Identifying and Handling Structural Incompleteness for Validation of Probabilistic Knowledge-Bases

Eugene Santos Jr., University of Connecticut; Sheila B. Banks, Calculated Insight; Scott M. Brown and David J. Bawcom, Air Force Institute of Technology

The PESKI (Probabilities, Expert Systems, Knowledge, and Inference) system attempts to address some of the problems in expert system design through the use of the Bayesian Knowledge Base (BKB) representation. Knowledge gathered from a domain expert is placed into this framework and inferencing under uncertainty is performed over it. However, by the nature of BKBs, not all knowledge is incorporated, i.e. the representation need not be a complete representation of all combinations and possibilities of the knowledge, as this would be impractical in many real-world systems. Therefore, inherent in such a system is the problem of incomplete knowledge, or gaps within the knowledge base where areas of lacking knowledge preclude or hinder arrival at a solution. Some of this knowledge is intentionally omitted because its not needed for inferencing, while other knowledge is erroneously omitted but necessary for valid results. Intentional omission, a strength of the BKB representation, allows for capturing only the relevant portions of knowledge critical to modeling an expert’s knowledge within a domain. This research proposes a method for handling the latter form of incompleteness administered through a graphical interface. The goal is to detect incompleteness and be corrected by a knowledge engineer in an intuitive fashion.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.