Elamur Research⌕
☰
home / papers / 2010 / UBIU: A Language-Independent System for Coreference Resolution

UBIU: A Language-Independent System for Coreference Resolution

Summary

The paper tackles coreference resolution for many languages, where machine-learned systems usually exist for only a few languages or a fixed set. The authors developed UBIU as an entry in SemEval-2010 Task 1, the shared task on resolving coreference across many languages. It pairs memory-based learning in TiMBL with a mention-pair feature set adapted from Rahman and Ng (2009). The features rely only on standard annotation layers (POS, morphology, dependency relations, named entities), and a feature gets a null value when a language lacks that layer.

The only language-specific component is a small set of regular expressions that find markable heads. A markable's full extent is reconstructed afterwards from dependency subtrees. In postprocessing, each mention keeps its highest-confidence antecedent, and pronouns are always forced to receive one.

An ablation shows that accurate mention boundaries matter far more than gold linguistic annotation. This suggests the system can run on automatically parsed data with little loss, which is useful for less-resourced languages.

Key points

  • Shared task and scale: evaluated on SemEval-2010 Task 1 across Catalan, Dutch, English, German, Italian and Spanish. Development took about 4 months, and adapting the head-finding module to a new language needs roughly 1 person-hour.
  • Classifier settings: TiMBL parameters were tuned jointly for the best average across all languages rather than per language. The chosen setting is k=3 with Overlap distance and gain ratio weighting.
  • Ablation on development data (Catalan, English, German, Spanish):
    • Gold markables raise mention identification by about 20+ F1 points (e.g., English 96.1 vs. 72.1).
    • Swapping gold linguistic annotation for automatic annotation changes the scores very little.
  • Official results (regular setting):
    • English is strongest (CEAF 53.6, B³ 58.7).
    • Dutch is weakest (CEAF 17.0), partly because its named-entity markables exclude determiners that the dependency subtree includes.
  • Postprocessing fallback for pronouns: when no candidate is classified as coreferent, the system picks the closest subject. If that fails, it takes the candidate TiMBL rejected with the lowest confidence.

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

14 papers of our library cite it (of the 140673 whose bibliographies we read); 3 of them have a page here, 11 not yet.

Cite

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

@inproceedings{zhekova2010ubiu,
  title = {{UBIU: A Language-Independent System for Coreference Resolution}},
  author = {Zhekova, Desislava and Kübler, Sandra},
  booktitle = {Proceedings of the 5th International Workshop on Semantic Evaluation},
  year = {2010},
  url = {https://aclanthology.org/S10-1019/}
}

References

14 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Walter Daelemans and Antal van den Bosch. 2005. Memory Based Language Processing. Cambridge University Press. find
  2. Walter Daelemans, Jakub Zavrel, Ko van der Sloot, and Antal van den Bosch. 2007. TiMBL: Tilburg memory based learner – version 6.1 – reference guide. Technical Report ILK 07-07, Induction of Linguistic Knowledge, Computational Linguistics, Tilburg University. find
  3. Sanda M. Harabagiu and Steven J. Maiorano. 2000. Multilingual coreference resolution. In Proceedings of ANLP 2000, Seattle, WA. find our card
  4. Sandra Kübler, Ryan McDonald, and Joakim Nivre. 2009. Dependency Parsing. Morgan Claypool. find
  5. Xiaoqiang Luo and Imed Zitouni. 2005. Multilingual coreference resolution with syntactic features. In Proceedings of HLT/EMNLP 2005, Vancouver, Canada. find
  6. Katja Markert and Malvina Nissim. 2005. Comparing knowledge sources for nominal anaphora resolution. Computational Linguistics, 31(3). find our card
  7. Ruslan Mitkov. 1998. Robust pronoun resolution with limited knowledge. In Proceedings of ACL/COLING 1998, Montreal, Canada. find our card
  8. Ruslan Mitkov. 1999. Multilingual anaphora resolution. Machine Translation, 14(3-4):281–299. find
  9. Vincent Ng and Claire Cardie. 2002. Improving machine learning approaches to coreference resolution. In Proceedings of ACL 2002, pages 104–111, Philadelphia, PA. find our card
  10. Massimo Poesio, Tomonori Ishikawa, Sabine Schulte im Walde, and Renata Vieira. 2002. Acquiring lexical knowledge for anaphora resolution. In Proceedings of LREC 2002, Las Palmas, Gran Canaria. find our card
  11. Altaf Rahman and Vincent Ng. 2009. Supervised models for coreference resolution. In Proceedings of EMNLP 2009, Singapore. find our card
  12. 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 multiple languages. In Proceedings of the 5th International Workshop on Semantic Evaluations (SemEval-2010), U… find our card
  13. Wee Meng Soon, Hwee Tou Ng, and Daniel Chung Yong Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational Linguistics, 27(4):521–544. find our card
  14. Holger Wunsch. 2009. Rule-Based and Memory-Based Pronoun Resolution for German: A Comparison and Assessment of Data Sources. Ph.D. thesis, Universität Tübingen. find

↑ back to top

This page in Markdown →