IREF-an interactive theory-driven knowledge refinement tool

A discussion is given on the integration of a domain-based conceptual model developed for interactive knowledge acquisition with machine-learning techniques, specifically explanation-based learning. The authors present the design of an interactive knowledge refinement tool which is based on an incomplete domain theory. In combination with a generalization component, the tool can actively suggest or confirm modifications to the knowledge base, rather than just accept editing commands passively. Furthermore, the interface presented to users of the tool (the conceptual model) reflects the semantics of the domain. Users are completely shielded from the implementation details of the underlying performance element and its knowledge representation. The work suggests that incorporating a machine-learning component in an interactive knowledge acquisition tool can yield fruitful returns