An Empirical Study of Lexical Variation Methods for Biomedical Information Retrieval

Abstract

The rapid growth in the volume of biomedical litera-ture has made it increasingly difficult to find precise information. To implement an accurate and fast bio-medical information retrieval (IR) system domain, we must deal with the variants of biomedical terms care-fully. In this paper, we focus on the lexical variants of query terms and propose eight lexical variation rules. Based on the biological IR tasks defined for the Ge-nomic Track, we design a series of experiments to examine the effects of these eight rules and combina-tions thereof. We evaluate the rules by three indicators: MAP, recall, and query term frequency. Our experi-ment results show that varying hyphenation signifi-cantly improves the performance of information re-trieval. In addition, the variation rules for Greek transcriptions also increase IR performance when a query contains terms with such transcriptions. How-ever, the rules that generate general variants tend to undermine MAP scores.