histlearn/notas-comunidade-ptbr

一个经过策划、丰富的葡萄牙语X（原Twitter）社区笔记子集，包含主题建模、命名实体识别，并复现了官方矩阵分解评分算法。数据集基于2026-04-07的官方Birdwatch公开快照，覆盖时间从2021-10-13到2026-04-05，包含142,448条葡萄牙语笔记（通过fastText语言检测，置信度≥0.80，平均0.98），涉及90,822条独特推文，536,373个实体（涵盖22种命名实体类型），以及899个细粒度主题（组织成38个宏观主题）。数据集适用于文本分类、标记分类等任务，专注于社区笔记、错误信息、巴西政治等领域。A curated and extensive subset of Portuguese X (formerly Twitter) Community Notes, which incorporates topic modeling, named entity recognition, and reproduces the official matrix factorization rating algorithm. This dataset is derived from the official Birdwatch public snapshot dated 2026-04-07, spanning the time frame from October 13, 2021 to April 5, 2026. It includes 142,448 Portuguese-language notes (detected via fastText with confidence scores ≥0.80, with an average confidence of 0.98), covering 90,822 unique Tweets, 536,373 entities across 22 named entity types, and 899 fine-grained topics grouped into 38 macro topics. This dataset is suitable for tasks including text classification and label classification, with a focus on domains such as Community Notes, misinformation, and Brazilian politics.