You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Chinese Patent Dataset (2003-2020) | 中国专利数据集

1. 数据集简介 (Introduction)

本数据集包含了从 2003年到2020年 期间发布的约 1800万份 中国专利数据。数据涵盖了发明专利、实用新型等多种类型,是研究中国技术演进、知识产权保护以及进行法律大模型训练的高质量语料。

This dataset contains approximately 18 million Chinese patent records published between 2003 and 2020. It is a high-quality corpus for studying technology evolution in China and training legal/LLM models.

2. 数据规模 (Dataset Summary)

  • 时间跨度: 2003 - 2020
  • 数据总量: 约 1,800 万条记录
  • 原始格式: JSON
  • 总体大小: 约 250 GB (未压缩)

3. 数据字段 (Data Fields)

每个 JSON 条目通常包含以下字段(请根据实际情况修改):

  • patent_id: 专利申请号/公开号
  • title: 专利标题
  • abstract: 摘要
  • claims: 首项权利要求
  • classification: IPC分类号
  • assignee: 专利权人
  • app_date: 申请日期

4. 使用场景 (Usage)

  • 大模型预训练 (Pretraining): 作为科技与法律领域的垂直语料。
  • 专利检索与推荐: 训练向量索引模型进行相似专利检索。
  • 技术趋势分析: 挖掘近 20 年中国技术热点的迁移。

5. 局限性 (Limitations)

  • 数据仅截至 2020 年,不包含近年的最新专利。
  • 文本内容来源于原始数据转换,建议使用前进行简单的清洗。

6. 许可协议 (License)

本项目采用 Apache License 2.0 协议开源。

Downloads last month
962