Pre-Training Strategies Using Contrastive Learning and Playlist Information for Music Classification and Similarity

TL;DR

This paper explores contrastive learning with playlist data as a weak supervision source to improve music representation models for classification and similarity tasks, showing that playlist-based pairing enhances performance.

Contribution

It introduces playlist-based pair generation methods for contrastive learning, extending prior metadata-based approaches, and demonstrates their effectiveness in music classification and similarity tasks.

Findings

Playlist-based pairs improve music similarity results.

Playlist pre-training outperforms artist-based pairing in classification.

Proposed methods achieve competitive or superior results across datasets.

Abstract

In this work, we investigate an approach that relies on contrastive learning and music metadata as a weak source of supervision to train music representation models. Recent studies show that contrastive learning can be used with editorial metadata (e.g., artist or album name) to learn audio representations that are useful for different classification tasks. In this paper, we extend this idea to using playlist data as a source of music similarity information and investigate three approaches to generate anchor and positive track pairs. We evaluate these approaches by fine-tuning the pre-trained models for music multi-label classification tasks (genre, mood, and instrument tagging) and music similarity. We find that creating anchor and positive track pairs by relying on co-occurrences in playlists provides better music similarity and competitive classification results compared to choosing tracks from the same artist as in previous works. Additionally, our best pre-training approach based on playlists provides superior classification performance for most datasets.