metadata

language:
  - fa
license: apache-2.0
size_categories:
  - 10K<n<100K
task_categories:
  - automatic-speech-recognition
pretty_name: Persian ASR Youtube (30 Second Chunk)
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: video_id
      dtype: string
    - name: segment_id
      dtype: int64
    - name: title
      dtype: string
    - name: transcription
      dtype: string
    - name: youtube_url
      dtype: string
  splits:
    - name: train
      num_bytes: 15011333947.12248
      num_examples: 32746
    - name: test
      num_bytes: 1868480465.95316
      num_examples: 4094
    - name: val
      num_bytes: 1876553690.74436
      num_examples: 4093
  download_size: 18614667732
  dataset_size: 18756368103.82
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: val
        path: data/val-*

How To Use

from datasets import load_dataset
train = load_dataset('pourmand1376/asr-farsi-youtube-chunked-30-seconds', split='train+val')
test =load_dataset('pourmand1376/asr-farsi-youtube-chunked-30-seconds', split='test')

+300 Hours ASR dataset generated from this kaggle dataset