Graph-aware predictions
AnalysisGNN combines symbolic score structure with neural prediction across harmonic and form tasks.
About
A research workspace for interactive music analysis.
The app is designed for analysts who want state-of-the-art AI music analysis, score context, and controlled correction loops in one place.
Human-in-the-loop refinement
Region corrections are applied immediately and can also become constraints when the model re-runs selected regions around expert edits.
Interoperable outputs
Corrected results can be exported as CSV, DCML harmonies TSV, Delta Lake data, or MusicXML with harmony annotations.
Corpus creation with AI-assisted correction
Generate first-pass labels, correct predictions directly from the score, and use those edits as controlled context for re-inference. This supports traceable corpus-building while keeping labels editable and exportable.
Why a website shell?
The analysis workspace remains a focused tool, while documentation and project context now live in proper pages with stable navigation.
Model scope and review
AnalysisGNN is designed to accelerate expert work, not replace it. Harmony and structure predictions are generated by an AI model, so not every label will be correct or uniquely determined. Treat the output as an interpretive proposal, review it against the score, and use corrections to keep model outputs, human decisions, and exported annotations traceable.
Research paper
Read the AnalysisGNN paper for the model design, multi-task setup, and evaluation context.
AnalysisGNN has received support from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101019375 "Whither Music?".