Network Visualization

Graphs and Maps

Methods based on various forms of graph theory are useful in a wide variety of fields, not the least within the natural and technological sciences. When it comes to using these methods in relation to social networks, they have proven to be very useful within the discipline of linguistics. Languages and discursive structures lend themselves well to thinking in terms of composition, structure, and relations. Graph theory also plays a very important part in sociology and social network analysis to map and analyze connections between individuals and groups, social categorizations, power structures, diffusion mechanisms, etc.

The exact layout of network maps may vary greatly, and, depending on what the analyst wants to display and on the character of the data, some layouts may be deemed more fitting than others in certain contexts. As Adam Perer argues, network visualization is not a straightforward process, since it poses a challenge regarding the choice of the best-suited ways and tools for performing the representation. Especially when one works with large networks, visualizations can often be quite chaotic and hard to read. While network images serve the purpose of making results of analyses of network data clearer and easier to interpret, cluttered maps with overlapping edges and node labels that are sometimes nearly illegible often undermine the strengths of the strategy. In those cases, editing the visualizations through various techniques for sizing, zooming, panning, or filtering can sometimes help overcome the problems.

[Page 620]As many networks lack any inherent or underlying spatial logic, the elements displayed are often positioned based on other considerations. One way of laying out the map is to do it manually, arranging the elements in some way that the researcher considers meaningful for his or her current purposes. While this strategy involves tedious work, especially as regards larger networks, most network visualization software available today uses different layout algorithms based on the primary design principle of clarity and readability.

There has been extensive work on automated graph layout, and in many of the most commonly used computer programs, a prominent role is played by the Kamada-Kawai (1989) and the Fruchterman-Reingold (1991) layout algorithms. These are force-directed algorithms designed to produce a network image, which is as aesthetically pleasing as possible. These approaches echo Jacob Moreno's idea that “the fewer the number of lines crossing, the better the sociogram.” Because of this, the relative positions of nodes in a network most often do not convey any substantial analytic information but rather serve aesthetic needs. More sophisticated examples of network visualization, however, may be more explanatory. When working with clustering, for example, the layout principle is commonly used to convey spatial proximity that is based on how cluster boundaries are delineated.

...