NomaDamas/split_search_qa

---
license: unknown
dataset_info:
- config_name: corpus
features:
- name: query_id
dtype: string
- name: snippets
dtype: string
- name: air_date
dtype: string
- name: category
dtype: string
- name: value
dtype: string
- name: round
dtype: string
- name: show_number
dtype: int32
- name: doc_id
dtype: string
- name: __index_level_0__
dtype: int64
splits:
- name: train
num_bytes: 6252715344
num_examples: 14120776
download_size: 3271155810
dataset_size: 6252715344
- config_name: qa_data
features:
- name: query_id
dtype: string
- name: question
dtype: string
- name: answer
dtype: string
- name: search_results
struct:
- name: related_links
sequence: string
- name: snippets
sequence: string
- name: titles
sequence: string
- name: urls
sequence: string
- name: doc_id
sequence: string
- name: __index_level_0__
dtype: int64
splits:
- name: train
num_bytes: 6503932619
num_examples: 173397
- name: test
num_bytes: 1830028629
num_examples: 43350
download_size: 5008413626
dataset_size: 8333961248
configs:
- config_name: corpus
data_files:
- split: train
path: corpus/train-*
- config_name: qa_data
data_files:
- split: train
path: qa_data/train-*
- split: test
path: qa_data/test-*
---
# preprocessed_SearchQA

The SearchQA question-answer pairs originate from J! Archive2, which comprehensively archives all question-answer pairs
from the renowned television show Jeopardy! The passages, sourced from Google search web page snippets.
We offer passage metadata, encompassing details like 'air_date,' 'category,' 'value,' 'round,' and 'show_number,'
enabling you to enhance retrieval performance at your discretion.
Should you require further details about SearchQA, please refer to below links.

[Github](https://github.com/nyu-dl/dl4ir-searchQA)<br>
[Paper](https://arxiv.org/abs/1704.05179)<br>

The dataset is derived from [searhQA](https://huggingface.co/datasets/search_qa).<br>
This preprocessed dataset is for RAG. For more information about our task, visit our [repository](https://github.com/NomaDamas/RAGchain)!<br>

Preprocess SearchQA dataset code for RAG benchmark. <br>
More information, refer to this link! [huggingface](https://huggingface.co/datasets/NomaDamas/search_qa_split)

原始信息汇总

数据集概述

数据集配置

• corpus

  • 特征
    • query_id: 字符串
    • snippets: 字符串
    • air_date: 字符串
    • category: 字符串
    • value: 字符串
    • round: 字符串
    • show_number: 整数 (int32)
    • doc_id: 字符串
    • __index_level_0__: 整数 (int64)
  • 分割
    • train: 字节数 6252715344, 样本数 14120776
  • 下载大小: 3271155810 字节
  • 数据集大小: 6252715344 字节

• qa_data

  • 特征
    • query_id: 字符串
    • question: 字符串
    • answer: 字符串
    • search_results: 结构体
      • related_links: 字符串序列
      • snippets: 字符串序列
      • titles: 字符串序列
      • urls: 字符串序列
    • doc_id: 字符串序列
    • __index_level_0__: 整数 (int64)
  • 分割
    • train: 字节数 6503932619, 样本数 173397
    • test: 字节数 1830028629, 样本数 43350
  • 下载大小: 5008413626 字节
  • 数据集大小: 8333961248 字节

数据文件

• corpus

  • train: corpus/train-*

• qa_data

  • train: qa_data/train-*
  • test: qa_data/test-*