An efficient and large-scale reasoning method for the semantic Web

Abstract

We present an extended version of the $\mathcal {C}\mathcal {E}\mathcal {D}\mathcal {A}\mathcal {R}$ taxonomic reasoner for large ontologies. This new version provides fuller support for TBox reasoning, checking consistency, and retrieving instances. The $\mathcal {C}\mathcal {E}\mathcal {D}\mathcal {A}\mathcal {R}$ system is based upon the $\mathcal {O}\mathcal {S}\mathcal {F}$ formalism. It is implemented on an entirely new architecture which includes several optimization techniques. We define a bidirectional mapping between $\mathcal {O}\mathcal {S}\mathcal {F}$ graph structures and the Resource Description Framework (RDF) allowing a translation from $\mathcal {O}\mathcal {S}\mathcal {F}$ queries into SPARQL for retrieving instances from an RDF triplestore. We carried out comparative performance evaluation experiments using $\mathcal {C}\mathcal {E}\mathcal {D}\mathcal {A}\mathcal {R}$ as well as well-known Semantic Web reasoners (such as FaCT++, Pellet, HermiT, TrOWL, and RacerPro) on very large public ontologies. For the same queries on the same ontologies, the results achieved by $\mathcal {C}\mathcal {E}\mathcal {D}\mathcal {A}\mathcal {R}$ were compared to those obtained by all the other reasoners. The results of experiments show that $\mathcal {C}\mathcal {E}\mathcal {D}\mathcal {A}\mathcal {R}$ consistently performs on a par with the fastest systems for concept classification, and several orders of magnitude more efficiently in terms of response time for Boolean query-answering over attributed concepts, as well as for ABox triplestore querying. The latter result is irrespective of the triplestore management used because the $\mathcal {C}\mathcal {E}\mathcal {D}\mathcal {A}\mathcal {R}$ reasoner uses its knowledge to optimize SPARQL queries before submitting them to the triplestore.