jealk/dk_retrieval_benchmark

---
language:
- da
size_categories:
- 10K<n<100K
pretty_name: Retsinformation DK Retrieval Benchmark
dataset_info:
- config_name: generated_questions
features:
- name: title_vejledning
dtype: string
- name: chunk_text
dtype: string
- name: url
dtype: string
- name: generated_question
dtype: string
splits:
- name: train
num_bytes: 263556
num_examples: 200
download_size: 48578
dataset_size: 263556
- config_name: retsinformation
features:
- name: url
dtype: string
- name: title
dtype: string
- name: html_content
dtype: string
- name: text_content
dtype: string
splits:
- name: train
num_bytes: 62646653
num_examples: 433
download_size: 20333540
dataset_size: 62646653
configs:
- config_name: generated_questions
data_files:
- split: train
path: generated_questions/train-*
- config_name: retsinformation
data_files:
- split: train
path: retsinformation/train-*
---

# Retsinformation retrieval benchmark

Datasets related to generating a Q & Chunk dataset based on guides (vejledninger) from retsinformation.dk to be used as a retrieval benchmark.

vejledninger_tekst.csv contains a dict with all vejledninger (scraped 8/11/23) from retsinformation.dk

chunks_id_text.csv contains text chunks of max 512 token len, based on splitting all the text from vejledninger_tekst.csv, along with a unique id

chunks_questions_100_samples.csv contains a sample of 200 auto-generated questions, based on the first 100 text chunks from the chunks_id_text.csv file, along with the matching text chunk.

原始信息汇总

Retsinformation DK Retrieval Benchmark 数据集概述

数据集信息

配置 generated_questions

• 特征:
  • title_vejledning: 字符串类型
  • chunk_text: 字符串类型
  • url: 字符串类型
  • generated_question: 字符串类型
• 分割:
  • train:
    • 字节数: 263556
    • 样本数: 200
• 下载大小: 48578 字节
• 数据集大小: 263556 字节

配置 retsinformation

• 特征:
  • url: 字符串类型
  • title: 字符串类型
  • html_content: 字符串类型
  • text_content: 字符串类型
• 分割:
  • train:
    • 字节数: 62646653
    • 样本数: 433
• 下载大小: 20333540 字节
• 数据集大小: 62646653 字节

配置信息

• 配置 generated_questions:
  • 数据文件路径: generated_questions/train-*
• 配置 retsinformation:
  • 数据文件路径: retsinformation/train-*