{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2024,9,9]],"date-time":"2024-09-09T14:40:19Z","timestamp":1725892819436},"reference-count":85,"publisher":"MIT Press","license":[{"start":{"date-parts":[[2021,3,18]],"date-time":"2021-03-18T00:00:00Z","timestamp":1616025600000},"content-version":"vor","delay-in-days":76,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2021,3,11]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>For natural language processing systems, two kinds of evidence support the use of text representations from neural language models \u201cpretrained\u201d on large unannotated corpora: performance on application-inspired benchmarks (Peters et al., 2018, inter alia), and the emergence of syntactic abstractions in those representations (Tenney et al., 2019, inter alia). On the other hand, the lack of grounded supervision calls into question how well these representations can ever capture meaning (Bender and Koller, 2020). We apply novel probes to recent language models\u2014 specifically focusing on predicate-argument structure as operationalized by semantic dependencies (Ivanova et al., 2012)\u2014and find that, unlike syntax, semantics is not brought to the surface by today\u2019s pretrained models. We then use convolutional graph encoders to explicitly incorporate semantic parses into task-specific finetuning, yielding benefits to natural language understanding (NLU) tasks in the GLUE benchmark. This approach demonstrates the potential for general-purpose (rather than task-specific) linguistic supervision, above and beyond conventional pretraining and finetuning. Several diagnostics help to localize the benefits of our approach.1<\/jats:p>","DOI":"10.1162\/tacl_a_00363","type":"journal-article","created":{"date-parts":[[2021,3,19]],"date-time":"2021-03-19T17:10:07Z","timestamp":1616173807000},"page":"226-242","update-policy":"http:\/\/dx.doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":11,"title":["Infusing Finetuning with Semantic Dependencies"],"prefix":"10.1162","volume":"9","author":[{"given":"Zhaofeng","family":"Wu","sequence":"first","affiliation":[{"name":"Paul G. Allen School of Computer Science & Engineering, University of Washington, United States. zfw7@cs.washington.edu"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hao","family":"Peng","sequence":"additional","affiliation":[{"name":"Paul G. Allen School of Computer Science & Engineering, University of Washington, United States. hapeng@cs.washington.edu"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Noah A.","family":"Smith","sequence":"additional","affiliation":[{"name":"Paul G. Allen School of Computer Science & Engineering, University of Washington, United States"},{"name":"Allen Institute for Artificial Intelligence, United States. nasmith@cs.washington.edu"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"281","published-online":{"date-parts":[[2021,3,11]]},"reference":[{"key":"2021060823401099800_bib1","article-title":"Universal conceptual cognitive annotation (UCCA)","volume-title":"Proceedings of ACL","author":"Abend","year":"2013"},{"key":"2021060823401099800_bib2","article-title":"Fine-grained analysis of sentence embeddings using auxiliary prediction tasks","volume-title":"Proceedings of ICLR","author":"Adi","year":"2017"},{"key":"2021060823401099800_bib3","article-title":"The Berkeley FrameNet project","volume-title":"Proceedings of ACL","author":"Baker","year":"1998"},{"key":"2021060823401099800_bib4","article-title":"The second PASCAL recognising textual entailment challenge","author":"Bar-Haim","year":"2006","journal-title":"Proceedings of the Second PASCAL Challenges Workshop on Recognising Textual Entailment"},{"key":"2021060823401099800_bib5","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D17-1209","article-title":"Graph convolutional encoders for syntax-aware neural machine translation","volume-title":"Proceedings of EMNLP","author":"Bastings","year":"2017"},{"issue":"1","key":"2021060823401099800_bib6","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1162\/coli_a_00367","article-title":"On the linguistic representational power of neural machine translation models","volume":"46","author":"Belinkov","year":"2020","journal-title":"Computational Linguistics"},{"key":"2021060823401099800_bib7","doi-asserted-by":"crossref","first-page":"49","DOI":"10.1162\/tacl_a_00254","article-title":"Analysis methods in neural language processing: A survey","volume":"7","author":"Belinkov","year":"2019","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2021060823401099800_bib8","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2020.acl-main.463","article-title":"Climbing towards NLU: On meaning, form, and understanding in the age of data","volume-title":"Proceedings of