README.md

metadata

dataset_info:
  features:
    - name: context
      dtype: string
    - name: question
      dtype: string
    - name: answer
      struct:
        - name: answer_end
          dtype: int64
        - name: answer_start
          dtype: int64
        - name: text
          dtype: string
  splits:
    - name: train
      num_bytes: 6245598
      num_examples: 9332
    - name: validation
      num_bytes: 503255
      num_examples: 764
    - name: test
      num_bytes: 553707
      num_examples: 844
  download_size: 2021465
  dataset_size: 7302560
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

I do not hold the copyright to this dataset; I merely restructured it to have the same structure as other datasets (that we are researching) to facilitate future coding and analysis. I refer to this link for the raw dataset.