Despite its wide usage among researchers, node-link diagrams are not subject to well-established graphical conventions like those found in geographical maps. One can easily misunderstand them, so the visualization should come with a cautionary text in the legend, stating that:
- Distances are not absolute but relative to local connections. In consequence, one should not compare two graphical distances.
- The representation may be rotated in every direction so the top, bottom, left and right positions have no particular meaning.
- Nodes at the center of the picture may not be central at all in the network.
Geographical conventions may nonetheless influence the design of node-link diagrams. When dealing with multiple data attributes, several authors (see (Boyack 2005) and (Klavans 2009)) distinguish the visual topology made of dots and lines to other visual variables. Like for geographical maps, the topology is then considered as the “base map“, while other variables are added as layers of information. In such cases, these visualizations are called “network maps”. Their comparison is facilitated because node and edge coordinates are the same for all maps. This approach is remarkably used in Scientometric studies (i.e. the study of science as a system), where maps of science represent the way scientific fields relate to each other through publications and co-authorship networks.
We have seen that the exploration of graphs is greatly enhanced by visualization. However when dealing with large graphs of hundred thousands of nodes and edges, reading a static picture is difficult and provides limited insights due to the density of nodes and links. One may want to focus on specific sub-graphs, or to compare maps colored by different attributes, or to filter the graph based on particular rules… Such tasks are supported by interactive features as we will see in the next blog post.
Don’t miss out the Graph Viz 101 series! Subscribe to the email alerts below (you can unsubscribe any time), or follow us on your favorite social network: Twitter, LinkedIn. Help us spread it to see everyone making better and useful graph visualizations!