图书情报工作 ›› 2019, Vol. 63 ›› Issue (10): 75-86.DOI: 10.13266/j.issn.0252-3116.2019.10.009

• 情报研究 • 上一篇    下一篇

基于指数随机图模型的专利引用关系形成机制研究——以奈拉滨药物为例

杨冠灿1, 刘占麟2, 李纲3   

  1. 1. 中国人民大学信息资源管理学院 北京 100872;
    2. 华盛顿大学工业工程系 西雅图 98105;
    3. 武汉大学信息管理学院 武汉 430072
  • 收稿日期:2018-08-05 修回日期:2018-11-27 出版日期:2019-05-20 发布日期:2019-05-20
  • 作者简介:杨冠灿(ORCID:0000-0002-1706-1884),讲师,博士,E-mail:yanggc@ruc.edu.cn;刘占麟,博士研究生;李纲(ORCID:0000-0001-5573-6400),教授,博士,博士生导师。
  • 基金资助:
    本文系国家自然科学基金项目"基于指数随机图模型的专利引用关系形成影响因素及机理研究"(项目编号:71403256)、国家自然科学基金项目"面向专利文本中实体关系抽取的远程监督方法研究"(项目编号:71704169)和国家自然科学基金重大项目"国家安全大数据综合信息集成与分析方法"(项目编号:71790612)研究成果之一。

Understanding Mechanisms of Patent Citation Formation Based on ERGM: A Case Study of the Nelarabine Drug

Yang Guancan1, Liu Zhanlin2, Li Gang3   

  1. 1. School of Information Resource Management of Renmin University of China, Beijing 1000872;
    2. Department of Industrial and Systems Engineering, University of Washington, Seattle 98105;
    3. School of Information Management of Wuhan University, Wuhan 430072
  • Received:2018-08-05 Revised:2018-11-27 Online:2019-05-20 Published:2019-05-20

摘要: [目的/意义] 专利引用关系形成问题是理解创新网络的一个重要问题。传统的回归模型对观察对象设定的独立性假设,无法将网络的结构效应因素整合到模型中来提供综合性的统计推断。指数随机图模型(ERGM,Exponential Random Graph Model)是一种创新性的统计推断方法,它能够将属性特征、自组织特征以及网络协同特征三种特征综合起来观察。[方法/过程] 以奈拉滨药物的专利引文网络作为研究对象,利用ERGM系统检验了影响专利引用关系的五种机制:专利属性的主效应;专利引用时间的差值效应;专利引用关系的聚敛效应;专利引用关系的传递效应;专利引用关系的网络协同效应。[结果/结论] 五种机制都在奈拉滨药物的专利引用关系的形成过程发挥了作用。但三种效应对于奈拉滨药物的专利引用关系的形成作用最为显著:共享发明人关系协同效应、共享家族关系协同效应、传递效应。一些辅助机制也会对专利引文关系形成产生影响,如引文时滞、权利要求数量和参考文献数量。

关键词: 专利引用关系形成, 指数随机图模型(ERGM), 奈拉滨, 统计网络模型

Abstract: [Purpose/significance] The Formation of patent citation is necessary to understand innovation networks. The independence assumption set by the Conventional regression model for observed objects cannot integrate the structural effect factors of the network into the model to provide comprehensive statistical inference. ERGMs (exponential random graph model) represent a methodological innovation of statistical inference for networks given their ability to model actor attributes along with endogenous self-organizational processes and exogenous network covariates.[Method/process] In this paper, ERGMs are applied to systematic inspect the five mechanisms affecting patent citation formation in a sample of Nelarabine drug. The five mechanisms contain main effect, difference effect of citation lag, and activity effect, transitivity effect and network covariates.[Result/conclusion] We find that five different types of mechanisms play diverse roles in patent citation formation. And three of effects among these mechanisms have significant impacts on citation formation of nelarabine drug:network covariates based on shared inventors and shared patent family membership, and transitivity effect. In addition, some aided mechanism play a supporting role on patent citation formation, such as difference of time lag, main effects of number of claims and reference.

Key words: patent citations formation, ERG (exponential random graph), nelarabine drug discovery, statistical network analysis

中图分类号: