Résumé
We consider the integration of heterogeneous data under a global RDF graph data model, based on the knowledge expressed in an ontology describing the concepts relevant for an application, and a set of entailment rules which characterize the logical relationships between these concepts. We consider both standard RDF Schema rules and user-specified ones. We propose a data integration architecture to compute certain query answers in this setting. Existing approaches to query answering in the presence of knowledge (expressed here in the ontology and the entailment rules) involve either the materialization of inferences in the data or the refor-mulation of the query. Both approaches have well-known drawbacks. We introduce a new approach to query answering , based on a reduction to view-based query answering. This approach avoids both materialization in the data and query reformulation. We define restrictions of our general architecture under which our method is correct, and formally prove its correctness.