A hybrid information retrieval system for medical field using MeSH ontology

Abstract

Using semantic relations between different terms beside their syntactical similarities in a search engine would result in systems with better overall precision. One major problem in achieving such systems is to find an appropriate way of calculating semantic similarity scores and combining them with those of classic methods. In this paper, we propose a hybrid approach for information retrieval in medical field using MeSH ontology. Our approach contains proposing a new semantic similarity measure and eliminating records with semantic score less than a specific threshold from syntactic results. Proposed approach in this paper outperforms VSM, graph comparison, neural network, Bayesian network and latent semantic indexing based approaches in terms of precision vs. recall. © 2009 Springer Berlin Heidelberg.