Elamur Research
☰
home / papers / 2020 / Bridging Anaphora Resolution as Question Answering

Bridging Anaphora Resolution as Question Answering

What it is about

The paper treats bridging anaphora resolution as question answering. Each anaphor is rephrased as a question such as "residents of what?" from its head and premodifiers, and a BERT-based span model (BARQA) finds the antecedent in the preceding context. The model needs no gold mentions. The authors also generate about 2.8 million "quasi-bridging" training pairs from parsed Gigaword: an "X of Y" or possessive NP gives an anaphor X and an antecedent Y in an earlier sentence. On a random sample of 100 pairs, 25% are rated fully acceptable and 37% acceptable but unnatural, and 38% are unacceptable. On ISNotes, the best setting (quasi-bridging pre-training, then BASHI fine-tuning) reaches 47.21% lenient accuracy. Using gold mentions for candidates, the strict accuracy is 50.08%, the new state of the art at the time.

Ideas

For Mova

The chapter is languages/en/35-reference-and-coreference.md. The code is the coref crate, mainly coref/src/sieve.rs and coref/src/mention.rs, with the UD parse from en/src/parse.rs. The measure is accuracy on bridging anaphors with the lenient and strict splits. ISNotes and BASHI are used only for evaluation, and their licences should be checked first. The BERT model is not for Mova. The question rewriting and the quasi-bridging data generation are the parts worth copying.

Library summary

The Research library's own summary of the paper, in our words — not the paper's text.

Written for the Research library (claude-opus-5-5) from the paper text. Original — aclanthology.org/2020.acl-main.132.pdf

Summary

The paper addresses bridging anaphora resolution: finding the antecedent of an anaphor that is linked to an earlier entity indirectly (e.g., "residents" referring to a set of buildings), not through coreference. Earlier systems scored anaphor–candidate pairs, focused mostly on noun-phrase semantics, and usually relied on gold mentions. Hou reformulates the task as span-based reading comprehension. Each anaphor becomes a short question such as "residents of what?", and a BERT-based QA model, BARQA, extracts the antecedent span from the preceding context. Because the model reads the whole context, it needs no candidate list. To compensate for the lack of large referential bridging corpora, the author builds a large automatically generated "quasi-bridging" dataset from parsed Gigaword. This dataset works better for pre-training than SQuAD. After pre-training on it and fine-tuning on a small in-domain set, BARQA sets new state-of-the-art results on ISNotes and BASHI. The work shows that QA framing offers a practical, mention-free way to perform context-sensitive bridging inference.

Key points

  • On BASHI, the earlier deterministic method of Hou (2018a) reached 29.94% strict accuracy. BARQA reached 32.27% using no mention information, 37.79% under lenient scoring, and 38.66% when system mentions and semantics were added.
  • On the ARRAU RST Test set, which mostly contains lexical bridging, BARQA trained on about 2,000 QA pairs from RST Train reached 34.59%, compared with 32.39% for the earlier method.
  • In the error analysis on ISNotes, BARQA resolved 36.36% of 407 context-dependent anaphors, versus 19.90% for MLN II + emb. However, it was weaker on 256 world-knowledge cases (71.88% vs. 88.28%).
  • Bridging often crosses sentence boundaries. In ISNotes, only about 26% of antecedents appear in the same sentence as the anaphor, and 23% are more than two sentences away.
  • The code and the QA datasets for ISNotes and BASHI (in SQuAD json format) are released at github.com/IBM/bridging-resolution.

Topics: Coreference resolution

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/2020.acl-main.132 · arXiv 2004.07898

@inproceedings{hou2020bridging,
  title = {{Bridging Anaphora Resolution as Question Answering}},
  author = {Hou, Yufang},
  booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
  year = {2020},
  doi = {10.18653/v1/2020.acl-main.132},
  eprint = {2004.07898},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.18653/v1/2020.acl-main.132}
}

