boffire commited on
Commit
4738bd7
·
verified ·
1 Parent(s): 20d2897

Update metadata files

Browse files
Files changed (4) hide show
  1. README.md +12 -27
  2. dataset_info.json +3 -6
  3. metadata.csv +0 -0
  4. metadata.jsonl +0 -0
README.md CHANGED
@@ -1,46 +1,31 @@
1
- ---
2
- license: cc-by-2.0
3
- language:
4
- - kab
5
- - ber
6
- pretty_name: kabyle synthetic audio
7
- tags:
8
- - kabyle
9
- - taqbaylit
10
- - taɣect
11
- - voice
12
- ---
13
  # Kabyle Parallel Corpus (OmniVoice × Tatoeba)
14
 
15
- Corpus parallèle de **998 phrases kabyles** avec audio généré par [OmniVoice](https://github.com/k2-fsa/OmniVoice).
16
 
17
  ## Statistiques
18
  - **Langue** : kabyle (`kab`)
19
- - **Phrases** : 998
20
- - **Durée totale** : 1949.3s
 
21
  - **Sampling rate** : 24000 Hz
22
- - **Source texte** : [Tatoeba](https://tatoeba.org) (phrases sans audio existant)
23
  - **Modèle TTS** : `k2-fsa/OmniVoice`
 
24
 
25
  ## Structure
26
  ```
27
- kabyle_corpus/
28
  ├── audio/ # Fichiers WAV
29
- ├── kab_00000001.wav
30
- │ └── ...
31
- ├── metadata.jsonl # Métadonnées (JSON Lines)
32
- ├── metadata.csv # Métadonnées (CSV)
33
  └── dataset_info.json # Info HuggingFace Datasets
34
  ```
35
 
36
- ## Utilisation avec HuggingFace `datasets`
37
  ```python
38
  from datasets import load_dataset
39
- ds = load_dataset("audiofolder", data_dir="kabyle_corpus")
40
- print(ds["train"][0]["text"])
41
- ds["train"][0]["audio"] # dict avec 'array' et 'sampling_rate'
42
  ```
43
 
44
  ## ⚖️ Licence
45
- Les textes proviennent de Tatoeba sous licence **CC BY 2.0 FR**.
46
- Les audios sont générés par OmniVoice.
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  # Kabyle Parallel Corpus (OmniVoice × Tatoeba)
2
 
3
+ Corpus parallèle de **997 phrases kabyles** avec audio généré par [OmniVoice](https://github.com/k2-fsa/OmniVoice).
4
 
5
  ## Statistiques
6
  - **Langue** : kabyle (`kab`)
7
+ - **Phrases totales** : 997
8
+ - **Nouvelles phrases (ce run)** : 987
9
+ - **Durée totale** : 1958.4s (32.6 min)
10
  - **Sampling rate** : 24000 Hz
11
+ - **Source texte** : [Tatoeba](https://tatoeba.org)
12
  - **Modèle TTS** : `k2-fsa/OmniVoice`
13
+ - **Dernière mise à jour** : 2026-05-10T10:49:01.178795
14
 
15
  ## Structure
16
  ```
17
+ kabyle_corpus_997/
18
  ├── audio/ # Fichiers WAV
19
+ ├── metadata.jsonl # Métadonnées
20
+ ── metadata.csv # Métadonnées CSV
 
 
21
  └── dataset_info.json # Info HuggingFace Datasets
22
  ```
23
 
24
+ ## Utilisation
25
  ```python
26
  from datasets import load_dataset
27
+ ds = load_dataset("audiofolder", data_dir=".")
 
 
28
  ```
29
 
30
  ## ⚖️ Licence
31
+ Textes : CC BY 2.0 FR (Tatoeba) | Audios : générés par OmniVoice
 
dataset_info.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "description": "Corpus parallèle kabyle (texte + audio) généré avec OmniVoice à partir de phrases Tatoeba sans enregistrement audio existant.",
3
  "citation": "@misc{kabyle_corpus_omnivoice, title={Kabyle Parallel Corpus (OmniVoice × Tatoeba)}, author={Generated via OmniVoice}, year={2026}}",
4
  "homepage": "https://tatoeba.org",
5
  "license": "CC BY 2.0 FR",
@@ -56,11 +56,8 @@
56
  "train": {
57
  "name": "train",
58
  "num_bytes": null,
59
- "num_examples": 998,
60
  "dataset_name": "kabyle_corpus"
61
  }
62
- },
63
- "download_size": null,
64
- "dataset_size": null,
65
- "size_in_bytes": null
66
  }
 
1
  {
2
+ "description": "Corpus parallèle kabyle (texte + audio) généré avec OmniVoice. 997 phrases.",
3
  "citation": "@misc{kabyle_corpus_omnivoice, title={Kabyle Parallel Corpus (OmniVoice × Tatoeba)}, author={Generated via OmniVoice}, year={2026}}",
4
  "homepage": "https://tatoeba.org",
5
  "license": "CC BY 2.0 FR",
 
56
  "train": {
57
  "name": "train",
58
  "num_bytes": null,
59
+ "num_examples": 997,
60
  "dataset_name": "kabyle_corpus"
61
  }
62
+ }
 
 
 
63
  }
metadata.csv CHANGED
The diff for this file is too large to render. See raw diff
 
metadata.jsonl CHANGED
The diff for this file is too large to render. See raw diff