Pre-scheduled Turn-Taking between Robots to Make Conversation Coherent

Abstract

Since a talking robot cannot escape from errors in recognizing user's speech in daily environment, its verbal responses are sometimes felt as incoherent with the context of conversation. This paper presents a solution to this problem that generates a social context where a user is guided to find coherency of the robot's utterances, even though its response is produced according to incorrect recognition of user's speech. We designed a novel turn-taking pattern in which two robots behave according to a pre-scheduled scenario to generate such a social context. Two experiments proved that participants who talked to two robots using that turn-taking pattern felt robot's responses to be more coherent than those who talked to one robot not using it; therefore, our proposed turn-taking pattern generated a social context for user's flexible interpretation of robot's responses. This result implies a potential of a multiple robots approach for improving the quality of human-robot conversation.