Conceptual Clustering of Explanations

ABSTRACT Inductive and explanation-based learning methods traditionally differ in the extent that they exploit background knowledge. Each is search intensive and sensitive to domain imperfections, albeit in different ways. Hybrid systems that abstract over explanations promise the advantages of each approach, but they require augmentation if their promise is to be fully realized. Conceptual clustering can be profitably applied to the abstraction and organization of explanations so that they may be efficiently and appropriately reused.