math_qa

  • Descriptif :

Un ensemble de données à grande échelle de problèmes de mots mathématiques et un résolveur de problèmes mathématiques neuronaux interprétable qui apprend à mapper les problèmes aux programmes d'exploitation.

Diviser Exemples
'test' 2 985
'train' 29 837
'validation' 4 475
  • Structure des fonctionnalités :
FeaturesDict({
   
'Problem': Text(shape=(), dtype=string),
   
'Rationale': Text(shape=(), dtype=string),
   
'annotated_formula': Text(shape=(), dtype=string),
   
'category': Text(shape=(), dtype=string),
   
'correct': Text(shape=(), dtype=string),
   
'correct_option': Text(shape=(), dtype=string),
   
'linear_formula': Text(shape=(), dtype=string),
   
'options': Text(shape=(), dtype=string),
})
  • Documentation des fonctionnalités :
Fonctionnalité Classe Forme Dtype Description
FonctionnalitésDict
Problème Texte chaîne
Raisonnement Texte chaîne
formule_annotée Texte chaîne
catégorie Texte chaîne
correct Texte chaîne
option_correcte Texte chaîne
formule_linéaire Texte chaîne
options Texte chaîne
  • Citation :
@misc{amini2019mathqa,
      title
={MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms},
      author
={Aida Amini and Saadia Gabriel and Peter Lin and Rik Koncel-Kedziorski and Yejin Choi and Hannaneh Hajishirzi},
      year
={2019},
      eprint
={1905.13319},
      archivePrefix
={arXiv},
      primaryClass
={cs.CL}
}