RA-MDS

RA-MDS数据集是由香港中文大学和腾讯AI实验室合作创建，专注于读者感知的多文档摘要问题。该数据集包含45个主题，每个主题涉及10篇新闻文档和平均215条读者评论，总计约19,000个标注的方面。数据集的创建过程涉及专家进行数据收集、方面标注和摘要撰写。RA-MDS数据集旨在通过结合新闻文档和读者评论，提高多文档摘要的质量，特别关注于捕捉读者兴趣和新闻内容的关键信息，以生成更全面和准确的摘要。The RA-MDS dataset was co-developed by The Chinese University of Hong Kong and Tencent AI Lab, focusing on reader-perceived multi-document summarization. It includes 45 topics, each paired with 10 news documents and an average of 215 reader comments, resulting in a total of approximately 19,000 annotated aspects. The construction of this dataset involved experts conducting data collection, aspect annotation and summary writing. The RA-MDS dataset aims to enhance the quality of multi-document summarization by integrating news documents and reader comments, with particular emphasis on capturing reader interests and critical information from news content to generate more comprehensive and accurate summaries.