KaraKaraWitch/TvTroper-2025
收藏官方服务:
资源简介:
TvTroper-2025是一个来自tvtropes.org的约708,000个页面的清洗和刷新快照。每个页面都有两种格式:原始HTML和经过Markdown清洗的版本。该数据集适用于语言模型的预处理或微调,专注于叙事模式、主题和流行文化知识的掌握。数据集包含 spoilers、非正式散文、内部笑话以及偶尔的NSFW语言。
TvTroper-2025 is an updated snapshot of ~708 k pages from tvtropes.org, available in both raw HTML and a Markdown-cleaned version. The dataset is intended for unsupervised pre-training or fine-tuning of language models focusing on narrative patterns, tropes, themes, and pop-culture knowledge. It includes spoilers, informal prose, in-jokes, and occasional NSFW language.
提供机构:
KaraKaraWitch


