README.md

---
dataset_info:
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
  - name: sentence
    dtype: string
  - name: source
    dtype: string
  splits:
  - name: train
    num_bytes: 5387295926.242
    num_examples: 28807
  - name: test
    num_bytes: 1290121556.452
    num_examples: 6268
  download_size: 6474496191
  dataset_size: 6677417482.693999
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
license: apache-2.0
task_categories:
- automatic-speech-recognition
language:
- wo
size_categories:
- 10K<n<100K
---


# Wolof Audio Dataset

The **Wolof Audio Dataset** is a collection of audio recordings and their corresponding transcriptions in Wolof. This dataset is designed to support the development of Automatic Speech Recognition (ASR) models for the Wolof language. It was created by combining three existing datasets:

- **ALFFA**: Available at [serge-wilson/wolof_speech_transcription](https://huggingface.co/datasets/serge-wilson/wolof_speech_transcription)
- **FLEURS**: Available at [vonewman/fleurs-wolof-dataset](https://huggingface.co/datasets/vonewman/fleurs-wolof-dataset)
- **Urban Bus Wolof Speech Dataset**: Available at [vonewman/urban-bus-wolof](https://huggingface.co/datasets/vonewman/urban-bus-wolof)
- **Kallama Dataset**: Available at [Moustapha91/wolof_kalama_TTS](https://huggingface.co/datasets/Moustapha91/wolof_kalama_TTS)

## Dataset Description

- **Language**: Wolof (`wo`)
- **Domain**: General speech and urban transportation
- **Data Type**: Audio recordings and transcriptions
- **Audio Format**: Varies (e.g., WAV, MP3)
- **Sampling Rate**: 16 kHz
- **Total Examples**: 24,346
  - **Training Set**: 20224 examples
  - **Test Set**: 4122 examples

### Features

- `audio`: An audio file containing speech in Wolof.
  - **Format**: Varies (e.g., WAV, MP3)
  - **Sampling Rate**: 16 kHz
- `sentence`: The textual transcription of the audio in Wolof.
- `source`: The origin of the example, either `'alffa'`, `'fleurs'`, `'urban_bus'`, or `'kallama'`

### Dataset Structure

#### Splits

The dataset is divided into two splits:

| Split  | Number of Examples |
|--------|---------------------|
| Train  | 28,807              |
| Test   | 6,268               |

## Usage Example

Here's how to load and use this dataset with the 🤗 Datasets library:

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("vonewman/wolof-audio-data")

# Access an example from the 'train' split
print(dataset['train'][0])

# Expected output:
# {
#   'audio': {
#     'path': '.../train/audio/<audio_file>',
#     'array': array([...]),
#     'sampling_rate': 16000
#   },
#   'sentence': 'Transcription of the audio in Wolof',
#   'source': 'alffa'  # or 'fleurs' or 'urban_bus' or 'kallama'
# }