Biomedical term disambiguation: An application to gene-protein name disambiguation

The huge volumes of biomedical texts available online drives the increasing need for automated techniques to analyze and extract knowledge from these repositories of information. Resolving the ambiguity in biological terms in these texts is an important step for developing efficient knowledge discovery techniques. In this paper, we present a new method for biomedical term disambiguation in biomedical texts. The method is based on machine learning and can be viewed as a word classification task. We evaluated the method on geneprotein name disambiguation using Medline abstracts from years 1999-2003 containing about 3000 to 6000 gene and protein names. The technique is effective in disambiguating gene and protein names, achieving impressive accuracy, precision, and recall, with accuracy approaching about 90%, and outperforming the recently published results on this problem. Our technique is also applicable for the general problem of named entity disambiguation.