Expert system benchmarks

Benchmarks for use with expert system shells are considered. Two approaches are pursued: running a shell with a real knowledge base on different machines, and running an artificial, stylized benchmark knowledge base with different shells on the same machine. The realistic benchmark uses PC Plus, a Lisp-based shell and development environment with an inference engine that uses backward chaining (forward chaining is also possible). Stylized knowledge bases have been used with different shells to compare loading and execution time, file size, and memory requirements. Some stylized-benchmark results are given.<>