侵权诉讼背景下的专利无效宣告影响因素研究  

Research on the Influencing Factors of Patent Invalidation under the Background of Infringement Litigation

作　　者：彭启宁 柳炳祥 付振康 冯广宇 贝汶瑜 PENG Qining;LIU Bingxiang;FU Zhenkang;FENG Guangyu;BEI Wenyu(Intellectual Property Information Service Center,Jingdezhen Ceramic University,Jingdezhen 333001;School of Information Engineering,Jingdezhen Ceramic University,Jingdezhen 333403;School of Information Management,Nanjing University,Nanjing 210023)

机构地区：[1]景德镇陶瓷大学知识产权信息服务中心,景德镇333001 [2]景德镇陶瓷大学信息工程学院,景德镇333403 [3]南京大学信息管理学院,南京210023

出　　处：《科技情报研究》2024年第1期75-89,共15页Scientific Information Research

基　　金：2022年度文化和旅游部提质培优计划专业研究生重点扶持项目(MLIS类)“中小型文化创意企业知识产权创造能力影响因素研究——以景德镇陶瓷文创企业为例”(编号:Mlis-003);江西省研究生创新基金项目“高校专利质量评价指标体系构建实证研究”(编号:YC2021-S532)。

摘　　要：[目的/意义]以侵权专利为切入点,探究不同因素对专利技术侵权无效宣告倾向的影响机制,对比分析同一领域不同的侵权主题下无效宣告影响因素存在的差异。[方法/过程]文章首先利用LDA主题模型对所选新兴产业领域的侵权主题进行细分,了解该领域侵权专利的不同侵权主题和侵权主题词;其次,利用统计相关性模型计算在不同侵权分类主题下的各类数据指标,对比分析无效宣告倾向的相关性;最后,通过构建多特征融合的随机森林模型,对不同侵权分类主题下的专利分别进行无效宣告分类的识别训练,并利用可解释机器学习中的LIME模型,对模型中计量指标特征影响程度进行解释。[结果/结论]根据主题分类后的相关性分析发现,不同主题分类下所选取的特征指标,不但在侵权后判定是否无效的整体影响程度各不相同,而且在不同分类结果的影响因素和影响程度排序也不相同,此外,不同分类主题所依赖的分类规则和分类指标存在显著的差异。[Purpose/significance]Taking patent infringement as the starting point,this paper explores the influence mechanism of different factors on the tendency of patent infringement declaration,then compares and analyzes the differences in the influencing factors of invalid declaration under different infringement themes in the same field.[Method/process]Firstly,this paper uses the LDA topic model to subdivide the infringement topics in the selected emerging industry field,and understands the different infringement topics and infringement keywords of the infringement patents in this field;Secondly,the statistical correlation model is used to calculate various data indicators under different infringement classification topics and comparatively analyze the correlation between invalid declaration tendencies.Finally,by constructing a multi-feature fusion random forest model,the patents under different infringement classification topics are identified and trained for invalid declaration classification,and the LIME model in machine learning can be explained.Explain the degree of influence of the measurement index features in the model.[Result/conclusion]According to the correlation analysis after topic classification,it is found that under different topic classifications,the selected feature indicators not only have different overall influence on the determination of invalidity after infringement,but also have different influencing factors and influence degree rankings in different classification results.There existed significant differences in the classification rules and classification indicators relied on by different classification topics.

关 键 词：LDA模型 随机森林模型 专利无效 Word2vec模型 可解释机器学习 

分 类 号：G358[文化科学—情报学]