Consistent query answering is a standard approach for producing meaningful query answers when data is inconsistent.
Recent work on consistent query answering in the presence of ontologies has shown this problem to be intractable in data complexity even for ontologies expressed in lightweight description logics. In order to better understand the source of this intractability, we investigate the complexity of consistent query answering for simple ontologies consisting only of class subsumption and class disjointness axioms. We show that for conjunctive queries with at most one quantified variable,
the problem is first-order expressible; for queries with at most two quantified variables, the problem has polynomial data complexity but may not be first-order expressible; and for three quantified variables, the problem may become
coNP-hard in data complexity. For queries having at most two quantified variables, we further identify a necessary and sufficient condition for first-order expressibility. In order to be able to handle arbitary conjunctive queries, we propose a novel inconsistency-tolerant semantics and show that
under this semantics, first-order expressibility is always guaranteed. We conclude by extending our positive results to
DL-Lite ontologies without inverse.
Activités de recherche
°
Ontologies °
Logique °
Complexité °
Raisonnement automatique
Equipe
°
Intelligence Artificielle et Systèmes d'Inférence
Contact
[aucun]