{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,13]],"date-time":"2026-11-13T23:59:36Z","timestamp":1794614376919,"version":"4.9.1"},"reference-count":57,"publisher":"MIT Press","issue":"4","license":[{"start":{"date-parts":[[2022,8,23]],"date-time":"2022-08-23T00:00:00Z","timestamp":1661212800000},"content-version":"vor","delay-in-days":234,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/"}],"content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2022,12,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>Recent years have witnessed increasing interest in developing interpretable models in Natural Language Processing (NLP). Most existing models aim at identifying input features such as words or phrases important for model predictions. Neural models developed in NLP, however, often compose word semantics in a hierarchical manner. As such, interpretation by words or phrases only cannot faithfully explain model decisions in text classification. This article proposes a novel Hierarchical Interpretable Neural Text classifier, called HINT, which can automatically generate explanations of model predictions in the form of label-associated topics in a hierarchical manner. Model interpretation is no longer at the word level, but built on topics as the basic semantic unit. Experimental results on both review datasets and news datasets show that our proposed approach achieves text classification results on par with existing state-of-the-art text classifiers, and generates interpretations more faithful to model predictions and better understood by humans than other interpretable neural text classifiers.1<\/jats:p>","DOI":"10.1162\/coli_a_00459","type":"journal-article","created":{"date-parts":[[2022,8,23]],"date-time":"2022-08-23T19:35:09Z","timestamp":1661283309000},"page":"987-1020","update-policy":"https:\/\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":14,"title":["Hierarchical Interpretation of Neural Text Classification"],"prefix":"10.1162","volume":"48","author":[{"given":"Hanqi","family":"Yan","sequence":"first","affiliation":[{"name":"Department of Computer Science, University of Warwick, UK. Hanqi.Yan@warwick.ac.uk"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lin","family":"Gui","sequence":"additional","affiliation":[{"name":"Department of Informatics, King\u2019s College London, UK. lin.1.gui@kcl.ac.uk"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yulan","family":"He","sequence":"additional","affiliation":[{"name":"Department of Informatics, King\u2019s College London, UK University of Warwick, UK The Alan Turing Institute, UK. yulan.he@kcl.ac.uk"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"281","published-online":{"date-parts":[[2022,12,1]]},"reference":[{"key":"2022120515175091100_bib1","doi-asserted-by":"publisher","first-page":"7590","DOI":"10.18653\/v1\/2020.acl-main.679","article-title":"The sensitivity of language models and humans to Winograd schema perturbations","volume-title":"ACL","author":"Abdou","year":"2020"},{"key":"2022120515175091100_bib2","first-page":"7786","article-title":"Towards robust interpretability with self-explaining neural networks","volume-title":"NIPS","author":"Alvarez-Melis","year":"2018"},{"key":"2022120515175091100_bib3","doi-asserted-by":"publisher","first-page":"169","DOI":"10.1162\/tacl_a_00261","article-title":"SECTOR: A neural model for coherent topic segmentation and classification","volume":"7","author":"Arnold","year":"2019","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2022120515175091100_bib4","doi-asserted-by":"publisher","first-page":"11396","DOI":"10.1609\/aaai.v35i13.17358","article-title":"Explaining a black-box by using a deep variational information bottleneck approach","volume-title":"Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference on Innovative Applications of Artificial Intelligence, IAAI 2021, The Eleventh Symposium on Educational Advances in Artificial Intelligence, EAAI 2021","author":"Bang","year":"2021"},{"key":"2022120515175091100_bib5","doi-asserted-by":"publisher","first-page":"2963","DOI":"10.18653\/v1\/P19-1284","article-title":"Interpretable neural predictions with differentiable binary variables","volume-title":"Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Volume 1: Long Papers","author":"Bastings","year":"2019"},{"key":"2022120515175091100_bib6","first-page":"993","article-title":"Latent Dirichlet allocation","volume":"3","author":"Blei","year":"2003","journal-title":"JMLR"},{"key":"2022120515175091100_bib7","first-page":"1877","article-title":"Language models are few-shot learners","volume-title":"NIPS","author":"Brown","year":"2020"},{"key":"2022120515175091100_bib8","doi-asserted-by":"publisher","first-page":"2031","DOI":"10.18653\/v1\/P18-1189","article-title":"Neural models for documents with