Providing contextualised knowledge often involves the difficult and time-consuming task of specifying the appropriate contexts in which the knowledge applies. This paper describes the Ripple Down Rules (RDR) knowledge acquisition and representation technique which does not attempt to define up front the possible context/s. Instead cases and the exception structure provide the context and rules provide the index by which to retrieve the case/s. KA is incremental. The domain expert locally patches rules as new cases are seen. Thus RDR is a hybrid case-based and rule-based approach. The use of Formal Concept Analysis to translate the RDR performance system into a formal context and uncover an explanation system in the form of an abstraction hierarchy further strengthens our emphasis on the combined use of cases and rules to provide contextualised knowledge.
展开▼
