We are developing chat-oriented dialogue systems that can talk with people naturally. Current dialogue systems sometimes respond to user utterances in a wrong way because of automatic speech recognition errors and insufficient system’s knowledge for dialogue topics. These errors make users to feel difficulties when they use such dialogue systems. To overcome the problem, we simultaneously improve the system’s performance and leverage human’s cognitive biases for multiple robots. When users talk with multiple robots those behaviors are coordinated in background, the robots can avoid dialogue breakdown and the users become easy to talk with them. Such robots that can talk naturally are natural interface between humans and information. Moreover, dialogues with the robots can help user to improve communication skills.
Please click the thumbnail image to open the full-size PDF file.