metadata","volume-title":"ACL","author":"Card","year":"2018"},{"issue":"1","key":"2022120515175091100_bib9","doi-asserted-by":"crossref","first-page":"419","DOI":"10.1609\/icwsm.v6i1.14321","article-title":"Visualizing topic models","volume":"6","author":"Chaney","year":"2021","journal-title":"Proceedings of the International AAAI Conference on Web and Social Media"},{"key":"2022120515175091100_bib10","doi-asserted-by":"publisher","first-page":"4236","DOI":"10.18653\/v1\/2020.emnlp-main.347","article-title":"Learning variational word masks to improve the interpretability of neural text classifiers","volume-title":"EMNLP","author":"Chen","year":"2020"},{"key":"2022120515175091100_bib11","doi-asserted-by":"publisher","first-page":"5578","DOI":"10.18653\/v1\/2020.acl-main.494","article-title":"Generating hierarchical explanations on text classification via feature interaction detection","volume-title":"Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL","author":"Chen","year":"2020"},{"key":"2022120515175091100_bib12","first-page":"882","article-title":"Learning to explain: An information-theoretic perspective on model interpretation","volume-title":"ICML, volume 80 of Proceedings of Machine Learning Research","author":"Chen","year":"2018"},{"key":"2022120515175091100_bib13","doi-asserted-by":"publisher","first-page":"3143","DOI":"10.18653\/v1\/2020.acl-main.286","article-title":"Towards interpretable clinical diagnosis with Bayesian Network Ensembles stacked on entity-aware CNNs","volume-title":"ACL","author":"Chen","year":"2020"},{"key":"2022120515175091100_bib14","doi-asserted-by":"publisher","first-page":"120","DOI":"10.1145\/3287560.3287572","article-title":"Bias in bios: A case study of semantic representation bias in a high-stakes setting","volume-title":"Proceedings of the Conference on Fairness, Accountability, and Transparency","author":"De-Arteaga","year":"2019"},{"key":"2022120515175091100_bib15","first-page":"4171","article-title":"BERT: Pre-training of deep bidirectional transformers for language understanding","volume-title":"NAACL","author":"Devlin","year":"2019"},{"key":"2022120515175091100_bib16","doi-asserted-by":"publisher","first-page":"4443","DOI":"10.18653\/v1\/2020.acl-main.408","article-title":"ERASER: A benchmark to evaluate rationalized NLP models","volume-title":"Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020","author":"DeYoung","year":"2020"},{"key":"2022120515175091100_bib17","first-page":"2454","article-title":"Towards a deep and unified understanding of deep neural models in NLP","volume-title":"ICML, volume 97 of Proceedings of Machine Learning Research","author":"Guan","year":"2019"},{"key":"2022120515175091100_bib18","doi-asserted-by":"publisher","first-page":"102160","DOI":"10.1016\/j.artmed.2021.102160","article-title":"Understanding patient reviews with minimum supervision","volume":"120","author":"Gui","year":"2021","journal-title":"Artificial Intelligence in Medicine"},{"issue":"4","key":"2022120515175091100_bib19","doi-asserted-by":"publisher","first-page":"1915","DOI":"10.1109\/TKDE.2020.2999489","article-title":"Multi task mutual learning for joint sentiment classification and topic detection","volume":"34","author":"Gui","year":"2022","journal-title":"IEEE Transactions on Knowledge and Data Engineering"},{"key":"2022120515175091100_bib20","doi-asserted-by":"publisher","first-page":"4198","DOI":"10.18653\/v1\/2020.acl-main.386","article-title":"Towards faithfully interpretable NLP systems: How should we define and evaluate faithfulness?","volume-title":"ACL","author":"Jacovi","year":"2020"},{"key":"2022120515175091100_bib21","doi-asserted-by":"publisher","first-page":"3543","DOI":"10.18653\/v1\/N19-1357","article-title":"Attention is not explanation","volume-title":"Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Volume 1 (Long and Short Papers)","author":"Jain","year":"2019"},{"key":"2022120515175091100_bib22","doi-asserted-by":"publisher","first-page":"3651","DOI":"10.18653\/v1\/P19-1356","article-title":"What does BERT learn about the structure of language?","volume-title":"ACL","author":"Jawahar","year":"2019"},{"key":"2022120515175091100_bib23","doi-asserted-by":"publisher","first-page":"3193","DOI":"10.18653\/v1\/2020.emnlp-main.258","article-title":"Cold-start and interpretability: Turning regular expressions into trainable recurrent neural networks","volume-title":"EMNLP","author":"Jiang","year":"2020"},{"key":"2022120515175091100_bib24","article-title":"Towards hierarchical importance attribution: Explaining compositional semantics for neural sequence models","volume-title":"8th International Conference on Learning Representations, ICLR 2020","author":"Jin","year":"2020"},{"key":"2022120515175091100_bib25","first-page":"3020","article-title":"Learning representations for counterfactual