ExpertGuide: A Conversational Case-Based Reasoning Tool for Developing Mentors in Knowledge Spaces

摘要

Case-based reasoning (CBR) has been used to improve knowledge management in corporate activities. It was initially used for problem solving and then for facilitating the distribution of knowledge and experiences. Since conversational CBR appeared, it has also been used to develop mentor systems in complex knowledge spaces. ExpertGuide was designed as a tool for developing WWW-based mentor systems. It was designed during the development of CBR systems used at NEC. Multilink retrieval provides a method for searching a case library from several viewpoints. Question selection by entropy finds most effective questions in discriminating cases. It does this by calculating the information gain of candidate questions. Indexing with scripts is a case-indexing method using Schank's scripts and is effective when situations and problems are hard to express and assess. ExpertGuide tightly integrates these techniques and provides a user interface for users who want advice from mentors. It is currently being used when building WWW-based mentor systems with scales ranging from division-wide to corporate-wide and nationwide.