Inference Engineering through Prototyping in Prolog

Abstract

Knowledge acquisition is a tedious process when developing medical expert systems. The lack of formal reasoning inherent to medical decision making is partially responsible for this. Prototyping in Prolog, allowing numerous quick successive tests and progressive refinements, is an attractive method for medical experts as wel as for knowledge engineers to develop an inference engine (IE) for a particular problem under study. The clinician confronted with a decision simulating system, rapidly recognizes the most important necessary feature not yet present. He is then anxious to find out the effect of incorporating the new feature and to apply the revised inference engine on the existing knowledge base (KB). Independence between the KB and the IE is not only required, but also enhanced by this approach. We applied this technique to build CASETIP (a Computer Advice System for the Empiric Therapy in Pneumonia) using Micro-Prolog on a 256 Kbyte microprocessor.