Generated Knowledge Work Documents

该数据集由25封电子邮件和会议记录组成，这些内容既包括真实案例，也包括通过一个可配置的多代理知识工作数据集生成器产生的文档。数据集涵盖了不同类型的文档（电子邮件和会议记录），并根据人类评分者在7点李克特量表上对文档真实性的评估来进行评价。规模上，该数据集包含了25份文档（分为5个类别，每个类别包含1份真实文档和4份生成文档）。任务方面，该数据集旨在进行文档生成和真实性评估。

This dataset consists of 25 emails and meeting transcripts, which include both real-world cases and documents generated by a configurable multi-agent knowledge work dataset generator. The dataset covers two types of documents: emails and meeting transcripts, and is evaluated based on human raters' assessments of document authenticity on a 7-point Likert scale. In terms of scale, the 25 documents are divided into 5 categories, with each category containing 1 real document and 4 generated documents. The dataset is designed for two core tasks: document generation and authenticity evaluation.