A multimodal end-of-turn prediction model: learning from parasocial consensus sampling

Virtual human, with realistic behaviors and social skills, evoke in users a range of social behaviors normally only seen in human face-to-face interactions. One of the key challenges in creating such virtual humans is to give them human-like conversational skills, such as turn-taking skill. In this paper, we propose a multimodal end-of-turn prediction model. Instead of recording face-to-face conversation data, we collect the turn-taking data using Parasocial Consensus Sampling (PCS) framework. Then we analyze the relationship between verbal and nonverbal features and turn-taking behaviors based on the consensus data and show how these features influence the time people use to take turns. Finally, we present a probabilistic multimodal end-ofturn prediction model, which enables virtual humans to make real-time turn-taking predictions. The result shows that our model achieves a higher accuracy than previous methods did.