openbookqa

  • Description:

The dataset contains 5,957 4-way multiple choice questions. Additionally, they provide 5,167 crowd-sourced common knowledge facts, and an expanded version of the train/dev/test questions where each question is associated with its originating core fact, a human accuracy score, a clarity score, and an anonymized crowd-worker ID.

Split Examples
'test' 500
'train' 4,957
'validation' 500
  • Feature structure:
FeaturesDict({
    'answerKey': ClassLabel(shape=(), dtype=tf.int64, num_classes=4),
    'clarity': tf.float32,
    'fact1': Text(shape=(), dtype=tf.string),
    'humanScore': tf.float32,
    'question': FeaturesDict({
        'choice_A': Text(shape=(), dtype=tf.string),
        'choice_B': Text(shape=(), dtype=tf.string),
        'choice_C': Text(shape=(), dtype=tf.string),
        'choice_D': Text(shape=(), dtype=tf.string),
        'stem': Text(shape=(), dtype=tf.string),
    }),
    'turkIdAnonymized': Text(shape=(), dtype=tf.string),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
answerKey ClassLabel tf.int64
clarity Tensor tf.float32
fact1 Text tf.string
humanScore Tensor tf.float32
question FeaturesDict
question/choice_A Text tf.string
question/choice_B Text tf.string
question/choice_C Text tf.string
question/choice_D Text tf.string
question/stem Text tf.string
turkIdAnonymized Text tf.string
  • Citation:
@article{mihaylov2018can,
  title={Can a suit of armor conduct electricity? a new dataset for open book question answering},
  author={Mihaylov, Todor and Clark, Peter and Khot, Tushar and Sabharwal, Ashish},
  journal={arXiv preprint arXiv:1809.02789},
  year={2018}
}