Analysis of the Acoustic Characteristics of Takarazuka Revue Member Names

The phenomenon where a sound itself conveys a specific impression is known as sound symbolism. Although sound symbolism has been observed in various domains, its comprehensive understanding remains elusive. Therefore, we are gathering and examining instances of sound symbolism to elucidate its broader context. In this study, we focused on the names of Takarazuka Revue members as examples, hypothesizing that their naming might exhibit a unique pattern. We explored their acoustic characteristics using machine learning techniques. Our study employed Convolutional Neural Networks (CNN) and Grad-CAM for Explainable Artificial Intelligence (XAI) for this investigation. The analysis revealed that male role names in the Takarazuka Revue tend to have fewer obstructive sounds compared to typical Japanese male names. Furthermore, it was observed that male role names have fewer resonant sounds than “musumeyaku (female role)” names.