Knowledge Acquisition

Abstract Knowledge acquisition is the process by which problem‐solving expertise is obtained from some knowledge source, usually a domain expert. This knowledge can then be implemented into an expert system program that can provide expert assistance to nonexperts when and where a human expert is not available. Traditionally knowledge acquisition is accomplished through a series of long and intensive interviews between a knowledge engineer, who is a computer specialist, and a domain expert, who has superior knowledge in the domain of interest. The difficulty, time, and cost of manual knowledge acquisition, however, have stimulated research in developing machine learning approaches that autonomously acquire knowledge from data sources without the assistance of humans. This article presents an in‐depth overview of knowledge acquisition by discussing the fundamental concepts of knowledge, types of knowledge, the process of knowledge acquisition, and knowledge acquisition methods. The article presents a classification that divides knowledge acquisition methods into three categories: manual methods, combined manual and automated methods, and automated methods. Manual methods discussed include interviews, task‐based methods, and expert self‐elicitation. Combined manual and automated methods discussed include expert‐driven and knowledge engineer‐driven approaches. Automated methods presented include learning by induction, neural networks, genetic algorithms, and analogical and case‐based reasoning. The article also addresses knowledge presentation and discusses eight of the most common representation methods: logic, production rules, frames, semantic networks, objects‐attributes‐value triplets, scripts, decision tables, and decision trees. The article concludes with a discussion of validation and verification of acquired knowledge.