Comparing terms, concepts and semantic classes in WordNet and the Unified Medical Language System

Objectives : The objective of this study is to compare how a general terminological system (WordNet) and a domain-specific one (UMLS) represent linguistic and knowledge phenomena at three different levels: terms, concepts, and semantic classes. Methods : For one general class (ANIMAL ) and one domain-specific class (HEALTH DISORDER ), the set of concepts corresponding to the class was established. Then, for each semantic class, the corresponding terms were mapped from one system to the other, both ways. Results : Only 2% of the domain-specific concepts from UMLS were found in WordNet, but 83% of the domain-specific concepts from WordNet were found in the UMLS. Concept overlap between the two systems varies from 48% to 97%. Discussion : Missing terms in both systems are discussed, as well as granularity and knowledge organization issues.