LinkedMDR: un modèle sémantique de représentation de corpus de documents multimédia

Projects, in the construction industry, involve the exchange of a large amount of information between several actors having different expertise and interests. Most of this information is unstructured, originated from different sources and dispersed across heterogeneous documents, thus producing implicit and explicit dependencies between them. This becomes very critical as it makes the annotation of the documents and the information retrieval more challenging at any stage of a building life cycle. In this work, we propose LinkedMDR: a novel ontology for Linked Multimedia Document Representation. Our ontology is based on the integration of the three standards addressing metadata and content representation: Dublin Core (DC), Text Encoding Initiative (TEI), and Moving Picture Experts Group (MPEG-7) together with the addition of new components offering more features especially in representing the collective knowledge of a document corpus. LinkedMDR is generic and offers, as well, a pluggable layer handling the particularities of a domain-specific knowledge. Experiments measure the efficiency and the effectiveness of our solution in comparison with the existing standards. MOTS-CLÉS : Documents hétérogènes; Modèle de documents; Ontologies; Système de Recherche d’Information