Reasoning in Description Logics by a Reduction to Disjunctive Datalog

Abstract

As applications of description logics proliferate, efficient reasoning with knowledge bases containing many assertions becomes ever more important. For such cases, we developed a novel reasoning algorithm that reduces a $\mathcal{SHIQ}$ knowledge base to a disjunctive datalog program while preserving the set of ground consequences. Queries can then be answered in the resulting program while reusing existing and practically proven optimization techniques of deductive databases, such as join-order optimizations or magic sets. Moreover, we use our algorithm to derive precise data complexity bounds: we show that $\mathcal{SHIQ}$ is data complete for NP, and we identify an expressive fragment of $\mathcal{SHIQ}$ with polynomial data complexity.