Datasets:
Upload ATTHack FR test (36634 rows, 27.1h, 16 kHz mono FLAC)
Browse files- README.md +102 -0
- data/test-00000-of-00002.parquet +3 -0
- data/test-00001-of-00002.parquet +3 -0
README.md
ADDED
|
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-nc-sa-4.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- automatic-speech-recognition
|
| 5 |
+
language:
|
| 6 |
+
- fr
|
| 7 |
+
size_categories:
|
| 8 |
+
- 10K<n<100K
|
| 9 |
+
pretty_name: ATTHack French — test split
|
| 10 |
+
tags:
|
| 11 |
+
- atthack
|
| 12 |
+
- french
|
| 13 |
+
- asr
|
| 14 |
+
- speech
|
| 15 |
+
- expressive-speech
|
| 16 |
+
- acted-speech
|
| 17 |
+
annotations_creators:
|
| 18 |
+
- expert-generated
|
| 19 |
+
source_datasets:
|
| 20 |
+
- extended|atthack
|
| 21 |
+
dataset_info:
|
| 22 |
+
features:
|
| 23 |
+
- name: id
|
| 24 |
+
dtype: string
|
| 25 |
+
- name: speaker_id
|
| 26 |
+
dtype: string
|
| 27 |
+
- name: axis
|
| 28 |
+
dtype: string
|
| 29 |
+
- name: sentence_id
|
| 30 |
+
dtype: string
|
| 31 |
+
- name: version
|
| 32 |
+
dtype: string
|
| 33 |
+
- name: transcript
|
| 34 |
+
dtype: string
|
| 35 |
+
- name: duration_sec
|
| 36 |
+
dtype: float32
|
| 37 |
+
- name: audio
|
| 38 |
+
dtype: audio
|
| 39 |
+
splits:
|
| 40 |
+
- name: test
|
| 41 |
+
num_examples: 36634
|
| 42 |
+
configs:
|
| 43 |
+
- config_name: default
|
| 44 |
+
data_files:
|
| 45 |
+
- split: test
|
| 46 |
+
path: data/test-*.parquet
|
| 47 |
+
---
|
| 48 |
+
|
| 49 |
+
# ATTHack — French test split
|
| 50 |
+
|
| 51 |
+
Split test du corpus **ATTHack** (IRCAM / OpenSLR 88) — corpus expressif
|
| 52 |
+
français, parole actée multi-axes (émotions, attitudes). Empaqueté en
|
| 53 |
+
Parquet shardé avec audio **FLAC 16 kHz mono** (resamplé depuis le 44.1 kHz
|
| 54 |
+
original pour cohérence avec Whisper / Canary).
|
| 55 |
+
|
| 56 |
+
## Contenu
|
| 57 |
+
|
| 58 |
+
- **36634** utterances
|
| 59 |
+
- Audio : **FLAC 16 kHz mono PCM_16** (resamplé depuis 44.1 kHz upstream)
|
| 60 |
+
- Langue : **français (fr)**
|
| 61 |
+
- Licence : **CC-BY-NC-SA-4.0** (héritée de ATTHack / OpenSLR 88)
|
| 62 |
+
- Durée totale : **27.12 h**
|
| 63 |
+
|
| 64 |
+
## Colonnes
|
| 65 |
+
|
| 66 |
+
| Colonne | Type | Description |
|
| 67 |
+
|---------|------|-------------|
|
| 68 |
+
| `audio` | Audio | dict `{path, array, sampling_rate}` — décodé auto |
|
| 69 |
+
| `id` | string | ID complet : `{speaker}_{axis}_{sentence}_{version}` |
|
| 70 |
+
| `speaker_id` | string | locuteur (ex `F01`, `M03`) |
|
| 71 |
+
| `axis` | string | axe expressif (ex `a1`, `a2`, …) |
|
| 72 |
+
| `sentence_id` | string | ID phrase (ex `s001`) |
|
| 73 |
+
| `version` | string | version d'enregistrement (ex `v01`) |
|
| 74 |
+
| `transcript` | string | transcription FR (référence WER) |
|
| 75 |
+
| `duration_sec` | float32 | durée en secondes |
|
| 76 |
+
|
| 77 |
+
## Utilisation
|
| 78 |
+
|
| 79 |
+
```python
|
| 80 |
+
from datasets import load_dataset
|
| 81 |
+
|
| 82 |
+
ds = load_dataset("ggfox00000/stt-atthack-test", split="test")
|
| 83 |
+
print(ds)
|
| 84 |
+
print(ds[0]["transcript"], ds[0]["speaker_id"], ds[0]["axis"])
|
| 85 |
+
print(ds[0]["audio"]["sampling_rate"], ds[0]["audio"]["array"].shape)
|
| 86 |
+
```
|
| 87 |
+
|
| 88 |
+
## Source
|
| 89 |
+
|
| 90 |
+
- ATTHack — IRCAM, https://www.openslr.org/88/
|
| 91 |
+
- Corpus expressif français acted speech, multi-locuteurs / multi-attitudes
|
| 92 |
+
|
| 93 |
+
## Licence
|
| 94 |
+
|
| 95 |
+
**CC-BY-NC-SA-4.0** (héritée de ATTHack upstream / OpenSLR 88) — usage non
|
| 96 |
+
commercial, partage à l'identique.
|
| 97 |
+
|
| 98 |
+
## Note format
|
| 99 |
+
|
| 100 |
+
Audio downsamplé 44.1 kHz → 16 kHz mono via `soxr.resample` (qualité HQ)
|
| 101 |
+
pour homogénéiser avec FLEURS / MLS / MediaSpeech (tous 16 kHz mono Whisper-
|
| 102 |
+
compatible). Les transcriptions et IDs sont préservés tels quels.
|
data/test-00000-of-00002.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f582ce527f3eb772adab9969d432f008e8153afbb91fd4c464789b3c7cc4dba
|
| 3 |
+
size 501993241
|
data/test-00001-of-00002.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e31194406acc8fe7e7b483e6bd740c3868ce1f29a371a7ad1eadfee610a18dce
|
| 3 |
+
size 498449739
|