[目的/意义] 对人工智能领域科研团队进行识别，并基于多个维度的指标提取领军科研团队，旨在丰富科研团队识别的流程与方法，为从科研团队视角分析人工智能领域脉络、前沿和主题提供依据。[方法/过程] 以Web of Science为数据来源，采集2009-2018年间人工智能学科领域所有科技论文的数据，通过算法设计与人工核查进行数据清洗；基于分数计数法构建全局合著网络，并利用社区探测算法动态调参、识别科研团队；进而基于多维度的指标提取出领军团队，并加以比较分析。[结果/结论] 从实践出发构造人工智能科技论文数据清洗的规则；构建基于合著关系识别人工智能科研团队的流程体系；提出通过消除边缘结点进行合著网络筛选，进而利用已知团队作为参考进行参数调整的思路；较为系统和准确地识别出全球人工智能科研团队，并基于发文量、被引量、h指数、中介中心度、接近中心度和加权点度中心度6个维度的指标提取出领军科研团队，同时，给出结合论文数据和实证调研对每个领军团队的示例性分析。
[Purpose/significance] This paper identifies the research team in the artificial intelligence field, and extracts the leading research team from multi-dimensional indicators, aiming to enrich the process and method of identification of the research team, and provide the basis for analyzing the context, frontier and theme of the field of artificial intelligence from the perspective of the research team.[Method/process] This paper was based on the publication data of the Web of Science category Computer Science, Artificial Intelligence from 2009 to 2018, and did data cleaning via programming and manual check. Global co-author network is constructed based on the fractional counting method, and the Louvain algorithm was used to dynamically tune and identify the research teams. Moreover, the leading research team was extracted based on different indicators with parameter adjustment.[Result/conclusion] From practical view, the study has constructed a set of rules for cleaning publication data of artificial intelligence field. The process of identifying artificial intelligence research teams based on co-authorship is constructed. The study proposes the method of tuning the parameter by eliminating edge nodes in the collaboration network and further taking the known research teams as baseline. The worldwide research teams of artificial intelligence field are systematically and accurately identified. The leading research teams are further extracted based on indicators of six dimensions, i.e. number of publications、number of citations、h index、weighted degree centrality、betweenness centrality、closeness centrality. Exemplary analysis is conducted on leading research teams of each dimension by combining the publication data and web information survey.
[1] 陈春花,杨映珊.基于团队运作模式的科研管理研究[J].科技进步与对策,2002(4):79-81.
[2] ACEDO F J, BARROSO C, CASANUEVA C, et al. Co-authorship in management and organizational studies:an empirical and network analysis[J]. Journal of management studies, 2006, 43(5):957-983.
[3] GREGORIO G, JINSEO P, CHARLES H, et al. Scientific authorships and collaboration network analysis on chagas disease:papers indexed in pubmed (1940-2009)[J]. Journal of the institute of tropical medicine in sao paulo, 2012, 54(4):219-228.
[4] 李亮, 朱庆华. 社会网络分析方法在合著分析中的实证研究[J].情报科学,2008, 26(4):549-555.
[5] 李纲,李春雅,李翔. 基于社会网络分析的科研团队发现研究[J].图书情报工作,2014,58(7):63-70,82.
[6] 沈耕宇,黄水清,王东波. 以作者合作共现为源数据的科研团队发掘方法研究[J].数据分析与知识发现, 2013,29(1):57-62.
[7] 吕璐成,赵亚娟,王学昭,等. 基于关联规则挖掘的研发团队识别方法[J].科技管理研究, 2016, 36(17):148-152,189.
[8] ANTONIO P, CARLOS O, FÉLIX M. Detecting, identifying and visualizing research groups in co-authorship networks[J]. Scientometrics, 2010, 82(2):307-319.
[9] 任妮, 周建农. 合著网络加权模式下科研团队的发现与评价研究[J].现代图书情报技术, 2015,31(9):68-75.
[10] 范丽鹏,余厚强,姜宇星,等.人工智能研究前沿识别与分析:基于高产机构对比研究视角[J].情报理论与实践,2019,42(9):16-21.
[11] TRAN H N, HUYNH T, DO T. Author name disambiguation by using deep neural network[C]//Asian conference on intelligent information and database systems. Phuket, Thailand:ACIIDS, 2014:123-132.
[12] GLANZEL W. National characteristics in international scientific co-authorship relations[J]. Scientometrics, 2001, 51(1):69-115.
[13] PRITYCHENKO B. Fractional authorship in nuclear physics[J]. Scientometrics, 2016, 106(1):461-468.
[14] 许治,陈丽玉,王思卉.高校科研团队合作程度影响因素研究[J].科研管理,2015,36(5):149-161.
