SamuelMauli/parity-juridico-dataset
收藏Hugging Face2026-04-28 更新2026-05-03 收录
下载链接:
https://hf-mirror.com/datasets/SamuelMauli/parity-juridico-dataset
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是一个用于巴西法律领域语义检索微调的数据集,内容涉及公共招标、TCU判例和Lei 14.133/21法律。数据集由Parity平台的知识库精心整理,包含训练集、评估集和交叉编码器用的配对数据。数据集涵盖了12个法律类别,如经济资格、技术资格、财政资格等。挖掘策略包括自我正例、类别正例和硬负例,以增强模型的微调学习效果。数据集还提供了一个基于此训练的模型,该模型在评估集上的三元组准确率从57.14%提升到了85.71%。
This dataset is a fine-tuning dataset for semantic retrieval in the Brazilian legal domain, covering public procurement, TCU jurisprudence, and Law 14.133/21. It is curated from the knowledge base of the Parity platform (parity.doublethree.com.br) and includes training sets, evaluation sets, and paired data for cross-encoders. The dataset encompasses 12 legal categories such as economic qualification, technical qualification, fiscal qualification, etc. The mining strategy includes self-positives, category-positives, and hard negatives to enhance fine-tuning learning. The dataset also provides a model trained on it, which improved triplet accuracy from 57.14% to 85.71% on the evaluation set.
提供机构:
SamuelMauli



