Using Bayesian Network for Conceptual Indexing: Application to Medical Document Indexing with UMLS Metathesaurus
Résumé
We describe a conceptual indexing method using UMLS meta-thesaurus. Concepts are automatically mapped from text us- ing MetaMap software tool for English, and a simplified mapping tool for other languages. The concepts and their semantic links given by UMLS are used to build a Bayesien network. Retrieval process is then an inference process of probabilities or weights. Dif- ferent types of relations are experimented in this model to evaluate their efficiency in retrieval.