cos_e

  • Descrizione :

Common Sense Explanations (CoS-E) consente ai modelli linguistici di addestramento di generare automaticamente spiegazioni che possono essere utilizzate durante l'addestramento e l'inferenza in un nuovo framework CAGE (Commonsense Auto-Generated Explanation).

Diviso Esempi
'train' 9.741
'validation' 1.221
  • Struttura delle caratteristiche :
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),
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
spiegazione_astrattiva Testo corda
Rispondere Testo corda
scelte Sequenza(Testo) (Nessuno,) corda
estrattiva_spiegazione Testo corda
id Testo corda
domanda Testo corda
  • Citazione :
@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"
}