Dealing with Inconsistencies and Updates in Description Logic Knowledge Bases

The main purpose of an "Ontology-based Information System" (OIS) is to provide an explicit description of the domain of interest, called ontology, and let all the functions of the system be based on such representation, thus freeing the users from the knowledge about the physical repositories where the real data reside. The functionalities that an OIS should provide to the user include both query answering, whose goal is to extract information from the system, and update, whose goal is to modify the information content of the system in order to reflect changes in the domain of interest. The "ontology" is a formal, high quality intentional representation of the domain, designed in such a way to avoid inconsistencies in the modeling of concepts and relationships. On the contrary, the extensional level of the system, constituted by a set of autonomous, heterogeneous data sources, is built independently from the conceptualization represented by the ontology, and therefore may contain information that is incoherent with the ontology itself. This dissertation presents a detailed study on the problem of dealing with inconsistencies in OISs, both in query answering, and in performing updates. We concentrate on the case where the knowledge base in the OISs is expressed in Description Logics, especially the logics of the DL-lite family. As for query answering, we propose both semantical frameworks that are inconsistency-tolerant, and techniques for answering unions of conjunctive queries posed to OISs under such inconsistency-tolerant semantics. As for updates, we present an approach to compute the result of updating a possibly inconsistent OIS with both insertion and deletion of extensional knowledge.