cos_e

  • תיאור :

Common Sense Explanations (CoS-E) מאפשר לאמן מודלים של שפות ליצור אוטומטית הסברים שניתן להשתמש בהם במהלך אימון והסקת מסקנות במסגרת חדשה של Commonsense Auto-Generated Explanation (CAGE).

לְפַצֵל דוגמאות
'train' 9,741
'validation' 1,221
  • מבנה תכונה :
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),
})
  • תיעוד תכונה :
תכונה מעמד צוּרָה Dtype תיאור
FeaturesDict
הסבר_מופשט טֶקסט חוּט
תשובה טֶקסט חוּט
בחירות רצף (טקסט) (אף אחד,) חוּט
מיצוי_הסבר טֶקסט חוּט
תְעוּדַת זֶהוּת טֶקסט חוּט
שְׁאֵלָה טֶקסט חוּט
  • ציטוט :
@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"
}