SUCRE: A Modular System for Coreference Resolution

Summary

Coreference resolution, which groups the mentions (markables) of a document into chains that point to the same entity, requires a lot of engineering. Much of that effort goes into designing and implementing features. Kobdani and Schütze present SUCRE, a modular toolkit that makes this work easier. It can handle noun, pronoun, or full coreference separately.

In preprocessing, the corpus is converted into a relational database with tables for words, markables, and links. Features are then specified in a regular feature definition language that combines words from markable pairs, such as first, last, head, or all words. The system also accepts features computed externally and can plug in outside classifiers. The authors describe the approach as language independent and usable for both knowledge-rich and knowledge-poor data.

In SemEval-2010 Task 1, SUCRE competed in the gold and regular closed tracks for six languages. It ranked first in the regular closed track for English and German on every metric. This suggests that flexible, declarative feature engineering can produce competitive coreference systems.

Key points

The minimal database model has three tables: Word, Markable, and Link. Optional word and markable feature columns hold attributes such as POS, gender, number, case, pronoun type, and semantic class.
The link feature language supports functions including exact and substring matching, edit distance, alias, word relation, parse tree path, and absolute value.
Training instances are generated following Soon et al. (2001). During decoding, candidate links are generated within a window of 100 markables. Chains are then built by best-first clustering, searching right-to-left from the end of the document.
Four classifiers are integrated: Decision-Tree, Naive-Bayes, SVM, and Maximum-Entropy. Decision-Tree gave the best results.
Evaluation used MUC, B³, CEAF, and BLANC. In the gold closed track for English and German, SUCRE led by large margins on MUC (+27.1 and +32.5) and BLANC (+9.5 and +9.0).

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

7 papers of our library cite it (of the 140673 whose bibliographies we read); 4 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{kobdani2010sucre,
  title = {{SUCRE: A Modular System for Coreference Resolution}},
  author = {Kobdani, Hamidreza and Schütze, Hinrich},
  booktitle = {Proceedings of the 5th International Workshop on Semantic Evaluation},
  year = {2010},
  url = {https://aclanthology.org/S10-1018/}
}

References

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

Amit Bagga and Breck Baldwin. 1998. Algorithms for scoring coreference chains. In In The First International Conference on Language Resources and Evaluation Workshop on Linguistics Coreference, pages 563–566. find our card
Thorsten Joachims. 2002. Learning to Classify Text Using Support Vector Machines, Methods, Theory, and Algorithms. Kluwer/Springer. find
Xiaoqiang Luo. 2005. On coreference resolution performance metrics. In HLT ’05: Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing, pages 25–32, Morristown, NJ, USA. Association for Computational Linguistics. find our card
Vincent Ng and Claire Cardie. 2002. Improving machine learning approaches to coreference resolution. In Proceedings of the ACL, pages 104–111. find our card
Marta Recasens and Eduard Hovy. in prep. BLANC: Implementing the Rand Index for Coreference Evaluation. Marta Recasens, Lluı́s Màrquez, Emili Sapena, M.Antònia Martı́, Mariona Taulé, Véronique Hoste, Massimo Poesio, and Yannick Versley. 2010. SemEval-2010 Task 1: Coreference resolution in multip… find
Wee Meng Soon, Hwee Tou Ng, and Daniel Chung Yong Lim. 2001. A machine learning approach to coreference resolution of noun phrases. In Computational Linguistics, pages 521–544. find our card
Josef Steinberger, Massimo Poesio, Mijail A. Kabad-jovb, and Karel Jezek. 2007. Two uses of anaphora resolution in summarization. In Information Processing and Management, Special issue on Summarization, pages 1663–1680. find
Yoshimasa Tsuruoka. 2006. A simple c++ library for maximum entropy classification. Tsujii laboratory, Department of Computer Science, University of Tokyo. find
Yannick Versley, Simone Paolo Ponzetto, Massimo Poesio, Vladimir Eidelman, Alan Jern, Jason Smith, and Xiaofeng Yang. 2008. Bart: A modular toolkit for coreference resolution. In Proceedings of the 46nd Annual Meeting of the Association for Computational Linguistics, pages 9–12. find our card
Marc Vilain, John Burger, John Aberdeen, Dennis Connolly, and Lynette Hirschman. 1995. A model-theoretic coreference scoring scheme. In MUC6 ’95: Proceedings of the 6th conference on Message understanding, pages 45–52, Morristown, NJ, USA. Association for Computational Linguistics. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/kobdani-2010-sucre-modular-system-coreference-resolution/

📄 SUCRE: A Modular System for Coreference Resolution