Automatic Classification and Visualization of UMLS Source Vocabularies through Semantic Group Profiles

The Unified Medical Language System® (UMLS) is a comprehensive terminology integration system designed to support the development of electronic information systems. The UMLS integrates 161 source vocabularies, though for a given purpose, a developer may not need every vocabulary. With the breadth of vocabularies available, there is a need for classifying the UMLS source vocabularies with respect to their content. We describe a technique for automatic content-based classification of source vocabularies using semantic group profiles. We also present different graphical representations of the source vocabularies, in order to facilitate the comparison and selection of sources. Finally, we compare the content-based classification to a manually created usage-based classification. Our classification can be easily produced for upcoming versions of the UMLS and is being integrated as part of the UMLS documentation.