inference","volume-title":"ICML","author":"Johansson","year":"2016"},{"key":"2022120515175091100_bib26","doi-asserted-by":"publisher","first-page":"3154","DOI":"10.18653\/v1\/2020.emnlp-main.255","article-title":"Interpretation of NLP models through input marginalization","volume-title":"EMNLP","author":"Kim","year":"2020"},{"key":"2022120515175091100_bib27","article-title":"Adam: A method for stochastic optimization","volume-title":"3rd International Conference on Learning Representations, ICLR 2015, Conference Track Proceedings","author":"Kingma","year":"2015"},{"key":"2022120515175091100_bib28","article-title":"Auto-encoding variational Bayes","volume-title":"2nd International Conference on Learning Representations, ICLR 2014, Conference Track Proceedings","author":"Kingma","year":"2014"},{"key":"2022120515175091100_bib29","doi-asserted-by":"publisher","first-page":"29","DOI":"10.1145\/3287560.3287590","article-title":"On human predictions with explanations and predictions of machine learning models: A case study on deception detection","volume-title":"Proceedings of the Conference on Fairness, Accountability, and Transparency","author":"Lai","year":"2019"},{"key":"2022120515175091100_bib30","doi-asserted-by":"publisher","first-page":"107","DOI":"10.18653\/v1\/D16-1011","article-title":"Rationalizing neural predictions","volume-title":"EMNLP","author":"Lei","year":"2016"},{"key":"2022120515175091100_bib31","article-title":"Understanding neural networks through representation erasure","author":"Li","year":"2016","journal-title":"CoRR"},{"key":"2022120515175091100_bib32","doi-asserted-by":"publisher","first-page":"2744","DOI":"10.18653\/v1\/D19-1276","article-title":"Specializing word embeddings (for parsing) by information bottleneck","volume-title":"Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, EMNLP-IJCNLP 2019","author":"Li","year":"2019"},{"issue":"6","key":"2022120515175091100_bib33","doi-asserted-by":"publisher","first-page":"1134","DOI":"10.1109\/TKDE.2011.48","article-title":"Weakly supervised joint sentiment-topic detection from text","volume":"24","author":"Lin","year":"2012","journal-title":"IEEE Transactions on Knowledge and Data Engineering"},{"issue":"10","key":"2022120515175091100_bib34","doi-asserted-by":"publisher","first-page":"36","DOI":"10.1145\/3233231","article-title":"The mythos of model interpretability","volume":"61","author":"Lipton","year":"2018","journal-title":"Communications of the ACM"},{"key":"2022120515175091100_bib35","first-page":"142","article-title":"Learning word vectors for sentiment analysis","volume-title":"ACL-HLT","author":"Maas","year":"2011"},{"key":"2022120515175091100_bib36","doi-asserted-by":"crossref","first-page":"1020","DOI":"10.1109\/ICDM.2012.110","article-title":"Learning attitudes and attributes from multi-aspect reviews","volume-title":"2012 IEEE 12th International Conference on Data Mining","author":"McAuley","year":"2012"},{"key":"2022120515175091100_bib37","doi-asserted-by":"publisher","first-page":"8538","DOI":"10.18653\/v1\/2020.acl-main.755","article-title":"Evaluating robustness to input perturbations for neural machine translation","volume-title":"ACL","author":"Niu","year":"2020"},{"key":"2022120515175091100_bib38","doi-asserted-by":"publisher","first-page":"9","DOI":"10.1145\/1651461.1651464","article-title":"Topic-dependent sentiment analysis of financial blogs","volume-title":"Proceedings of the 1st International CIKM Workshop on Topic-Sentiment Analysis for Mass Opinion, TSA \u201909","author":"O\u2019Hare","year":"2009"},{"key":"2022120515175091100_bib39","doi-asserted-by":"publisher","first-page":"4782","DOI":"10.18653\/v1\/2020.acl-main.432","article-title":"Learning to deceive with attention-based explanations","volume-title":"Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020","author":"Pruthi","year":"2020"},{"key":"2022120515175091100_bib40","doi-asserted-by":"publisher","first-page":"4902","DOI":"10.18653\/v1\/2020.acl-main.442","article-title":"Beyond accuracy: Behavioral testing of NLP models with checklist","volume-title":"ACL","author":"Ribeiro","year":"2020"},{"key":"2022120515175091100_bib41","first-page":"8116","article-title":"Interpretations are useful: Penalizing explanations to align neural networks with prior knowledge","volume-title":"ICML","author":"Rieger","year":"2020"},{"issue":"2","key":"2022120515175091100_bib42","doi-asserted-by":"publisher","first-page":"336","DOI":"10.1007\/s11263-019-01228-7","article-title":"Grad-CAM: Visual explanations from deep networks via gradient-based localization","volume":"128","author":"Selvaraju","year":"2020","journal-title":"IJCV"},{"key":"2022120515175091100_bib43","doi-asserted-by":"publisher","first-page":"2931","DOI":"10.18653\/v1\/P19-1282","article-title":"Is attention interpretable?","volume-title":"Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Volume 1: Long Papers","author":"Serrano","year":"2019"},{"key":"2022120515175091100_bib44","article-title":"Hierarchical