Efficient detection of potential inconsistency in taxonomic knowledge with uncertainty

Abstract

The authors present the logical framework for detection of potential conflicts in knowledge bases with uncertainty. In the solution, it is assumed that the uncertainty measure is modeled by the possibilistic necessity measure. The method presented allows the modeling of the effect of a user defined certainty threshold for belief propagation, and utilization of a partially inconsistent knowledge base. An efficient computation method is presented which is applicable for knowledge in a certain simple form, typically satisfied by a taxonomic knowledge base. The deductive system implemented by this method deals properly with cycles, and is both sound and complete.<>