• Description:

L'ensemble de données sur la justification du film contient des justifications humaines annotées pour les critiques de films.

Diviser Exemples
'test' 199
'train' 1 600
'validation' 200
  • Caractéristiques:
    'evidences': Sequence(Text(shape=(), dtype=tf.string)),
    'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=2),
    'review': Text(shape=(), dtype=tf.string),
  • citation:
    title = {ERASER: A Benchmark to Evaluate Rationalized NLP Models},
    author = {Jay DeYoung and Sarthak Jain and Nazneen Fatema Rajani and Eric Lehman and Caiming Xiong and Richard Socher and Byron C. Wallace}
  author    =  {Omar F. Zaidan  and  Jason Eisner  and  Christine Piatko},
  title     =  {Machine Learning with Annotator Rationales to Reduce Annotation Cost},
  booktitle =  {Proceedings of the NIPS*2008 Workshop on Cost Sensitive Learning},
  month     =  {December},
  year      =  {2008}