Expert system shells — selecting the most appropriate development environment

Abstract

Positive and negative features of commercially available expert system shells are reviewed from the user's perspective. This review does not present the features of any particular shell. Requirements and desirable features for expert system shells are dependent upon specific application needs and user expertise. Users fall into three general categories: (1) students learning about expert systems, (2) domain experts involved in prototyping, testing, and extending a system, or (3) knowledge engineers with familiarity and skill in knowledge representation methodology and building programs. All shells provide four basic functions: (1) a knowledge representation scheme, (2) an inference or search mechanism, (3) a means of describing a problem, and (4) status determination during problem solving. Despite this commonality, shells lie on a continuum of relatively simple languages to very elaborate development environments, each having its own purposes and strengths at different stages of the expert system development cycle. Users should carefully select the shell most suited to their needs and level of expertise. User satisfaction will be maximized through consideration of the following important shell features: (1) knowledge representation schemes, (2) knowledge engineering tools, (3) inference engine and problem solving strategies, (4) compatability and portability, (5) user interface, (6) performance and productivity, and (7) other advanced features provided by expert system technology.