Traditional social network analysis (SNA) [1] mainly studies the relationship between nodes and edges, and all nodes are of the same type of data. However, some network nodes may be of different types, such as paper collaboration relationships, which include both author nodes and paper nodes, as well as conference nodes and institution nodes. All these nodes are mixed together to form a social network, and each type of node is called a mode. It is difficult to study the relationship between modes using traditional graph analysis methods. The method of studying this type of social network is called multimodal social network analysis (mSNA). This article from IEEE VAST 2013 [2] designed a visual analysis method for multimodal social networks through cooperation with sociologists. First, the problem model was abstracted for mSNA, then the visual analysis system MMGraph was designed, and finally, based on the feedback from sociologists, the process of how researchers collaborate with domain experts was summarized.

Figure 1 System interface

Design Process

The design of the system is completed through five steps: early design, iterative tool development, formative qualitative evaluation, iterative tool refinement and summative qualitative evaluation.

In the initial design process, the researchers first asked sociologists about the questions they were interested in, and then designed a system prototype based on the questions. When entering the iterative system development phase, they were inspired by jigsaw[3] and designed parallel node-link bands (PNLBs), as shown in Figure 1 above. They displayed each node pattern set in the network as a strip, and only drew the edges between the patterns. The nodes in each pattern can be sorted according to the node attributes such as the age, name and degree of the node, and a node and its neighbor nodes can be displayed in a floating window. For high-dimensional nodes, they also designed the drawing of parallel coordinate axes, called Open Sesame, which can perform high-dimensional filtering operations on the nodes of the pattern.

Using this tool, they conducted a qualitative evaluation and selected 5 sociologists to complete 4 basic tasks. Everyone completed the tasks smoothly and gave feedback on their experience using the system. Sociologists believed that adding more graph metrics on the basis of PNLBs would be more helpful. Therefore, in the iterative system optimization phase, they defined three new types of polymorphic graph metrics:

1. Multimodal degree centrality : For a node, count the number of edges it connects to nodes inside (outside) the model. This scale can be used to measure the connectivity of a node. For example, in the paper collaboration graph, the number of papers published by an author (outside the model), the number of times an author collaborates with others (inside the model)
2. Multimodal betweenness centrality . For a node, count the number of times the shortest path between any two nodes that are not in the same mode passes through it. This metric can be used to measure the importance of a node in connecting other modes. For example, in the paper collaboration graph, papers with high values ​​of this metric are connected to more authors, and these authors no longer have collaborations except for this paper.
3. Multimodal closeness centrality . For a node, count the sum of the shortest path lengths from it to all nodes in other modes. This metric measures the distance from a node to other modes.

With these newly added scales, the design process entered the final stage, summarizing the qualitative evaluation. Three domain experts conducted the evaluation and used the system to freely analyze data networks. They were very satisfied with the new features. The three graph metrics scales were very helpful in finding special nodes in the network and discovering the relationship between different patterns.

Summarize

This article proposes a complete set of problem abstraction and design process for polymorphic social network analysis, implements the MMGraph visual analysis system, and continuously improves the system through communication with domain experts. It provides a workflow for us to analyze and design visualization systems and how to cooperate with domain experts.

References

[1] John Scott. Social Network Analysis: A Handbook. SAGE Publications , London, UK, 1991
[2] Ghani, Sohaib, Bum Chul Kwon, Seungyoon Lee, Ji Soo Yi, and Niklas Elmqvist. Visual Analytics for Multimodal Social Network Analysis: A Design Study with Social Scientists. IEEE Transactions on Visualization and Computer Graphics , 19(12):2032-2041, 2013
[3] Stasko, John, Carsten Görg, and Zhicheng Liu. Jigsaw: supporting investigative analysis through interactive visualization. Information Visualization , 7(2):118–132, 2008

via:/vis.pku.edu.cn/