PhD Dissertation Defense by CHENG Zhiyong

Rapid advances in mobile devices and cloud-based music streaming services have brought about a fundamental change in the way people consume music. Mobile devices become the mainstream platforms allowing people to enjoy favorite music anytime and anywhere. User’s short-term music preferences are influenced by user’s music tastes and local contexts, such as environment factors (e.g., location) and user states (e.g., activity and mood). As mobile devices are designed for personal media consumption and usually used on the move, it is important to consider user’s personal music tastes and the local contextual factors in music retrieval and recommendation.

In this dissertation, we analyze the influence of different factors (i.e., venue types and user’s demographic information) on user’s music tastes and preferences by leveraging large-scale social music data, and then exploit the influence of context factors and user’s music preferences to improve the accuracy of music retrieval and recommender systems. In particular, we propose (1) a VenueMusic system, which recommends suitable music tracks to different venue types; (2) a use-information-aware music retrieval system, which exploits user’s age and/or gender information in text-based music retrieval; and (3) a personalized text-based music retrieval system, which considers user’s personal music preference in music retrieval. In these systems, novel topic models are proposed to capture user’s music preferences under different contexts in latent semantic music spaces. To validate the effectiveness of the proposed systems, we construct several large-scale music datasets and conduct experiments to compare the proposed systems with other competitors. The experimental results show the advantages of exploiting contextual information in music recommendation and retrieval, and demonstrate the superiority of our systems in terms of search accuracy. Finally, we discuss a few interesting and promising research directions in the area of user-centered music retrieval and recommendation.