EMINES/Tamazight-Speech-to-Arabic-Text

Tamazight-Arabic Speech Recognition Dataset

Overview

This is the EMINES organization-hosted version of the Tamazight-Arabic Speech Recognition Dataset, synchronized with the original dataset. It contains ~15.5 hours of Tamazight speech (Tachelhit dialect) paired with Arabic transcriptions, designed for developing ASR and translation systems.

Quick Start

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("EMINES/Tamazight-Speech-to-Arabic-Text")

# Example usage
for example in dataset["train"]:
    audio = example["audio"]
    text = example["text"]
    metadata = example["metadata"]

Dataset Description

The dataset contains speech segments in Tamazight (Tachelhit dialect) paired with Modern Standard Arabic transcriptions. It serves as a foundation for:

• Speech recognition systems for Tamazight
• Speech-to-text translation models
• Linguistic research and language preservation

Dataset Statistics

• Total Examples: 20,344 audio segments
• Training Set: 18,309 examples (~8.9GB)
• Test Set: 2,035 examples (~992MB)
• Total Duration: ~15.5 hours
• Audio Format: WAV
• Text Format: Arabic script

Data Fields

{
    'audio': {
        'path': str,        # Path to the audio file
        'array': np.array,  # The audio signal
        'sampling_rate': int # Sampling rate of the audio
    },
    'text': str,           # Arabic transcription
    'metadata': {
        'duration': float,  # Duration in seconds
        'start_time': float, # Start time in original video
        'end_time': float,  # End time in original video
        'language': str,    # Language code
        'source': str      # Source video identifier
    }
}

Community & Development

This is an actively growing community project. We welcome contributions in:

1. Data

   • New audio-text pairs (Tachelhit)
   • Other Tamazight dialects (Tarifit, Central Atlas)
   • Transcription validation

2. Technical Improvements

   • Open-source subtitle extraction solutions
   • Pipeline optimizations
   • Preprocessing techniques

The source code for the data processing pipeline will be available soon!

Future Development

The dataset is expanding to include:

• Additional Tamazight dialects
• Improved data processing tools
• Community-contributed enhancements

Limitations and Biases

• Current focus on Tachelhit dialect
• Source material limited to movie content
• Variable audio quality
• Potential OCR-related transcription errors

Additional Information

Citation

If you use this dataset, please cite both the organization version and the original dataset:

@dataset{tamazight_asr_2024,
  author       = {Contributors},
  title        = {Tamazight-Arabic Speech Recognition Dataset},
  year         = {2025},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/EMINES/Tamazight-Speech-to-Arabic-Text} \url{https://huggingface.co/datasets/SoufianeDahimi/Tamazight-ASR-Dataset-v2}}
}

License

Apache 2.0

Links

• Original Dataset
• GitHub Repository
• Issue Tracker

Acknowledgments

This dataset was initially created by Soufiane Dahimi and is now maintained by the community under the EMINES organization umbrella.