lamm-mit motif-clustering
Clusters melodic motifs using scikit-learn (KMeans or hierarchical clustering) with optional UMAP projection. Takes motif JSON (from motif-detection) and groups them by interval similarity.
Usage
python3 skills/motif-clustering/scripts/motif_clustering.py --query "bach motifs" --n-clusters 8 --method kmeans
python3 skills/motif-clustering/scripts/motif_clustering.py --query "bach motifs" --method hierarchical
Output
{
"method": "kmeans",
"n_clusters": 8,
"clusters": [
{
"cluster_id": 0,
"n_members": 12,
"centroid_intervals": [2, -1, 2],
"genre_distribution": {"baroque": 8, "folk": 4},
"top_motifs": ["m_0001", "m_0012"]
}
],
"silhouette_score": 0.47,
"inertia": 234.5
}
Dependencies
- scikit-learn
- music21