A Study of Dialect Identification Method Using Phoneme Texts

Abstract

In recent years, various studies have been conducted on communication robots for the elderly to provide mental health care due to the problem of a shortage of nursing care personnel as society ages. Among these studies is one on a robot that speaks a dialect. Since there are many regional dialects in Japan, and many elderly people living in regions speak dialects on a daily basis, the authors believe that if a communication robot for the elderly can automatically identify the dialect of the speaker and communicate in that dialect, it will be possible to have a more friendly conversation. If a communication robot for the elderly can automatically identify the dialect of a speaker and communicate using that dialect, it will be possible to have a more friendly conversation. In this paper, we propose an effective text segmentation method as a way to improve the accuracy of dialect identification, and report the results of using this method to identify data from various regions.