---
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.