变压器专利分类数据

变压器专利分类数据，包括变压器的专利信息，根据人工和算法规则对变压器的专利信息进行标引、分类，建立多级技术分支分类专题数据库，有助于领域内的创新主体通过工程技术用语快速检索该技术领域内的专利信息。基于分类检索信息，企业可结合专利具体信息进一步分析变压器专利技术的发展趋势，例如按照年份申请量分析领域发展情况等，引导企业自身研发方向，避免重复研发及专利纠纷产生。1、数据来源：专利检索人员在企业自建数据库中根据关键词、申请人、分类号等条件构建检索式进行检索，得到专利数据。
2、数据清洗：针对检索得到专利数据，对专利的著录项目、标题、说摘要、权利要求书、说明书、说明书附图内容进行摘取与清洗。
3、数据标引：
3.1 技术分支拆解：专利标引人员对专利数据进行技术分支分解，分为多级技术分支。
3.2 技术分支扩展：专利标引人员针对各级技术分支关键词进行概念含义扩展，扩展方向包括：中英文上位、下位、同义词、近义词概念等。
3.3技术分支标引：专利标引人员对专利信息，通过Word2Vec算法（CBOW类型）将专利内容处理为专利词向量数据；在专利内容词向量数据的基础上，利用多类别支持向量机(Multi-class SVM)进行训练，训练建立各级技术分支与各级技术分支关键词以及扩展关键词相对应的专利词向量数据之间映射关系，并基于已训练的多类别支持向量机(Multi-class SVM)模型对单件专利进行技术分支标引，标引其技术分支分类。

This is a transformer patent classification dataset that collects patent information related to transformers. Patent information is indexed and classified based on manual and algorithmic rules, and a thematic database of multi-level technical branch classifications is established. This resource allows innovators in the field to quickly retrieve patent documents within this technical domain using standardized engineering technical terminology. Based on the classified retrieval results, enterprises can further analyze the development trends of transformer patent technologies by leveraging specific patent details—for instance, evaluating the field's development trajectory via annual application counts—to guide their internal R&D directions and prevent redundant research and patent disputes.

1. Data Source: Patent search professionals construct retrieval queries using criteria including keywords, applicants, and classification numbers in the enterprise's self-built database, to obtain targeted patent data.
2. Data Cleaning: For the retrieved patent datasets, core fields including bibliographic items, title, abstract, claims, specification, and contents of the specification drawings are extracted and cleaned.
3. Data Indexing:
3.1 Technical Branch Decomposition: Patent indexers decompose the collected patent data into multi-level technical branches.
3.2 Technical Branch Expansion: Patent indexers perform conceptual expansion on the keywords of each level of technical branches, covering Chinese and English superordinate, subordinate, synonym, and near-synonym concepts.
3.3 Technical Branch Indexing: For the patent information, patent indexers first convert the patent content into patent word vector data using the Word2Vec algorithm (CBOW type). Subsequently, a Multi-class Support Vector Machine (Multi-class SVM) model is trained on the patent word vector data to establish the mapping between each level of technical branches and the patent word vectors corresponding to technical branch keywords and their expanded variants. Finally, the pre-trained Multi-class SVM model is used to conduct technical branch indexing for individual patents, assigning their corresponding technical branch classification tags.

搜集汇总

背景与挑战

背景概述

该数据集包含7832条变压器专利信息，采用xlsx格式，每年更新一次。数据集通过人工和算法规则对专利信息进行标引和分类，建立多级技术分支分类专题数据库，主要用于帮助创新主体快速检索专利信息，分析技术发展趋势，引导研发方向。