面向科学文献同名消歧的可视化分析方法  

Visual Analysis for Name Disambiguation of Academic Papers

作　　者：张鹏宇 张勇 崔言杰 尹宝才 Zhang Pengyu;Zhang Yong;Cui Yanjie;Yin Baocai(Beijing Artificial Intelligence Institute,Faculty of Information Technology,Beijing University of Technology,Beijing 100124)

机构地区：[1]北京工业大学信息学部北京人工智能研究院,北京100124

出　　处：《计算机辅助设计与图形学学报》2022年第11期1659-1672,共14页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(62072015,U19B2039);中国高等教育学会专项课题(2020XXHYB16)。

摘　　要：直观地展示科研团队中的合作关系,能够辅助用户在科学文献管理工作中更好地进行文献同名消歧.由于科学文献作者合作关系复杂,难以通过可视化进行直观展示,因此设计并实现了面向科学文献的同名消歧可视化分析方法.首先,根据文献合著者存在的合作网络生成合作关系图,揭示科研团队中作者的合作关系;其次,为了展示不同作者研究方向之间的相关性,设计了合作关系图和发文期刊图之间的可视化联动;最后,通过结合深度学习模型分别对文献和作者进行分类,实现了从作者和团队任意主体出发的交叉分析与连贯推理.基于北京工业大学4000篇论文构成的真实数据进行案例分析,并邀请了科研管理人员和学生通过实验完成度和李克特量表进行评价,验证了所提方法的有效性.The visual presentation of academic social networks can help users better perform the name disambiguation work in academic paper management.The complex network among authors brings a great challenge to visual interaction.In order to help users better perform the name disambiguation work,a name disambiguation visualization analysis method for academic papers is designed and implemented.Proposed method first generates a cooperative relationship graph based on the cooperative network of co-authors,which is used to reveal the cooperative relationship of authors in the scientific research team.Then,to show the correlation between the research directions of different authors,the visual linkage between the collaboration graph and the published journal graph is designed.Finally,through the combination of the deep learning model,papers and authors are classified respectively to achieve the cross-analysis and coherent reasoning starting from the author or team.The system is based on 4000 actual papers for case studies and professionals and students are invited to use and evaluate the system,proving the effectiveness in solving name disambiguation.

关 键 词：可视化分析 网络分析 同名消歧 多视图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]