Logic Journal of IGPL · 2000 · 359 citations · 30 references
ComputingComputer ScienceKnowledge RepresentationLogicTheory Of ComputationAutomated ReasoningConstraint SatisfactionDescription LogicsModel TheoryDecision ProceduresDescription Logic ReasoningExpressive Description LogicsInverse RolesNumber RestrictionsRole HierarchiesTransitive RolesWorst-case Complexity
Description Logics are knowledge‑representation formalisms that build complex concepts and roles from atomic ones, and while expressive role constructors are useful, they often pose computational challenges. The authors present an algorithm for deciding satisfiability in ALC extended with transitive, inverse, and functional roles under general concept inclusion axioms and role hierarchies, aiming for practical implementability. They develop this algorithm, analyze decidability limits—showing that relaxing role restrictions in number restrictions leads to undecidability—and introduce optimisation techniques that enable efficient implementations despite high worst‑case complexity. The study shows that ALC with transitive and inverse roles remains PSPACE‑complete and that the optimised decision procedures perform well on real‑life problems.
Description Logics (DLs) are a family of knowledge representation formalisms mainly characterised by constructors to build complex concepts and roles from atomic ones. Expressive role constructors are important in many applications, but can be computationally problematical. We present an algorithm that decides satisfiability of the DL ALC extended with transitive and inverse roles and functional restrictions with respect to general concept inclusion axioms and role hierarchies; early experiments indicate that this algorithm is well-suited for implementation. Additionally, we show that ALC extended with just transitive and inverse roles is still in PSPACE. We investigate the limits of decidability for this family of DLs, showing that relaxing the constraints placed on the kinds of roles used in number restrictions leads to the undecidability of all inference problems. Finally, we describe a number of optimisation techniques that are crucial in obtaining implementations of the decision procedures, which, despite the high worst-case complexity of the problem, exhibit good performance with real-life problems.
30
A machine program for theorem-proving
Martin Davis, George Logemann, Donald Loveland · Communications of the ACM · 1962 · 3.2K citations · Full text
Attributive concept descriptions with complements
Manfred Schmidt-Schauß, Gert Smolka · Artificial Intelligence · 1991 · 964 citations
Peter F. Patel‐Schneider · 1998 · 530 citations