Elamur Research⌕
☰
home / papers / 2012 / Domain Adaptation of Coreference Resolution for Radiology Reports

Domain Adaptation of Coreference Resolution for Radiology Reports

Summary

General-purpose coreference tools such as BART, the Stanford Deterministic Coreference Resolution System and GATE were designed for newswire-style text and fit radiology reports poorly. In these reports, definite noun phrases (for example, mentions of an anatomical structure) usually point to something visible in the accompanying image rather than to earlier text. Repeated mentions of the same concept are also already unified when information extraction pipelines map them to vocabularies like UMLS or RadLex. The authors therefore narrow the task. Only demonstrative pronouns and non-pleonastic "it" mentions are linked, each to one previously tagged medical concept or to NULL. Their resolver is unsupervised and heuristic, inspired by Haghighi and Klein (2009). It discards candidates outside a short window, with mismatched grammatical number or with a different semantic category, then selects the nearest remaining one. On 150 manually annotated reports it reaches a pairwise F1 of 58.66, with precision noticeably higher than recall. The study shows that redefining coreference around what clinical extraction actually requires gives a practical, precision-oriented component for radiology NLP.

Key points

  • Data: 300 reports were sampled from a proprietary Northwestern University Medical School collection of more than 100,000 de-identified reports spanning 9 years. They cover 100 CT Chest, 100 Ultrasound Breast and 100 MR Brain reports. Half were used for validation and half were hand-annotated for testing.

  • Pronoun types: The validation set contained no personal, possessive or indefinite pronouns. Only demonstratives and referential "it" appeared.

  • Algorithm: Mentions are ordered by head position. Candidate antecedents are filtered in sequence:

    • they must fall within the previous two sentences;
    • they must agree in number;
    • they must share a semantic category (same UMLS semantic type or same RadLex parent).

    The closest surviving candidate is chosen, and links are propagated transitively.

  • Results: Pairwise precision is 74.90, recall 48.22 and F1 58.66. This falls within the 39.8–67.3 F1 range that Haghighi and Klein (2009) report for various methods and test sets. Precision was deliberately prioritized, so most errors are missed links rather than wrong ones.

  • Code availability: The paper says source code is available on request, while its conclusion describes the system as publicly available. The annotated data is not released.

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

5 papers of our library cite it (of the 140673 whose bibliographies we read); 2 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{apostolova2012domain,
  title = {{Domain Adaptation of Coreference Resolution for Radiology Reports}},
  author = {Apostolova, Emilia and Tomuro, Noriko and Mongkolwat, Pattanasak and Demner-Fushman, Dina},
  booktitle = {BioNLP: Proceedings of the 2012 Workshop on Biomedical Natural Language Processing},
  year = {2012},
  url = {https://aclanthology.org/W12-2414/}
}

References

11 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. J. Castano, J. Zhang, and J. Pustejovsky. 2002. Anaphora resolution in biomedical literature. In International Symposium on Reference Resolution. Citeseer. find
  2. D.H. Cunningham, D.D. Maynard, D.K. Bontcheva, and M.V. Tablan. 2002. GATE: A Framework and Graphical Development Environment for Robust NLP Tools and Applications. find
  3. C. Gasperin, N. Karamanis, and R. Seal. 2007. Annotation of anaphoric relations in biomedical full-text articles using a domain-relevant scheme. In Proceedings of DAARC, volume 2007. Citeseer. find
  4. C. Gasperin. 2006. Semi-supervised anaphora resolution in biomedical texts. In Proceedings of the Workshop on Linking Natural Language Processing and Biology: Towards Deeper Biological Literature Analysis, pages 96–103. Association for Computational Linguistics. find
  5. A. Haghighi and D. Klein. 2009. Simple coreference resolution with rich syntactic and semantic features. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 3-Volume 3, pages 1152–1161. Association for Computational Linguistics. find our card
  6. V. Ng and C. Cardie. 2002. Improving machine learning approaches to coreference resolution. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 104–111. find our card
  7. K. Raghunathan, H. Lee, S. Rangarajan, N. Chambers, M. Surdeanu, D. Jurafsky, and C. Manning. 2010. A multi-pass sieve for coreference resolution. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 492–501. Association for Computational Linguistics. find our card
  8. W.M. Soon, H.T. Ng, and D.C.Y. Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational Linguistics, 27(4):521– 544. find our card
  9. J. Su, X. Yang, H. Hong, Y. Tateisi, J. Tsujii, M. Ashburner, U. Leser, and D. Rebholz-Schuhmann. 2008. Coreference resolution in biomedical texts: a machine learning approach. Ontologies and Text Mining for Life Sciences 08. find
  10. Y. Versley, S.P. Ponzetto, M. Poesio, V. Eidelman, A. Jern, J. Smith, X. Yang, and A. Moschitti. 2008. Bart: A modular toolkit for coreference resolution. In Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies: Demo Session, pages 9–… find our card
  11. J. Zheng, W.W. Chapman, R.S. Crowley, and G.K. Savova. 2011. Coreference resolution: A review of general methodologies and applications in the clinical domain. Journal of biomedical informatics. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/apostolova-2012-domain-adaptation-coreference-resolution-radiology/