poquad_v2 / README.md
mzasada's picture
Update README.md
1255ba9
metadata
license: apache-2.0
dataset_info:
  features:
    - name: id
      dtype: string
    - name: title
      dtype: string
    - name: context
      dtype: string
    - name: question
      dtype: string
    - name: is_impossible
      dtype: string
    - name: answers
      sequence:
        - name: text
          dtype: string
        - name: answer_start
          dtype: int32
        - name: generative_answer
          dtype: string
  splits:
    - name: train
      num_bytes: 66007405
      num_examples: 56618
    - name: validation
      num_bytes: 8303914
      num_examples: 7060
  download_size: 16398725
  dataset_size: 74311319
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
task_categories:
  - question-answering
language:
  - pl
size_categories:
  - 10K<n<100K

Orignal dataset is https://huggingface.co/datasets/clarin-pl/poquad

As orignal dataset is mainly for SQuADv1 type of models, this copy also unlocks access to SQuADv2 type of features in poquad like empty questions.