README.md

metadata

dataset_info:
  - config_name: female-01
    features:
      - name: audio
        dtype: audio
      - name: sentence
        dtype: string
      - name: source
        dtype: string
      - name: __index_level_0__
        dtype: int64
    splits:
      - name: train
        num_bytes: 6428695607.035
        num_examples: 26035
      - name: test
        num_bytes: 1628876735.954
        num_examples: 6509
    download_size: 7348734127
    dataset_size: 8057572342.989
  - config_name: male-01
    features:
      - name: audio
        dtype: audio
      - name: sentence
        dtype: string
      - name: source
        dtype: string
      - name: __index_level_0__
        dtype: int64
    splits:
      - name: train
        num_bytes: 6428695607.035
        num_examples: 26035
      - name: test
        num_bytes: 1628876735.954
        num_examples: 6509
    download_size: 7444446024
    dataset_size: 8057572342.989
configs:
  - config_name: female-01
    data_files:
      - split: train
        path: female-01/train-*
      - split: test
        path: female-01/test-*
  - config_name: male-01
    data_files:
      - split: train
        path: male-01/train-*
      - split: test
        path: male-01/test-*

Voice Synthetic Dataset

A Dhivehi text-to-speech dataset containing professionally recorded audio samples with their corresponding transcriptions.

Dataset Configuration

• Audio Format: WAV files
• Language: Dhivehi (dv)
• Source: cv
• Split Ratio:
  • Train: 80%
  • Test: 20%
• Structure:
  • audio: Path to the audio file
  • sentence: Corresponding Dhivehi text transcription
  • source: Source of the audio data

Data Format

{
    'audio': {
        'path': 'path/to/audio.wav',
        'array': array(...),
        'sampling_rate': 22050
    },
    'sentence': 'ދިވެހި ޓެކްސްޓް'  # Dhivehi text
    'source': 'cv'  # Source identifier
}