Regional Classification of Traditional Japanese Folk Songs

- Classification Using Cluster Analysis -

Abstract

In this study, we focus on the melodies of Japanese folk songs, and examine the basic structures of Japanese folk songs that represent the characteristics of different regions. We sample the five largest song genres within the music corpora of the Nihon Min-yo Taikan (Anthology of Japanese Folk Songs), consisting of 202,246 tones from 1,794 song pieces from 45 prefectures in Japan. Then, we calculate the probabilities of 24 transition patterns that fill the interval of the perfect fourth pitch, which is the interval that maintains most of the frequency for one-step and two-step pitch transitions within 11 regions, in order to determine the parameters for cluster analysis. As a result, we successively classify the regions into two basic groups, eastern Japan and western Japan, which corresponds to geographical factors and cultural backgrounds, and also match accent distributions in the Japanese language.