An Autonomous Dialogue Android that Acquires Human-Item Knowledge in a Shopping Mall

Abstract

The purpose of this research is to develop a dialogue robot that can acquire knowledge about human society autonomously in a real world. In this study, we define empirical data as image data obtained from a camera (visual information) and dialogue data obtained from a microphone (auditory information). The robot acquires knowledge about the co-occurrence between humans and items, that is, whether it is rare for humans to have the items from empirical data. The proposed system not only acquires knowledge but generates speech based on the acquired knowledge. We implemented the proposed method on an android, a human-like robot. We conducted an experiment in which the android was placed in a real world (shopping mall) and interacted with customers. The experimental results suggest that our system can acquire knowledge in terms of the percentage of affirmative responses to the robot's questions.