| Abstract | In the ontology classification task, consequence-based reasoners are typically significantly faster while tableau-based reasoners can process more expressive DL languages. However, both of them have difficulty to classify some available large and complex ALCHOI ontologies with complete results in acceptable time. We present a prototype hybrid reasoning system WSReasoner, which is built upon and takes advantages of both types of reasoners to provide efficient classification service. In our proposed approach, we approximate the target ontology O by a weakened version Owk and a strengthened version Ostr , both are in a less expressive DL ALCH and classified by a consequence-based main reasoner. Classification of Owk produces a subset of subsumptions of ontology O and the target of the classification of Ostr is to produce a superset of subsumptions of O. Additional subsumptions derived from Ostr may be unsound, so they are further verified by a tableau-based assistant reasoner. For the ALCHOI ontologies in our experiment, except for one for which WSReasoner has not obtained the result, the number of subsumptions derived from WSReasoner is no fewer than from the reasoners that could finish the classification. Moreover, WSReasoner takes less time than tableau-based reasoners when the ALCHOI ontologies are large. |
|---|
