PubTermVariants: biomedical term variants and their use for PubMed search

Term normalization is frequently used in information retrieval task to reduce variant word forms to a common form. The most general term normalization technique used in practice is stemming, however it has been found to not be completely reliable. Here we present PubTermVariants, a high-quality data-driven resource of term variant pairs that can improve search results in PubMed. For a given pair, we consider two terms to be variants if they stem to the same form, pass the hypergeometric test, and pass the morpho-semantic test. We perform manual evaluation of a subset of PubTermVariants that confirms the high quality of the candidate pairs. We further present experiments that demonstrate their usefulness for PubMed search.