Ontology Graph Based Query Expansion for Biomedical Information Retrieval

Query expansion based biomedical information retrieval has been studied for over two decades; most of the studies focus only on taking advantage of one vocabulary: MeSH. We propose a completely different approach utilizing an arbitrary number of controlled vocabularies from Metathesaurus. Experiment shows that our ontology based query expansion scheme achieves 8.2% and 17.7% improvement compared with schemes using pseudo relevance feedback query expansion and using no query expansion respectively. The average improvement is 24.8% in comparison to all other existing strategies. Furthermore, we identify that generalized biomedical concepts are the reason for performance degradation.