Compilers and knowledge dictionaries for expert systems: inference engines of the future

摘要

For the most part, the current technology of expert system inference engines only provides basic coverage of its rule base, i.e. they do not go beyond the simple matching of symptoms to rules. Today's inference engines do not provide any dynamic reorganization or substitution when certain symptoms are either unavailable as session input or not included in the knowledge base. An expert system prototype CompilerlKnowledge Dictionary Expert System (CKDES) is developed to demonstrate how load, compile, sort, and grouping techniques can both simplify and improve the organization/ reorganization of knowledge bases. The CKDES prototype can take “ill-defined” (unorganized) text knowledge, load (scan) it in, and reorganize it into a workable and efficient knowledge base. A “knowledge dictionary” is generated during the knowledge base load1compile process. A “knowledge dictionary” (KD) is here defined as a repository (file) containing the definitions of all symptoms and diagnoses used by the knowledge base. Inference engines of the future can use this KD to dynamically substitute for unavailable symptoms and reorganize decision tree constructs. The CKDES prototype is tested in a nursing diagnostic application. Results of this experiment are presented.