Using description logics for knowledge intensive case-based reasoning

Abstract

In this paper we argue that description logics with their object-oriented representation based on a declarative semantics and their powerful inferences are a good base for building similarity-based systems. But in existing description logic systems it is not possible to formulate and use knowledge about concrete domains (e.g. data types like numbers, strings, sets of symbols). Based on Baader and Hanschke's theoretical work on “admissible concrete domains” we realized Ctl, an extensible description logic system that is able to integrate such concrete domains via a generic interface to existing implementations of such data types. Initially, we coupled a CLP(R)-system in order to realize sound and complete inferences over systems of linear inequalities. This concrete domain is especially useful within the area of second-level corporate support, a domain whose requirements initiated our investigations on description logics and which we will use for our illustrating examples.