基于加权网络的专利技术主题分析

Analysis of Patent Technology Themes Based on Weighted Network

在专利网络研究中,通过设定阈值构建无权无向网络并将其用于专利内容分析的方法得到广泛研究,但其中阈值设定随意性大、主观性强,最优阈值的选择和阈值优化往往是非常困难的事情。鉴于此,提出一种基于加权网络分析专利技术主题的方法,该方法将专利文本相似度设定为边的权重,从而构建专利加权网络,通过网络特性分析技术主题的发展。以中国电商发明专利为例,研究表明:(1)通过专利加权网络节点单位权重及其分布,发现电商专利技术主题凝聚性不强;(2)通过节点强度指标,发现电商数据分析类专利已经成为研究热点;(3)通过权重筛选,发现一个风筝型聚集社团,该社团特性显示电商专利发展有一个局部突出的技术点。案例研究进一步表明专利加权网络的优势在于将文本相似度引入专利网络构建中,以权重角度分析专利之间的确定关系,从而发现更多的网络特性,揭示专利技术主题的发展。

The method of constructing unauthorized and undirected networks by setting thresholds has been widely studied in the research of patent networks.However,the threshold setting is arbitrary and subjective,both the selection of the optimal threshold and threshold optimization are very difficult to carry out.In view of this,a method of analyzing patent technology themes based on weighted network is proposed.This method sets the patent text similarity as the weight of edges,thereby constructing a patent weighted network and analyzing the development of patent technology themes through network characteristics.Taking Chinese e-commerce invention patents as the study case,the research shows that:(1)the cohesion of e-commerce patent technology themes is not strong from the research of unit weights and their distributions of patent weighted network nodes;(2)the e-commerce data analysis patents have become a research hotspot from the study of node strength indicators;(3)through weight screening,the research finds a kite-type agglomeration community whose characteristics show that there is a local outstanding technical point in the development of e-commerce patents.The case study further shows that the advantage of the patent weighted network is to introduce text similarity into the construction of the patent network,and to analyze the definite relationship between patents from the perspective of weight,so as to discover more network characteristics and reveal the development of patent technology themes.