日本某企业磁材专利布局策略数据

通过对日本某磁性材料企业产品专利数据进行收集并利用LLM算法进行处理形成数据，数据包含产品分类、技术问题、技术效果等核心发明要点信息。这一专利数据是同类企业制定产品开发策略的重要参考资料。企业可以通过分析竞争对手的专利申请情况，了解其研发实力和战略意图。这有助于企业制定相应的竞争策略，如加强某项技术的研发、寻找差异化发展途径等。通过有效的专利数据，企业可以了解行业技术发展趋势、竞争对手的技术动向以及潜在的合作对象，这有助于企业及时调整研发方向，保持技术领先地位，实现技术创新和市场竞争优势的提升。（1）利用规则匹配算法，对专利说明书的内容进行大文本字段级解析，把说明书的内容拆分成标题摘要、技术领域、背景技术、发明内容、附图说明、具体实施方式等具体信息，储存于列表中；（2）运用RAG（检索增强生成）算法，把与需要提取的信息密切相关的专利信息作为上下文信息发送给基于神经网络计算的large language model，LLM模型，模型对专利涉及的发明要点等信息进行分析提取。标题摘要部分信息处理后得到产品信息，背景技术部分信息经处理得到市场需求信息，发明内容和具体实施方式部分信息处理得到技术问题和技术效果。（3）结合人工判断优化，利用大语言模型的能力，对专利的发明要点进行聚合和分类。（4）修正后的专利提取和分类结果数据，可继续用于模型的迭代训练和调优，提高模型对专利进行信息提取和分类的准确性。

This dataset is constructed by collecting product patent data from a Japanese magnetic materials enterprise and processing it with LLM algorithms. The dataset includes core invention key points such as product classification, technical problems, and technical effects.

This patent dataset serves as an important reference for peer enterprises when formulating product development strategies. Enterprises can analyze the patent filings of their competitors to gain insights into their R&D capabilities and strategic intentions. This helps enterprises formulate corresponding competitive strategies, such as intensifying R&D on specific technologies, seeking differentiated development paths, and so on. With high-quality patent data, enterprises can grasp industry technology development trends, competitors' technological developments, and potential cooperation partners. This enables enterprises to timely adjust their R&D directions, maintain technological leadership, and achieve improvements in technological innovation and market competitive advantages.

(1) Use rule-based matching algorithms to perform large-text field-level parsing on the content of patent specifications, splitting the specification content into specific sections including title and abstract, technical field, background art, summary of the invention, brief description of the drawings, and detailed description of the embodiments, then storing these sections in a list;
(2) Apply RAG (Retrieval-Augmented Generation) algorithms, taking patent information closely related to the target information to be extracted as context and sending it to the neural network-based Large Language Model (LLM), which analyzes and extracts key invention-related information from the patent. Processing the title and abstract section yields product information; processing the background art section yields market demand information; and processing the summary of the invention and detailed description of the embodiments sections yields technical problems and technical effects.
(3) Combine manual judgment for optimization, leveraging the capabilities of large language models to aggregate and classify the key invention points of the patents;
(4) The revised patent extraction and classification result data can be further used for iterative training and tuning of the model, improving the accuracy of the model in patent information extraction and classification.

搜集汇总

数据集介绍

特点

该数据集包含558条日本某磁性材料企业的专利数据，涵盖产品分类、技术问题、技术效果等核心发明要点信息，每季度更新。数据可用于分析竞争对手的研发动向和制定产品开发策略。