Research

We describe our approach to developing KA tools that derive expectations from the KB in order to guide users during KB creation. Through an analysis of the KB creation task, we were able to detect several sources for such expectations. The expectations result from enforcing constraints in theknowledge representation system, lookingformissing pieces of knowledge in the KB, and working out incrementally the inter-dependencies among the different components of the KB. As the user defines new KB elements (i.e., new concepts, new relations, new problem-solving knowledge), the KA tool can form increasingly more frequent and more reliable expectations. We implemented a KA tool called EMeD that uses these sources of expectations to support users in adding problem-solving knowledge. Our preliminary evaluation shows an average time savings of 30% to enter the new knowledge. We believe it will be even higher for users who are not experienced knowledge engineers.