- Açıklama :
QASC, cümle kompozisyonuna odaklanan, soru cevaplayan bir veri setidir. İlkokul bilimi ile ilgili 9,980 8 yollu çoktan seçmeli sorudan oluşur (8,134 tren, 926 dev, 920 test) ve 17 milyon cümlelik bir külliyatla birlikte gelir.
Ana sayfa : https://allenai.org/data/qasc
Kaynak kodu :
tfds.question_answering.qasc.Qasc
Sürümler :
-
0.1.0
(varsayılan): Sürüm notu yok.
-
İndirme boyutu :
1.54 MiB
Veri kümesi boyutu :
6.61 MiB
Otomatik önbelleğe alınmış ( belgeler ): Evet
Bölmeler :
Bölmek | Örnekler |
---|---|
'test' | 920 |
'train' | 8.134 |
'validation' | 926 |
- Özellikler :
FeaturesDict({
'answerKey': Text(shape=(), dtype=tf.string),
'choices': Sequence({
'label': Text(shape=(), dtype=tf.string),
'text': Text(shape=(), dtype=tf.string),
}),
'combinedfact': Text(shape=(), dtype=tf.string),
'fact1': Text(shape=(), dtype=tf.string),
'fact2': Text(shape=(), dtype=tf.string),
'formatted_question': Text(shape=(), dtype=tf.string),
'id': Text(shape=(), dtype=tf.string),
'question': Text(shape=(), dtype=tf.string),
})
Denetimli anahtarlar (bkz
as_supervised
doc ):None
Şekil ( tfds.show_examples ): Desteklenmez.
Örnekler ( tfds.as_dataframe ):
- alıntı :
@article{allenai:qasc,
author = {Tushar Khot and Peter Clark and Michal Guerquin and Peter Jansen and Ashish Sabharwal},
title = {QASC: A Dataset for Question Answering via Sentence Composition},
journal = {arXiv:1910.11473v2},
year = {2020},
}