[3I6-GS-11] AI and Society:

[3I6-GS-11-01] Designing an AI Chatbot Utilizing Dialects

A Comparative Study of Conversation Experiences in the Toyama Dialect and Standard Japanese

〇Kaito Makino¹, Daisuke Uriu¹ (1. Shibaura Institute of Technology)

Keywords:chatbot, dialect, familiarity

Many AI chatbots use standard language, ensuring accuracy but often feeling formal and cold. This study explores how dialects enhance chatbot interactions by adding warmth and familiarity. We developed a Toyama dialect AI chatbot by constructing a dataset and training an existing language model. The chatbot, with a simple personality, operates on the LINE platform. To evaluate its impact, we conducted conversation experiments with eight Toyama dialect speakers and eleven standard language speakers, comparing a Toyama dialect model and a standard language model. Participants were interviewed to assess their impressions. Results showed that the Toyama dialect model created a friendly and warm impression, enhancing user experience. Even standard language speakers found it more engaging in certain situations. This study highlights the potential of dialect-based chatbots to improve interaction quality, making AI conversations feel more natural and personable.