Metaphorical affect sensing in an intelligent conversational agent

Abstract

We report new developments of an affect detection component on the processing of several different types of metaphorical affective expressions. The component has been embedded in an intelligent conversational AI agent interacting with human users under loose scenarios. The detected affective states also play an important role in producing emotional animation for users' avatars. Evaluation for the affect detection component is provided. Our work contributes to the conference themes on human-robots interactionbased on affect sensing, affective computing, learning and children, narrative storytelling and evaluation of affective social interaction.