Executable conceptual structures

Abstract

In the conventional knowledge based systems, there is a distinct separation between the domain knowledge and the inference engine that utilises these domain knowledge for problem solving. The knowledge acquisition process for engineering such a system is a major bottle-neck for the liberal exploitation of knowledge based systems. The author proposes to reduce, if not eliminate this bottle-neck by eliminating the distinction between the knowledge base and the inference engine, while maintaining the declarative and procedural nature of the domain knowledge. This is achieved with the development of an extended form of knowledge representation scheme, known as the Executable Conceptual Structures. This paper outlines two forms of Executable Conceptual Structures. They are the Actor Graphs and Problem Maps, respectively [1]. Actor Graphs are the atomic components of a larger and more complex structure called the Problem Maps. Actor Graphs represent only a single action, while the Problem Maps represent plans and associated strategic knowledge in the form of Conceptual Rules for plan selections and/or plan disambiguation during run time. Executable Conceptual Structures are the result of the merger of two fundamental principles, yet related in some ways. They are conceptual graphs [2] and the object oriented principles. The synthesis of the above two principles, enabled the development of an extended knowledge representation technique, well suited for representing both the declarative knowledge and its associated and/or related procedural knowledge. This paper also outlines two advancements requisite for the development of Executable Conceptual Structures: Intelligent Control Script and the Actor Paradigm. This paper further outlines the formal definition of these Executable Conceptual Structures in terms of lambda expressions. The development of Actor Graphs and Problem Maps are indeed a major breakthrough in themselves as well as for related research areas in Artificial Intelligence, mainly “automated programming”, “knowledge acquisition” and “expert systems” to outline a few. This paper concludes by outlining three limiting factors associated with the current implementation of the Executable Conceptual Structures.