cos_e

  • الوصف :

تسمح تفسيرات الحس المشترك (CoS-E) بتدريب النماذج اللغوية لتوليد التفسيرات تلقائيًا التي يمكن استخدامها أثناء التدريب والاستدلال في إطار عمل جديد للشرح المُنشأ تلقائيًا (CAGE) من Commonsense.

انشق، مزق أمثلة
'train' 9741
'validation' 1،221
  • هيكل الميزة :
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),
})
  • وثائق الميزة :
ميزة فصل شكل نوع وصف
الميزات
التفسير التجريدي نص سلسلة
إجابه نص سلسلة
اختيارات تسلسل (نص) (لا أحد،) سلسلة
استخلاصي_تفسير نص سلسلة
هوية شخصية نص سلسلة
سؤال نص سلسلة
  • الاقتباس :
@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"
}