融合结构数据和语义的专利技术主题识别研究——以非小细胞肺癌治疗领域为例

Patented Technology Topic Recognition Fusing Structural Data and Semantics——take the field of non-small cell lung cancer as an example

摘要:

针对专利技术主题识别方法存在缺少语义语境、可解释性弱和主题界定模糊等问题,提出了一种融合专利结构数据和文本语义的技术主题识别分析方法一并解决上述问题,有助于领域人员把握技术研究内容,为研发决策提供科学支持。方法以专利IPC作为结构数据改进纯文本主题建模获取由IPC和专家分类意见指导的主题词向量,同时使用word2vec获取语义词向量,将结果进行向量拼接,进而获得易于解释的精准技术主题,满足细粒度分析要求。最后,以非小细胞肺癌治疗领域作为实证研究,证实了该方法的科学性和实用性。

Abstract:

Aiming at the problems of lack of semantic context, weak interpretability and vague topic definition in the patent technology subject identification method, this study proposes a technology subject identification and analysis method that integrates patent structure data and text semantics to solve the above problems, which will help to solve the above problems. As researchers grasp the technical research content, provide scientific support for R D decision-making. In this paper, patent IPC is used as structural data to improve the topic modeling of plain text to obtain topic word vectors guided by IPC and expert classification opinions. At the same time, word2vec is used to obtain semantic word vectors, and the two results are vector spliced to obtain accurate technical topics that are easy to explain. , to meet the fine-grained analysis requirements. Finally, taking the field of non-small cell lung cancer treatment as an empirical study, the scientificity and practicality of the method are confirmed.