README.md

metadata

dataset_info:
  features:
    - name: text
      dtype: string
    - name: timestamp
      dtype: string
    - name: url
      dtype: string
    - name: zh_id
      dtype: int64
  splits:
    - name: train
      num_bytes: 8494000105
      num_examples: 4856777
  download_size: 5360940712
  dataset_size: 8494000105
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
language:
  - zht
  - zh
task_categories:
  - text-generation
  - fill-mask
pretty_name: C4-zhtw
size_categories:
  - 100M<n<1B

Dataset Card for C4-zhtw

• Traditional Chinese subset of the C4 dataset.
• De-duplicated with MinHash.
• Is suggested to filter the dataset with NLU models before any serious use.