Related Experiment Videos
Disambiguating ambiguous biomedical terms in biomedical narrative text: an unsupervised method.
H Liu1, Y A Lussier, C Friedman
1Computer Science Division, Graduate School and University Center, City University of New York, New York, New York 10016, USA. hol7001@dmi.columbia.edu
Journal of Biomedical Informatics
|April 30, 2002
Summary
This study introduces an unsupervised method for biomedical word sense disambiguation (WSD). The approach efficiently assigns correct meanings to ambiguous terms, achieving high accuracy.
Area of Science:
- Biomedical Informatics
- Computational Linguistics
Background:
- Natural Language Processing (NLP) is increasingly used in biomedicine for information extraction and concept indexing.
- Current biomedical word sense disambiguation (WSD) relies on manual, handcrafted rules, which are time-consuming, difficult to maintain, and often incomplete.
Purpose of the Study:
- To develop an efficient, unsupervised method for biomedical word sense disambiguation (WSD).
- To overcome the limitations of handcrafted rules in WSD for the biomedical domain.
Main Methods:
- A two-phase unsupervised approach was developed to build a WSD classifier.
- Phase one automatically generates a sense-tagged corpus for an ambiguous biomedical term.
- Phase two derives a WSD classifier using the automatically generated corpus as training data.
Main Results:
- The developed WSD classifiers achieved approximately 97% overall accuracy.
- Individual ambiguous terms demonstrated over 90% accuracy in sense disambiguation.
- The unsupervised method proved effective and efficient for biomedical WSD.
Conclusions:
- The proposed two-phase unsupervised method offers an effective solution for biomedical word sense disambiguation.
- This approach significantly improves upon traditional handcrafted rule-based systems.
- The method demonstrates high accuracy and applicability across different biomedical text genres.