Elamur Research⌕
☰
home / papers / 2018 / Neural Cross-Lingual Coreference Resolution And Its Application To Entity Linking

Neural Cross-Lingual Coreference Resolution And Its Application To Entity Linking

Summary

The paper asks whether a coreference resolver trained only on English can work on other languages without annotations in those languages. The authors build an entity-centric neural model that gradually merges partial entities. It first merges names with names, then nominals with names in the same sentence, and finally nominals across sentences. Each candidate entity pair is represented by combining mention-pair vectors. These vectors are built from language-independent features (string match, acronym, distance, mention and entity types, first-name mismatch, speaker cues) plus averaged word embeddings. An attention layer weights the mention pairs according to their name/nominal type combination. Shared multilingual embeddings come from monolingual CBOW word2vec vectors, with Chinese and Spanish projected into the English space using the regression approach of Mikolov et al. (2013b). On the TAC 2015 Tri-Lingual EDL data, the English-trained model nearly matches models trained on Spanish and Chinese. Plugged into the zero-shot entity linker of Sil et al. (2018), it beats the top 2015 TAC system. This shows that both coreference and linking can be transferred across languages without target-language supervision.

Key points

  • Data: TAC 2015 Tri-Lingual EL dataset (newswire and discussion forums; types PER, ORG, GPE, LOC, FAC).
    • Training documents: En 168, Es 129, Zh 147.
    • Test documents: 167, 167 and 166 respectively.
  • English baseline: On the En test set, the model reaches a CoNLL score of 85.2, versus 77.2 for Clark and Manning (2016). Both systems were run with gold mention boundaries.
  • Cross-lingual coreference: The English-trained model scores only 0.5 CoNLL points below the Spanish-trained model on Es. It is 0.3 points below the Chinese-trained model on Zh.
  • Entity linking accuracy: EL + En Coref reaches 81.1 on Es and 83.9 on Zh.
    • This beats TAC 2015 Rank 1 (80.4 / 83.1).
    • It also beats linking without coreference (78.1 / 81.3).
  • Comparison with S IL 18: The reported S IL 18 results (82.3 / 84.4) rely on in-language coreference training data. The fully zero-shot pipeline comes within 1.2 points of them on Es and 0.5 points on Zh.

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: Named entity recognition · Coreference resolution

Cited by

14 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 10 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/p18-2063 · arXiv 1806.10201

@inproceedings{kundu2018neural,
  title = {{Neural Cross-Lingual Coreference Resolution And Its Application To Entity Linking}},
  author = {Kundu, Gourab and Sil, Avi and Florian, Radu and Hamza, Wael},
  booktitle = {Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)},
  year = {2018},
  doi = {10.18653/v1/p18-2063},
  eprint = {1806.10201},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.18653/v1/p18-2063}
}