ACL","author":"Bender","year":"2020"},{"key":"2021060823401099800_bib9","article-title":"The fifth PASCAL recognizing textual entailment challenge","volume-title":"Proceedings of TAC","author":"Bentivogli","year":"2009"},{"key":"2021060823401099800_bib10","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/S17-2001","article-title":"SemEval-2017 task 1: Semantic textual similarity multilingual and crosslingual focused evaluation","volume-title":"Proceedings of SemEval","author":"Cer","year":"2017"},{"key":"2021060823401099800_bib11","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/K19-2007","article-title":"HIT-SCIR at MRP 2019: A unified pipeline for meaning representation parsing via efficient training and effective encoding","volume-title":"Proceedings of MRP","author":"Che","year":"2019"},{"key":"2021060823401099800_bib12","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P17-1152","article-title":"Enhanced LSTM for natural language inference","volume-title":"Proceedings of ACL","author":"Chen","year":"2017"},{"key":"2021060823401099800_bib13","first-page":"1396","article-title":"On the shortest arborescence of a directed graph","volume":"14","author":"Chu","year":"1965","journal-title":"Science Sinica"},{"key":"2021060823401099800_bib14","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/W19-4828","article-title":"What does BERT look at? an analysis of BERT\u2019s attention","volume-title":"Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","author":"Clark","year":"2019"},{"key":"2021060823401099800_bib15","article-title":"ELECTRA: Pre-training text encoders as discriminators rather than generators","volume-title":"Proceedings of ICLR","author":"Clark","year":"2020"},{"key":"2021060823401099800_bib16","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D18-1217","article-title":"Semi-supervised sequence modeling with cross-view training","volume-title":"Proceedings of EMNLP","author":"Clark","year":"2018"},{"issue":"2\u20133","key":"2021060823401099800_bib17","doi-asserted-by":"crossref","first-page":"281","DOI":"10.1007\/s11168-006-6327-9","article-title":"Minimal recursion semantics: An introduction","volume":"3","author":"Copestake","year":"2005","journal-title":"Research on Language and Computation"},{"key":"2021060823401099800_bib18","article-title":"(accessed September 1, 2020)","author":"Csernai","year":"2017","journal-title":"First Quora Dataset Release: Question Pairs"},{"key":"2021060823401099800_bib19","article-title":"The PASCAL recognising textual entailment challenge","volume-title":"Proceedings of the First International Conference on Machine Learning Challenges: Evaluating Predictive Uncertainty Visual Object Classification, and Recognizing Textual Entailment","author":"Dagan","year":"2005"},{"key":"2021060823401099800_bib20","article-title":"Generating typed dependency parses from phrase structure parses","volume-title":"Proceedings of LREC","author":"de Marneffe","year":"2006"},{"key":"2021060823401099800_bib21","article-title":"BERT: Pre-training of deep bidirectional transformers for language understanding","volume-title":"Proceedings of NAACL","author":"Devlin","year":"2019"},{"key":"2021060823401099800_bib22","article-title":"Fine-tuning pretrained language models: Weight initializations, data orders, and early stopping","author":"Dodge","year":"2020","journal-title":"arXiv preprint arXiv:2002.06305"},{"key":"2021060823401099800_bib23","article-title":"Automatically constructing a corpus of sentential paraphrases","volume-title":"Proceedings of the Third International Workshop on Paraphrasing","author":"Dolan","year":"2005"},{"key":"2021060823401099800_bib24","article-title":"Deep biaffine attention for neural dependency parsing","volume-title":"Proceedings of ICLR","author":"Dozat","year":"2017"},{"key":"2021060823401099800_bib25","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P18-2077","article-title":"Simpler but more accurate semantic dependency parsing","volume-title":"Proceedings of ACL","author":"Dozat","year":"2018"},{"key":"2021060823401099800_bib26","doi-asserted-by":"crossref","first-page":"233","DOI":"10.6028\/jres.071B.032","article-title":"Optimum branchings","volume":"71B","author":"Edmonds","year":"1967","journal-title":"Journal of Research of the National Bureau of Standards"},{"key":"2021060823401099800_bib27","doi-asserted-by":"crossref","DOI":"10.3115\/992628.992688","article-title":"Three new probabilistic models for dependency parsing: An exploration","volume-title":"Proceedings of COLING","author":"Eisner","year":"1996"},{"key":"2021060823401099800_bib28","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P17-2012","article-title":"Learning to parse and translate improves neural machine translation","volume-title":"Proceedings of