jhu-clsp/news21-instructions

---
language:
- en
multilinguality:
- monolingual
task_categories:
- text-retrieval
source_datasets:
- trec-news-2021
task_ids:
- document-retrieval
config_names:
- corpus
tags:
- text-retrieval
- instruction-retrieval
dataset_info:
- config_name: top_ranked
features:
- name: qid
dtype: string
- name: pid
dtype: string
splits:
- name: top_ranked
num_examples: 32000
- config_name: qrels_og
features:
- name: query-id
dtype: string
- name: corpus-id
dtype: string
- name: score
dtype: float64
splits:
- name: test
num_examples: 4277
- config_name: qrels_changed
features:
- name: query-id
dtype: string
- name: corpus-id
dtype: string
- name: score
dtype: float64
splits:
- name: test
num_examples: 4277
- config_name: corpus
features:
- name: _id
dtype: string
- name: title
dtype: string
- name: text
dtype: string
splits:
- name: corpus
num_examples: 30921
- config_name: queries
features:
- name: _id
dtype: string
- name: text
dtype: string
- name: instruction_og
dtype: string
- name: instruction_changed
dtype: string
- name: keywords
dtype: string
- name: short_query
dtype: string
splits:
- name: queries
num_examples: 32
configs:
- config_name: qrels_og
data_files:
- split: test
path: qrels_og/test.jsonl
- config_name: qrels_changed
data_files:
- split: test
path: qrels_changed/test.jsonl
- config_name: corpus
data_files:
- split: corpus
path: corpus.jsonl
- config_name: queries
data_files:
- split: queries
path: queries.jsonl
- config_name: top_ranked
data_files:
- split: top_ranked
path: top_ranked.jsonl
license: mit
---

原始信息汇总

数据集概述

基本信息

• 语言: 英语
• 多语言性: 单语
• 任务类别: 文本检索
• 源数据集: TREC News 2021
• 任务ID: 文档检索
• 配置名称: corpus
• 标签: 文本检索, 指令检索

数据集配置详情

配置: top_ranked

• 特征:
  • qid: 字符串
  • pid: 字符串
• 分割:
  • top_ranked: 32000个样本

配置: qrels_og

• 特征:
  • query-id: 字符串
  • corpus-id: 字符串
  • score: 浮点数
• 分割:
  • test: 4277个样本

配置: qrels_changed

• 特征:
  • query-id: 字符串
  • corpus-id: 字符串
  • score: 浮点数
• 分割:
  • test: 4277个样本

配置: corpus

• 特征:
  • _id: 字符串
  • title: 字符串
  • text: 字符串
• 分割:
  • corpus: 30921个样本

配置: queries

• 特征:
  • _id: 字符串
  • text: 字符串
  • instruction_og: 字符串
  • instruction_changed: 字符串
  • keywords: 字符串
  • short_query: 字符串
• 分割:
  • queries: 32个样本