AnalysisGNN: Unified Music Analysis with Graph Neural Networks

2025-09-08

AI Summary

Existing music analysis methods are typically domain-specific and struggle to uniformly process heterogeneous symbolic score data with inconsistent annotations, exhibiting limited robustness against annotation inconsistencies and domain shifts. Method: We propose the first unified music analysis framework based on graph neural networks, integrating multi-task learning, logit fusion, a weighted loss function, and a novel non-harmonic note prediction module—designed to filter out interfering notes and enhance label consistency—alongside a data shuffling strategy to improve cross-domain generalization. Contribution/Results: Experiments demonstrate that our framework achieves performance on par with traditional static methods across multiple heterogeneous corpora, while significantly improving cross-domain stability and robustness. To our knowledge, this is the first approach to systematically address both annotation inconsistency and domain shift in multi-domain music analysis.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction
• Natural Language Processing: Safety and Robustness

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Recent years have seen a boom in computational approaches to music analysis, yet each one is typically tailored to a specific analytical domain. In this work, we introduce AnalysisGNN, a novel graph neural network framework that leverages a data-shuffling strategy with a custom weighted multi-task loss and logit fusion between task-specific classifiers to integrate heterogeneously annotated symbolic datasets for comprehensive score analysis. We further integrate a Non-Chord-Tone prediction module, which identifies and excludes passing and non-functional notes from all tasks, thereby improving the consistency of label signals. Experimental evaluations demonstrate that AnalysisGNN achieves performance comparable to traditional static-dataset approaches, while showing increased resilience to domain shifts and annotation inconsistencies across multiple heterogeneous corpora.

Problem

• Unifies music analysis across domains with GNN
• Integrates heterogeneously annotated symbolic music datasets
• Improves consistency by excluding non-chord tones

Innovation

• Graph neural network with multi-task loss
• Non-Chord-Tone module excludes passing notes
• Logit fusion integrates heterogeneous datasets