Development of a methodology for knowledge elicitation for building expert systems
Abstract
A critical first step in building expert systems is the extraction of knowledge from human experts. The inability to elicit this knowledge efficiently and accurately has considerably impeded the development of expert systems. This thesis develops a conceptual framework and methodology for knowledge elicitation. The conceptual framework is established by generalizing Newell's and Simon's (1972) problem space concept and integrating it with Kelly's (1955) theory of personal constructs. This framework models, in a domain-independent manner, the structure of human problem solving knowledge and the context in which problems are solved. The structure of human problem solving knowledge is delineated by defining the knowledge structures associated with each component of the problem space. A model of the context is outlined by identifying and defining the components of a domain's problem space. A Structured Methodology for Eliciting Expertise (SMEE) is derived using the above conceptual framework. The structure of human problem solving knowledge, specified in the conceptual framework, determines the knowledge which is elicited by SMEE and the sequence in which it is extracted. The model of the context, provided by the conceptual framework, is employed by SMEE to elicit that knowledge which is used in an automated manner by the expert. SMEE is a structured multi-phase methodology that extracts knowledge in a domain-independent manner and overcomes many of the limitations inherent in the existing methods for knowledge elicitation. Several parts of the methodology which could be automated are implemented as a computer program in Turbo-Pascal, on a micro-computer. SMEE was used to elicit knowledge from operators of a flexible manufacturing system, in the domain of tool wear detection and monitoring. Reliability and validity evaluations performed on the extracted knowledge establish the validity of this approach.
Degree
Ph.D.
Advisors
Salvendy, Purdue University.
Subject Area
Industrial engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.