Exploring Context-Sensitive Query Reformulation in a Biomedical Digital Library

We propose a novel semantic query expansion technique that enables inference of contextual information in queries and user information. In the present study, we detect and map bio entities such as gene, protein, and disease in a query to concept tuples, and incorporate user context data based on the PubMed query logs and user profile into the algorithm. In objective evaluation, we can see a concept tuple aided with UMLS concepts adds semantic information to the initial query. In subjective evaluation, we find that in a context-enabled search environment, where context terms that the users are interested in are combined into their initial search terms, users tend to assign higher relevance scores to the retrieval results by these queries.