Thursday, June 09, 2005

Social Networking and Graphic Depictions of Human Relations

Some of the largest data sets we work with can be represented by graphs: the web graph, the internet graph and telephone call graphs have all driven work on handling, visualizing and analysing massive graphs. In a way these can all be thought of as social graphs -- where nodes are "individuals" of some sort, and edges link individuals that share some connection. When the individuals are humans, these graphs have a particular resonance. As an undergrad I did a final year project on graph drawing algorithms, and used as one of the example graphs a network of the 'social interactions' of my colleagues. Here's a few other interesting graphs I've seen recently:


  • Exploring Enron. The collected emails from Enron visualized and searchable / filterable to explore who knew what when.
  • Jefferson High School A graph of relationships at a midwest high school. (Despite the URL, this should be safe to view while at work, unless your cow orkers are offending by graphs).
  • Social networking sites. Frienster, orkut and the rest seem to have fallen from the brief period of popularity. That doesn't stop new sites from being set up, the latest asking people to input edges attesting to their past personal relations. I'm linking to a news story about it, since the web filters here disallow me from visiting the site itself...


Another graph we love to explore is the collaboration graph. Thanks to the DBLP there's a lot of data conveniently arranged in electronic form. One can download this and explore t with the DBL browser. But I haven't seen many other interesting uses of this data. It would be great to see a good visualization of this graph, highlighting various cliques and quasi-cliques in the graph. We often check our Erdos Number, but who is the center of this graph? That is, who has the minimum distance to every other researcher, or smallest average distance to others? This is of interest since although Kevin Bacon is usually taken as the center of the film star graph, Sean Connery has a lower average distance. Is there one giant component, or are there many small components? There's plenty of fun to be had with large graphs like this, if you have the time or the tools.
Post a Comment

Disqus for The Geomblog