Type extensibility of a knowledge representation system with powersets

Abstract

This paper deals with KRSS (knowledge representation systems) close to frame systems including description logic and object-based systems. The main relation that leads to inferences is subsumption which is usually computed over knowledge descriptions that refer to data types. Although subsumption between knowledge terms is well-defined, its implementation on external data types depends upon the host language which is used for the actual implementation of data types. As a consequence, no KRS is able to integrate a new (external) data type such that its values can be safely involved in subsumption and further inferences. This is the problem addressed in this paper. The solution we propose relies on the design of a polymorphic type system connected to both the Krs and the host language. It is designed so that it can extend the KRS with any type implementation available in the host language (standard atomic types, library, package, etc.). Meanwhile, the values of the new type get safely involved in the Krs reasoning processes. The presented type system prevents the incompleteness of subsumption due to its incomplete processing on external data.