Abstract
This paper develops a methodology for benchmarking knowledge-based systems that practitioners may use to perform a comparative analysis of expert system shells. A program utilizing a deliberate instructional mix was found to be the most suitable and accurate way to measure the value of shells. The benchmark is intended for rule-based shells, such as CLIPS, VP-Expert and Ibis. The methodology for the approach is a generic rule-based algorithm that is easily adaptable to meet the language requirements of individual shells. We present results for three shells.
