Module extraction in expressive ontology languages via datalog reasoning

摘要: Module extraction is the task of computing a (preferably small) fragment M an ontology O that preserves class entailments over signature interest Σ. Extracting modules minimal size well-known to be computationally hard, and often algorithmically infeasible, especially for highly expressive languages. Thus, practical techniques typically rely on approximations, where provably captures relevant entailments, but not guaranteed minimal. Existing approximations ensure all second-order w.r.t. Σ, which stronger condition than required in many applications, may lead unnecessarily large practice. In this paper we propose novel approach module reduced reasoning problem datalog. Our generalises existing elegant way. More importantly, it allows are tailored preserve only specific kinds thus significantly smaller. evaluation wide range ontologies confirms feasibility benefits our