Selective Source for Query Expansion in Health Information Retrieval

Abstract

Query expansion aims to solve the vocabulary mismatch problem by adding new terms to the original query or
reweighting existing query terms. Since there are external sources available, it challenges to select the right source for
each query expansion. Useful terms of the right expansion source result in increasing the rank of relevance documents.
We propose the selective source expansion in health information retrieval framework. Our source selection method is based
on the relative entropy of two probability distributions. These probability distributions estimate from pairs of terms in the
query and pairs of terms in the collection, respectively, instead of the individual query term. The proposed framework
improves retrieval performance from baseline retrieval, traditional query expansion, and existing selective query
framework as well.