Analysis of Speech Dialogue to Detect Active Conversation and Lapse in Conversation toward Development of Conversation Support Robot after Co-occurrence and Mutual Assistance Matching

Abstract

The formation of human relationships is important to form communities. As a supporting system to strengthen the bonding among community members, we firstly tried to connect users by co-occurrence and mutual assistance matching to form relationships. Then, we brought them together. At the time of this experiment, users can easily carry on animated conversations if a common topic is suggested to them. In this study, we utilized a questionnaire to acquire user's personal characteristics for co-occurrence and mutual assistance matching. Next, we conducted a conversation experiment via Skype to investigate activation and inactivation during user's conversations by using speech dialogue and the LF/HF ratio calculated from user's heart rates. During the conversation experiment, matched users made conversations for 10 min on each common topic based on their interests suggested. The result showed that users could easily have animated conversations when a common topic was suggested to them. In addition, we investigated acoustic features, the overlap of utterances, and the number of utterances, backchannel feedbacks, and laughter.