cos_e

  • Descriptif :

Les explications de sens commun (CoS-E) permettent aux modèles de langage de formation de générer automatiquement des explications qui peuvent être utilisées pendant la formation et l'inférence dans un nouveau cadre d'explications auto-générées de sens commun (CAGE).

Diviser Exemples
'train' 9 741
'validation' 1 221
  • Structure des fonctionnalités :
FeaturesDict({
    'abstractive_explanation': Text(shape=(), dtype=string),
    'answer': Text(shape=(), dtype=string),
    'choices': Sequence(Text(shape=(), dtype=string)),
    'extractive_explanation': Text(shape=(), dtype=string),
    'id': Text(shape=(), dtype=string),
    'question': Text(shape=(), dtype=string),
})
  • Documentation des fonctionnalités :
Caractéristique Classer Forme Dtype La description
FonctionnalitésDict
explication_abstraite Texte chaîne de caractères
réponse Texte chaîne de caractères
les choix Séquence (texte) (Aucun,) chaîne de caractères
explication_extractive Texte chaîne de caractères
identifiant Texte chaîne de caractères
question Texte chaîne de caractères
  • Citation :
@inproceedings{rajani2019explain,
     title = "Explain Yourself! Leveraging Language models for Commonsense Reasoning",
    author = "Rajani, Nazneen Fatema  and
      McCann, Bryan  and
      Xiong, Caiming  and
      Socher, Richard",
      year="2019",
    booktitle = "Proceedings of the 2019 Conference of the Association for Computational Linguistics (ACL2019)",
    url ="https://arxiv.org/abs/1906.02361"
}