ACL","author":"Eriguchi","year":"2017"},{"key":"2021060823401099800_bib29","doi-asserted-by":"crossref","DOI":"10.3115\/1654536.1654538","article-title":"The third PASCAL recognizing textual entailment challenge","volume-title":"Proceedings of the ACL-PASCAL Workshop on Textual Entailment and Paraphrasing","author":"Giampiccolo","year":"2007"},{"key":"2021060823401099800_bib30","article-title":"Assessing BERT\u2019s syntactic abilities","author":"Goldberg","year":"2019","journal-title":"arXiv preprint arXiv:1901.05287"},{"issue":"5\u20136","key":"2021060823401099800_bib31","doi-asserted-by":"crossref","first-page":"367","DOI":"10.1016\/j.compbiolchem.2004.09.006","article-title":"Comparing two k-category assignments by a k-category correlation coefficient","volume":"28","author":"Gorodkin","year":"2004","journal-title":"Computational Biology and Chemistry"},{"key":"2021060823401099800_bib32","article-title":"Announcing Prague Czech-English dependency treebank 2.0.","volume-title":"Proceedings of LREC","author":"Hajic","year":"2012"},{"key":"2021060823401099800_bib33","article-title":"A structural probe for finding syntax in word representations","volume-title":"Proceedings of NAACL","author":"Hewitt","year":"2019"},{"key":"2021060823401099800_bib34","doi-asserted-by":"crossref","DOI":"10.24963\/ijcai.2018\/796","article-title":"Visualisation and \u2018diagnostic classifiers\u2019 reveal how recurrent and recursive neural networks process hierarchical structure","volume-title":"Proceedings of IJCAI","author":"Hupkes","year":"2018"},{"key":"2021060823401099800_bib35","article-title":"Who did what to whom?: A contrastive study of syntacto-semantic dependencies","volume-title":"Proceedings LAW","author":"Ivanova","year":"2012"},{"key":"2021060823401099800_bib36","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2020.acl-main.197","article-title":"SMART: Robust and efficient fine-tuning for pre-trained natural language models through principled regularized optimization","volume-title":"Proceedings of ACL","author":"Jiang","year":"2020"},{"key":"2021060823401099800_bib37","article-title":"Semi- supervised classification with graph convolutional networks","volume-title":"Proceedings of ICLR","author":"Kipf","year":"2016"},{"key":"2021060823401099800_bib38","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D19-1445","article-title":"Revealing the dark secrets of BERT","volume-title":"Proceedings of EMNLP","author":"Kovaleva","year":"2019"},{"key":"2021060823401099800_bib39","article-title":"Syntactic structure distillation pretraining for bidirectional encoders","author":"Kuncoro","year":"2020","journal-title":"arXiv preprint arXiv:2005.13482"},{"key":"2021060823401099800_bib40","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/K19-2004","article-title":"SJTU-NICT at MRP 2019: Multi-task learning for end-to-end uniform semantic graph parsing","volume-title":"Proceedings of MRP","author":"Li","year":"2019"},{"key":"2021060823401099800_bib41","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/N19-1112","article-title":"Linguistic knowledge and transferability of contextual representations","volume-title":"Proceedings of NAACL","author":"Liu","year":"2019"},{"key":"2021060823401099800_bib42","article-title":"RoBERTa: A robustly optimized BERT pretraining approach","author":"Liu","year":"2019","journal-title":"arXiv preprint arXiv:1907.11692"},{"key":"2021060823401099800_bib43","article-title":"Decoupled weight decay regularization","volume-title":"Proceedings of ICLR","author":"Loshchilov","year":"2019"},{"key":"2021060823401099800_bib44","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D17-1159","article-title":"Encoding sentences with graph convolutional networks for semantic role labeling","volume-title":"Proceedings of EMNLP","author":"Marcheggiani","year":"2017"},{"issue":"2","key":"2021060823401099800_bib45","first-page":"313","article-title":"Building a large annotated corpus of English: The Penn Treebank","volume":"19","author":"Marcus","year":"1993","journal-title":"Computational Linguistics"},{"key":"2021060823401099800_bib46","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P19-1334","article-title":"Right for the wrong reasons: Diagnosing syntactic heuristics in natural language inference","volume-title":"Proceedings of ACL","author":"McCoy","year":"2019"},{"key":"2021060823401099800_bib47","doi-asserted-by":"crossref","DOI":"10.3115\/1220575.1220641","article-title":"Non-projective dependency parsing using spanning tree algorithms","volume-title":"Proceedings of NAACL","author":"McDonald","year":"2005"},{"key":"2021060823401099800_bib48","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P16-1105","article-title":"End-to-end relation extraction using LSTMs on sequences and tree structures","volume-title":"Proceedings