Team Aviz

Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: New Results

Keywords : Visualization, Social Networks, Matrix, Graph Layout.

Alternative Visualizations of Social Networks

Participants : Nathalie Henry [ correspondant ] , Jean-Daniel Fekete.

Figure 2. MatLink and NodeTrix visualizations of a social network

Social networks analysis and visualization is becoming more and more important, due to the development of online communities on the Web, but also to the increase of security-related threats such as terrorist attacks and spreading of epidemics.

Visualizing large or dense social networks is simply not possible using current node-link diagram representations. We have shown that the matrix representation was a good alternative to node-link diagrams. However, it has not received as much attention as node-link diagrams in the past and the research community needs to design good navigation and layout methods to improve it.

Last year, we proposed two enhancements to matrix visualization: hybrid representations using links overlaid on top of a matrix (MatLink, Fig.  2 ). We have shown that the MatLink representation improved the performance compared to the traditional matrix representation for tasks related to path-finding.

To further improve social network visualization, we have introduced the NodeTrix hybrid representation  [3] is meant for small-world networks that are globally sparse but locally dense. NodeTrix mixes node-link diagrams for the global sparse structure and matrices for the local dense structure, effectively using each representation for its strength and avoiding their weaknesses.

This year, we tackled the problem of non-hierarchical aggregation in networks since, in social networks, important actors frequently belong to several groups. We have explored several mechanisms to duplicate actors in NodeTrix  [9] and showed that duplicating a node implied several tradeoffs that need to be considered depending on the analytical task.


Logo Inria