open-index/arctic
收藏资源简介:
Arctic Shift Reddit Archive 是一个全面的Reddit存档数据集,包含自2005年12月至2026年2月的所有公开子版块的评论和提交内容。数据集以Parquet格式组织,分为评论和提交两部分,每月一个文件,大月份则分为多个约200 MB的分片。数据集总计包含11.1亿条数据(8.9亿条评论,2.2亿条提交),压缩后大小为1001.8 GB。该数据集适用于语言模型训练、情感分析、社区研究和信息检索等多种自然语言处理任务。数据集支持多种工具和库的直接使用,如DuckDB、`datasets`、`pandas`和`huggingface_hub`。
The Arctic Shift Reddit Archive is a comprehensive dataset containing all public subreddit comments and submissions from December 2005 through February 2026. Organized as monthly Parquet files, the dataset is split into comments and submissions, with larger months divided into multiple ~200 MB shards. In total, the dataset includes 11.1B items (8.9B comments, 2.2B submissions) in 1001.8 GB of compressed Parquet. It is ideal for various NLP tasks such as language model training, sentiment analysis, community research, and information retrieval. The dataset is designed for easy access and supports direct usage with tools like DuckDB, `datasets`, `pandas`, and `huggingface_hub`.




