| --- |
| language: |
| - ha |
| license: cc-by-4.0 |
| task_categories: |
| - automatic-speech-recognition |
| tags: |
| - audio |
| - speech |
| - hausa |
| - african-voices |
| pretty_name: AfricanVoices Hausa Train |
| size_categories: |
| - 10K<n<100K |
| --- |
| |
| # AfricanVoices Hausa -- Train Split |
|
|
| Hausa speech dataset from [AfricanVoices.io](https://africanvoices.io). |
|
|
| ## Usage |
|
|
| ```python |
| from datasets import load_dataset |
| ds = load_dataset("suleiman2003/afri-temp-data4", split="train") |
| print(ds[0]) |
| # {'audio': Audio(...), 'transcript': '...', 'gender': '...', ...} |
| ``` |
|
|
| ## Structure |
|
|
| Each batch is in its own subdirectory under `train/`: |
| ``` |
| train/ |
| batch_1/ |
| *.flac + metadata.csv |
| batch_2/ |
| *.flac + metadata.csv |
| ... |
| ``` |
|
|
| Audio files are FLAC format. Metadata includes: transcript, speaker_id, |
| gender, age_group, education, duration, language, snr, domain. |
|
|
