A multiple‐observation approach in knowledge acquisition for expert systems: A case study

This article describes (i) the elicitation component of knowledge acquisition (KA) in the domain of cartographic cataloging of single‐sheet maps produced by three U.S. map‐making agencies; and (ii) the performance evaluation results that demonstrate the effectiveness of the elicitation method. A multiple‐observation approach was used to elicit knowledge that human experts use when solving descriptive cataloging problems. The use of the elicitation techniques is discussed in view of problem identification goals and their applicability for this study. A body of published knowledge was integrated with a body of the elicited personal knowledge to form the knowledge‐base of an experimental cataloging advisor—Mapper. The study demonstrates that library students, who are map cataloging novices, can produce significantly better cataloging answers when assisted by the Mapper than can those who were not assisted by the Mapper. Better scores of Mapper students are attributed mainly to the elicited personal knowledge. © 1992 John Wiley & Sons, Inc.