MusicXML tokenisation improves symbolic music AI understanding and creation
Does Notation Matter? MusicXML Tokenisation for Symbolic Music Generation and Understanding
SoundMachine Learning
Summary
Music can be written down in many ways, from simple digital notes to full detailed sheet music. The authors studied if using richer, more detailed music formats helps AI programs learn and create music better. They made a new method to turn detailed music sheets (MusicXML) into data AI can understand. Their tests showed that AI using this richer format did better both at tasks like guessing the composer and suggesting musical parts, and at generating new music that people preferred. This means that the way music is represented can affect how well AI understands and makes music.
What this means in practice
- •For music software developers: Build music generation tools that use rich MusicXML format for better understanding and creation quality.
- •For digital music platform teams: Improve music analysis features like composer recognition and difficulty estimation by leveraging MusicXML-based models.
Authors
Andrea Poltronieri, Matteo Spanio, Simone Chieppa, Xavier Serra, Martín Rocamora
Abstract
Symbolic music representations encode musical content at varying levels of completeness, from performance-oriented formats such as MIDI to fully notated scores such as MusicXML. Yet the impact of this representational richness on deep learning models remains poorly understood. While tokenisation has been extensively studied for MIDI, score-based formats have received considerably less attention, and MusicXML in particular remains largely unexplored despite being the de-facto standard for score interchange. In this paper, we introduce the first general-purpose tokenisation framework for MusicXML, and conduct a systematic empirical study to determine whether notational richness confers a measurable advantage for deep learning on symbolic music. We train 26 autoregressive and masked language models, comparing MusicXML tokenisation against MIDI and ABC-based representations, probing the effect of different tokenisation strategies within MusicXML, and testing whether music-grammar-aware tokenisation outperforms treating scores as plain text. We evaluate music understanding tasks using linear probing across composer classification, difficulty estimation, and accompaniment suggestion, and generative tasks using objective statistical metrics and a human pairwise preference study. Our results show that MusicXML-tokenised models outperform MIDI and text-based counterparts on understanding and generative tasks.