They are doing topic modeling on a text corpus. This one was a small collection of blog posts, journal articles and essays about "what is/are the digital humanities." Check out the MALLET toolkit for more on the specific tool.
I'm probably not doing it justice, but as I understand it they are basically finding statistically significant recurring patterns of word clusters (based on a seed number, so you say "what are the 5 clusters of words"), then feed those results into a graphing/visualization algorithm (Gephi). Obviously the result is highly mediated and in this case the data set was very small (50 documents), so he can't make any significant claims. But I'm arguing (for my dissertation proposal) that there is some potential here and more work, done by me :), is needed.
I'm just getting (re)familiar with this stuff, it's been nearly a decade since I did all that bioinformatics stuff at Wes. My data mining chops are pretty rusty.
Some good humanistic discussions here:http://mith.umd.edu/topic-modeling-in-the-humanities-an-overview/http://www.stanford.edu/~mjockers/cgi-bin/drupal/node/39