Mood Classification Using Listening Data
This work addresses the problem of automatic mood prediction for music recommendation, but it is incremental as it primarily compares existing feature types on a new dataset.
The study tackled mood classification in music by comparing listening-based and content-based features, finding that matrix factorization of listening data outperforms audio content embeddings on a dataset of 67k tracks annotated with 188 moods.
The mood of a song is a highly relevant feature for exploration and recommendation in large collections of music. These collections tend to require automatic methods for predicting such moods. In this work, we show that listening-based features outperform content-based ones when classifying moods: embeddings obtained through matrix factorization of listening data appear to be more informative of a track mood than embeddings based on its audio content. To demonstrate this, we compile a subset of the Million Song Dataset, totalling 67k tracks, with expert annotations of 188 different moods collected from AllMusic. Our results on this novel dataset not only expose the limitations of current audio-based models, but also aim to foster further reproducible research on this timely topic.