Gigaword headline generation corpus
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/harvardnlp/sent-summary
下载链接
链接失效反馈官方服务：
资源简介：
该数据集旨在用于从新闻文章中生成语法连贯、语义正确的抽象性标题。为确保数据质量，研究人员手动筛选了高质量的新闻文章和标题。该数据集的规模分别为训练集500,000条、验证集30,000条以及测试集30,000条，其任务是新闻标题生成（NHG）。
This dataset is intended for generating grammatically coherent and semantically correct abstractive headlines from news articles. To guarantee data quality, researchers manually curated high-quality news articles and their corresponding headlines. The dataset comprises 500,000 training samples, 30,000 validation samples, and 30,000 test samples, with the targeted task being News Headline Generation (NHG).
提供机构：
Harvard NLP
