政务公开数据集

为构建政务大语言模型和领域知识库,我们从各官方信息发布网站中收集并整理了一个包含文档和问答对数据的综合性政务公开信息数据集.数据集中的大部分数据源自各政府门户网站及多个政务信息公开平台,部分问答对数据由ChatGPT 3.5生成,并经人工筛选精炼得到.

To construct government affairs large language models and domain knowledge bases, we collected and curated a comprehensive government public information dataset containing documents and question-answer pair data from various official information release websites. Most of the data in the dataset originates from various government portals and multiple government information disclosure platforms, while some of the question-answer pair data was generated by ChatGPT 3.5 and refined through manual screening.

搜集汇总

背景与挑战

背景概述

政务公开数据集是一个综合性政务信息数据集，包含1900篇政务文档和10503条问答对，用于构建政务大语言模型。问答对分为官方问答和口语化问答两类，分别覆盖政务服务咨询的核心内容和市民政务咨询场景。