Self-organization in social tagging systems

Abstract

Individuals often imitate each other to fall into the typical group, leading to a self-organized state of typical behaviors in a community. In this paper, we model self-organization in social tagging systems and illustrate the underlying interaction and dynamics. Specifically, we introduce a model in which individuals adjust their own tagging tendency to imitate the average tagging tendency. We found that when users are of low confidence, they tend to imitate others and lead to a self-organized state with active tagging. On the other hand, when users are of high confidence and are stubborn for changes, tagging becomes inactive. We observe a phase transition at a critical level of user confidence when the system changes from one regime to the other. The distributions of post length obtained from the model are compared to real data which show good agreements.

Recommender systems

“…Tags therefore represent a simple yet promising tool to provide reasonable recommendations and solve some outstanding problems in recommender systems, e.g.the cold-start problem [264]. The social impact [265] and dynamical properties of folksonomies [266,267] are expected to be applied to obtain trustworthy and real-time recommendations in tagging systems. In addition, the hypergraph [92] theory is considered to fully utilize the complete network structure of tagging platforms without using any hybrid methods and losing any information, which gives promises for generally more reliable recommendations.…”

Tag-Aware Recommender Systems: A State-of-the-Art Survey

“…(v) the multi-layered network [104] consists of user social interactions, tag co-occurrence relations and user-item-tag ternary information can be considered to describe the hierarchical structure of social tagging systems, and thus the Social Network Analysis (SNA) [105] and social influence [106,107,108] based techniques can be used to provide more substantial recommendations, and social predictions [109,110] as well; (vi) most tagging platforms are dynamical systems and evolve over time [111,112], thus the study of human dynamics [113] in analyzing the temporal behaviors and interests can provide real-time recommendations [114,115].…”

Latent geometry of bipartite networks

“…Hypergraphs, in their turn, are further generalizable to multipartite hypergraphs, where hyperedges may connect several nodes of different type. Recently, tripartite hypergraphs have been proposed to model tagged social networks, also known as folksonomies [24][25][26][27][28].…”