Elamur Research⌕
☰
home / papers / 2005 / Improving Pronoun Resolution Using Statistics-Based Semantic Compatibility Information

Improving Pronoun Resolution Using Statistics-Based Semantic Compatibility Information

Summary

This paper asks whether automatically gathered semantic statistics can help resolve pronouns, especially neutral ones like "it." Earlier studies found that corpus-derived predicate-argument counts gave little benefit. The authors test two factors that might change this: where the statistics come from and how the learner uses them.

They score how well each antecedent candidate fits the pronoun's context using three relation types: possessive-noun, subject-verb and verb-object. Counts come either from a shallow-parsed large corpus or from exact-match Google queries, expanded with inflected forms and determiners. Each score is expressed as a raw frequency or normalized by the candidate's own frequency. The score is added as a feature to two C5 decision-tree setups: the single-candidate (SC) model, which judges one candidate at a time, and the twin-candidate (TC) model, which compares pairs of candidates.

On MUC data, web statistics combined with the TC model work best for neutral pronouns, clearly beating the usual corpus plus SC setup. This shows that noisy semantic knowledge becomes useful when it has wider coverage and is used to rank candidates against each other.

Key points

  • The best configuration, Web+TC, reaches 79.2% success on neutral pronoun (N-Pron) resolution. That is up to 11.5% higher than the Corpus+SC combination used in earlier work such as Kehler et al. (2004).
  • Adding the web-based feature raises N-Pron success by about 4.9% under SC and 5.3% under TC. Corpus-based features give smaller gains, roughly 2–3%.
  • TC outperforms SC by more than 5% success on neutral pronouns and more than 8% on personal pronouns. It also applies the semantic feature more reliably.
  • The semantic feature does not help with personal pronouns. Their human-category constraint is already captured by the NE and Pron features.
  • In the feature analysis, the web-based SemMag feature alone achieves 61.2% success, compared with 65.3% for FirstNP, the strongest single feature. SemMag also appears near the top of the learned TC decision tree.

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

21 papers of our library cite it (of the 140673 whose bibliographies we read); 14 of them have a page here, 7 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/1219840.1219861

@inproceedings{yang2005improving,
  title = {{Improving Pronoun Resolution Using Statistics-Based Semantic Compatibility Information}},
  author = {Yang, Xiaofeng and Su, Jian and Tan, Chew Lim},
  booktitle = {Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL'05)},
  year = {2005},
  doi = {10.3115/1219840.1219861},
  url = {https://doi.org/10.3115/1219840.1219861}
}

References

17 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. S. Abney. 1996. Partial parsing via finite-state cascades. In Workshop on Robust Parsing, 8th European Summer School in Logic, Language and Information, pages 8–15. find
  2. D. Bean and E. Riloff. 2004. Unsupervised learning of contextual role knowledge for coreference resolution. In Proceedings of 2004 North American chapter of the Association for Computational Linguistics annual meeting. find our card
  3. I. Dagan and A. Itai. 1990. Automatic processing of large corpora for the resolution of anahora references. In Proceedings of the 13th International Conference on Computational Linguistics, pages 330–332. find
  4. J. Hobbs. 1978. Resolving pronoun references. Lingua, 44:339–352. find our card
  5. A. Kehler, D. Appelt, L. Taylor, and A. Simma. 2004. The (non)utility of predicate-argument frequencies for pronoun interpretation. In Proceedings of 2004 North American chapter of the Association for Computational Linguistics annual meeting. find
  6. F. Keller and M. Lapata. 2003. Using the web to obtain freqencies for unseen bigrams. Computational Linguistics, 29(3):459–484. find
  7. R. Mitkov. 1998. Robust pronoun resolution with limited knowledge. In Proceedings of the 17th Int. Conference on Computational Linguistics, pages 869–875. find our card
  8. N. Modjeska, K. Markert, and M. Nissim. 2003. Using the web in machine learning for other-anaphora resolution. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 176–183. find our card
  9. V. Ng and C. Cardie. 2002. Improving machine learning approaches to coreference resolution. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 104–111, Philadelphia. find our card
  10. M. Poesio, R. Mehta, A. Maroudas, and J. Hitzeman. 2004. Learning to resolve bridging references. In Proceedings of 42th Annual Meeting of the Association for Computational Linguistics. find our card
  11. J. R. Quinlan. 1993. C4.5: Programs for machine learning. Morgan Kaufmann Publishers, San Francisco, CA. find our card
  12. W. Soon, H. Ng, and D. Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational Linguistics, 27(4):521–544. find our card
  13. M. Strube and C. Muller. 2003. A machine learning approach to pronoun resolution in spoken dialogue. In Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics, pages 168–175, Japan. find our card
  14. M. Strube. 1998. Never look back: An alternative to centering. In Proceedings of the 17th Int. Conference on Computational Linguistics and 36th Annual Meeting of ACL, pages 1251– 1257. find
  15. X. Yang, G. Zhou, J. Su, and C. Tan. 2003. Coreference resolution using competition learning approach. In Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics, Japan. find our card
  16. G. Zhou and J. Su. 2002. Named Entity recognition using a HMM-based chunk tagger. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, Philadelphia. find
  17. G. Zhou, J. Su, and T. Tey. 2000. Hybrid text chunking. In Proceedings of the 4th Conference on Computational Natural Language Learning, pages 163–166, Lisbon, Portugal. find

↑ back to top

This page in Markdown →