Elamur Research⌕
☰
home / papers / 2004 / A General-Purpose, Off-the-shelf Anaphora Resolution Module: Implementation and Preliminary Evaluation

A General-Purpose, Off-the-shelf Anaphora Resolution Module: Implementation and Preliminary Evaluation

Summary

The paper addresses a gap in NLP tooling. Robust algorithms for resolving pronouns and definite descriptions exist, but in 2004 there was no domain-independent anaphora resolution component that developers could drop into a pipeline the way they use taggers or parsers. The authors present GuiTAR (General Tool for Anaphora Resolution), a modular Java system. It reads a minimal XML input format, MAS-XML, which is derived from the GNOME mark-up scheme, and returns the same document with coreference links added as separate <ante> elements.

Preprocessing modules convert raw text or other XML into MAS-XML. Resolution algorithms sit behind an abstract API, so each one can be replaced independently. The current version includes the MARS pronoun algorithm and a partial Vieira–Poesio method for definite descriptions.

The system was evaluated on hand-annotated texts from the GNOME corpus. The error analysis points to three needs:

  • lexical knowledge, to handle antecedents that use different heads or synonymous modifiers;
  • discourse-new detection;
  • better choice among competing same-head antecedents.

The work matters because it supports task-based evaluation of anaphora resolution inside real applications.

Key points

  • Input/output design: MAS-XML marks nominal expressions with <ne> elements carrying CAT, PER, NUM and GEN attributes. It also uses <mod>, <nphead>, <w> (with POS) and <s> tags. The output reuses the MATE/GNOME <ante>/<anchor> notation.
  • Implementation: The raw-text preprocessing is built on Edinburgh's LT-XML tools plus a heuristic syntactic extractor. The discourse model follows Centering notions (utterances, forward-looking centers Cf), and the resolution entry point is the resolveAnaphor() method.
  • Data: The evaluation used museum and pharmaceutical texts from the GNOME corpus, containing 3354 nominal expressions in 28 NP types. The identity relation makes up 55.95% of the 2075 annotated anaphoric relations.
  • Corpus statistics: About 95% of personal pronouns and 32% of definite descriptions (the-np) are anaphoric through identity.
  • Coverage gaps: Demonstratives and proper nouns are not yet handled. The pronoun algorithm was only lightly tuned, with several MARS antecedent indicators switched off.

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

15 papers of our library cite it (of the 140673 whose bibliographies we read); 5 of them have a page here, 10 not yet.

Cite

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

@inproceedings{poesio2004general,
  title = {{A General-Purpose, Off-the-shelf Anaphora Resolution Module: Implementation and Preliminary Evaluation}},
  author = {Poesio, Massimo and Kabadjov, Mijail A.},
  booktitle = {Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC'04)},
  year = {2004},
  doi = {10.63317/2hw2na6qsij8},
  url = {https://doi.org/10.63317/2hw2na6qsij8}
}

References

15 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Alonso, L. and M. Fuentes, 2003. Integrating Cohesion and Coherence for Text Summarization. In Proc. of the EACL Student Session, pages 1–8, Budapest. find
  2. Boguraev, B. and C. Kennedy, 1997. Salience-based content characterisation of text documents. In Proc. of ACL/EACL Workshop on Summarization, pages 2–9. find
  3. Brew, C. and D. McKelvie and R. Tobin and H. Thompson and A. Mikheev, The XML Library LT XML Version 1.2, LT-XML Byron, D., 2002. Resolving Pronominal Reference to Abstract Entities. In Proc. of the 40th ACL.. find
  4. Byron, D. and J. Tetreault, 1999. A Flexible Architecture for Reference Resolution. In Proc. of the 9th EACL.. find
  5. Gaizauskas, R. and K. Humphreys, 2000. Quantitative Evaluation of Coreference Algorithms in an Information Extraction System. In S. Botley and T. McEnery (eds.), Corpus-based and Computational Approaches to Discourse Anaphora. Amsterdam / New York: John Benjamins, pages 145–169. find
  6. Grosz, B., J. Aravind, and W. Scott, 1995. Centering: A Framework for Modelling the Local Coherence of Discourse. Computational Linguistics, 21(2). find
  7. Hobbs, J.R., 1978. Resolving pronoun references. Lingua, 44, pages 311–338. find our card
  8. Kennedy, C. and B. Boguraev, 1996. Anaphora for everyone: Pronominal anaphora resolution without a parser. In Proc. of the 16th COLING, Budapest. find our card
  9. Morton, T. Coreference for NLP Applications. Proc. 38 th ACL Ng, V. and C. Cardie, 2002. Improving Machine Learning Approaches to Coreference Resolution. In Proc of ACL McKelvie, D., C. Brew, and H. Thompson, 1997. Using SGML as a Basis for Data-Intensive NLP. In Proc. of ANLP, Washington D.C. find
  10. Mitkov, R., 2002. Anaphora Resolution. find
  11. Longman Poesio, M. and R. Vieira, 1998. A corpus-based investigation of definite description use. Computational Linguistics, 24(2):183–216. Also available as Research Paper CCS-RP-71, Centre for Cognitive Science, University of Edinburgh. find
  12. Poesio, M., F. Bruneseaux and L. Romary, 1999. The MATE meta-scheme for coreference in dialogues in multiple languages. In Proc. of the ACL Workshop on Standards for Discourse Tagging. Maryland. find our card
  13. Poesio, M., Ishikawa, T., Sabine Schulte im Walde, and R. Vieira. 2002. Acquiring Lexical Information for Anaphora Resolution. Proc. of LREC Poesio, M., 2004. The MATE/GNOME annotation scheme for anaphora deixis, revisited. Proc. of SIGDIAL. find
  14. Vieira, R. and M. Poesio, 2000. An empirically-based system for processing definite descriptions. Computational Linguistics, 26(4). find
  15. Watson, R., J. Preiss, and E.J. Briscoe, 2003. The Contribution of Domain-independent Robust Pronominal Anaphora Resolution to Open-Domain Question-Answering. In Proc of Int. Symposium on Reference Resolution, Venezia find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/poesio-2004-general-purpose-off-shelf-anaphora/