infini-news-corpus
收藏资源简介:
INFINI-NEWS语料库是一个从Common Crawl的CC-NEWS存档(2021-2025年)中提取的大规模多语言新闻数据集,专为时间序列NLP研究、新闻理解和大型语言模型训练而设计。该数据集包含826,718,662篇新闻文章,来自全球57,928个独特的新闻域名,涵盖136种语言。数据按年份组织,共35,051个JSONL文件,总字数达6450亿,总字符数9.4万亿。主要语言包括英语(35.2%)、西班牙语(9.5%)、俄语(7.1%)等。每篇文章包含URL、发布日期、下载日期、来源域名、标题、描述、正文、语言代码、国家顶级域名、作者列表和图片URL等字段。数据集适用于文本生成、文本分类和标记分类等NLP任务,仅供研究使用,需遵守CC-BY 4.0许可和原始出版商的版权规定。
The INFINI-NEWS Corpus is a large-scale multilingual news dataset extracted from the CC-NEWS archive (2021–2025) of Common Crawl, designed specifically for time-series NLP research, news understanding, and large language model (LLM) training. This dataset contains 826,718,662 news articles sourced from 57,928 unique news domains worldwide, covering 136 languages. Organized by year, it consists of 35,051 JSONL files, with a total word count of 645 billion and a total character count of 9.4 trillion. Its dominant languages include English (35.2%), Spanish (9.5%), Russian (7.1%), among others. Each article includes fields such as URL, publication date, download date, source domain, title, description, full text, language code, country code top-level domain (ccTLD), author list, and image URL. This dataset is applicable to NLP tasks including text generation, text classification and token classification, and is for research use only. Users must comply with the CC-BY 4.0 license and copyright regulations of the original publishers.




