Authors: Roderick Murray-Smith, Danial Boland, Beatrix Vad, John Williamson, Peter Berg Steffensen
Type: Conference paper
Conference: 32nd International Conference on Machine Learning, Lille, France, 2015
Title: Machine Learning for Music Discovery Workshop at the 32nd International Conference on Machine Learning, Lille, France, 2015.
Abstract: We present the design and evaluation of an interactive tool for music exploration, with musical mood and genre inferred directly from tracks. It uses probabilistic representations of multivariable predictions of subjective characteristics of the music to give users subtle, nuanced visualisations of the 2D map. These explicitly represent the uncertainty and overlap among features and support music exploration and casual playlist generation. A longitudinal trial in users’ homes showed that probabilistic highlighting of subjective features led to more focused exploration in mouse activity logs, and 6 of 8 users preferred the probabilistic highlighting.