Abstract
Motivational interviewing (MI) is a counseling method for motivating behavioral change of the client . Although the number of studies ofn MI is increasing, there is no MI dataset in Japanese, nor is there a dialogue system for MIusing it. In this study, we developed an MI dataset corpus and a dialogue system using the dataset. With the cooperation of experts, First, we recorded MI sessions between MI experts and clients to create the data and created a corpus. Analyzing the collected corpus, we found that Next, we created a dialogue model. Since reflection and summarizationing are important in MI. Therefore,, we created a summary generation model specialized for this purpose to summarize client’s utterances and a basic dialogue model thatfor generatesing other types of utterancesconversations. A dialogue system By integrating these two models, we builtd a dialogue system was constructed. We, and conducted a preliminary evaluation of the system, and found that was conducted. Thethe dialogues generated by the system presented some typical characteristics of MI.
