metadata

language:
  - pa
task_categories:
  - automatic-speech-recognition
pretty_name: Alvyn Abranches Punjabi Youtube Dataset
dataset_info:
  features:
    - name: id
      dtype: string
    - name: audio
      dtype: audio
    - name: sentence
      dtype: string
  splits:
    - name: train
      num_bytes: 10393183807
      num_examples: 23673
  download_size: 10356864899
  dataset_size: 10393183807
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Punjabi Youtube Dataset

This dataset consists of 80% of Sikhism based Educational Video content downloaded from Youtube. This dataset is weakly labeled and does not have any punctuations in it nor it guarentees that the first word or the last word is not skipped.
In case you are training whisper-large-v3 then only this data would be useful.

YouTube Videos: 999
Total Time: "179:55:06", "hours:minutes:seconds"