Jess: Implementing A High-Performance
Symbolic Reasoning Engine In Java

Ernest J. Friedman-Hill
Sandia National Laboratories

2:00pm October 29, 1999
Lecture Hall V, ECS

The old notion of the Expert System has enjoyed a recent resurgence of popularity, both in traditional applications like help desks and classifiers and as reasoning portions of Agents. Jess (the Java Expert System Shell) is a Java implementation of the Rete algorithm, an efficient pattern-matching technique for Expert Systems (or more generally for production systems) that has been applied to all of these problems, and more. Jess is extremely fast, outperforming some other well-known Rete implementations (including those written in C) by as much as several orders of magnitude on the Miranker benchmarks. This presentation will focus on the techniques used to achieve this speed in Java, a language often itself criticized for its performance disadvantages. I will also describe some of Jess's other novel features, again with an eye towards the unique implementation issues and opportunities presented by the Java environment.