Verification and Validation of Knowledge Bases Using Test Cases Generated by Restriction Rules

Knowledge based systems have been developed to solve many problems. Their main characteristic consists on the use of a knowledge representation of a specific domain to solve problems in such a way that it emulates the reasoning of a human specialist. As conventional systems, knowledge based systems are not free of failures. This justifies the need for validation and verification for this class of systems. Due to the lack of techniques which can guarantee their quality and reliability, this paper proposes a process to support validation of specific knowledge bases. In order to validate the knowledge base, restriction rules are used. These rules are elicit and represented as If Then Not rules and executed using a backward chaining reasoning process. As the result of this process test cases are created and submitted to the knowledge base in order to prove whether there are inconsistencies in the domain representation. Two main advantages can be highlighted here: the use of restriction rules which are considered as meta-knowledge (these rules improve the knowledge representation power of the system) and a process that can generate useful test cases (test cases are usually difficult and expensive to be created)