iTOChat: a Mobile Bot Recommending Engineering Departments and Open Campus Events

Naoki Eguchi, Isshin Nakao, Kotaro Saito, Juntao Wang, Hiromu Motomatsu, Tsunenori Mine

Abstract: Chatbots are widely used in education, medicine, business, and other fields, among which natural language processing such as ChatGPT, especially chatbots based on large-scale language models (LLMs), are expected to flexibly answer users' complex questions. In this study, we present iTOChat, a mobile bot that uses ChatGPT to recommend departments and open campus events to students visiting Kyushu University, an open campus of the School of Engineering, by inputting students' personal information such as hobbies and future plans. Although iTOChat is expected to make accurate recommendations to students, this is not easy because ChatGPT does not contain detailed information about engineering departments and open campus events. Therefore, we developed a method to supplement the missing information in ChatGPT by embedding reasons and conditions for recommending departments in the ChatGPT prompts. This paper describes the features and objectives of the developed system and evaluates the satisfaction of students who used iTOChat at an open campus.