경주

  • 설명 :

Race는 28,000개 이상의 구절과 거의 100,000개의 질문이 있는 대규모 읽기 이해 데이터 세트입니다. 데이터 세트는 중학생 및 고등학생을 위해 설계된 중국의 영어 시험에서 수집되었습니다. 데이터 세트는 기계 이해를 위한 훈련 및 테스트 세트로 제공될 수 있습니다.

FeaturesDict({
    'answers': Sequence(Text(shape=(), dtype=string)),
    'article': Text(shape=(), dtype=string),
    'example_id': Text(shape=(), dtype=string),
    'options': Sequence(Sequence(Text(shape=(), dtype=string))),
    'questions': Sequence(Text(shape=(), dtype=string)),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
답변 시퀀스(텍스트) (없음,)
기사 텍스트
example_id 텍스트
옵션 시퀀스(시퀀스(텍스트)) (없음, 없음)
질문 시퀀스(텍스트) (없음,)
@article{lai2017large,
    title={RACE: Large-scale ReAding Comprehension Dataset From Examinations},
    author={Lai, Guokun and Xie, Qizhe and Liu, Hanxiao and Yang, Yiming and Hovy, Eduard},
    journal={arXiv preprint arXiv:1704.04683},
    year={2017}
}

레이스/높음(기본 구성)

  • 데이터 세트 크기 : 52.39 MiB

  • 분할 :

나뉘다
'dev' 1,021
'test' 1,045
'train' 18,728

인종/중간

  • 데이터 세트 크기 : 12.51 MiB

  • 분할 :

나뉘다
'dev' 368
'test' 362
'train' 6,409