Elamur Research⌕
☰
home / papers / 2019 / Rewarding Coreference Resolvers for Being Consistent with World Knowledge

Rewarding Coreference Resolvers for Being Consistent with World Knowledge

Summary

Resolving coreference often requires world knowledge, such as knowing that "the Sunshine State" refers to Florida. Small annotated corpora rarely contain such links. The authors fine-tune a neural coreference resolver with reinforcement learning, using a knowledge-based signal instead of extra annotation. The resolver processes random Wikipedia documents, and each mention is replaced by its antecedent. Stanford OpenIE then extracts subject-relation-object triples from the rewritten text. Universal Schema models score how plausible these triples are given Wikidata and Wikipedia, and the normalized score becomes a policy-gradient reward. Three reward models, each capturing a different view of entity relatedness, produce three policies. These are merged through multi-task RL in the style of DisTraL into Coref-Distill, which is then tuned further with a distilled reward. Starting from the Lee et al. (2018) system, the method gives small but statistically significant gains on OntoNotes and WikiCoref. The work shows that freely available, multilingual knowledge bases can partly replace costly coreference annotation, which is promising for low-resource languages.

Key points

  • Three reward models were trained: RE-KG on 12M Wikidata triples (1M expanded with relation aliases), RE-Text on 2M OpenIE triples from 50,000 Wikipedia summaries, and RE-Joint on 60K Wikipedia triples grounded in Wikidata. The weight-interpolated RE-Distill reached F1 0.88, compared with 0.83 for the best single model.
  • On the CoNLL F1 metric, Coref-Distill improved over the re-trained Lee et al. (2018) baseline from 72.60 to 73.10 on OntoNotes and from 57.49 to 58.14 on WikiCoref.
  • Fine-tuning raised precision on OntoNotes (75.62 vs. 73.80) but slightly reduced recall (70.75 vs. 71.34). Spurious links and singleton clusters explain why the F1 gain is only marginal.
  • Qualitatively, the fine-tuned model detects more named-entity mentions. It also begins linking eventive noun phrases (e.g., "This attempt" with "releasing"), a pattern absent from both OntoNotes and the baseline's predictions.
  • The code is released at github.com/rahular/coref-rl. The authors argue that Wikidata's size and multilinguality make the approach portable to other languages.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Coreference resolution

Cited by

7 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 5 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/d19-1118 · arXiv 1909.02392

@inproceedings{aralikatte2019rewarding,
  title = {{Rewarding Coreference Resolvers for Being Consistent with World Knowledge}},
  author = {Aralikatte, Rahul and Lent, Heather and Gonzalez, Ana Valeria and Hershcovich, Daniel and Qiu, Chen and Sandholm, Anders and Ringaard, Michael and Søgaard, Anders},
  booktitle = {Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)},
  year = {2019},
  doi = {10.18653/v1/d19-1118},
  eprint = {1909.02392},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.18653/v1/d19-1118}
}

