Elamur Research⌕
☰
home / papers / 2012 / A Mixed Deterministic Model for Coreference Resolution

A Mixed Deterministic Model for Coreference Resolution

Summary

This CoNLL-2012 shared task system tackles coreference resolution for English (closed track) and Chinese (closed and open tracks). It is a hybrid. Most decisions come from hand-written deterministic rules in the spirit of the Stanford multi-pass approach, and a small machine-learning component filters out non-referring mentions.

The pipeline has two stages. First, candidate mentions (NPs, pronouns, named entities) are taken from parse trees. Pleonastic English "it" is removed using regular expressions plus a C4.5 decision tree. Second, mentions are grouped by rule families: NP–NP matching (exact, partial, alias/abbreviation, WordNet-based similarity) and pronoun–NP linking driven by gender, number, distance and speaker. The resulting clusters are merged into chains.

For Chinese, gender, number and named-entity information were not provided. In the closed track the authors induce them from the training data with patterns. In the open track they use external dictionaries and the LTP parser.

Official averaged F1 scores were 58.68 (English), 60.69 (Chinese closed) and 61.02 (Chinese open). Errors from parsing and from external tools turned out to limit performance considerably.

Key points

  • Classifying "it" with C4.5 used POS, parse bit, predicate-argument and neighbouring-word features. The training data held 9697 instances, 4043 of them referential, and the classifier reached about 0.68 overall F on the development data.
  • Pronouns fall into seven rule categories with paired English and Chinese forms. There are extra Chinese-specific rules, for example for 双方 and 其.
  • Chinese open-track resources were appellation, gender, geographical-name and temporal-word dictionaries, a pinyin-to-character converter, and LTP (for POS, parse and NEs).
  • Using gold parses, Chinese test F1 rises to 66.46 (closed) and 66.38 (open), which shows how much parse quality matters. The open resources added little.
  • With gold mentions, F1 reaches 69.88 for English and 76.05 for Chinese, so the rule-based resolution stage still has room for improvement.

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

4 papers of our library cite it (of the 140673 whose bibliographies we read); each has a page here.

Cite

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

@inproceedings{yuan2012mixed,
  title = {{A Mixed Deterministic Model for Coreference Resolution}},
  author = {Yuan, Bo and Chen, Qingcai and Xiang, Yang and Wang, Xiaolong and Ge, Liping and Liu, Zengjian and Liao, Meng and Si, Xianbo},
  booktitle = {Joint Conference on EMNLP and CoNLL - Shared Task},
  year = {2012},
  url = {https://aclanthology.org/W12-4507/}
}

References

13 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Bo Yuan, Qingcai Chen, Xiaolong Wang, Liwei Han. 2009. Extracting Event Temporal Information based on Web. 2009 Second International Symposium on Knowledge Acquisition and Modeling, pages.346-350 find
  2. Cicero Nogueira dos Santos, Davi Lopes Carvalho. 2011. Rule and Tree Ensembles for Unrestricted Coreference Resolution. Proceedings of the 15th Conference on Computational Natural Language Learning: Shared Task, pages 51–55. find
  3. Emili Sapena, Llu´ıs Padr´o and Jordi Turmo. 2011. RelaxCor Participation in CoNLL Shared Task on Coreference Resolution. Proceedings of the 15th Conference on Computational Natural Language Learning: Shared Task, pages 35–39. find our card
  4. Heeyoung Lee, Yves Peirsman, Angel Chang, Nathanael Chambers, Mihai Surdeanu, Dan Jurafsky. 2011. Stanford’s Multi-Pass Sieve Coreference Resolution System at the CoNLL-2011 Shared Task. Proceedings of the 15th Conference on Computational Natural Language Learning: Shared Task, pages 28–34, Portland… find our card
  5. Liu Ting, Che Wanxiang, Li Zhenghua. 2011. Language Technology Platform. Journal of Chinese Information Processing. 25(6): 53-62 find
  6. Jun Lang, Bing Qin, Ting Liu, Sheng Li. 2007. Intra-document Coreference Resolution: The state of the art. Journal of Chinese Language and Computing. 17 (4):227-253 find
  7. Kai-Wei Chang Rajhans Samdani. 2011. Inference Protocols for Coreference Resolution. Proceedings of the 15th Conference on Computational Natural Language Learning: Shared Task, pages 40–44, Portland, Oregon. find our card
  8. Karthik Raghunathan, Heeyoung Lee, Sudarshan Rangarajan, Nathanael Chambers, Mihai Surdeanu, Dan Jurafsky, Christopher Manning. 2010. A Multi-Pass Sieve for Coreference Resolution. In EMNLP. find our card
  9. Kong Fang, Zhu Qiaoming and Zhou Guodong. 2012(a). Anaphoricity determination for coreference resolution in English and Chinese languages. Computer Research and Development (Chinese). find
  10. Kong Fang and Zhou Guodong. 2012(b). Tree kernel-based pronoun resolution in English and Chinese languages. Journal of Software (Chinese). Accepted: 23(8). find
  11. Sameer Pradhan, Lance Ramshaw, Mitchell Marcus, Martha Palmer, Ralph Weischedel and Nianwen Xue.2011. CoNLL-2011 Shared Task: Modeling Unrestricted Coreference in OntoNotes. Proceedings of the Fifteenth Conference on Computational Natural Language Learning (CoNLL 2011). Portland, OR. find our card
  12. Sameer Pradhan and Alessandro Moschitti and Nianwen Xue and Olga Uryupina and Yuchen Zhang. 2012. CoNLL-2012 Shared Task: Modeling Multilingual Unrestricted Coreference in OntoNotes. Proceedings of the Sixteenth Conference on Computational Natural Language Learning (CoNLL 2012). Jeju, Korea. find our card
  13. Shane Bergsma and Dekang Lin. 2006. Bootstrapping Path-Based Pronoun Resolution Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 33–40, Sydney Wang Houfeng. 2002. Survey: Computational Models and Technologies in Anaphora Resoluti… find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yuan-2012-mixed-deterministic-model-coreference-resolution/