References

35 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Minneapolis, USA, 2–7 June 2019, pages 4171–4186. find
  2. Kerstin Eckart, Arndt Riester, and Katrin Schweitzer. 2012. A discourse information radio news database for linguistic analysis. In Christian Chiarcos, Sebastian Nordhoff, and Sebastian Hellmann, editors, Linked Data in Linguistics, pages 65–76. Springer Berlin Heidelberg. find
  3. Kari Fraurud. 1990. Definiteness and the processing of noun phrases in natural discourse. Journal of Semantics, 7:395–433. find
  4. Claire Gardent and Hélène Manuélian. 2005. Création d’un corpus annoté pour le traitement des descriptions définies. Traitement Automatique des Langues, 46(1):115–140. find
  5. Matt Gardner, Jonathan Berant, Hannaneh Hajishirzi, Alon Talmor, and Min Sewon. 2019. Question answering is a format; when is it useful? arXiv preprint arXiv:909.11291. find
  6. Yufang Hou. 2016. Incremental fine-grained information status classification using attention-based LSTMs. In Proceedings of the 26th International Conference on Computational Linguistics, Osaka, Japan, 11–16 December 2016, pages 1880–1890. find
  7. Yufang Hou. 2018a. A deterministic algorithm for bridging anaphora resolution. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, 31 October– 4 November 2018, pages 1938–1948. find
  8. Entity-relation extraction as multi-turn question answering. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, 28 July–2 August 2019, pages 1340– 1350. find
  9. Yufang Hou. 2018b. Enhanced word representations for bridging anaphora resolution. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, New Orleans, Louisiana, 1–6 June 2018, pages 1–7. find
  10. Katja Markert, Yufang Hou, and Michael Strube. 2012. Collective classification for fine-grained information status. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics, Jeju Island, Korea, 8–14 July 2012, pages 795–804. find our card
  11. Yufang Hou, Katja Markert, and Michael Strube. 2013a. Cascading collective classification for bridging anaphora recognition using a rich linguistic feature set. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, Seattle, Wash., 18–21 October 2013, pages 814– 820. find
  12. Yufang Hou, Katja Markert, and Michael Strube. 2013b. Global inference for bridging anaphora resolution. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Atlanta, Georgia, 9–14 June 2013, pages 907–917. find our card
  13. Yufang Hou, Katja Markert, and Michael Strube. 2014. A rule-based system for unrestricted bridging resolution: Recognizing bridging anaphora and finding links to antecedents. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, Doha, Qatar, 25–29 October 2014, pages 2082–2093. find
  14. Yufang Hou, Katja Markert, and Michael Strube. 2018. Unrestricted bridging resolution. Computational Linguistics, 44(2):237–284. find our card
  15. Mandar Joshi, Danqi Chen, Yinhan Liu, Daniel S. Weld, Luke Zettlemoyer, and Omer Levy. 2019. Spanbert: Improving pre-training by representing and predicting spans. arXiv preprint arXiv:1907.10529. ↗ our card
  16. Vid Kocijan, Ana-Maria Cretu, Oana-Maria Camburu, Yordan Yordanov, and Thomas Lukasiewicz. 2019. A surprisingly robust trick for the Winograd Schema Challenge. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, 28 July–2 August 2019, pages 4837– 4842. find
  17. Emmanuel Lassalle and Pascal Denis. 2011. Leveraging different meronym discovery methods for bridging resolution in French. In Proceedings of the 8th Discourse Anaphora and Anaphor Resolution Colloquium (DAARC 2011), Faro, Algarve, Portugal, 6–7 October 2011, pages 35–46. find
  18. Xiaoya Li, Fan Yin, Zijun Sun, Xiayu Li, Arianna Yuan, Duo Chai, Mingxin Zhou, and Jiwei Li. 2019. find
  19. Bryan McCann, Nitish Shirish Keskar, Caiming Xiong, and Richard Socher. 2018. The natural language decathlon: Multitask learning as question answering. arXiv preprint arXiv:1806.08730. ↗ our card
