Revision of Production System Rule-Bases

We describe CLIPS-R, a theory revision system for the revision of CLIPS rule-bases. CLIPS-R differs from previous theory revision systems in that it operates on forward chaining production systems. Revision of production system rulebases is important because production systems can perform a variety of tasks such as monitoring and design in addition to classification tasks that have been addressed by previous research. We show that CLIPS-R can take advantage of a variety of user specified constraints on the correct processing of instances, such as ordering constraints on the displaying of information, and the contents of the final fact list. In addition, we show that CLIPS-R can operate as well as existing systems when the only constraint on processing an instance is the correct classification of the instance.

• Type: Conference
• Year: 1994
• Where: ICML
• Authors: Patrick M. Murphy, Michael J. Pazzani
• Keywords: ICML 1994, Machine Learning, Production Systems, Theory Revision, User Specified Constraints