Knowledge Base Refinement as Improving an Incorrect and Incomplete Domain Theory
Abstract
The ODYSSEUS program automates knowledge base refinement by improving a domain theory. This paper describes the techniques used by ODYSSEUS to address three types of domain theory pathologies: incorrectness, inconsistency, and incompleteness. In ODYSSEUS, an incomplete domain theory is extended by the Metarule Chain Completion Method. This method exploits the use of an explicit meta-level representation of the strategy knowledge for a generic problem class (e.g., heuristic classification) that is separate from the domain theory (e.g., medicine) to be improved. Our work implements and compares the extension of an incompleted domain theory using case-based inductive learning and explanation- based apprenticeship learning; in the latter, learning occurs by completing failed explanations of observed human problem-solving actions. Extending an incomplete domain theory and correcting an incorrect domain theory both use the Confirmation Decision Procedure Method, which validates arbitrary instantiated tuples of the knowledge base by the use of an underlying domain theory. Lastly, the consistency of the knowledge base is improved by use of the Sociopathic Reduction Algorithm. (kr)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1990
- Accession Number
- ADA224441
Entities
People
- David C. Wilkins
Organizations
- University of Illinois Urbana–Champaign