← DatasetsGénération de texte, Classification de texte, Classification tabulaire, Embeddings · mis à jour il y a 4 sem
GH
gnakalehacker/rap-ivoire-v0-1
⌗ datasetDataset structuré de métadonnées et d’annotations sur le Rap Ivoire : 872 morceaux, 5 322 segments alignés, 44,39 heures représentées, captions bilingues, caractéristiques musicales et splits sans fuite connue d’artistes.
Vues
93
Téléchargements
12
Likes
0
Clones
0
Fichiers
13
Taille
18 Mo
| # | entier 181 | texte · longueurs 870 |
|---|---|---|
1 | # Mandia Rap Ivoire v0.1 | |
2 | Mandia Rap Ivoire is a structured dataset describing contemporary | |
3 | Ivorian rap music. | |
4 | This public release contains **metadata and annotations only**. | |
5 | It does not redistribute source recordings or separated audio stems. | |
6 | ## Dataset size | |
7 | - 872 usable tracks | |
8 | - 5,322 aligned segments | |
9 | - approximately 44.39 hours represented | |
10 | - target segment duration: approximately 30 seconds | |
11 | ## Segment categories | |
12 | - 5,289 `paired_music` | |
13 | - 26 `beat_only` | |
14 | - 7 `audiovisual_context` | |
15 | `paired_music` segments are suitable for the core paired | |
16 | vocal/instrumental training manifest. | |
17 | `beat_only` segments contain useful instrumental material. | |
18 | `audiovisual_context` segments were manually preserved because they | |
19 | represent authentic music-video context, transitions, scene setup, | |
20 | spoken introductions or related audiovisual material, but they are | |
21 | excluded from the core music-training recommendation. | |
22 | ## Splits | |
23 | The canonical v0.1 split is artist-aware and based on connected | |
24 | components formed from reliable identified artist entities. | |
25 | - Train: 707 tracks, 4,256 segments, ~35.486 h |