---
dataset_info:
- config_name: default
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 48000
  - name: duration
    dtype: float64
  - name: speaker
    dtype: string
  - name: text
    dtype: string
  - name: weak_transcription
    dtype: string
  - name: text_from_weak_transcription
    dtype: bool
  - name: filename
    dtype: string
  - name: speaker_embedding
    sequence: float16
  splits:
  - name: train
    num_bytes: 1031251593.0036386
    num_examples: 9791
  download_size: 1130099555
  dataset_size: 1031251593.0036386
- config_name: denoised
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 48000
  - name: duration
    dtype: float32
  - name: full_text
    dtype: string
  - name: verses
    sequence:
    - name: verse_id
      dtype: string
    - name: text
      dtype: string
  splits:
  - name: train
    num_bytes: 1332368261
    num_examples: 626
  download_size: 1300523557
  dataset_size: 1332368261
- config_name: raw
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 48000
  - name: duration
    dtype: float32
  - name: full_text
    dtype: string
  - name: speech_with_music
    dtype: bool
  - name: verses
    sequence:
    - name: verse_id
      dtype: string
    - name: text
      dtype: string
  splits:
  - name: train
    num_bytes: 1233358863.0
    num_examples: 626
  download_size: 1189075024
  dataset_size: 1233358863.0
- config_name: utterance-semi-transcription
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 48000
  - name: duration
    dtype: float64
  - name: speaker
    dtype: string
  - name: full_text
    dtype: string
  - name: transcription_first_half
    dtype: string
  - name: transcription_second_half
    dtype: string
  - name: full_transcription
    dtype: string
  splits:
  - name: train
    num_bytes: 1205323718.845
    num_examples: 11041
  download_size: 1150601541
  dataset_size: 1205323718.845
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
- config_name: denoised
  data_files:
  - split: train
    path: denoised/train-*
- config_name: raw
  data_files:
  - split: train
    path: raw/train-*
- config_name: utterance-semi-transcription
  data_files:
  - split: train
    path: utterance-semi-transcription/train-*
language:
- bm
---
# Bible Bambara Audio Dataset

## Overview

This dataset contains audio recordings of Bible passages in Bambara language along with their transcriptions. The dataset consists of approximately **42.7** hours of audio content, making it a valuable resource for speech processing tasks in Bambara language.

## Project

This dataset is part of a larger initiative to preserve and digitize audio & texts in Bambara language, making them accessible in both text and audio formats. The audio recordings provide high-quality readings of biblical texts, offering a resource for both linguistic research and accessibility purposes.

## Bambara Language

Bambara (Bamanankan) is a Mande language spoken primarily in Mali by millions of people as both a first and second language. It serves as a lingua franca in Mali and is also spoken in neighboring countries (Burkina Faso, Ivory Coast, etc.). Bambara is written in both Latin script and N'Ko script, and has a rich oral tradition integral to Malian culture.

## Dataset Structure

### Data Fields

The dataset includes the following fields:

- **audio**: Audio file containing the spoken Bible passage (48kHz sampling rate)
- **full_text**: Complete transcription of the audio passage in Bambara
- **speech_with_music**: Boolean indicating whether the audio contains background music
- **duration**: Length of the audio clip in seconds
- **verses**: A sequence of verse objects containing:
  - **verse_id**: Identifier of the verse
  - **text**: Text content of the individual verse

## Usage

This dataset can be used for various speech technology applications:

- **Speech Recognition**: Training ASR models for Bambara
- **Text-to-Speech**: Development of TTS systems for Bambara
- **Speech Analysis**: Studying prosody and phonetics of Bambara

### Loading the Dataset

```python
from datasets import load_dataset
dataset = load_dataset("djelia/bible-bambara-audio", "raw")
```

## Dataset Creation

The audio recordings and transcriptions were collected from Bible.is website.

## Additional Information

- **Source**: https://live.bible.is/bible/BAMLSB/DEU/1
- **Curated by**: @oza75 at Djelia
- **Total Duration**: 42.67 hours
- **Audio Format**: WAV files sampled at 48kHz
- **Language**: Bambara