Extending the WARREN Abstract Machine to Feature Prolog

Abstract

Inheritance hierarchies are employed in knowledge representation and object-oriented programming in the sense of representing taxonomic information. Feature Prolog provides a useful tool to represent taxonomic information in Logic in a simple and natural way. In our approach, inheritance hierarchies are built-up from feature types, that are record-like structures, ordered by subtyping. The presence of feature types reduces the deduction tree and avoids unnecessary backtracking. In Feature Prolog there are feature terms besides the common Prolog terms - used to denote subsets of feature types. The integration of feature terms into the Prolog inference mechanism needs an extension of SLD-resolution with feature unification, that is unification respecting the taxonomic information of the feature types. We describe an extension of the abstract Prolog instruction set, known as WARREN Abstract Machine, for inheritance hierarchies.