Résumé
Domain specific ontologies are invaluable but their development faces many challenges. In most cases, domain knowledge bases are built with very limited scope without considering the benefits of including domain knowledge to a general ontology. Furthermore, most existing resources lack meta-information about association strength (weights) and annotations (frequency information like frequent, rare ... or relevance information like pertinent or irrelevant). In this paper, we are presenting a semantic resource for radiology built over an existing general semantic lexical network (JeuxDeMots). This network combines weight and annotations on typed relations between terms and concepts. Some inference mechanisms are applied to the network to improve its quality and coverage. We extend this mechanism to relation annotation. We describe how annotations are handled and how they improve the network by imposing new constraints especially those founded on medical knowledge. 1 Introduction For more than two decades, medical practice and bio-medical research have benefited from the availability of biomedical ontologies (Bodenreinder, 2008). These resources are used for semantic analysis such as entity recognition (i.e., the identification of biomedical entities in texts as name of genes, disease, etc.), and relation extraction (i.e., the identification of semantic relationships among biomedical entities like for instance interaction between proteins). In the framework of the UMLS project, which interrelates some 60 controlled vocabularies, an upper-level ontology, the UMLS semantic network (Lomax, 2004) has been built. In the field of radiology, such a semantic network is used to facilitate or automate the analysis of radiologist reports in order to extract recommended courses of action or to trigger warning systems to improve patient management (Yetisgen-Yildiz and al., 2013). There exist reference on-tologies in biomedical domain (UMLS), but they might not be suited to a particular domain like radiology because result sets are too large and too complex (Mejino 2008). To solve this problem, the Radiology Society of North America (RSNA) has created reference ontology for radiology RadLex (Rubin, 2008). RadLex and its derivatives rely on English and are not considered medically complete (Hong, 2012).