This paper discusses the dynamics of conversation exchanges, specifically focusing on interruptions. It highlights the importance of equipping virtual agents with the ability to manage these exchanges for more realistic interactions. The study scrutinizes multimodal signals such as pitch, loudness, and facial expressions to differentiate between types of exchanges. The paper emphasizes the role of interruptions in regulating dialogue rhythm, expressing interest, and reinforcing engagement. The goal is to develop an Embodied Conversational Agent (ECA) capable of engaging in natural interactions with humans by managing different types of exchanges.
Publication date: 9 Nov 2023
Project Page: https://arxiv.org/abs/2311.04747v1
Paper: https://arxiv.org/pdf/2311.04747
