Enrichissement sémantique associé à la détection de la négation et des antécédents familiaux dans un entrepôt de données hospitalier

(Background) We developed in previous work a biomedical data warehouse for mining clinical reports by full text search engine. But the lack of consideration of negation and family history generate noise. (Methods) We developed an algorithm to detect the French notion of negation in medical reports and he context of family history. The semantic enrichment reflects this new semantic information. We integrate in the data warehouse schema and in data mining algorithm this new information to improve the performance of the search engine. (Results) the integration of the detection of negation and family history significantly improves the precision of the search engine despite a slight decrease in recall. (Conclusion) The syntactic and semantic enrichment has been successfully integrated into the data warehouse. Mots-clés : Entrepôt de données ; fouille de données ; Traitement automatique du langage