The inherent vagueness and ambiguity of non-monotonic reasoning makes it impossible to formulate detailed specifications to validate KBS performance by using traditional test-case-based approach. In this paper, we present a practical validation technique for non-monotonic KBSs, which utilizes automatically generated test cases instead. We show how such test cases and other useful validation knowledge can be acquired from operational theories, which are functionally equivalent to the original non-monotonic theory, and how this validation knowledge can be used by domain experts to evaluate the validity of the knowledge model.
Subjects: 1.7 Expert Systems; 3.3 Nonmonotonic Reasoning
Submitted: Feb 13, 2006