An Experiment in Direct Knowledge Acquisition

Peter W. Mullarkey

LQMS is a knowledge-based system that identifies and explains anomalies in data acquired from multiple sensors. The knowledge base was built by a sequence of domain experts. Its prototype performed with a high level of accuracy and that performance has been incrementally and significantly improved during development and field testing. Several points are developed in this paper. (1) The combination of an intuitive model (sufficient for the task) and powerful, graphical development tools allowed the domain experts to build a large, high performance system. (2) The Observation-Situation-Blation representation illustrates an intermediate point on the simplicity-expressiveness spectrum, which is understandable to the domain experts, while being expressive enough for the diagnostic task. (3) The system was designed as a workbench for the domain experts. This enticed them to become more directly involved, and, resulted in a better system. (4) The use of an integrated knowledge base edit-tracking system was important to the project in several ways: it reassured computer-naive experts that they could not damage the overall system, which increased their productivity; and, it also allowed experts located in various places around the world to compare, contrast, and integrate changes in a structured way.

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.