RobotsMali/bam-asr-all

All Bambara ASR Dataset

This dataset aims to gather all publicly available Bambara ASR datasets. It is primarily composed of the Jeli-ASR dataset (available at RobotsMali/jeli-asr), along with the Mali-Pense data curated and published by Aboubacar Ouattara (available at oza75/bambara-tts). Additionally, it includes 1 hour of audio recently collected by the RobotsMali AI4D Lab, featuring children's voices reading some of RobotsMali GAIFE books. This dataset is desihgned for automatic speech recognition (ASR) task primarily.

Important Notes

1. Please note that this dataset is currently in development and is therefore not fixed. The structure, content, and availability of the dataset may change as improvements and updates are made.
2. The Dataset viewer has been disabled for this dataset since it uses a custom loading script. You can safely load this dataset and all its features as a HF dataset object though (see usage section)

---

Directory Structure

bam-asr-all/
|
├── README.md
├── metadata.csv
├── manifests/
│   ├── jeli-asr-rmai-test-manifest.json
│   ├── jeli-asr-rmai-train-manifest.json
│   ├── oza-bam-asr-test-manifest.json
│   └── oza-bam-asr-train-manifest.json
│   └── oza-mali-pense-train-manifest.json
│   └── reading-tutor-train-manifest.json
│   └── train-manifest.json # jeli-asr-rmai-train-manifest.json + oza-bam-asr-train-manifest.json
│   └── test-manifest.json # jeli-asr-rmai-test-manifest.json + oza-bam-asr-test-manifest.json
│
├── french-manifests/
│   ├── jeli-asr-rmai-test-french-manifest.json
│   ├── jeli-asr-rmai-train-french-manifest.json
│   ├── oza-bam-asr-test-french-manifest.json
│   └── oza-bam-asr-train-french-manifest.json
│   └── oza-mali-pense-train-french-manifest.json
│
├── jeli-asr-rmai/         |
│   ├── train/             |
│   └── test/              |
│                          | These two subset are combined as jeli-asr
├── bam-asr-oza/           |
│   ├── train/             |
│   └── test/              |
|
├── oza-mali-pense/
│   ├── train/
|
├── rt-data-collection/

manifests Directory

This directory contains the manifest files used for training speech recognition (ASR) and text-to-speech (TTS) models. Those are JSON files:

Each line in the manifest files is a JSON object with the following structure:

{
  "audio_filepath": "bam-asr-all/rt-data-collection/zctn7pFmtmR45FKym7d5.wav", 
  "duration": 10.24, 
  "text": "ni birituban dɔ tun bɛ se ka piyano fɔ, a tun bɛ fara sɛ ka dɔnkilida la."
}

• audio_filepath: The relative path to the corresponding audio file.
• duration: The duration of the audio file in seconds.
• text: The transcription of the audio in Bambara.

3. french-manifests/

This directory contains French equivalent manifest files for the dataset. The structure is similar to the manifests/ directory but with French transcriptions

---

Dataset Details

• Total Duration: 37.41 hours
• Number of Samples: 38,769
  • Training Set: 37,306 samples
  • Testing Set: 1,463 samples

Subsets:

• Oza's Bambara-ASR: ~29 hours (clean subset).
• Jeli-ASR-RMAI: ~3.5 hours (filtered subset).
• oza-tts-mali-pense: ~4 hours
• reading-tutor-data-collection: ~1 hour

---

Usage

The manifest files are specifically created for training Automatic Speech Recognition (ASR) models in NVIDIA NeMo framework, but they can be used with any other framework that supports manifest-based input formats or reformatted for other use cases.

To use the dataset, simply load the manifest files (train-manifest.json and test-manifest.json) in your training script. The file paths for the audio files and the corresponding transcriptions are already provided in these manifest files You can also load it directly in a HuggingFace dataset object.

Downloading the Dataset:

# Clone dataset repository maintaining directory structure for quick setup with Nemo
git clone --depth 1 https://huggingface.co/datasets/RobotsMali/bam-asr-all

OR

from datasets import load_dataset

# Load the dataset into Hugging Face Dataset object
dataset = load_dataset("RobotsMali/bam-asr-all", trust_remote_code=True)
# Note: You can also download only a specific subset if you with

Known Issues

This dataset also has most of the issues of Jeli-ASR, plus a few samples with missing french translations

---

Citation

If you use this dataset in your research or project, please credit the creators of these datasets.

• Jeli-ASR dataset: Jeli-ASR Dataset. 
• Oza's Bambara-ASR dataset: oza75/bambara-asr
• Oza's Bambara-TTS dataset: oza75/bambara-tts