research results

Overview:

This study aimed to design a richer music experience to ‘deepen’ and ‘broaden’ the music experience in music subscription services. Specifically, the study targeted the music experience in music subscriptions and conducted an analytical study that collaborated between the content aspect and the aspect of human cognitive characteristics, such as the actual relationship between the structure of the music and preferences. At the same time, a design method for the automatic generation of thumbnail images of songs and an interactive, interactive recommendation system were developed.

　First, the content characteristics of hit songs were analyzed quantitatively, focusing on chord progressions, and changes before and after the music streaming era were traced. In addition, the actual usage of song recommendation was analyzed using a social survey approach from the perspective of ‘how many unknown songs with different characteristics from those we have listened to before have we encountered’. Through these results, measures to reflect them in the new design of the music experience were examined. Second, we focused on the thumbnail images of songs, which play an important role in the selection of songs in music subscription services and developed and verified a method for automatically generating prompts to effectively generate images based on the characteristics of songs, using an AI-based image generation model. For interactive recommendation, we proposed a song recommendation method that predicts the next song by learning the relationships between songs, sequences and repetitions in the input sequence, and obtained results showing better performance than comparative methods on all metrics.

Future plan

・To incorporate the results of the content analysis into the recommendation method.
・To verify how the recommendation method can enhance the depth and breadth of the music experience.
・To develop interactive appreciation using generative AI as a new approach to interactive recommendation methods.

▼Click on the image to enlarge