SamuelMauli/parity-juridico-dataset-v3

该数据集是一个巴西法律领域的纯文本数据集，包含100%的法律领域文本，没有通用NLI噪声或非法律句子。数据集包括195个三元组、445个对和957个原始法律文本。与v2版本相比，移除了非法律内容，增加了严格的法律正则表达式过滤，并添加了来自公共API和TCU的数据集。数据集来源包括parity-v1、pierreguillou-lener-lm、legal-bench-br和parity-editais。数据集主要用于文本检索和句子相似性任务，适用于葡萄牙语（巴西）的法律、采购、判例等领域。

This dataset is a pure Brazilian legal domain text dataset, containing 100% legal domain text without generic NLI noise or non-legal sentences. The dataset includes 195 triplets, 445 pairs, and 957 raw legal texts. Compared to v2, non-legal content has been removed, strict legal regex filtering has been added, and datasets from public APIs and TCU have been included. Data sources include parity-v1, pierreguillou-lener-lm, legal-bench-br, and parity-editais. The dataset is primarily used for text retrieval and sentence similarity tasks, applicable to Portuguese (Brazil) in legal, procurement, jurisprudence, and other fields.