Elamur Research⌕
☰
home / papers / 2003 / Coreference Resolution Using Competition Learning Approach

Coreference Resolution Using Competition Learning Approach

Summary

The paper argues that standard supervised coreference systems fall short because they score each antecedent candidate on its own. In this single-candidate setup, a classifier's confidence value is used to rank competitors, but it does not model how candidates compare with one another. Yang et al. instead propose a twin-candidate ("competition learning") model. Each training instance pairs an anaphor with one coreferent and one non-coreferent candidate, and the classifier, trained with C5.0, learns which of the two is preferable. At resolution time, every candidate is compared with every other one. The candidate that wins the most pairwise contests is chosen, and ties go to the nearest one. A candidate filter, built from a single-candidate classifier, removes implausible NPs first to cut computation and noise. The authors also show formally that the single-candidate model is a special case of the twin-candidate one. On the MUC-6 and MUC-7 data, the approach beats a single-candidate baseline, mostly through better pronoun resolution. This suggests that learning preferences directly works better than ranking by isolated confidence scores.

Key points

  • Overall MUC-score F-measure rises over the single-candidate baseline (Ng and Cardie, 2002a): from 69.4 to 71.3 on MUC-6 and from 58.7 to 60.2 on MUC-7.
  • Pronoun resolution shows the largest gains:
    • F-measure is about 3.2 points higher on MUC-6 and 1.6 points higher on MUC-7 than the baseline.
    • It is about 21.1 and 13.1 points higher than the twin-candidate system of Connolly et al. (1997), which lacks candidate filtering.
  • Non-pronoun resolution improves only slightly. Precision is high (90.4 on MUC-6, 87.6 on MUC-7), but recall is low (51.3 and 39.7). The authors attribute this to strong cues such as alias, apposition and string matching dominating these cases.
  • Compared with Connolly et al., the method adds a candidate filter, uses round-robin win counting instead of a greedy pairwise elimination, and includes richer features (Appositive, Name Alias, String-matching).
  • Future work includes a looser filter combined with anaphoricity determination, and more syntactic features such as grammatical role and parallelism.

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

44 papers of our library cite it (of the 140673 whose bibliographies we read); 23 of them have a page here, 21 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/1075096.1075119

@inproceedings{yang2003coreference,
  title = {{Coreference Resolution Using Competition Learning Approach}},
  author = {Yang, Xiaofeng and Zhou, Guodong and Su, Jian and Tan, Chew Lim},
  booktitle = {Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics},
  year = {2003},
  doi = {10.3115/1075096.1075119},
  url = {https://doi.org/10.3115/1075096.1075119}
}

References

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

  1. Chinatsu Aone and Scott W.Bennett. 1995. Evaluating automated and manual acquisition of anaphora resolution strategies. In Proceedings of the 33 rd Annual Meeting of the Association for Computational Linguistics, Pages 122-129. find our card
  2. D.Bean and E.Riloff. 1999. Corpus-Based identification of non-anaphoric noun phrases. In Proceedings of the 37 th Annual Meeting of the Association for Computational Linguistics, Pages 373-380. find our card
  3. Brennan, S, E., M. W. Friedman and C. J. Pollard. 1987. A Centering approach to pronouns. In Proceedings of the 25 th Annual Meeting of The Association for Computational Linguistics, Page 155-162. find
  4. Dennis Connolly, John D. Burger and David S. Day. 1997. A machine learning approach to anaphoric reference. New Methods in Language Processing, Page 133-144. find
  5. Joseph F. McCarthy. 1996. A trainable approach to coreference resolution for Information Extraction. Ph.D. thesis. University of Massachusetts. find
  6. Ruslan Mitkov. 1998. Robust pronoun resolution with limited knowledge. In Proceedings of the 17 th Int. Conference on Computational Linguistics (COLING-ACL'98), Page 869-875. find our card
  7. Ruslan Mitkov. 1999. Anaphora resolution: The state of the art. Technical report. University of Wolverhampton, Wolverhampton. find
  8. MUC-6. 1995. Proceedings of the Sixth Message Understanding Conference (MUC-6). Morgan Kaufmann, San Francisco, CA. find
  9. MUC-7. 1998. Proceedings of the Seventh Message Understanding Conference (MUC-7). Morgan Kaufmann, San Francisco, CA. find
  10. Vincent Ng and Claire Cardie. 2002a. Improving machine learning approaches to coreference resolution. In Proceedings of the 40 rd Annual Meeting of the Association for Computational Linguistics, Pages 104-111. find our card
  11. Vincent Ng and Claire Cardie. 2002b. Identifying anaphoric and non-anaphoric noun phrases to improve coreference resolution. In Proceedings of 19th International Conference on Computational Linguistics (COLING-2002). find
  12. J R. Quinlan. 1993. C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo, CA. find our card
  13. Wee Meng Soon, Hwee Tou Ng and Daniel Chung Yong Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational Linguistics, 27(4), Page 521-544. find our card
  14. Michael Strube. Never look back: An alternative to Centering. 1998. In Proceedings of the 17th Int. Conference on Computational Linguistics and 36th Annual Meeting of ACL, Page 1251-1257 find
  15. Joel R. Tetreault. 2001. A Corpus-Based evaluation of Centering and pronoun resolution. Computational Linguistics, 27(4), Page 507-520. find our card
  16. M. Vilain, J. Burger, J. Aberdeen, D. Connolly, and L.Hirschman. 1995. A model-theoretic coreference scoring scheme. In Proceedings of the Sixth Message understanding Conference (MUC-6), Pages 42-52. find our card
  17. GD Zhou and J. Su, 2000. Error-driven HMM-based chunk tagger with context-dependent lexicon. In Proceedings of the Joint Conference on Empirical Methods on Natural Language Processing and Very Large Corpus (EMNLP/ VLC'2000). find
  18. GD 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, P473-478. find

↑ back to top