Music exploration by impression based interaction

Abstract

Search and recommendation systems help users find their favorites among an abundant amount of songs available on distributors such as iTunes. Nevertheless, it is still hard for us to efficiently retrieve the right music. This paper pro- poses an interactive music exploration system to help them reach their favorite songs by means of interactive feedbacks submitted by the user. The recommendation component of our system is based on the feedback component accepts the user's feedbacks in terms of impression features such as "more ballad-like" and "less pop-like". The proposed concepts have been implemented as an online demonstration system using a real world dataset.