - based regularisation methods such as mixup , which generate virtual training samples , have proven to be effective for various tasks and modalities .", "entity": "similarity measures", "output": "hyperbolic space", "neg_sample": ["similarity measures is done by using OtherScientificTerm", "interpolation - based regularisation methods such as mixup , which generate virtual training samples , have proven to be effective for various tasks and modalities ."], "relation": "used for", "id": "2022.acl-short.1378", "year": 2022, "rel_sent": "DMix leverages the hyperbolic space as a similarity measure among input samples for a richer encoded representation .", "forward": false, "src_ids": "2022.acl-short.1378_5305"}
+{"input": "text representation learning is done by using Material| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "text representation learning", "output": "nlp papers", "neg_sample": ["text representation learning is done by using Material", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.1705", "year": 2022, "rel_sent": "In this short paper , we formally analyze several recent NLP papers proposing text representation learning using DPText and reveal their false claims of being differentially private .", "forward": false, "src_ids": "2022.acl-short.1705_5306"}
+{"input": "nlp papers is used for Task| context: as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods . one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees . despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp .", "entity": "nlp papers", "output": "text representation learning", "neg_sample": ["nlp papers is used for Task", "as privacy gains traction in the nlp community , researchers have started adopting various approaches to privacy - preserving methods .", "one of the favorite privacy frameworks , differential privacy ( dp ) , is perhaps the most compelling thanks to its fundamental theoretical guarantees .", "despite the apparent simplicity of the general concept of differential privacy , it seems non - trivial to get it right when applying it to nlp ."], "relation": "used for", "id": "2022.acl-short.1705", "year": 2022, "rel_sent": "In this short paper , we formally analyze several recent NLP papers proposing text representation learning using DPText and reveal their false claims of being differentially private .", "forward": true, "src_ids": "2022.acl-short.1705_5307"}
+{"input": "multiple synonyms matching network is used for Task| context: existing methods usually apply label attention with code representations to match related text snippets .", "entity": "multiple synonyms matching network", "output": "automatic icd coding", "neg_sample": ["multiple synonyms matching network is used for Task", "existing methods usually apply label attention with code representations to match related text snippets ."], "relation": "used for", "id": "2022.acl-short.1768", "year": 2022, "rel_sent": "Code Synonyms Do Matter : Multiple Synonyms Matching Network for Automatic ICD Coding.", "forward": true, "src_ids": "2022.acl-short.1768_5308"}
+{"input": "code representation learning is done by using OtherScientificTerm| context: automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs ) . existing methods usually apply label attention with code representations to match related text snippets .", "entity": "code representation learning", "output": "synonyms", "neg_sample": ["code representation learning is done by using OtherScientificTerm", "automatic icd coding is defined as assigning disease codes to electronic medical records ( emrs ) .", "existing methods usually apply label attention with code representations to match related text snippets ."], "relation": "used for", "id": "2022.acl-short.1768", "year": 2022, "rel_sent": "Then , we propose a multiple synonyms matching network to leverage synonyms for better code representation learning , and finally help the code classification .", "forward": false, "src_ids": "2022.acl-short.1768_5309"}