Songformer: Scaling Music Structure Analysis With Heterogeneous Supervision

Abstract

Music structure analysis (MSA) underpins music understanding and controllable generation, yet progress has been limited by small, inconsistent corpora. We present SongFormer, a scalable framework that learns from heterogeneous supervision. SongFormer (i) fuses short- and long-window self-supervised learning representations to capture both fine-grained and long-range dependencies, and (ii) introduces a learned source embedding to enable training with partial, noisy, and schema-mismatched labels. To support scaling and fair evaluation, we release SongFormDB, the largest MSA corpus to date (over 14k songs spanning languages and genres), and SongFormBench, a 300-song expert-verified benchmark. On SongFormBench, SongFormer sets a new state of the art in strict boundary detection (HR.5F) and achieves the highest functional label accuracy, while remaining computationally efficient; it surpasses strong baselines and Gemini 2.5 Pro on these metrics and remains competitive under relaxed toleranc

Stats

• citations0
• S2 citations—
• github stars132
• HF likes0
• heat score4.25
• arxiv keyhao2025songformer

Related papers

• Supervised Metric Learning For Music Structure Features (2021)0.00
• Convolutive Block-matching Segmentation Algorithm With Application To Music Structure Analysis (2022)0.00
• Ssm-net: Feature Learning For Music Structure Analysis Using A Self-similarity-matrix Based Loss (2022)0.00
• Supervised And Unsupervised Learning Of Audio Representations For Music Understanding (2022)0.00
• Songmass: Automatic Song Writing With Pre-training And Alignment Constraint (2020)11.39
• Deep Audio-visual Singing Voice Transcription Based On Self-supervised Learning Models (2023)0.00
• Learning Music Audio Representations Via Weak Language Supervision (2021)10.07
• Singmos-pro: An Comprehensive Benchmark For Singing Quality Assessment (2025)0.00