Deconstructing Jazz Piano Style Using Machine Learning

2025-04-07

AI Summary

This study addresses the problem of interpretable modeling of jazz piano performance style. We propose a novel four-domain disentangled multi-input neural network architecture that maps stylistic characteristics onto four musicologically meaningful dimensions: melody, harmony, rhythm, and dynamics. Methodologically, the approach integrates MIR-based feature extraction (chroma, tempo, velocity), attention mechanisms, and model-agnostic interpretability techniques (LIME/SHAP). Evaluated on an 84-hour curated dataset of recordings by 20 canonical jazz pianists, the model achieves 94% classification accuracy. Key contributions include: (1) the first theory-driven, music-informed framework for stylistic disentanglement, bridging machine learning and music analysis; (2) open-sourcing of the trained model and an interactive web-based exploration platform; and (3) generation of empirically verifiable style attribution maps, enabling hypothesis-driven validation of multiple music-theoretic claims regarding jazz piano idioms.

Technology Category

• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Machine Learning: Deep Neural Architectures and Foundation Models
• Application Domains: Humanities & Computational Social Science

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Artistic style has been studied for centuries, and recent advances in machine learning create new possibilities for understanding it computationally. However, ensuring that machine-learning models produce insights aligned with the interests of practitioners and critics remains a significant challenge. Here, we focus on musical style, which benefits from a rich theoretical and mathematical analysis tradition. We train a variety of supervised-learning models to identify 20 iconic jazz musicians across a carefully curated dataset of 84 hours of recordings, and interpret their decision-making processes. Our models include a novel multi-input architecture that enables four musical domains (melody, harmony, rhythm, and dynamics) to be analysed separately. These models enable us to address fundamental questions in music theory and also advance the state-of-the-art in music performer identification (94% accuracy across 20 classes). We release open-source implementations of our models and an accompanying web application for exploring musical styles.

Problem

• Identify jazz musicians using machine learning models
• Analyze melody, harmony, rhythm, and dynamics separately
• Improve music performer identification accuracy to 94%

Innovation

• Multi-input architecture for four musical domains
• Supervised-learning models for jazz musician identification
• Open-source models and web application released