Elamur Research
☰
home / papers / 2002 / Improving Machine Learning Approaches to Coreference Resolution

Improving Machine Learning Approaches to Coreference Resolution

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)" does not allow redistributing the text itself. Original — aclanthology.org/P02-1014.pdf

Summary

This paper addresses noun phrase coreference resolution, the task of grouping the NPs in a document that refer to the same real-world entity. It builds on the decision-tree approach of Soon et al. (2001), which uses only 12 surface features, and improves it in two ways. The first is three changes to the learning framework: best-first clustering, which picks the most confident antecedent instead of the nearest one; a matching change in how positive training instances are chosen; and splitting the string-match feature by NP type. The second is a larger feature set, growing from 12 to 53 lexical, semantic, knowledge-based and grammatical features. The authors train C4.5 and RIPPER on the MUC-6 and MUC-7 corpora. The framework changes consistently raise precision. Using all 53 features hurts performance, mainly because common noun resolution gets worse. A hand-picked subset that removes low-precision rules gives the best results reported up to that point. The work shows that learning-based coreference depends on how the classifier, instance selection and clustering interact, and on choosing linguistic knowledge selectively rather than applying it as hard constraints.

Key points

  • Final F-measures were 70.4 on MUC-6 (RIPPER) and 63.4 on MUC-7 (C4.5). The original Soon et al. system scored 62.6 and 60.4.
  • The authors' re-implementation of the Soon baseline reached 66.3 on MUC-6 and 61.2 on MUC-7 with C4.5. Classifiers were trained on the 30 MUC "dry run" texts and tested on the formal evaluation texts, with all NPs found automatically.
  • The expanded set includes 26 new grammatical features. Only four of them (AGREEMENT, ANIMACY, BINDING, MAXIMALNP) appeared in the learned models.
  • Hand-selected subsets of 22–26 features raised precision a lot at a small cost in recall. Common noun resolution was still the weakest part.
  • In the final tree, ALIAS replaces SOON_STR at the root, and the system allows pronoun antecedents in earlier sentences, which the Soon tree largely blocked.

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

220 papers of our library cite it (of the 140673 whose bibliographies we read); 16 of them have a page here, 204 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/1073083.1073102

@inproceedings{ng2002improving,
  title = {{Improving Machine Learning Approaches to Coreference Resolution}},
  author = {Ng, Vincent and Cardie, Claire},
  booktitle = {Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics},
  year = {2002},
  doi = {10.3115/1073083.1073102},
  url = {https://doi.org/10.3115/1073083.1073102}
}

References

14 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. C. Aone and S. W. Bennett. 1995. Evaluating Automated and Manual Acquisition of Anaphora Resolution Strategies. In Proceedings of the 33rd Annual Meeting of the Association for Computational Linguistics, pages 122–129. find
  2. W. Cohen. 1995. Fast Effective Rule Induction. In Proceedings of the Twelfth International Conference on Machine Learning. find
  3. R. Grishman. 1995. The NYU System for MUC-6 or Where’s the Syntax? In Proceedings of the Sixth Message Understanding Conference (MUC-6). find
  4. S. Harabagiu, R. Bunescu, and S. Maiorano. 2001. Text and Knowledge Mining for Coreference Resolution. In Proceedings of the Second Meeting of the North America Chapter of the Association for Computational Linguistics (NAACL-2001), pages 55–62. find
  5. S. Lappin and H. Leass. 1994. An Algorithm for Pronominal Anaphora Resolution. Computational Linguistics, 20(4):535–562. find our card
  6. D. Lin. 1995. University of Manitoba: Description of the PIE System as Used for MUC-6. In Proceedings of the Sixth Message Understanding Conference (MUC-6). find
  7. J. McCarthy and W. Lehnert. 1995. Using Decision Trees for Coreference Resolution. In Proceedings of the Fourteenth International Conference on Artificial Intelligence, pages 1050–1055. find
  8. R. Mitkov. 1997. Factors in anaphora resolution: they are not the only things that matter. A case study based on two different approaches. In Proceedings of the ACL’97/EACL’97 Workshop on Operational Factors in Practical, Robust Anaphora Resolution. find
  9. MUC-6. 1995. Proceedings of the Sixth Message Understanding Conference (MUC-6). Morgan Kaufmann, San Francisco, CA. find
  10. MUC-7. 1998. Proceedings of the Seventh Message Understanding Conference (MUC-7). Morgan Kaufmann, San Francisco, CA. find
  11. J. R. Quinlan. 1993. C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo, CA. find our card
  12. C. Sidner. 1979. Towards a Computational Theory of Definite Anaphora Comprehension in English Discourse. PhD Thesis, Massachusetts Institute of Technology. find
  13. W. M. Soon, H. T. Ng, and D. C. Y. Lim. 2001. A Machine Learning Approach to Coreference Resolution of Noun Phrases. Computational Linguistics, 27(4):521–544. find our card
  14. 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 45– 52, San Francisco, CA. Morgan Kaufmann. find our card