DomainSum
收藏资源简介:
DomainSum是一个用于抽象文本摘要中细粒度领域转移的分层基准数据集,由夏威夷大学马诺阿分校创建。该数据集包含五个不同领域的数据,每个领域在体裁、风格和主题三个层次上进行了细分,总数据量超过110万条。数据集的创建过程结合了高质量的公开数据集,并通过系统化的分析方法对不同层次的领域转移进行了详细研究。DomainSum主要应用于自然语言处理中的文本摘要任务,旨在解决模型在不同领域间迁移时的性能问题。
DomainSum is a hierarchical benchmark dataset for fine-grained domain transfer in abstractive text summarization, developed by the University of Hawaii at Manoa. It contains data from five distinct domains, with each domain subdivided along three dimensions: genre, style, and topic, and the total number of instances exceeds 1.1 million. The dataset was constructed by incorporating high-quality publicly available datasets, and systematic analytical approaches were employed to carry out in-depth investigations into domain transfers at different hierarchical levels. DomainSum is primarily applied to text summarization tasks in natural language processing, with the objective of mitigating performance degradation when models transfer across diverse domains.
DomainSum
数据集概述
- 名称: DomainSum
数据集详情
- 描述: 该数据集的README文件内容为空,无法提供更多详细信息。




