Bridging the gap between knowledge engineering and efficient implementation in an intelligent tutoring system

Abstract

Symbolic manipulation languages, such as LISP and Prolog, are advantageous in that they can describe a problem space symbolically. However, they are inefficient when handling large rule-based systems. In order to create practical and usable implementations, two systems are often implemented. One system is the knowledge engineering prototype in which rules are formulated symbolically and manipulated as such. The other is a production version that guarantees realistic response time and execution profile to an end user. To create the second version it is usually necessary to recode the system in a traditional non-symbolic language like C or C++. This paper describes an automatic approach to transforming a symbolic knowledge base into an efficient non-symbolic representation resulting in an extremely efficient and practical implementation.