README.md

---
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-data5", 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.