Wu, Hsuan WeiHsuan WeiWuPan, Szu HanSzu HanPanHAN-PANG HUANG2024-01-052024-01-052023-01-019798350306354https://scholars.lib.ntu.edu.tw/handle/123456789/638227This paper describes the design and the implementation of a trilingual conversation system that is capable of speaking Chinese, English, and Hokkien (Taiwanese). The system is designed to fulfill the emotional support and companionship needs of the elderly population in Taiwan, which is expected to increase significantly in the coming years. To achieve this goal, the system is implemented on a humanoid robot called "Mobi", which allows for natural and fluent interaction with the user. The conversation system is based on a modular design approach, which utilizes the GPT model with Speech recognition and Speech synthesis to achieve real-time interactive conversation in three different languages and provide a user-friendly simple interface that allows the user to switch between languages and modes. Overall, this trilingual conversation system holds great potential for providing emotional support and companionship to the elderly population, multilingual interaction to break language barriers, and can be applied to various fields which will significantly impact the improvement of life quality and work efficiency.GPT Model | Speech Recognition | Speech Synthesis | Trilingual Conversation System | Web Crawlingconference paper10.1109/CACS60074.2023.103261612-s2.0-85179847541https://api.elsevier.com/content/abstract/scopus_id/85179847541