Abstract
This paper proposes a transition model of topics for a human-robot communication. In general, a robot should have scenarios on topics based on a current state and a personal preference. The topics used in the communication include the information on seasons, voice and the surrounding environments on the person and robot. In order to realize the flexible communication with a person, we propose a Boltzmann selection for the transition of topics used in the conversation. Moreover, the robot learns the personal preference by human response and human state. Topics are changed with the learning of a suitable personal preference. Next, we conduct experiments of communication between a person and the robot.