of ACL","author":"Miwa","year":"2016"},{"key":"2021060823401099800_bib49","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/K19-2001","article-title":"MRP 2019: Cross-framework meaning representation parsing","volume-title":"Proceedings of MRP","author":"Oepen","year":"2019"},{"key":"2021060823401099800_bib50","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/S15-2153","article-title":"Semeval 2015 task 18: Broad-coverage semantic dependency parsing","volume-title":"Proceedings of SemEval","author":"Oepen","year":"2015"},{"key":"2021060823401099800_bib51","doi-asserted-by":"crossref","DOI":"10.3115\/v1\/S14-2008","article-title":"SemEval 2014 task 8: Broad-coverage semantic dependency parsing","volume-title":"Proceedings SemEval","author":"Oepen","year":"2014"},{"key":"2021060823401099800_bib52","article-title":"Discriminant-based MRS banking.","volume-title":"Proceedings of LREC","author":"Oepen","year":"2006"},{"issue":"1","key":"2021060823401099800_bib53","doi-asserted-by":"crossref","first-page":"71","DOI":"10.1162\/0891201053630264","article-title":"The proposition bank: An annotated corpus of semantic roles","volume":"31","author":"Palmer","year":"2005","journal-title":"Computational Linguistics"},{"key":"2021060823401099800_bib54","article-title":"Improving natural language inference with a pretrained parser","author":"Pang","year":"2019","journal-title":"arXiv preprint arXiv:1909.08217"},{"key":"2021060823401099800_bib55","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D16-1244","article-title":"A decomposable attention model for natural language inference","volume-title":"Proceedings of EMNLP","author":"Parikh","year":"2016"},{"key":"2021060823401099800_bib56","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P17-1186","article-title":"Deep multitask learning for semantic dependency parsing","volume-title":"Proceedings of ACL","author":"Peng","year":"2017"},{"key":"2021060823401099800_bib57","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P18-1173","article-title":"Backpropagating through structured argmax using a SPIGOT","volume-title":"Proceedings of ACL","author":"Peng","year":"2018"},{"key":"2021060823401099800_bib58","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/N18-1135","article-title":"Learning joint semantic parsers from disjoint data","volume-title":"Proceedings of NAACL","author":"Peng","year":"2018"},{"key":"2021060823401099800_bib59","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/N18-1202","article-title":"Deep contextualized word representations","volume-title":"Proceedings of NAACL","author":"Peters","year":"2018"},{"key":"2021060823401099800_bib60","volume-title":"Head-Driven Phrase Structure Grammar","author":"Pollard","year":"1994"},{"issue":"140","key":"2021060823401099800_bib61","first-page":"1","article-title":"Exploring the limits of transfer learning with a unified text-to-text transformer","volume":"21","author":"Raffel","year":"2020","journal-title":"Journal of Machine Learning Research"},{"key":"2021060823401099800_bib62","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D16-1264","article-title":"SQuAD: 100,000+ questions for machine comprehension of text","volume-title":"Proceedings of EMNLP","author":"Rajpurkar","year":"2016"},{"key":"2021060823401099800_bib63","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-319-93417-4_38","article-title":"Modeling relational data with graph convolutional networks","volume-title":"European Semantic Web Conference","author":"Schlichtkrull","year":"2018"},{"key":"2021060823401099800_bib64","article-title":"Enhanced english universal dependencies: An improved representation for natural language understanding tasks","volume-title":"Proceedings of LREC","author":"Schuster","year":"2016"},{"key":"2021060823401099800_bib65","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/W16-2209","article-title":"Linguistic input features improve neural machine translation","volume-title":"Proceedings of the First Conference on Machine Translation","author":"Sennrich","year":"2016"},{"key":"2021060823401099800_bib66","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P16-1162","article-title":"Neural machine translation of rare words with subword units","volume-title":"Proceedings ACL","author":"Sennrich","year":"2016"},{"key":"2021060823401099800_bib67","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D16-1159","article-title":"Does string-based neural MT learn source syntax?","volume-title":"Proceedings of EMNLP","author":"Shi","year":"2016"},{"key":"2021060823401099800_bib68","article-title":"Recursive deep models for semantic compositionality over a sentiment treebank","volume-title":"Proceedings of EMNLP","author":"Socher","year":"2013"},{"key":"2021060823401099800_bib69","article-title":"UDPipe 