论文部分内容阅读
【目的】探索基于两种相似度矩阵的专利引文耦合分析方法以识别研究前沿。【方法】基于原始观测值和余弦距离两种相似度算法,建立专利相似度矩阵,利用社会网络分析得到研究前沿簇,并进行簇类命名,从而得到研究前沿。并利用Innography数据库的脑机接口领域专利对以上方法进行案例研究。【结果】发现两种相似度矩阵中,基于原始观测值算法得到6个研究前沿簇,涉及6类BCI研究内容;基于余弦距离算法得到9个研究前沿簇,涉及8类BCI研究内容,两者的FID重合率均为43%。【局限】本文侧重于两种算法的结果,即研究前沿数量、重合度和内容进行比较,缺少对于算法本身特性的比较。【结论】基于这两种相似度算法的引文耦合法均可识别出领域的研究前沿,余弦距离相似度算法能识别出更多数量的研究前沿,且比基于原始观测值相似度算法的识别结果更全面。
【Objective】 To explore a patent citation coupling analysis method based on two similarity matrixes to identify the research frontier. 【Method】 Based on the two similarity algorithms of original observation and cosine distance, a patent similarity matrix is established, and the research frontier clusters are obtained by social network analysis and the cluster names are named, leading to the research frontier. And use Innography database in the field of brain machine interface patent case studies on the above methods. 【Result】 In the two similarity matrices, six research frontier clusters based on the original observation algorithm were found, covering six kinds of BCI research contents. Based on the cosine distance algorithm, nine research frontier clusters were obtained, involving eight kinds of BCI research contents, both The FID coincidence rate was 43%. [Limitations] This article focuses on the results of two algorithms, that is, the number of frontiers, degree of coincidence and content of the study, the lack of comparison of the characteristics of the algorithm itself. 【Conclusion】 Citation coupling methods based on these two similarity algorithms all can identify the research frontier in the field. The cosine distance similarity algorithm can identify a larger number of research frontier than the original observation similarity algorithm More comprehensive.