README.md

metadata

license: apache-2.0
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: gender
      dtype: string
    - name: speaker_id
      dtype: string
    - name: transcription
      dtype: string
  splits:
    - name: train
      num_bytes: 222401245.94
      num_examples: 2654
    - name: test
      num_bytes: 90421105.904
      num_examples: 1138
  download_size: 313908599
  dataset_size: 312822351.844
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

Notes

• Same format as authentic dataset here
• Added train split (70%) and test split (30%)
• authentic data of the same split could be found on authentic dataset with split
• synthesized using Openai tts (self-funded)
  • based on transcription of the anthentic dataset
  • synthesized using TTS for mandarin, with no external prompt

For non-comercial use only