Refinement complements verification and validation
Knowledge based systems are being applied in ever increasing numbers. The development of knowledge acquisition tools has eased the “Knowledge Acquisition Bottleneck”. More recently there has been a demand for mechanisms to assure the quality of knowledge based systems. Checking the contents of the knowledge base and the performance of the knowledge based systems at various stages throughout its life cycle is an important component of quality assurance. Hence, the demand now is for verification and validation tools. However, traditionally, verification and validation have identified possible faults in the knowledge base. In contrast, this paper advocates the use of knowledge refinement to correct identified faults in parallel with the ongoing verification and validation, thus easing the progress towards correct knowledge based systems. An automated refinement tool is described which uses the output from verification and validation tools to assemble evidence from which the refinement process can propose repairs. It is hoped that automated refinement in parallel with validation and verification may ease the “Knowledge V &V Bottleneck”.