Incremental and Persistent Reasoning in FaCT++

ABSTRACT. Reasoning in complex DLs is very expensive. However it is possible to identify groups of scenarios in which the ontology in question is not changed at all (e.g., query answering) or the changes are negligible comparing to the whole ontology (e.g., manual editing). To support these scenarios, FaCT++ implement the two techniques. In persistent reasoning after classification the reasoner saves its internal state together with computed information (e.g., class taxonomy) on a hard drive, so on the next run with the same ontology it could save significant amount of work by loading this information instead of computing it. Within the incremental reasoning mode the reasoner accepts change and identify a (usually small) portion of internal state that is affected by the change. Affected areas (e.g. parts of a class hierarchy) is then reasoned about using new information. This approach is in a number of cases more efficient than complete reload of an ontology.