Guys! Christopher Olah (a mathy person with whom I was previously unacquainted) just posted a pretty nifty set of visualizations of the Harry Potter, Naruto, and Twilight on Fanfiction.net (those were the top 3 fandoms as of mid-2010).
Be sure to click through to the interactive graphs (e.g., HP) and the rec lists based on applying the Google PageRank algorithm (e.g., HP)
He also used collaborative filtering (attn fffinnagain) to generate personalized recs for users, based on what similar users have liked. (You can only get recs from his work right now if you're a popular FFNet author, but it's a good proof of concept of how collaborative filtering could work more broadly.)
I'm not sure of a few aspects of methodology -- e.g., how he classified pairings. (One of the banes of my existence in trying to do FFN stats is the fact that ships and relationship categories are not consistently tagged.) But it's pretty cool stuff.