Automated Deduction Techniques for Classification in Description Logic Systems

Mechanical theorem provers are becoming increasingly more powerful, and we believe that it is time to examine whether certain tasks that have formerly been accomplished by other means can now be performed efficiently by a theorem prover. One such task is classification in description logic-based knowledge representation systems. Description logic systems provide a formalism for expressing knowledge based on concepts and roles. Subsumption checking is one important reasoning faculty offered by such systems. In this article we use a theorem prover coupled with a finite-model finder to perform subsumption checking. This approach is complete and sound for description logic systems whose underlying logic has the finite model property. The performance is compared with several other well-known description logic systems. Some efficient strategies to compute the subsumption hierarchy, known as classification, are also described.