A Knowledge Compilation Method Through Conversion of Symbolic Rules and Facts into Functions

Search this article

Abstract

This paper describes a practical knowledge compilation method which is based on conversion of knowledge from rules and facts in a symbolic data form to those in a functional form through a semi-automatic translation process. Knowledge is doubly compiled at a declarative/procedural level of knowledge representation and at a Lisp/machine code level of computer language. The proposed method is applied to a kno1wledge-based system for automatic pipe route planning in the area of industrial plant layout design. The inference program based on the proposed method solves route priority assignment problems more than 50 times efficiently than a reference inference program in logic programming style. The results suggest that the proposed knowledge compilation method should be a useful tool through its coordination of the requirement of inference efficiency in system utilization and that of knowledge transparency in system construction and maintenance.

This paper describes a practical knowledge compilation method which is based on conversion of knowledge from rules and facts, in a symbolic data form, to those in a functional form through a semi-automatic translation process. Knowledge is doubly compiled at a declarative/procedural level of knowledge representation and at a Lisp/machine code level of computer language. The proposed method is applied to a kno1wledge-based system for automatic pipe route planning in the area of industrial plant layout design. The inference program, based on the proposed method solves route priority assignment problems more than 50 times efficiently than a reference inference program in logic programming style. The results suggest that the proposed knowledge compilation method should be a useful tool, through its coordination of the requirement of inference efficiency in system utilization and that of knowledge transparency in system construction and maintenance.

Journal

Details 詳細情報について

Report a problem

Back to top