Cooperative dialog system for decision support and its response generation

This paper describes an outline of "cooperative dialog system for decision support" and a framework of its cooperative response generation. The system consists of "utterance interpreter, dialog manager, decision supporter, user model, and database", and handles decision support based on data retrieval. The system is designed to achieve two types of cooperation, the cooperation in interactions and cooperation in decision support, and cooperative response generation concerns cooperation in interactions. First, cooperative responses are classified into precondition-related responses, information-related responses, reason-related responses, and responses in the form of a question. Next, an utterance model representing the user's intentions is introduced, and the knowledge description needed for response generation is explained. Finally, a response generation method using the utterance model and knowledge description is proposed. The method covers classified responses. Moreover, it is domain-independent and can be easily implemented because of its simplicity.<<ETX>>