Generalized Feature Graphs: A Kernel for Description Logic Implementations

If feature graphs are extended in a simple way they provide a basis for implementing arbitrary description logic formalisms. We describe the extension and the resulting language and we briefly show how to use it as a kernel for description logics. 1 Introduction A common characteristics of the feature graphs introduced by Rounds and Kasper [ Rounds and Kasper, 1986 ] and all extensions thereof is the fact that the graphs must be rooted. The root node automatically represents the object depicted by the feature graph (the depictee). Hence, feature graphs always describe the depictee with the help of other objects which are functionally dependent on the depictee. This mechanism was sufficient in the original application of feature graphs for representing grammatical information in unification grammars. Later, feature graphs found their way to knowledge representation and have been integrated with description logic formalisms [ Nebel and Smolka, 1990 ] . However, description logics prov..