Lexically-Suggested Hyponymic Relations Among Medical Terms and Their Representation in the UMLS.
Bodenreider O, Burgun A, Rindflesch TC
Proc. of TIA'2001; Terminology and Artificial Intelligence. 2001;:11-21.
Abstract:
Objective: Among the various methods for identifying thesaurus relations from text corpora, methods based on head modifier relation are interesting in the context of medical terminologies, especially for those terms which differ from one another by only one modifier. Adjectival modifiers play a particular role because they usually introduce a hyponymic relation. This study focuses on comparing lexically-suggested hyponymic relations among medical terms to inter-concept relationships represented in the Unified Medical Language System (UMLS) Metathesaurus. Methods: Adjectival modifiers were identified from 63,000 medical terms from the UMLS Metathesaurus, and transformed terms were generated by removing them from the original terms. Candidate hyponymic relations were then tested against inter-concept relationships recorded in the UMLS Metathesaurus. Results: In 50% of the cases, suggested hyponymic relations were present in the UMLS Metathesaurus. In 25% of the cases, the original term and the transformed terms were "siblings" in the UMLS. In the remaining 25%, no relationship was recorded in the UMLS between these two terms. The lack of relationships observed in the UMLS Metathesaurus is discussed. Additional methods for automatically assessing the suggested hyponymic relations are proposed. Further research directions are briefly presented.
Bodenreider O, Burgun A, Rindflesch TC. Lexically-Suggested Hyponymic Relations Among Medical Terms and Their Representation in the UMLS.
Proc. of TIA'2001; Terminology and Artificial Intelligence. 2001;:11-21.