Elamur Research⌕
☰
home / papers / 2019 / The Referential Reader: A Recurrent Entity Network for Anaphora Resolution

The Referential Reader: A Recurrent Entity Network for Anaphora Resolution

Summary

Most coreference systems use the mention-pair approach, which scores every pair of spans. This costs a lot of computation and annotated data, and it does not match how people read, which is roughly word by word. The authors propose the Referential Reader, which processes text strictly left to right and keeps a small fixed set of memory cells for entities. At each token, learned gates decide whether to update a cell (marking coreference with what it already holds), overwrite the least salient cell with a new entity, or skip the token. Salience grows with reuse and fades over time. Because all memory operations are differentiable, the model trains end-to-end on gold pronoun–name links. Its recurrent state also allows next-word prediction, so it can be pretrained as a language model on unlabeled text. On the GAP benchmark, the incremental model beats strong mention-pair and rule-based baselines. This shows that online, memory-based anaphora resolution is practical and can benefit from semi-supervised learning.

Key points

  • Evaluation and pretraining data: Evaluated on GAP (Webster et al., 2018) with the official script. Language-model pretraining used a 3.8M-token corpus built from the Wikipedia articles behind GAP instances.
  • Main results (two memory cells, strictly incremental): LM + coref training reaches F1 72.1, with F1 M 72.8, F1 F 71.4 and bias 0.98. This exceeds retrained Lee et al. (2017) at 67.0 and Parallelism+URL at 70.6.
  • Effect of pretraining and BERT: LM pretraining adds 3.2 F1 over coref-only training (68.9). Adding BERT embeddings raises F1 to 78.8, averaged over 5 seeds, but bias shifts slightly to 0.96.
  • Robustness tests: On a test set made by concatenating pairs of GAP instances, F1 drops only to 70.2, so the model reuses its two cells effectively. Using four cells gives F1 71.4 on the original test set.
  • Implementation details: Gumbel-softmax makes the choice of cell to overwrite differentiable. Salience half-lives are 4 entity mentions and 30 non-entity tokens. Code is available at github.com/liufly/refreader.

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

13 papers of our library cite it (of the 140673 whose bibliographies we read); 3 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/p19-1593 · arXiv 1902.01541

@inproceedings{liu2019referential,
  title = {{The Referential Reader: A Recurrent Entity Network for Anaphora Resolution}},
  author = {Liu, Fei and Zettlemoyer, Luke and Eisenstein, Jacob},
  booktitle = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics},
  year = {2019},
  doi = {10.18653/v1/p19-1593},
  eprint = {1902.01541},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.18653/v1/p19-1593}
}

References

20 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jianpeng Cheng, Li Dong, and Mirella Lapata. 2016. Long short-term memory-networks for machine reading. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 551–561. find our card
  2. Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. In Proceedings of the NIPS 2014 Deep Learning and Representation Learning Workshop. find
  3. Kevin Clark and Christopher D. Manning. 2015. Entity-centric coreference resolution with model stacking. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages… find our card
  4. 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,… find
  5. Bhuwan Dhingra, Qiao Jin, Zhilin Yang, William Cohen, and Ruslan Salakhutdinov. 2018. Neural models for reasoning over multiple mentions using coreference. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologie… find our card
  6. Barbara J Grosz, Scott Weinstein, and Aravind K Joshi. 1995. Centering: A framework for modeling the local coherence of discourse. Computational linguistics, 21(2):203–225. find our card
  7. Mikael Henaff, Jason Weston, Arthur Szlam, Antoine Bordes, and Yann LeCun. 2017. Tracking the world state with recurrent entity networks. In Proceedings of the 5th International Conference on Learning Representations, Toulon, France. find
  8. Eric Jang, Shixiang Gu, and Ben Poole. 2017. Categorical reparameterization with gumbel-softmax. In Proceedings of the 5th International Conference on Learning Representations. find our card
  9. Yangfeng Ji, Chenhao Tan, Sebastian Martschat, Yejin Choi, and Noah A. Smith. 2017. Dynamic entity representations in neural language models. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 1830– 1839. Association for Computational Linguistics. find
  10. Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980. find
  11. Sosuke Kobayashi, Naoaki Okazaki, and Kentaro Inui. 2017. A neural language model for dynamically representing the meanings of unknown words and entities in a discourse. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 473– 483… find
  12. 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, Copenhagen, Denmark. Association for Computational Linguistics. find our card
  13. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word representation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1532–1543, Doha, Qatar. Association for Computational Linguistics. find
  14. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. https://d4mucfpksywv.cloudfront.net/better-language-models/ language-models.pdf. ↗
  15. David Schlangen, Timo Baumann, and Michaela At-terer. 2009. Incremental reference resolution: The task, metrics for evaluation, and a Bayesian filtering model that is sensitive to disfluencies. In Proceedings of the SIGDIAL 2009 Conference: The 10th Annual Meeting of the Special Interest Group on Di… find
  16. Sainbayar Sukhbaatar, Arthur Szlam, Jason Weston, and Rob Fergus. 2015. End-to-end memory networks. In Proceedings of Advances in Neural Information Processing Systems, pages 2440–2448, Montréal, Canada. find
  17. Michael K Tanenhaus, Michael J Spivey-Knowlton, Kathleen M Eberhard, and Julie C Sedivy. 1995. Integration of visual and linguistic information in spoken language comprehension. Science, 268(5217):1632–1634. find
  18. Kellie Webster, Marta Recasens, Vera Axelrod, and Jason Baldridge. 2018. Mind the GAP: A balanced corpus of gendered ambiguous pronouns. Transactions of the Association for Computational Linguistics, 6:605–617. find our card
  19. Jason Weston, Sumit Chopra, and Antoine Bordes. 2015. Memory networks. In Proceedings of the 3rd International Conference on Learning Representations, San Diego, USA. find
  20. Zichao Yang, Phil Blunsom, Chris Dyer, and Wang Ling. 2017. Reference-aware language models. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 1850–1859. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2019-referential-reader-recurrent-entity-network/