Elamur Research⌕
☰
home / papers / 2011 / The Taming of Reconcile as a Biomedical Coreference Resolver

The Taming of Reconcile as a Biomedical Coreference Resolver

Summary

The authors adapt Reconcile, a general supervised coreference engine originally built for newswire, to biomedical text for the Protein Coreference (COREF) supporting task of the BioNLP Shared Task 2011. The task data come from the Genia-Medco coreference corpus. Systems must find gene/protein-related mentions and output anaphor–antecedent pairs.

The team made several changes to Reconcile:

  • They swapped in the Genia Tagger and the Enju parser for preprocessing.
  • They dropped features that did not fit the domain or data, mostly WordNet-based and paragraph features.
  • They replaced Reconcile's poorly performing mention detector with CRF++ sequence taggers built on typical biomedical NER features. One variant uses a single model for all mentions; the other trains separate anaphor and antecedent models.

Three clustering strategies were compared, and coreference chains were converted back into pairs. Using separate detectors with most-recent-first clustering, the system ranked first among six teams. The work shows that domain-specific mention detection, rather than the linking step, is the main obstacle to biomedical coreference.

Key points

  • COREF data: 800 training, 150 development and 260 test files. Training has 2,313 links over 4,367 mentions, and anaphors are mostly short pronouns such as "that" and "which."
  • The CRF++ mention detector MD-II (separate anaphor/antecedent models) reached 81.31% precision and 64.78% recall. The Genia NP chunker managed only 6.58% precision.
  • On the dev set, MUC reached 87.32% with gold mentions but only about 49.65% with automatic mentions. Most-recent-first clustering beat best-first and Reconcile's default single-link clustering.
  • MD-II scored roughly 5 points higher than the single-model MD-I across all link modes. Weak antecedent recall was the main bottleneck.
  • The final test submission achieved 34.05% F (73.26% precision, 22.18% recall) on protein links and 25.41% on atom links, the best result in the task.

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

5 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 3 not yet.

Cite

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

@inproceedings{kim2011taming,
  title = {{The Taming of Reconcile as a Biomedical Coreference Resolver}},
  author = {Kim, Youngjun and Riloff, Ellen and Gilbert, Nathan},
  booktitle = {Proceedings of BioNLP Shared Task 2011 Workshop},
  year = {2011},
  url = {https://aclanthology.org/W11-1813/}
}

References

18 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jenny Finkel, Shipra Dingare, Christopher D Manning, Malvina Nissim, Beatrice Alex, and Claire Grover. 2005. Exploring the Boundaries: Gene and Protein Identification in Biomedical Text. BMC Bioinformatics. 6:S5. find
  2. Jin-Dong Kim, Sampo Pyysalo, Tomoko Ohta, Robert Bossy, and Jun’ichi Tsujii. 2011. Overview of BioNLP Shared Task 2011. Proceedings of the BioNLP 2011 Workshop Companion Volume for Shared Task, Portland, Oregon, June. find
  3. Taku Kudoh. 2007. CRF++. http://crfpp.sourceforge.net/. ↗
  4. Ryan McDonald and Fernando Pereira. 2005. Identifying Gene and Protein Mentions in Text Using Conditional Random Fields. BMC Bioinformatics. 6:S6. find
  5. Yusuke Miyao and Jun'ichi Tsujii. 2008. Feature Forest Models for Probabilistic HPSG Parsing. Computational Linguistics, 34(1):35–80. find our card
  6. Ngan L. T. Nguyen and Jin-Dong Kim. 2008. Exploring Domain Differences for the Design of Pronoun Resolution Systems for Biomedical Text. Proceedings of COLING 2008:625-632 find
  7. Ngan L. T. Nguyen, Jin-Dong Kim, and Jun’ichi Tsujii. 2008. Challenges in Pronoun Resolution System for Biomedical Text. Proceedings of LREC 2008. find
  8. Ngan L. T. Nguyen, Jin-Dong Kim, and Jun’ichi Tsujii. 2011. Overview of the Protein Coreference task in BioNLP Shared Task 2011. Proceedings of the BioNLP 2011 Workshop Companion Volume for Shared Task, Portland, Oregon, June. find
  9. Massimo Poesio and Mijail A. Kabadjov. 2004. A general-purpose, off-the-shelf anaphora resolution module: implementation and preliminary evaluation. Proceedings of LREC 2004. find our card
  10. Long Qiu, Min-Yen Kan, and Tat-Seng Chua. 2004. A Public Reference Implementation of the Rap Anaphora Resolution Algorithm. Proceedings of LREC 2004. find our card
  11. Veselin Stoyanov, Claire Cardie, Nathan Gilbert, Ellen Riloff, David Buttler, and David Hysom. 2010a. Reconcile: A Coreference Resolution Platform. Tech Report. Cornell University. find
  12. Veselin Stoyanov, Claire Cardie, Nathan Gilbert, Ellen Riloff, David Buttler, and David Hysom. 2010b. Coreference Resolution with Reconcile. Proceedings of ACL 2010. find our card
  13. Veselin Stoyanov, Nathan Gilbert, Claire Cardie, and Ellen Riloff. 2009. Conundrums in Noun Phrase Coreference Resolution: Making Sense of the State-of-the-Art. Proceedings of ACL-IJCNLP 2009. find our card
  14. Yoshimasa Tsuruoka and Jun'ichi Tsujii. 2005. Bidirectional Inference with the Easiest-First Strategy for Tagging Sequence Data. Proceedings of HLT/EMNLP 2005:467-474. find
  15. Yannick Versley, Simone P. Ponzetto, Massimo Poesio, Vladimir Eidelman, Alan Jern, Jason Smith, Xiaofeng Yang, and Alessandro Moschitti. 2008. BART: A modular toolkit for coreference resolution. Proceedings of LREC 2008. find our card
  16. Xiaofeng Yang, Jian Su, Guodong Zhou, and Chew Lim Tan. 2004a. A NP-Cluster Based Approach to Coreference Resolution. Proceedings of COLING 2004:226-232. find
  17. XiaoFeng Yang, GuoDong Zhou, Jian Su, and Chew Lim Tan. 2004b. Improving Noun Phrase Coreference Resolution by Matching Strings. Proceedings of IJCNLP 2004:226-333. find
  18. GuoDong Zhou, Dan Shen, Jie Zhang, Jian Su, and SoonHeng Tan. 2005. Recognition of Protein/Gene Names from Text Using an Ensemble of Classifiers. BMC Bioinformatics. 6:S7. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/kim-2011-taming-reconcile-as-biomedical-coreference/