DIVERSESUMM
收藏资源简介:
DIVERSESUMM数据集由Salesforce AI Research创建,旨在支持从多篇新闻文章中总结多样化信息的新任务。该数据集包含245个新闻故事,每个故事由10篇文章组成,并附有人工验证的参考。数据集通过一个基于GPT-3.5-Turbo的管道进行自动多样化信息发现和人工验证阶段,确保数据完整性。DIVERSESUMM特别关注于捕捉和总结新闻报道中的不同观点和信息,为理解和总结同一新闻事件的复杂性提供了挑战和机会。
The DIVERSESUMM dataset was developed by Salesforce AI Research to support a novel task of summarizing diversified information from multiple news articles. This dataset contains 245 news stories, each consisting of 10 articles paired with human-validated reference summaries. The dataset utilizes a GPT-3.5-Turbo-based pipeline for automated diversified information discovery and manual verification phases to ensure data integrity. DIVERSESUMM specifically focuses on capturing and summarizing diverse perspectives and information within news coverage, providing both challenges and opportunities for understanding and summarizing the complexity of the same news event.




