Increasing UMLS Coverage and Reducing Ambiguity via Automated Creation of Synonymous Terms

Background: Although extensive synonymy is one of the greatest strengths of the UMLS Metathesaurus, much research has nonetheless focused on identifying and measuring gaps in UMLS synonymy. This paper proposes a methodology for further extending the UMLS’ already rich synonymy by semi-automatically creating new strings not in the UMLS, and including them as additional synonymous strings within existing UMLS concepts. Results: In this paper we present our methodology for identifying missing UMLS synonymy and semi-automatically creating synonyms to fill these gaps. We created an enhanced Metathesaurus supplemented by these strings, and improved the performance on both biomedical literature and clinical text of two well known named-entity-recognition applications at the US National Library of Medicine, MetaMap and the Medical Text Indexer (MTI). Conclusions: Our methods propose first steps toward extending the already rich synonymy of the UMLS by filling in some synonymy gaps. We further theorize that some of the newly created strings could also be used to extend the Medical Subject Headings (MeSH) entry terms, and thereby enhance MEDLINE indexing and PubMed queries by better reflecting how authors actually refer to biomedical concepts in the literature.