Assisting the Translation of SNOMED CT into French

The objective of this study is to evaluate to approaches assisting the translation of SNOMED CT into French. Two types of approaches were combined: a concept-based one, which relies on conceptual information of the UMLS Metathesaurus and a lexical-based one, which relieson NLP techniques. In addition to the French terminologies (whether included in UMLS or not). Using the concept-based approach, a set of 156,157 (39.4%) SNOMED CT terms were translated to at least one French term from UMLS. Expanded to the French terms from UMLS terminologies translated by CISMeF, 2,548 (+0.7%) additional SNOMED CT terms were translated to at least one French term. Using the lexical-based approach, a set of 145,737 (36.8%) SNOMED CT terms were translated to at least one French term from HeTOP. The qualitative evaluation showed that 44% of the translations were rated as "relevant". Overall, the two approaches have provided the translation of 168,750 (42.6%) SNOMED CT terms into French using different bilingual terminological sources included in UMLS or in HeTOP.