This paper discusses the use of Large Language Models (LLMs) to enhance the conversational abilities of social robots. The authors propose a fully automated conversation system that uses LLMs to generate expressive behaviours consistent with the personality of the robot. The system includes a text-to-speech engine capable of varying delivery styles and a library of physical actions for the robot. The study also includes a pilot where volunteers interacted with a social robot using the proposed system. The feedback was largely positive, highlighting the robot’s empathy, naturalness, and entertainment value.
Publication date: 20 Feb 2024
Project Page: https://doi.org/10.1145/3610978.3640562
Paper: https://arxiv.org/pdf/2402.11571
