KNOWLEDGE ACQUISITION IN RULE-BASED SYSTEMS—KNOWLEDGE ABOUT REPRESENTATIONS AS A BASIS FOR SYSTEM CONSTRUCTION AND MAINTENANCE1

Recent research efforts aimed at task-oriented systems have emphasized the importance of large stores of domain-specific knowledge as a basis for high performance. But assembling the required knowledge base is a difficult task that often extends over several years, and involves numerous modification to the knowledge base. Given the difficulty of making even small changes to a program, this presents a challenging problem in system construction. We have studied this issue in the context of TIERESIAS, a program designed to function as an assistant in the task of building large knowledge bases. TIERESIAS facilitates the interactive transfer of expertise from a human expert to the knowledge base of the system, in a dialog conducted in a restricted subset of natural language. One such knowledge transfer task involves teaching the system about a new conceptual primitive from which new inference rules can be built. We show that by providing a program with a store of knowledge about its own representations, this acquisition of new concepts can be carried out in a high-level dialog that transfers information efficiently. The necessary knowledge about representations includes both structural and organizational information, and is specified in a data structure schema, a device used to describe representations.