References

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

  1. Gabor Angeli, Melvin Jose Johnson Premkumar, and Christopher D. Manning. 2015. Leveraging linguistic structure for open domain information extraction. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Lang… find our card
  2. Sören Auer, Christian Bizer, Georgi Kobilarov, Jens Lehmann, Richard Cyganiak, and Zachary Ives. 2007. Dbpedia: A nucleus for a web of open data. In Proceedings of the 6th International The Semantic Web and 2Nd Asian Conference on Asian Semantic Web Conference, ISWC’07/ASWC’07, pages 722–735, Berlin… find
  3. Antoine Bordes, Jason Weston, Ronan Collobert, and Yoshua Bengio. 2011. Learning structured embeddings of knowledge bases. In AAAI. find
  4. Kai-Wei Chang, Wen tau Yih, Bishan Yang, and Christopher Meek. 2014. Typed tensor decomposition of knowledge bases for relation extraction. In EMNLP. find
  5. Kevin Clark and Christopher D Manning. 2016. Deep reinforcement learning for mention-ranking coreference models. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2256–2262. find our card
  6. Pascal Denis and Jason Baldridge. 2009. Global joint models for coreference resolution and named entity classification. Procesamiento del Lenguaje Natural, 42. find
  7. Ali Emami, Adam Trischler, Kaheer Suleman, and Jackie Chi Kit Cheung. 2018. A generalized knowledge hunting framework for the Winograd schema challenge. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Student Research Workshop, pa… find
  8. Abbas Ghaddar and Philippe Langlais. 2016. Wikicoref: An english coreference-annotated corpus of wikipedia articles. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016), Portorož, Slovenia. European Language Resources Association (ELRA), European Lan… find our card
  9. Alex Graves and Jürgen Schmidhuber. 2005. Frame-wise phoneme classification with bidirectional lstm and other neural network architectures. Neural Networks, 18(5-6):602–610. find
  10. Kenton Lee, Luheng He, Mike Lewis, and Luke Zettlemoyer. 2017. End-to-end neural coreference resolution. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 188–197. find our card
  11. Kenton Lee, Luheng He, and Luke Zettlemoyer. 2018. Higher-order coreference resolution with coarse-to-fine inference. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 687–… find our card
  12. Adam Lerer, Ledell Wu, Jiajun Shen, Timothee Lacroix, Luca Wehrstedt, Abhijit Bose, and Alex Peysakhovich. 2019. PyTorch-BigGraph: A Large-scale Graph Embedding System. In Proceedings of the 2nd SysML Conference, Palo Alto, CA, USA. find
  13. Yankai Lin, Zhiyuan Liu, and Maosong Sun. 2015. Modeling relation paths for representation learning of knowledge bases. In EMNLP. find
  14. Christopher D. Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven J. Bethard, and David Mc-Closky. 2014. The Stanford CoreNLP natural language processing toolkit. In Association for Computational Linguistics (ACL) System Demonstrations, pages 55–60. find our card
  15. Yuning Mao, Xiang Ren, Jiaming Shen, Xiaotao Gu, and Jiawei Han. 2018. Building a large-scale annotated chinese corpus. In ACL. find
  16. Sebastian Martschat and Michael Strube. 2014. Recall error analysis for coreference resolution. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 2070–2081, Doha, Qatar. Association for Computational Linguistics. find our card
  17. Yuanliang Meng and Anna Rumshisky. 2018. Triad-based neural network for coreference resolution. In COLING. find
  18. Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. Deep contextualized word representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technolog… find
  19. Roman Prokofyev, Alberto Tonon, Michael Luggen, Loic Vouilloz, Djellel Eddine Difallah, and Philippe Cudré-Mauroux. 2015. Sanaphor: Ontology-based coreference resolution. In International Semantic Web Conference. find
  20. Sebastian Riedel, Limin Yao, Andrew McCallum, and Benjamin M. Marlin. 2013. Relation extraction with matrix factorization and universal schemas. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 74… find
  21. Yee Teh, Victor Bapst, Wojciech M. Czarnecki, John Quan, James Kirkpatrick, Raia Hadsell, Nicolas Heess, and Razvan Pascanu. 2017. Distral: Robust multitask reinforcement learning. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neu… find
  22. Kristina Toutanova, Danqi Chen, Patrick Pantel, Hoifung Poon, Pallavi Choudhury, and Michael Gamon. 2015. Representing text for joint embedding of text and knowledge bases. In EMNLP. find
  23. Patrick Verga, David Belanger, Emma Strubell, Benjamin Roth, and Andrew McCallum. 2016. Multilingual relation extraction using compositional universal schema. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technolo… find
  24. Patrick Verga and Andrew McCallum. 2016. Row-less universal schema. In AKBC. find
  25. Jheng-Long Wu and Wei-Yun Ma. 2017. A deep learning framework for coreference resolution based on convolutional neural network. 2017 IEEE 11th International Conference on Semantic Computing (ICSC), pages 61–64. find
  26. Bishan Yang and Tom Michael Mitchell. 2017. Leveraging knowledge bases in lstms for improving machine reading. In ACL. find
  27. Xiangrong Zeng, Shizhu He, Kang Liu, and Jian Zhao. 2018. Large scaled relation extraction with reinforcement learning. In AAAI. find
  28. Hongming Zhang, Yan Song, and Yangqiu Song. 2019. Incorporating context and external knowledge for pronoun coreference resolution. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Sho… find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/aralikatte-2019-rewarding-coreference-resolvers-being-consistent/