References

31 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Tomas Mikolov, Quoc V Le, and Ilya Sutskever. 2013b. Exploiting similarities among languages for machine translation. arXiv preprint arXiv:1309.4168. find
  2. Waleed Ammar, George Mulcaire, Yulia Tsvetkov, Guillaume Lample, Chris Dyer, and Noah A Smith. 2016. Massively multilingual word embeddings. arXiv preprint arXiv:1602.01925. find our card
  3. Vincent Ng and Claire Cardie. 2002. Improving machine learning approaches to coreference resolution. In ACL. find our card
  4. Eric Bengtson and Dan Roth. 2008. Understanding the value of features for coreference resolution. In EMNLP. find our card
  5. Jian Ni, Georgiana Dinu, and Radu Florian. 2017. Weakly supervised cross-lingual named entity recognition via effective annotation and representation projection. In ACL. find
  6. Anders Björkelund and Jonas Kuhn. 2014. Learning structured perceptrons for coreference resolution with latent antecedents and non-local features. In ACL. find our card
  7. Mark Palatucci, Dean Pomerleau, Geoffrey E Hinton, and Tom M Mitchell. 2009. Zero-shot learning with semantic output codes. In NIPS. find
  8. Kevin Clark and Christopher D Manning. 2015. Entity-centric coreference resolution with model stacking. In ACL. find our card
  9. Sameer Pradhan, Alessandro Moschitti, Nianwen Xue, Olga Uryupina, and Yuchen Zhang. 2012. Conll-2012 shared task: Modeling multilingual unrestricted coreference in ontonotes. In EMNLP-CoNLL. find our card
  10. Kevin Clark and Christopher D Manning. 2016. Improving coreference resolution by learning entity-level distributed representations. In ACL. find our card
  11. Greg Durrett, David Leo Wright Hall, and Dan Klein. 2013. Decentralized entity-level modeling for coreference resolution. In ACL. find our card
  12. Sameer Pradhan, Lance Ramshaw, Mitchell Marcus, Martha Palmer, Ralph Weischedel, and Nianwen Xue. 2011. Conll-2011 shared task: Modeling unrestricted coreference in ontonotes. In CoNLL. find our card
  13. Greg Durrett and Dan Klein. 2014. A joint model for entity analysis: Coreference, typing, and linking. Transactions of the Association for Computational Linguistics, 2. find our card
  14. Karthik Raghunathan, Heeyoung Lee, Sudarshan Rangarajan, Nathanael Chambers, Mihai Surdeanu, Dan Jurafsky, and Christopher Manning. 2010. A multi-pass sieve for coreference resolution. In EMNLP. find our card
  15. Manaal Faruqui and Chris Dyer. 2014. Improving vector space word representations using multilingual correlation. In EACL. find
  16. Altaf Rahman and Vincent Ng. 2009. Supervised models for coreference resolution. In EMNLP. find our card
  17. Avirup Sil, Georgiana Dinu, and Radu Florian. 2015. The ibm systems for trilingual entity discovery and linking at tac 2015. In TAC. find
  18. Eraldo Rezende Fernandes, Cı́cero Nogueira Dos Santos, and Ruy Luiz Milidiú. 2012. Latent structure perceptron with feature induction for unrestricted coreference resolution. In EMNLP-CoNLL. find our card
  19. Avirup Sil and Radu Florian. 2016. One for all: Towards language independent named entity linking. In ACL. find
  20. Hannaneh Hajishirzi, Leila Zilles, Daniel S Weld, and Luke Zettlemoyer. 2013. Joint coreference resolution and named-entity linking with multi-pass sieves. In EMNLP. find our card
  21. Avirup Sil, Gourab Kundu, Radu Florian, and Wael Hamza. 2018. Neural cross-lingual entity linking. In AAAI. find
  22. Heng Ji, Joel Nothman, Ben Hachey, and Radu Florian. 2015. Overview of tac-kbp2015 tri-lingual entity discovery and linking. In TAC. find
  23. Richard Socher, Milind Ganjoo, Christopher D Manning, and Andrew Ng. 2013. Zero-shot learning through cross-modal transfer. In NIPS. find
  24. Heng Ji, Joel Nothman, Ben Hachey, et al. 2014. Overview of tac-kbp2014 entity discovery and linking tasks. In TAC. find
  25. Chen-Tse Tsai and Dan Roth. 2016. Cross-lingual wikification using multilingual embeddings. In HLT-NAACL. find
  26. Kenton Lee, Luheng He, Mike Lewis, and Luke Zettlemoyer. 2017. End-to-end neural coreference resolution. arXiv preprint arXiv:1707.07045. find our card
  27. Shyam Upadhyay, Manaal Faruqui, Chris Dyer, and Dan Roth. 2016. Cross-lingual models of word embeddings: An empirical comparison. In ACL. find
  28. Sebastian Martschat and Michael Strube. 2015. Latent structures for coreference resolution. Transactions of the Association for Computational Linguistics, 3:405–418. find our card
  29. Sam Wiseman, Alexander M Rush, and Stuart M Shieber. 2016. Learning global features for coreference resolution. In NAACL. find our card
  30. Sam Joshua Wiseman, Alexander Matthew Rush, Stuart Merrill Shieber, and Jason Weston. 2015. Learning anaphoricity and antecedent ranking features for coreference resolution. In ACL. find our card
  31. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013a. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/kundu-2018-neural-cross-lingual-coreference-resolution/