Erik Schmidt

Filip Korzeniowski, Oriol Nieto, Matthew McCallum et al.

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.

Massimo Quadrana, Marta Reznakova, Tao Ye et al.

Finding the music of the moment can often be a challenging problem, even for well-versed music listeners. Musical tastes are constantly in flux, and the problem of developing computational models for musical taste dynamics presents a rich and nebulous problem space. A variety of factors all play some role in determining preferences (e.g., popularity, musicological, social, geographical, generational), and these factors vary across different listeners and contexts. In this paper, we leverage a massive dataset on internet radio station creation from a large music streaming company in order to develop computational models of listener taste evolution. We delve deep into the complexities of this domain, identifying some of the unique challenges that it presents, and develop a model utilizing recurrent neural networks. We apply our model to the problem of next station prediction and show that it not only outperforms several baselines, but excels at long tail music personalization, particularly by learning the long-term dependency structure of listener music preference evolution.

Jordi Pons, Oriol Nieto, Matthew Prockup et al.

The lack of data tends to limit the outcomes of deep learning research, particularly when dealing with end-to-end learning stacks processing raw data such as waveforms. In this study, 1.2M tracks annotated with musical labels are available to train our end-to-end models. This large amount of data allows us to unrestrictedly explore two different design paradigms for music auto-tagging: assumption-free models - using waveforms as input with very small convolutional filters; and models that rely on domain knowledge - log-mel spectrograms with a convolutional neural network designed to learn timbral and temporal features. Our work focuses on studying how these two types of deep architectures perform when datasets of variable size are available for training: the MagnaTagATune (25k songs), the Million Song Dataset (240k songs), and a private dataset of 1.2M songs. Our experiments suggest that music domain assumptions are relevant when not enough training data are available, thus showing how waveform-based models outperform spectrogram-based ones in large-scale data scenarios.