Learning improved integrity constraints and schemas from exceptions in databases and knowledge bases

In order to set up efficient storage structures and detect errors in the data being stored, the designers of data and knowledge base management systems must initially build schema and integrity constraints that anticipate the kind of facts that are to be stored. The purpose of this chapter is to demonstrate some ways in which a KBMS can be endowed with the ability to refine its schema on the basis of the data actually stored so far. In particular, due to the unpredictability and evolution of the natural world, as well as possible errors made by designers, a KBMS must be tolerant of occasional deviations from the norms set out during design — in other words, it must accommodate exceptions. Two techniques are presented through which a computer system can suggest modifications and additions to the current definitions and semantic integrity constraints of a knowledge base by “ learning from the exceptions encountered. One schema refinement method is based on generalizing from example exceptions to form descriptions of classes of similar objects. The other technique refines integrity constraints by” explaining the occurrence of exceptions in terms of a detailed theoretical world-model and then generalizing to those cases where similar explanations hold.