metadata
dataset_info:
features:
- name: condition
dtype: string
- name: group
sequence: string
- name: __index_level_0__
dtype: int64
splits:
- name: train
num_bytes: 736350
num_examples: 2759
- name: validation
num_bytes: 84745
num_examples: 345
- name: test
num_bytes: 92647
num_examples: 345
download_size: 503833
dataset_size: 913742
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
介绍
quora的这个检索数据集并不怎么好, query和corpus都时问题,而且没有label为0的qerl 然后才一共出产了3449条数据, 有点少了. 这不是我理想的数据集.
一共囊括了11750句语料,3449条query,平均1个query对应3.4个corpus,组合的话,能有14000条数据.