In this paper, we propose a new framework of knowledge revision,
Similarity-Driven Knowledge Revision. For an object-oriented knowledge base
KB, our revision is triggered when a
similarity between sort concepts detected from
KB does not fit a user's intuition. We revise
KB into a knowledge base from which such an undesirable similarity is not detected and in which the logical semantics of
KB is still preserved. An observation of undesirable similarity is due to an
over-general typing of variable in
KB. In order to modify the typing, we introduce a notion of
extended sorts that can be viewed as a sort concept not appearing explicitly in the original knowledge base. If a variable typing with some sort is considered over-general, the typing is modified by replacing it with more specific extended sort. Such an extended sort can be efficiently identified by forward reasoning with
SOL-resolution from the original knowledge base. Some experimental results show that the use of SOL-resolution can drastically improve the computational efficiency.
View full abstract