中文摘要：

[目的／意义]在科学研究中，科研团队通过学术交流互动推动着科学进步。以计算语言学领域为例，识别科学领域中科研团队的角色并探究其特征。[方法／过程]通过构建机构作者合作网络，运用社群识别算法发现科研团队，结合论文引用关系构建科研团队引证网络，基于蝴蝶结模型和网络位置理论划分出领导者、中介者、追随者和孤立者等4种角色的科研团队。[结果／结论]不同角色的科研团队在成员数量、发文量、合作强度等3个方面具有不同的特征。如领导者角色的团队数量最少而平均规模较大，追随者角色团队数量最多而团队规模较小，中介者团队合作密度与团队的发文量之间存在着显著的负相关关系，孤立者角色的团队与其他团队几乎不存在引证和被引关系。

英文摘要：

[ Purpose/significance ] In scientific research, research groups promote the advancement of science through academic exchanges and interaction. This paper recognizes the roles of scientific research groups and explores their features based on a case study of computational linguistics. [ Method/process]Firstly, combining community identifica- tion algorithm, this paper constructed the corporate author cooperative network to identify research groups. Secondly, the citation network of research groups was built based on citation relationship. Then, this paper identified leaders, agents, followers and isolated four roles of the research teams by combining the Bow-Tie model and the network positional relation- ship theory. [ Result/conclusion] Lastly, the different features of four roles of research groups were revealed by the num- ber of members, issued amount and cooperation intensity. For example, the number of leader role is the least and the av- erage is larger. The number of follower role is the largest and average is smaller. There is a significant correlation between the density and the results in agents. The groups of isolated have no correlation with others.