Current music streaming and recommendation services, while useful, suffer from unidimensional interaction (sequential track lists) and biases (position and popularity bias). This paper introduces EmoMTB, an intelligent audiovisual music exploration system. EmoMTB uses a city metaphor, where tracks are visualized as colored cubes in buildings, with similar tracks clustered together. Users navigate this 'music city' via smartphone, exploring both familiar and unfamiliar genres. The system integrates an emotion-aware recommendation system, re-ranking suggestions based on user-identified emotion or collective emotion from EmoMTB's Twitter channel. Evaluation involved quantifying clustering homogeneity, measuring emotion prediction accuracy, and conducting a user survey.

Alessandro B Melchiorre, Markus Schedl, David Penz, Christian Ganhör, Oleg Lesota, Vasco Fragoso, Florian Fritzl, Emilia Parada-Cabaleiro, Franz Schubert