Datasets:
Search is not available for this dataset
audio audioduration (s) 0.44 11.3 |
|---|
End of preview. Expand in Data Studio
Persian Voice Atlas
A Persian (Farsi) narrated-speech dataset with emotion and category annotations, intended for text-to-speech, speech recognition, and speech-emotion research.
Status: work in progress. Speaker S01 is uploaded first; additional speakers (S02, ...) and the full metadata file will follow.
Audio
- Format: WAV
- Sample rate: [FILL IN, e.g. 22050 Hz / 44100 Hz]
- Channels: [mono / stereo]
- Organised by speaker:
S01/S01_SCRIPT_0001.wav, etc.
Metadata schema
The metadata file (added later) uses these columns:
| column | description |
|---|---|
file_name |
path to the clip, relative to the repo root |
id |
clip identifier (e.g. SCRIPT_0001) |
duration |
clip length in seconds |
speaker_id |
speaker label (S01, S02, ...) |
gender |
speaker gender |
age_group |
speaker age bracket |
transcript_raw |
original Persian transcript |
transcript_normalized |
normalised Persian transcript |
emotion |
emotion label (e.g. happy, neutral) |
emotion_intensity |
low / medium / high |
nonverbal_tokens |
non-verbal sounds in the clip |
category |
content category (e.g. greeting, weather) |
notes |
free-text notes |
status |
recording status |
flag_reason |
QA flag, if any |
License
[CHOOSE A LICENSE deliberately - e.g. CC BY 4.0, CC BY-NC 4.0, or custom - as it governs how others may use the data.]
- Downloads last month
- 78