README.md

metadata

dataset_info:
  features:
    - name: clean
      dtype: string
  splits:
    - name: train
      num_bytes: 2675756722.6081944
      num_examples: 2135664
    - name: test
      num_bytes: 19573123.978550296
      num_examples: 13703
    - name: valid
      num_bytes: 35044138.97279745
      num_examples: 22533
  download_size: 942170210
  dataset_size: 2730373985.559542
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: valid
        path: data/valid-*

Fully diacritized arabic sentences gathered and cleaned from:

• 1st Dataset
• and 2nd Dataset where the Tashkeela part was removed from the first dataset and replaced with that of the second one.