  20. Courtney Napoles, Matthew Gormley, and Benjamin Van Durme. 2012. Annotated Gigaword. In Proceedings of the Joint Workshop on Automatic Knowledge Base Construction & Web-scale Knowledge Extraction (AKBC-WEKEX) Montréal, Québec, Canada, 7-8 June 2012, pages 95–100. find
  21. Anna Nedoluzhko, Jiřı́ Mı́rovskỳ, and Petr Pajas. 2009. The coding scheme for annotating extended nominal coreference and bridging anaphora in the Prague dependency treebank. In Proceedings of the Third Linguistic Annotation Workshop at ACL-IJCNLP 2009, Suntec, Singapore, 6–7 August 2009, pages 108– 111. find
  22. Malvina Nissim, Shipara Dingare, Jean Carletta, and Mark Steedman. 2004. An annotation scheme for information status in dialogue. In Proceedings of the 4th International Conference on Language Resources and Evaluation, Lisbon, Portugal, 26–28 May 2004, pages 1023–1026. find
  23. Juri Opitz and Anette Frank. 2018. Addressing the Winograd Schema Challenge as a sequence ranking task. In Proceedings of the First International Workshop on Language Cognition and Computational Models, pages 41–52. find
  24. Robert Parker, David Graff, Junbo Kong, Ke Chen, and Kazuaki Maeda. 2011. English Gigaword Fifth Edition. LDC2011T07. find
  25. Massimo Poesio. 2004. The MATE/GNOME proposals for anaphoric annotation, revisited. In Proceedings of the 5th SIGdial Workshop on Discourse and Dialogue, Cambridge, Mass., 30 April – 1 May 2004, pages 154–162. find
  26. Massimo Poesio, Yulia Grishina, Varada Kolhatkar, Nafise Sadat Moosavi, Ina Rösiger, Adam Roussel, Fabian Simonjetz, Alexandra Uma, Olga Uryupina, Juntao Yu, and Heike Zinsmeister. 2018. Anaphora resolution with the ARRAU corpus. In Proceedings of the Workshop on Computational Models of Reference, Anaphora and Coreference. New Orleans, Louisiana, June 6, 2018, pages 11–22. find
  27. Massimo Poesio, Rahul Mehta, Axel Maroudas, and Janet Hitzeman. 2004. Learning to resolve bridging references. In Proceedings of the 42nd Annual Meeting of the Association for Computational Linguistics, Barcelona, Spain, 21–26 July 2004, pages 143–150. find
  28. Massimo Poesio and Renata Vieira. 1998. A corpus-based investigation of definite description use. Computational Linguistics, 24(2):183–216. find
  29. Massimo Poesio, Renata Vieira, and Simone Teufel. 1997. Resolving bridging references in unrestricted text. In Proceedings of the ACL Workshop on Operational Factors in Practical, Robust Anaphora Resolution for Unrestricted Text, Madrid, Spain, July 1997, pages 1–6. find
  30. Altaf Rahman and Vincent Ng. 2012. Resolving complex cases of definite pronouns: The Winograd Schema Challenge. In Proceedings of the 2012 Conference on Empirical Methods in Natural Language Processing and Natural Language Learning, Jeju Island, Korea, 12–14 July 2012, pages 777–789. find our card
  31. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. SQuAD: 100, 000+ questions for machine comprehension of text. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, Austin, Texas, USA, 1–4 November 2016, pages 2383–2392. find our card
  32. Ina Rösiger. 2018. BASHI: A corpus of wall street journal articles annotated with bridging links. In Proceedings of the 11th International Conference on Language Resources and Evaluation, Miyazaki, Japan, 7–12 May 2018, pages 382–388. find
  33. Ina Rösiger, Arndt Riester, and Jonas Kuhn. 2018. Bridging resolution: Task definition, corpus resources and rule-based experiments. In Proceedings of the 27th International Conference on Computational Linguistics, Santa Fe, New-Mexico, USA, 20– 26 August 2018, pages 3516–3528. find our card
  34. Ralph Weischedel, Martha Palmer, Mitchell Marcus, Eduard Hovy, Sameer Pradhan, Lance Ramshaw, Nianwen Xue, Ann Taylor, Jeff Kaufman, Michelle Franchini, Mohammed El-Bachouti, Robert Belvin, and Ann Houston. 2011. OntoNotes release 4.0. LDC2011T03, Philadelphia, Penn.: Linguistic Data Consortium. find
  35. Wei Wu, Fei Wang, Arianna Yuan, Fei Wu, and Jiwei Li. 2020. Coreference resolution as query-based span prediction. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, Seattle, Wash., 5–10 July 2020. find