Elamur Research⌕
☰
home / papers / 2009 / Global Learning of Noun Phrase Anaphoricity in Coreference Resolution via Label Propagation

Global Learning of Noun Phrase Anaphoricity in Coreference Resolution via Label Propagation

Summary

The paper investigates whether deciding up front if a noun phrase is anaphoric can reliably help coreference resolution. Earlier attempts to add this filtering step produced mixed results. The authors treat anaphoricity as a global, graph-based semi-supervised problem. They apply label propagation (Zhu and Ghahramani 2002), in which labeled NPs from training texts and unlabeled NPs from a test document form the vertices of one graph. Labels spread along weighted edges until the assignments stabilize, so every NP in a document is labeled jointly rather than one at a time. Edge weights come from one of two similarity functions: a feature-based RBF kernel or a convolution tree kernel over parse structures.

On the ACE 2003 corpus (NWIRE, NPAPER and BNEWS), the graph-based approach beats an SVM that classifies each NP independently. When used as a preprocessing filter for an SVM coreference system, it raises overall F1. The tree kernel helps with both pronouns and definite NPs, whereas the RBF kernel mainly helps pronouns. This suggests that syntactic structure matters most for spotting anaphoric definite descriptions.

Key points

  • Data: All experiments use ACE 2003 (NWIRE, NPAPER, BNEWS domains). The baseline coreference system follows the single-candidate feature set of Yang et al. (2003), using an SVM with an RBF kernel.
  • Anaphoricity accuracy, RBF kernel: Label propagation outperforms per-instance SVM classification. On NWIRE, accuracy on anaphoric/non-anaphoric NPs is 74.8/84.4% versus 71.3/80.2% for the SVM.
  • Anaphoricity accuracy, tree kernel: The best parse-tree scheme (DET) reaches 79.2/87.8% on NWIRE, compared with 76.5/82.3% for an SVM using the same structure. Removing antecedent-candidate, predicate or first-right-sibling information each costs roughly 3–5 points.
  • Coreference F1 gains: Overall F1 goes from 59.4/62.1/55.5 (baseline) to 60.7/63.8/57.4 with the RBF kernel and 61.8/65.2/59.6 with the tree kernel. Most of the improvement comes from higher precision.
  • Definite NPs: With the RBF kernel, definite-NP F1 barely changes. With the tree kernel, it rises by about 1.3–1.9 points (e.g., 40.2 → 42.1 on NWIRE).

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

10 papers of our library cite it (of the 140673 whose bibliographies we read); 3 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.

@inproceedings{zhou2009global,
  title = {{Global Learning of Noun Phrase Anaphoricity in Coreference Resolution via Label Propagation}},
  author = {Zhou, GuoDong and Kong, Fang},
  booktitle = {Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing},
  year = {2009},
  doi = {10.3115/1699571.1699640},
  url = {https://doi.org/10.3115/1699571.1699640}
}

References

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

  1. Bean D. and Riloff E. (1999). Corpus-based Identification of Non-Anaphoric Noun Phrases. ACL’1999:373-380. find our card
  2. Bergsma S., Lin D.K. and Goebel R. (2008). Distributional Identification of Non-Referential Pronouns. ACL’2008: 10-18. find
  3. Charniak E. (2001). Immediate-head Parsing for Language Models. ACL’2001: 129-137. find our card
  4. Cherry C. and Bergsma S. (2005). An expectation maximization approach to pronoun resolution. CoNLL’2005:88-95. find our card
  5. Collins M. and Duffy N. (2001). Convolution kernels for natural language. NIPS’2001: 625-632. find
  6. Denber M. (1998). Automatic Resolution of Anaphora in English. Technical Report, Eastman Kodak Co. find
  7. Denis P. and Baldridge J. (2007). Joint determination of anaphoricity and coreference using integer programming. NAACL-HLT’2007:236-243. find
  8. Evans R. (2001). Applying machine learning toward an automatic classification of it. Literary and Linguistic Computing, 16(1):45.57. find
  9. Joachims T. (1998). Text Categorization with Support Vector Machine: learning with many relevant features. ECML-1998: 137-142. find
  10. Kennedy C. and Boguraev B. (1996). Anaphora for everyone: pronominal anaphora resolution without a parser. COLING’1996: 113-118. find our card
  11. Lappin S. and Leass H.J. (1994). An algorithm for pronominal anaphora resolution. Computational Linguistics, 20(4):535.561. find our card
  12. Ng V. and Cardie C. (2002a). Identifying anaphoric and non-anaphoric noun phrases to improve coreference resolution. COLING’2002:730-736. find
  13. Ng V. and Cardie C. (2002b). Improving machine learning approaches to coreference resolution. ACL’2002: 104-111 find our card
  14. Ng V. (2004). Learning Noun Phrase Anaphoricity to Improve Conference Resolution: Issues in Representation and Optimization. ACL’2004: 151-158 find our card
  15. Paice C.D. and Husk G.D. (1987). Towards the automatic recognition of anaphoric features in English text: the impersonal pronoun it. Computer Speech and Language, 2:109-132. find
  16. Soon W.M., Ng H.T. and Lim D. (2001). A machine learning approach to coreference resolution of noun phrase. find
  17. Computational Linguistics, 2001, 27(4):521-544. find
  18. Strube M. and Muller C. (2003). A machine learning approach to pronoun resolution in spoken dialogue. ACL’2003: 168-175 find our card
  19. Vieira R. and Poesio M. (2000). An empirically based system for processing definite descriptions. Computational Linguistics, 27(4): 539–592. find
  20. Vilain M., Burger J., Aberdeen J., Connolly D. and Hirschman L. (1995). A model theoretic coreference scoring scheme. MUC-6: 45–52. find our card
  21. Yang X.F., Zhou G.D., Su J. and Chew C.L. (2003). Coreference Resolution Using Competition Learning Approach. ACL’2003:177-184 find our card
  22. Yang X.F., Su J. and Tan C.L. (2005). A Twin-Candidate Model of Coreference Resolution with Non-Anaphor Identification Capability. IJCNLP’2005:719-730. find our card
  23. Yang X.F., Su J. and Tan C.L. (2006). Kernel-based pronoun resolution with structured syntactic knowledge. COLING-ACL’2006 : 41-48. find our card
  24. Yang X.F., Su J., Lang J., Tan C.L., Liu T. and Li S. (2008). An Entity-Mention Model for Coreference Resolution with Inductive Logic Programming. ACL’2008: 843-851. find our card
  25. Zhou G.D. and Su. J. (2000). Error-driven HMM-based chunk tagger with context-dependent lexicon. EMNLP-VLC’2000: 71–79 find
  26. Zhou G.D. and Su J. (2002). Named Entity recognition using a HMM-based chunk tagger. In ACL’2002:473–480. find
  27. Zhou G.D., Kong F. and Zhu Q.M. (2008). Context-sensitive convolution tree kernel for pronoun resolution . IJCNLP’2008:25-31. find our card
  28. Zhu X. and Ghahramani Z. (2002). Learning from Labeled and Unlabeled Data with Label Propagation. CMU CALD Technical Report.CMU-CALD-02-107. find

↑ back to top

This page in Markdown →