Datasets:
audio unknown | text string | condition string | voice string | command_type string |
|---|---|---|---|---|
[
82,
73,
70,
70,
6,
231,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
76,
73,
83,
84,
26,
0,
0,
0,
73,
78,
70,
79,
73,
83,
70,
84,
14,
0,
0,
0,
... | sélectionner la suivante | low_rms | ALAVOINE_Nadege.mp3 | selection |
[
82,
73,
70,
70,
178,
161,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
76,
73,
83,
84,
26,
0,
0,
0,
73,
78,
70,
79,
73,
83,
70,
84,
14,
0,
0,
0... | fin du champ | low_rms | ALAVOINE_Nadege.mp3 | navigation |
"UklGRgbzAABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGH(...TRUNCATED) | reprendre à prilocaine | low_rms | ALG2019_0035_ALG2019_0035.mp3 | edition |
"UklGRlw4AQBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGE(...TRUNCATED) | sélectionner la phrase précédente | low_rms | ALG2019_0035_ALG2019_0035.mp3 | selection |
"UklGRlzEAABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGE(...TRUNCATED) | sélectionner mot | low_rms | ALG2019_0057_ALG2019_0057.mp3 | selection |
"UklGRrJJAQBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGF(...TRUNCATED) | reprendre à eporatio | low_rms | ALG2019_0057_ALG2019_0057.mp3 | edition |
"UklGRrLtAABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGF(...TRUNCATED) | ouvrir la parenthèse | low_rms | ALG2019_0058_ALG2019_0058.mp3 | ponctuation |
"UklGRgbnAABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGH(...TRUNCATED) | premier champ | low_rms | ALG2019_0058_ALG2019_0058.mp3 | navigation |
"UklGRrIJAgBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGF(...TRUNCATED) | chiffre romain quatre | low_rms | ALG2019_0131_ALG2019_0131.mp3 | chiffres |
"UklGRgbzAABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAATElTVBoAAABJTkZPSVNGVA4AAABMYXZmNTguNzYuMTAwAGRhdGH(...TRUNCATED) | effacer zoledronate | low_rms | ALG2019_0131_ALG2019_0131.mp3 | edition |
End of preview. Expand in Data Studio
FR-Vocalcommands-dictation-app
Commandes vocales de l'application de dictee PraxyDICTE, en francais : formatage, edition, selection, navigation, majuscules, ponctuation et valeurs medicales (guide PraxyDICTE LITE v1.6.20). Les commandes de selection et d'edition sont enrichies de noms de medicaments (liste du repo de fine-tuning Qwen) pour coller au vocabulaire reel des utilisateurs.
| element | valeur |
|---|---|
| clips audio | 49092 |
| textes distincts | 6849 |
| duree | 37.03 h |
| voix | Coqui XTTS v2, 300 voix FR clonees |
| conditions audio | normal (XTTS natif), low_rms (volume -23 dB), speedup (atempo 1.2) |
| format | WAV mono : 24 kHz pour normal, 16 kHz pour low_rms et speedup |
| texte | minuscules, accents conserves, ponctuation dictee incluse |
Structure
audio/<condition>/<shard>/<index>.wav
metadata.jsonl (audio_filepath, text, duration, voice, command_type, condition, index)
parquet/ (audio binaire + text + condition)
Colonnes de metadata.jsonl
audio_filepath: chemin relatif au repo (manifest NeMo compatible) ;text: commande prononcee, minuscules, sans ponctuation finale ;duration: duree du clip en secondes ;voice: fichier de voix de reference utilise par XTTS ;command_type:formatage,edition,selection,navigation,majuscules,ponctuation,chiffres,reco_medicale,insertion;condition:normal,low_rmsouspeedup;index: index du texte dans le corpus source (identique entre conditions).
Usage NeMo
manifest = "metadata.jsonl" # audio_filepath est relatif au repo
- Downloads last month
- 399