考虑标签层级结构的专利文本分类算法研究

A Study of Patent Text Classification Algorithm Considering Tag Hierarchy Structure

针对海量的中文专利文本,为提高人工分类的效率,减少由分类人员主观知识和客观因素影响导致的错误分类,本研究提出一种融合标签层次结构信息的专利文本分类模型。以2017年中国专利申请数据为实验数据集,针对国际专利分类号的层次结构信息构建一个全局的层级多标签分类模型,并在专利文本表征中融入专利标签的层次结构信息。实验结果表明,在中文专利文本分类领域融入标签的层次结构信息有助于提升模型性能。

For the massive Chinese patent text,in order to improve the efficiency of manual classification and reduce the misclassification caused by the influence of subjective knowledge and objective factors of classifiers,this study proposes a patent text classification model incorporating labeled hierarchical structure information.Taking the 2017 Chinese patent application data as the experimental dataset,a global hierarchical multi-label classification model is constructed for the hierarchical structure information of international patent classification classification numbers,and the hierarchical structure information of patent labels is incorporated in the patent text representation.The experimental results show that incorporating the hierarchical structure information of labels in this area of Chinese patent text categorization helps to improve the performance of the model.