Adapting Behaviour of Socially Interactive Robot Based on Text Sentiment

Abstract

The assessment of emotion is a complex process that encompasses information from various channels or sources, e.g., verbal, para-linguistic, non-verbal or even textual cues. Understanding the emotion of an interaction partner and exhibiting proper bodily reactions play a crucial role in establishing an engaging interaction. The evaluation of emotion, in general, is significant as far as emotionally intelligent human-robot interaction is concerned. In this paper, a text-based interaction system has been introduced to establish affective communication. A chatbot system has been implemented on a humanoid robot named ROBIN. The sentiment of the user input texts has been analyzed with MITIE machine learning toolkit. The robot considers emotional states of the users’ text utterances and reacts with appropriate speech, gesture, and facial expressions. This triggers real-time affective behavior from the robot’s perspective. The proposed system has been tested on several interaction scenarios to validate the approach from a human-centered perspective.