Raveform: A Dataset of Metrical and Functional Structure Annotations for EDM Tracks in DJ Mixes

Research on disc jockey (DJ) mixes remains limited in the field of music information retrieval (MIR) despite the long history and cultural significance of DJing. This is largely due to the lack of comprehensive datasets and studies on DJ mixes. In this article, we introduce Raveform, a novel dataset containing 4,902 DJ mix links and 56,873 individual track links, including 1,423 tracks with detailed structural annotations manually annotated by three domain experts. To ensure consistency, we introduce an electronic dance music (EDM)‑specific structural vocabulary and propose a framework that reconciles the subjective perspectives of the domain experts through standardized annotation guidelines. Using the dataset, we analyze the structural characteristics of DJ mixes in terms of segment length, flow, and sub‑band energy. Furthermore, we train and evaluate deep learning models on metrical and functional structure analysis tasks. We show that the proposed dataset tailored to DJ mixes is effective for the tasks, especially when combined with existing resources. The Raveform dataset and associated models are publicly available at https://mir-aidj.github.io/raveform/.