FAME
收藏资源简介:
FAME数据集是由德国哥廷根大学的研究团队通过MIMIC框架生成的,包含800个会议记录,其中英语500个,德语300个。这些会议记录覆盖了14种不同类型的会议,如项目更新、头脑风暴等,并以300篇维基百科文章作为知识来源。数据集通过模拟多智能体对话,生成具有实际知识基础和参与者个人特征的会议记录,旨在为会议总结研究提供一个新的、可扩展的数据代理。
The FAME dataset was generated by the research team from the University of Göttingen, Germany, using the MIMIC framework. It contains 800 meeting transcripts, among which 500 are in English and 300 in German. These transcripts cover 14 distinct types of meetings, such as project updates and brainstorming sessions, and utilize 300 Wikipedia articles as knowledge sources. The dataset is constructed by simulating multi-agent dialogues to generate meeting transcripts with practical knowledge bases and individual characteristics of participants, aiming to provide a novel and scalable data resource for meeting summarization research.




