SOFTWARE FOR EXPERT SYSTEMS DEVELOPMENT.

Although some expert systems utilize artificial intelligence languages such as LISP and PROLOG, many rely on commercially available expert systems development tools. Three categories of tools are examined: shells, programming environments, and rule-induction software. They are compared by implementing a simple set of rules with commercially available software. Expert systems shells are often inexpensive and easy to use, but system developers are generally unable to modify a shell's control structure. Also, a shell's reliance on rules for representing inferential knowledge is often inefficient and awkward. Knowledge engineering programming environments remove these restrictions by allowing the developer to influence the rule-based inference procedure and by using frames and inheritance relations. Programming environments have relatively high hardware and software costs and require that the expert systems developer have some programming expertise. Rule induction tools use case examples to derive rules and have potential value as aids in knowledge acquisition.