Evaluating explainable harmonic decision-making on music corpora
In brief
- EMH generates chords from melodies using explicit, interpretable rules.
- Exact agreement with human annotations: 20.5% for chords, 47.6% for harmonic functions.
- 46% of human chords fall outside the model's vocabulary, limiting direct comparisons.
A study tests an explainable chord generation model (EMH) on annotated musical excerpts from HookTheory. The rule-based model achieves 20.5% exact agreement with human annotations. Results distinguish representation limits from divergent harmonic choices without favoring the human solution.
This summary gives you the essentials. The full article, its figures and quotes are on cs.AI updates on arXiv.org.
Read the original articleAI-generated summary of the source, checked by rules. Report an error