EXPERT-SYSTEM VALIDATION THROUGH KNOWLEDGE-BASE REFINEMENT

Knowledge base (KB) refinement is a suitable technique to Support expe rt system (ES) validation. When used for validation, KB refinement sho uld be guided not only by the number of errors to solve but also by th e importance of those errors. Most serious errors should be solved fir st, even causing other errors of lower importance but assuring a neat validity gain. These are the bases for IMPROVER, a KB refinement tool designed to support ES validation. IMPROVER refines ES for medical dia gnosis with this classification of error importance: false negative > false positive > ordering mismatch. IMPROVER has been used to support the validation of PNEUMON-IA, a real ES on the medical domain. After r efinement, the ES validity has increased substantially. Detailed evide nce of this improvement is provided, as well as examples of how the re finement process was pet-formed. (C) 1996 John Wiley & Sons, Inc.