SOCIAL TAGGING DYNAMICS UNDER SYSTEM RECOMMENDATION AND RESOURCE MULTIDIMENSIONALITY

• Published in: Journal of systems science and systems engineering Vol. 25; no. 3; pp. 271 - 286
• Main Authors: Xia, Haoxiang, Zhao, Xiaowei, Liu, Huiyu
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2016; Springer Nature B.V; Institute of Systems Engineering, Dalian University of Technology, Dalian, 116024, China%Institute of Systems Engineering, Dalian University of Technology, Dalian, 116024, China; School of Software Technology, Dalian University of Technology, Dalian, 116620, China%Institute of Systems Engineering, Dalian University of Technology, Dalian, 116024, China; Dalian Branch of China Unicom Ltd., Dalian, 116001, China
• Subjects: Combinatorial analysis; Complexity; Dynamic models; Economic Theory/Quantitative Economics/Mathematical Methods; Engineering; Operations Research/Decision Theory; Power law; Tagging; Tags; 动态模型; 多维性; 幂律分布; 影响系统; 标签系统; 生成模型; 社会性; 资源; tag usage patterns; dynamic model; Social tagging systems
• ISSN: 1004-3756; 1861-9576
• DOI: 10.1007/s11518-016-5299-z

Social tagging systems have attracted plenty of research endeavors recently. The dynamic models of tag generation or tag usage are one of the key subjects of inquiry. However, the existing models do not well explain the ＂staged＂ power-law distribution of tag usage frequencies as observed in various social tagging systems. To cope with this, a new tag-generation model is proposed in this paper, which is based on a preferential selection mechanism influenced by the combinatorial effects of system recommendation and resource multidimensionality. Furthermore, to validate the model, the simulative results under different parameter combinations are compared with the distributions of tag usage frequencies in datasets from three famous social tagging systems, namely Delicious.com, Last.fin and Flickr. For different categories of resources of the three systems, three tag usage patterns can be identified, namely the power-law distribution with two plateaus, the power-law distribution with one plateau, and the standard power-law distribution. All the three patterns can be well fitted and explained by the proposed model.