2.0 prototype at CoNLL 2018 UD shared task","volume-title":"Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies","author":"Straka","year":"2018"},{"key":"2021060823401099800_bib70","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/K19-2012","article-title":"\u00daFAL MRPipe at MRP 2019: UDPipe goes semantic in the meaning representation parsing shared task","volume-title":"Proceedings of MRP","author":"Straka","year":"2019"},{"key":"2021060823401099800_bib71","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D18-1548","article-title":"Linguistically-informed self-attention for semantic role labeling","volume-title":"Proceedings of EMNLP","author":"Strubell","year":"2018"},{"key":"2021060823401099800_bib72","doi-asserted-by":"crossref","DOI":"10.3115\/1596324.1596352","article-title":"The CoNLL 2008 shared task on joint parsing of syntactic and semantic dependencies","volume-title":"Proceedings of CoNLL","author":"Surdeanu","year":"2008"},{"key":"2021060823401099800_bib73","article-title":"Shallow syntax in deep water","author":"Swayamdipta","year":"2019","journal-title":"arXiv preprint arXiv:1908.11047"},{"key":"2021060823401099800_bib74","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D18-1412","article-title":"Syntactic scaffolds for semantic structures","volume-title":"Proceedings of EMNLP","author":"Swayamdipta","year":"2018"},{"key":"2021060823401099800_bib75","article-title":"What do you learn from context? Probing for sentence structure in contextualized word representations","volume-title":"Proceedings of ICLR","author":"Tenney","year":"2019"},{"key":"2021060823401099800_bib76","article-title":"Graph attention networks","volume-title":"Proceedings of ICLR","author":"Veli\u010dkovi\u0107","year":"2018"},{"key":"2021060823401099800_bib77","article-title":"Exploring and predicting transferability across NLP tasks","volume-title":"Proceedings of EMNLP","author":"Tu","year":"2020"},{"key":"2021060823401099800_bib78","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D19-1534","article-title":"Do NLP models know numbers? Probing numeracy in embeddings","volume-title":"Proceedings of EMNLP","author":"Wallace","year":"2019"},{"key":"2021060823401099800_bib79","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/W18-5446","article-title":"GLUE: A multi-task benchmark and analysis platform for natural language understanding","volume-title":"Proceedings of EMNLP","author":"Wang","year":"2018"},{"key":"2021060823401099800_bib80","doi-asserted-by":"crossref","first-page":"625","DOI":"10.1162\/tacl_a_00290","article-title":"Neural network acceptability judgments","volume":"7","author":"Warstadt","year":"2019","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2021060823401099800_bib81","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/N18-1101","article-title":"A broad-coverage challenge corpus for sentence understanding through inference","volume-title":"Proceedings of NAACL","author":"Williams","year":"2018"},{"key":"2021060823401099800_bib82","article-title":"Question answering on freebase via relation extraction and textual evidence","volume-title":"Proceedings of ACL","author":"Kun","year":"2016"},{"key":"2021060823401099800_bib83","article-title":"Syntax-aware opinion role labeling with dependency graph convolutional networks","volume-title":"Proceedings of ACL","author":"Bo","year":"2020"},{"key":"2021060823401099800_bib84","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v34i05.6510","article-title":"Semantics-aware BERT for language understanding","volume-title":"Proceedings of AAAI","author":"Zhang","year":"2020"},{"key":"2021060823401099800_bib85","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v34i05.6511","article-title":"Sg-net: Syntax-guided machine reading comprehension.","volume-title":"Proceedings of AAAI","author":"Zhang","year":"2020"}],"container-title":["Transactions of the Association for Computational Linguistics"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/direct.mit.edu\/tacl\/article-pdf\/doi\/10.1162\/tacl_a_00363\/1924049\/tacl_a_00363.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"http:\/\/direct.mit.edu\/tacl\/article-pdf\/doi\/10.1162\/tacl_a_00363\/1924049\/tacl_a_00363.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,12,22]],"date-time":"2022-12-22T00:35:54Z","timestamp":1671669354000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/tacl\/article\/doi\/10.1162\/tacl_a_00363\/98087\/Infusing-Finetuning-with-Semantic-Dependencies"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021]]},"references-count":85,"URL":"https:\/\/doi.org\/10.1162\/tacl_a_00363","relation":{},"ISSN":["2307-387X"],"issn-type":[{"value":"2307-387X","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2021]]},"published":{"date-parts":[[2021]]}}}