Elamur Research⌕
☰
home / papers / 2010 / Creating a Coreference Resolution System for Italian

Creating a Coreference Resolution System for Italian

Summary

The authors build what they describe as the first full-scale coreference resolution (CR) system for Italian. They adapt BART, an open-source modular CR toolkit originally built for English, and the system works on raw text rather than on pre-annotated mentions. Most of BART's machinery carries over unchanged. They chose a Soon et al. (2001)-style pairwise setup with a maximum entropy classifier and a standard set of mention-pair features. The Italian-specific work went into two places:

  • Language plugin: an alias module with hand-made lists of company and person designators, location name-variant patterns mined from the training data, and looser abbreviation rules.
  • Preprocessing pipeline: built on the mention tagger of Biggio et al. (2009) and TextPro, with gold training mentions reduced to their heads and deeply embedded mentions removed.

Evaluation uses the ICab corpus of "Adige" newspaper articles. On gold mentions, the retrained English system reaches a MUC F-score in the high forties. The Italian adaptations raise this to about 56. On raw text, the full system reaches 56.1 MUC F. The work shows that a modular CR platform can be ported to new languages cheaply. It also finds that shallow processing beats full parsing for Italian.

Key points

  • Data: ICab contains four days of "Adige" news articles, about 350,000 tokens. Three days were used for training and one for testing.
  • Gold mentions (MUC F): the language-agnostic baseline scored 47.9 and the Italian-specific system 56.4. On CEAF, the scores were 63.6 and 66.3.
  • Aliasing alone (MUC F): the Italian-specific alias module scored 36.0, compared with 28.3 for the generic one.
  • End-to-end system on raw text: with the chunk-based shallow pipeline, the system reaches MUC F 56.1 (R 45.8, P 72.3) and CEAF 63.3.
  • Parsing pipeline: an alternative using MaltParser, trained on a converted Torino University Treebank of about 60,000 tokens, scored lower at MUC F 53.8. The authors link this mainly to the treebank being small relative to ICab.

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

9 papers of our library cite it (of the 140673 whose bibliographies we read); 3 of them have a page here, 6 not yet.

Cite

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

@inproceedings{poesio2010creating,
  title = {{Creating a Coreference Resolution System for Italian}},
  author = {Poesio, Massimo and Uryupina, Olga and Versley, Yannick},
  booktitle = {Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10)},
  year = {2010},
  doi = {10.63317/2mzb8my546vf},
  url = {https://doi.org/10.63317/2mzb8my546vf}
}

References

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

  1. Chinatsu Aone and Scott Bennett. 1995. Evaluating automated and manual acquisition of anaphora resolution strategies. In Proc. find our card
  2. Silvana Marianela Bernaola Biggio, Claudio Giuliano, Massimo Poesio, Yannick Versley, Olga Uryupina, and Roberto Zanoli. 2009. Local entity detection and recognition task. In Proceedings of Evalita-2009. find
  3. Cristina Bosco and Vincenzo Lombardo. 2006. Comparing linguistic information in treebank annotations. In LREC 2006. find
  4. Sanda Harabagiu and Steven Maiorano. 2000. Multilingual coreference resolution. In Sixth Applied Natural Language Processing Conference (ANLP-NAACL 2000). find our card
  5. Xiaoqiang Luo and Imed Zitouni. 2005. Multi-lingual coreference resolution with syntactic features. In Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing. find our card
  6. Bernardo Magnini, Emanuele Pianta, Manuela Speranze, Valentina Bartalesi Lenzi, and Rachele Sprugnoli. 2007. Italian content annotation bank (i-cab). Technical report, FBK-IRST. find
  7. Ruslan Mitkov, Lamia Belguith, and Malgorzata Stýs. 1998. Multilingual robust anaphora resolution. In EMNLP 1998. find our card
  8. Joakim Nivre, Johan Hall, Jens Nilsson, Atanas Chanev, Gulsen Eryigit, Sandra Kübler, Svetoslav Marinov, and Erwin Marsi. 2007. Maltparser: A language-independent system for data-driven dependency parsing. Natural Language Engineering, 13(2):95–135. find
  9. Frankie Patman and Paul Thompson. 2003. Names: A new frontier in text mining. In Proceedings of the 1st NSF/NIJ Symposium, pages 27–38. find
  10. Emanuele Pianta, Christian Girardi, and Roberto Zanoli. 2007. The textpro tool suite. Technical report, FBK-IRST. find
  11. Wee Meng Soon, Hwee Tou Ng, and Daniel Chung Yong Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational Linguistics (Special Issue on Computational Anap hora Resolution), 27(4):521–544. find our card
  12. Yannick Versley, Simone Paolo Ponzetto, Massimo Poesio, Vladimir Eidelman, Alan Jern, Jason Smith, Xiaofeng Yang, and Alessandro Moschitti. 2008. BART: A modular toolkit for coreference resolution. In Proceedings of the Linguistic Coreference Workshop at the International Conference on Language Reso… find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/poesio-2010-creating-coreference-resolution-system-italian/