File size: 444 Bytes
a231cf6 c4083a9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
---
dataset_info:
features:
- name: sentence1
dtype: string
- name: sentence2
dtype: string
splits:
- name: train
num_bytes: 14498836027
num_examples: 6458670
download_size: 8956015300
dataset_size: 14498836027
---
# Dataset Card for "wiki-span"
This dataset is constructed by sampling 25%-50% of each wikipedia record twice, as positive pairs. It can be used to train unsupervised sentence representation models. |