Piecewise Synonym Construction Using WordNet for UMLS Source Integration
Kuo-Chuan Huang, James Geller, Michael H Halper · Medinfo 2007: Proceedings of the 12th World Congress on Health (Medical) Informatics; Building Sustainable Health Systems · 2007
Source vocabulary integration forms the basis of the UMLS's content and its planned continued growth. When integrating a new term, it is essential to recognize situations where the concept expressed by that term already exists in the UMLS, even if it is represented by a different term (i.e., a synonym). A preliminary study has shown that even when using the UMLS's rich set of synonyms, a new term for an existing concept may not match any of them. A process for automatically constructing additional, potential synonyms from multi-word source terms is presented. This process consists of decomposing the source term into words, finding separate synonyms for each in WordNet, and then recombining those synonyms into multi-word terms. We call such synonyms piecewise synonyms, and we present a precise formulation for their construction. The methodology is tested in the context of a (partial) reintegration of the 'Minimum Standard Terminology' into the UMLS. The results show that additional 'source term / UMLS term' matches are gained.