Upload
hollie-jasmine-harrington
View
218
Download
0
Embed Size (px)
Citation preview
23/4/21 Dep. phys., Univ. Fribourg 1
Social Tagging Networks: Structure, Dynamics & Applications
Collaborators: Chuang LIU (刘闯 ) , Yi-Cheng ZHANG (张翼成 ) Tao ZHOU (周涛 )
Zi-Ke ZHANG (张子柯 )
Department of Physics, University of Fribourg, Switzerland
Dep. phys., Univ. Fribourg 223/4/21
Outline
Structure and Dynamics of Social Tagging Networks What is SNT? Hypergraph Strutures Dynamics and emergent properties
Applications in Personalized Recommendation (PR) Why Recommendation? How Tags benefit PR?
Conclusions & Discussion
Dep. phys., Univ. Fribourg 323/4/21
What is social tagging networks?
Dep. phys., Univ. Fribourg 423/4/21
What is social tagging networks?
Dep. phys., Univ. Fribourg 523/4/21
From Graph to Hypergraph
Bipartite Network [EPL, 90 (2010) 48006]
Hyper Network [from wikipedia.org]
Unipartite Network
Dep. phys., Univ. Fribourg 623/4/21
Hypergraph structures in STN
Zhang and Liu, J. Stat. Mech. (2010) P10005
Hyperedge: basic unit Hyper Network
Dep. phys., Univ. Fribourg 723/4/21
Two roles of social tags
Zhang and Liu, J. Stat. Mech. (2010) P10005
RolesRole1: an accessorial tool helping
users organize resources: Fig. (a)Role2: a bridge that connects users
and resources: Fig. (b)
Dep. phys., Univ. Fribourg 823/4/21
Dynamics and evolution of social tagging networks (1/3) At each time step, a random user can either:
Choose an item(resource), and annotate it with a relevant or random tag with probability p (Role 1)
or choose a tag, and find a relevant or random item with probability 1-p (Role 2)
Dep. phys., Univ. Fribourg 923/4/21
Dynamics of social tagging networks (2/3) HyperDegree Distribution:
Clustering Coefficient:
Dep. phys., Univ. Fribourg 1023/4/21
Dynamics of social tagging networks (3/3) Average Distance:
Dep. phys., Univ. Fribourg 1123/4/21
How to be personalized?Social influence Content-based recommendationNetwork-based recommendation
Zhang et al. Physica A 389 (2010) 179
Applications in Personalized Recommendation (PR) Why Recommendation?
Information overload!
Dep. phys., Univ. Fribourg 1223/4/21
Method I&II:Tripartite Hybrid(Role 1)
Zhang et al. Physica A 389 (2010) 179
Item-user: [Method I, PRE 76 (2007) 046115]
Item-tag: [Method II]
Linear Hybrid:
Dep. phys., Univ. Fribourg 1323/4/21
Method III: Tag-driven(Role 2)
Zhang et al. Accepted by EPL
Method III:
Dep. phys., Univ. Fribourg 1423/4/21
Algorithm Performance (1/3)
Dep. phys., Univ. Fribourg 1523/4/21
Algorithm Performance (2/3)
Dep. phys., Univ. Fribourg 1623/4/21
Algorithm Performance (3/3)
Dep. phys., Univ. Fribourg 1723/4/21
Conclusions and Discussion
Conclusions Structure and Dynamics The roles of social tags Application in Personalized recommendation
Discussion Recommendation with full hypergraph structure Multi-scale recommendations (semantic-based) Recommendation with community structures