Finally, interaction can be used not only to explore a dataset, but also to command the other steps of the processing chain. For example, one may filter the network according to a given query based on the properties of nodes and edges, such as “display the nodes of degree greater than 10” (Adar 2006). One may also acquire new data by interacting with the representation, as it is the case on visual Web crawlers: crawlers are programs which grab the content of Web pages by recursively visiting the hyperlinks of given Web pages. One can encode Web pages as nodes, and hyperlinks as edges. The corresponding node-link diagram represents the Web explored by the crawler. One could then ask for the crawler to visit the hyperlinks of a Web page by double-clicking on its corresponding node. The crawler would therefore retrieve the new Web pages and scan the new hyperlinks available, to update the visualization.
Interaction techniques are therefore essential to explore large networks, to speed up analysis tasks, and to integrate visualization in the data processing chain. However, time-varying networks raise specific challenges that we will cover in a following blog post