图书情报工作 ›› 2018, Vol. 62 ›› Issue (10): 94-105.DOI: 10.13266/j.issn.0252-3116.2018.10.013

• 知识组织 • 上一篇    下一篇

基于非负矩阵分解的技术主题演化分析

王园园, 赵亚娟   

  1. 中国科学院文献情报中心 北京 100190 中国科学院大学经济管理学院 北京 100190
  • 收稿日期:2017-10-30 修回日期:2018-01-22 出版日期:2018-05-20 发布日期:2018-05-20
  • 通讯作者: 赵亚娟(ORCID:0000-0003-3501-8131),研究员,博士,硕士生导师,通讯作者,E-mail:zhaoyj@mail.las.ac.cn
  • 作者简介:王园园(ORCID:0000-0003-1079-0766),硕士研究生。

Evolution Analysis of Technological Topic: An Approach Based on Non-negative Matrix Factorization

Wang Yuanyuan, Zhao Yajuan   

  1. National Science Library, Chinese Academy of Sciences, Beijing 100190 School of Economics and Management, University of Chinese Academy of Sciences, Beijing 100190
  • Received:2017-10-30 Revised:2018-01-22 Online:2018-05-20 Published:2018-05-20

摘要: [目的/意义]分析技术主题演化过程可以梳理技术发展脉络,对于发展创新、预测技术发展趋势具有重要意义,但是从语义角度分析技术主题演化轨迹的研究较少。因此,从语义的角度出发,分析技术主题演化过程。[方法/过程]提出基于非负矩阵分解的改进的动态非负矩阵分解模型对专利文本进行动态主题建模,并利用TextRank算法抽取名词短语进行标注,增强所抽取技术主题的可解释性。在此基础上,利用词向量的方式计算技术演化轨迹,并进行可视化展示。[结果/结论]对2002年、2005年、2008年、2011年和2014年的五方专利进行实证分析,识别出65个技术主题及其演化轨迹,表明方法的可行性。

关键词: 技术主题演化, 非负矩阵分解, 主题模型, 动态主题分析

Abstract: [Purpose/significance] Analyzing the evolution of technological topic makes it possible for us to track the development of technology, which is essential for improving innovation activity and forecasting development trends of technology. However, to our knowledge, scholars pay less attention to the semantic perspective of technological topic. Therefore, this paper intends to analyze the evolution of technological topic from the perspective of semantic.[Method/process] This paper proposed a dynamic topic model based on non-negative matrix factorization, and labeled the technology topics with noun phrases extracted by TextRank algorithm, which enhances the interpretability. Then, the study computed and visualized the evolutionary trajectory of technological topics with word embedding.[Result/conclusion] This paper uses five countries' (China, America, Japan, South Korea, Europe) patent data in 2002, 2005, 2008, 2011 and 2014 to test our model. During the course of the experiment,our method extracted evolutionary trajectories of 65 technical topics, which verified the effectiveness of our method.

Key words: technological topic evolution, non-negative matrix factorization (NMF), topic model, dynamic topic analysis

中图分类号: