burkimbia/speech-dataset-public

Moore Language Speech Dataset

Dataset Overview

150 hours of Moore language speech data for advancing Burkina Faso AI community and speech technology development.The corpus includes diverse speakers and contexts to ensure robust model training capabilities

Key Statistics

• Total Duration: ~ 150 hours
• Audio Format: WAV files sampled at 16kHz
• Language: Moore (Mooré)
• Quality: Manually validated by contributors

Data Fields

• audio: Audio recordings (16kHz sampling rate)
• full_text: Moore transcription
• duration: Duration in seconds
• speakers: Speaker diarization information
• source: Recording metadata

Technical Processing

Audio Processing

• Deezer Spleeter: Music removal and audio cleaning
• Format: Standardized to 15kHz WAV

Speaker Analysis

• pyannote.audio: Speaker diarization for multi-speaker content

Processing Tools

• https://github.com/sawadogosalif/audio_processing_playground/
• https://github.com/anyantudre/MooreSpeechCorpora (for bible.com)

Contributors and Credits

Dataset Preparation

• Prepared by: @sawadogosalif Salif SAWADOGO
• Credit: @anyantudre Alban NYANTUDRE for MooreSpeechCorpora tool for bible.com

Contributors

• @sheickydollar
• Ismael SIMPORE
• Boukaré OUEDRAOGO
• Adama OUARMA

Usage Guidelines

⚠️ Dataset Gating Notice

This dataset is gated and should NOT be used for commercial purposes.

Permitted Uses

• Academic research and education
• Non-commercial AI model development
• Open-source community projects

Prohibited Uses

• Commercial speech recognition services
• Proprietary product development
• Revenue-generating applications
• Closed-source commercial software

Access Requirements

Users must:

1. Acknowledge non-commercial use restriction
2. Provide information about intended research use
3. Agree to proper attribution
4. Respect Moore community cultural heritage

Basic Usage

# pip install "datasets[audio]<4.0.0"
from datasets import load_dataset, DownloadConfig
dataset = load_dataset("sawadogosalif/speech-dataset-public", split="train", download_config=DownloadConfig(token=os.environ["HF_TOKEN"])))

Research Applications

• Automatic Speech Recognition (ASR) for Moore
• Speaker Diarization
• Cross-lingual Speech Processing
• Low-resource Language Technologies
• African Language NLP

License

Non-commercial license designed to:

• Support open research and education
• Protect Moore language community interests
• Ensure ethical use of linguistic resources
• Promote responsible AI development

By using this dataset, you agree to respect these terms and contribute positively to the Burkina Faso AI community.

---

Advancing AI technologies for African languages through collaborative, high-quality linguistic resources.