An Empirical Study of Word Sense Disambiguation for Biomedical Information Retrieval System

Abstract

Document representation is an important stage to ensure the indexation of biomedical document. The ordinary way to represent a text is a bag of words BoW, This Representation suffers from the lack of sense in resulting representations ignoring all semantics that reside in the original text; instead of, the Conceptualization using background knowledge enriches document representation models. Three strategies can be used in order to realize the conceptualization task: Adding Concept, Partial Conceptualization, and Complete Conceptualization. While searching polysemic term corresponding senses in semantic resources, multiple matches are detected then introduce some ambiguities in the final document representation, three strategies for Disambiguation can be used: First Concept, All Concepts and Context-Based. SenseRelate is a well-known Context-Based algorithm, which uses a fixed window size and taking into consideration the distance weight on how far the terms in the context are from the target word. This may impact negatively on the yielded concepts or senses, we propose a simple modified version of SenseRelate algorithm namely NoDistanceSenseRelate, which simply ignore the distance that is the terms in the context will have the same distance weight. In order to evaluate the effect of the conceptualization strategies and Disambiguation strategies in the indexing process, in this study, several experiments have been conducted using OHSUMED corpus on a biomedical information retrieval system. The obtained results using OHSUMED corpus show that the Context-Based methods (SenseRelate and NoDistanceSenseRelate) outperform the others ones when applying Adding Concept Conceptualization strategy results using Biomedical Information retrieval system. The obtained results prove the evidence of adding the sense of concepts to the Term Representation in the IR process.