SUMMZOO
收藏arXiv2023-05-28 更新2024-06-21 收录
下载链接：
https://github.com/microsoft/UniSumm
下载链接
链接失效反馈官方服务：
资源简介：
SUMMZOO是一个综合性的少样本摘要基准数据集，由浙江大学和微软研究院合作创建。该数据集包含8个摘要任务，覆盖新闻、学术论文、会议等多种领域，旨在通过多样性和鲁棒性原则，为少样本摘要模型的评估提供一个公平且有效的平台。数据集中的每个任务都提供了多个少样本训练集，以增强评估的稳健性。
SUMMZOO is a comprehensive few-shot summarization benchmark dataset co-created by Zhejiang University and Microsoft Research. It consists of 8 summarization tasks spanning multiple domains including news, academic papers, conference proceedings, etc. It aims to provide a fair and effective platform for evaluating few-shot summarization models based on the principles of diversity and robustness. Each task in the dataset provides multiple few-shot training sets to enhance the robustness of the evaluation.
提供机构：
浙江大学
创建时间：
2022-11-18