interpretations for neural network predictions","volume-title":"ICLR","author":"Singh","year":"2019"},{"key":"2022120515175091100_bib45","doi-asserted-by":"publisher","first-page":"169","DOI":"10.18653\/v1\/2020.acl-srw.23","article-title":"Exploring interpretability in event extraction: Multitask learning of a neural event classifier and an explanation decoder","volume-title":"ACL","author":"Tang","year":"2020"},{"key":"2022120515175091100_bib46","first-page":"3959","article-title":"Learning dynamic hierarchical topic graph with graph convolutional network for document classification","volume-title":"Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, volume 108 of Proceedings of Machine Learning Research","author":"Wang","year":"2020"},{"key":"2022120515175091100_bib47","doi-asserted-by":"publisher","first-page":"11","DOI":"10.18653\/v1\/D19-1002","article-title":"Attention is not not explanation","volume-title":"Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, EMNLP-IJCNLP 2019","author":"Wiegreffe","year":"2019"},{"key":"2022120515175091100_bib48","doi-asserted-by":"publisher","first-page":"4166","DOI":"10.18653\/v1\/2020.acl-main.383","article-title":"Perturbed masking: Parameter-free probing for analyzing and interpreting BERT","volume-title":"ACL","author":"Wu","year":"2020"},{"key":"2022120515175091100_bib49","doi-asserted-by":"publisher","first-page":"3055","DOI":"10.1145\/3442381.3450045","article-title":"Graph topic neural network for document representation","volume-title":"Proceedings of the Web Conference 2021, WWW \u201921","author":"Xie","year":"2021"},{"key":"2022120515175091100_bib50","doi-asserted-by":"publisher","first-page":"3364","DOI":"10.18653\/v1\/2021.acl-long.261","article-title":"Position bias mitigation: A knowledge-aware graph model for emotion cause extraction","volume-title":"Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, ACL\/IJCNLP 2021, (Volume 1: Long Papers)","author":"Yan","year":"2021"},{"key":"2022120515175091100_bib51","first-page":"5754","article-title":"XLNet: Generalized autoregressive pretraining for language understanding","volume-title":"NIPS","author":"Yang","year":"2019"},{"key":"2022120515175091100_bib52","doi-asserted-by":"publisher","first-page":"1480","DOI":"10.18653\/v1\/N16-1174","article-title":"Hierarchical attention networks for document classification","volume-title":"NAACL","author":"Yang","year":"2016"},{"key":"2022120515175091100_bib53","first-page":"12822","article-title":"Understanding interlocking dynamics of cooperative rationalization","volume-title":"Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021, NeurIPS 2021","author":"Yu","year":"2021"},{"key":"2022120515175091100_bib54","doi-asserted-by":"publisher","first-page":"256","DOI":"10.18653\/v1\/2020.emnlp-main.18","article-title":"KERMIT: Complementing transformer architectures with encoders of explicit syntactic interpretations","volume-title":"EMNLP","author":"Zanzotto","year":"2020"},{"key":"2022120515175091100_bib55","doi-asserted-by":"publisher","first-page":"8045","DOI":"10.18653\/v1\/2020.acl-main.717","article-title":"Learning interpretable relationships between entities, relations and concepts via Bayesian structure learning on open domain facts","volume-title":"ACL","author":"Zhang","year":"2020"},{"key":"2022120515175091100_bib56","doi-asserted-by":"publisher","first-page":"846","DOI":"10.18653\/v1\/2020.acl-main.78","article-title":"Interpretable operational risk classification with semi-supervised variational autoencoder","volume-title":"ACL","author":"Zhou","year":"2020"},{"key":"2022120515175091100_bib57","first-page":"6803","article-title":"Towards interpretable natural language understanding with explanations as latent variables","volume-title":"Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020","author":"Zhou","year":"2020"}],"container-title":["Computational Linguistics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/direct.mit.edu\/coli\/article-pdf\/48\/4\/987\/2061762\/coli_a_00459.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/direct.mit.edu\/coli\/article-pdf\/48\/4\/987\/2061762\/coli_a_00459.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,12,5]],"date-time":"2022-12-05T15:19:08Z","timestamp":1670253548000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/coli\/article\/48\/4\/987\/112768\/Hierarchical-Interpretation-of-Neural-Text"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022]]},"references-count":57,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2022,12,1]]},"published-print":{"date-parts":[[2022,12,1]]}},"URL":"https:\/\/doi.org\/10.1162\/coli_a_00459","relation":{},"ISSN":["0891-2017","1530-9312"],"issn-type":[{"value":"0891-2017","type":"print"},{"value":"1530-9312","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2022]]},"published":{"date-parts":[[2022]]}}}