Constructing and Refining Knowledge Bases: A Collaborative Apprenticeship Multistrategy Learning Approach

摘要

This research has developed a theory, methodology and learning agent shell for development of knowledge bases and knowledge-based agents, by domain experts, with limited assistance from knowledge engineers. The main feature of this approach is that a subject matter expert communicates his or her expertise to the learning agent in a very natural way, similar to how the expert would communicate it to a human apprentice while solving problems in cooperation. Starting from an initial ontology, an expert may teach the agent how to solve a certain type of problem by providing a concrete example, helping the agent to understand the solution, supervising the agent as it attempts to solve new problems, and correcting its mistakes. Through such natural interactions, the agent will be guided in learning complex problem solving rules, and in extending and correcting its knowledge base. This research has been done in the context of the DARPA High Performance Knowledge Bases program, where it has been applied to two challenge problems, the Workaround challenge problem, and the Course of Action challenge problem. The agent development approach and the two developed agents have been evaluated in several intensive studies, and have demonstrated very good results.