Belle Tseng from NEC has a terrific visual metaphor for a way to think about blog clustering. She shows us a mountain range, then cuts it off just below the peak with a level plane. The mountain peaks now look disconnected. If you move that plane lower, it’s clear that they’re all part of the same mountain.
This is the same strategy that her group is using for “Tomographic Clustering” of blog posts. Put the rank of a blog (through something like PageRank) on the y-axis of a graph. Draw connections between the different blogs showing which link to one another. Set the threshold fairly high and you don’t have much connection amongst the A-list. Let the threshold drop a bit lower – considering lower ranked blogs – and the blogs look much more interconnected. In other words, the “peaks” are authorities and the “valleys” are connectors. Very interesting, though hard to explain without visuals, so check out her paper.
Natalie Glance, one of the conference organizers, has a great paper on the US political blogosphere. She and a collaborator found a set of liberal and conservative blogs – self-identified – and did a close study of their linking behavior, looking to see a) whether the communities linked to each other and b) whether they linked internally, ala an “echo chamber”.
The results seem to indicate that blogs are part of some sort of “cyberbalkanization”. 91% of links to blogs linked to blogs of a similar persuasion. Conservative blogs were more likely to link to other conservative blogs than liberals to liberals – 82% of conservative blogs are linked to at least once in the set, while only 67% of liberal blogs are. Despite this disparity, both classes of blogs have roughly similar average numbers of outgoing links.
Using technorati, blogpulse and other ranking engines, Natalie and her collaborator found the conservative and liberal “a-lists”. They harvested posts for the top 20 in each list – 12,470 liberal posts, 10,414 conservative posts. Extracting all the links, excluding self-links, they found 1,511 left-left citations, 2,110 right-right, 347 right-left and slightly fewer left-right. While there were fewer conservative posts, there were more right-right links, implying that conservative bloggers link to each other significantly more than liberal bloggers, at least amongst the A-list.
Natalie then used a set of linguistic analysis tools – link and phrase correlations. There were no significant differences between the two communities – in other words, there wasn’t evidence that either the left or right blogospheres were more self-similar as regarding using the same phraseology or linking to the same sites.
Half of the posts in this set of A-list posts cite mainstream media sources. Across these sources, some are evenly cited from the left and the right – the New York Times, the Washington Post. The right showed a strong preference for Fox News, the National Review, the NY Post and the Washington Times; the left preferred the LA Times, Salon and the Boston Globe.
While that’s pretty predictable, the results regarding the mentions of specific political names are a little surprising. The left mentions Donald Rumsfeld and Zell Miller; the right mentions Dan Rather, Michael Moore, and Terry McAuliffe. In other words, there’s some evidence that people would rather tear down than build up…
Evidently, lefties find The Onion funnier than the right – 50 links from the left, 14 from the right. And there’s a set of comics – Day by Day, Cox and Forkum – that (as a card-carrying liberal) I’ve simply never encountered.
Very psyched to read the whole paper sometime when jetlag hasn’t left me half dead.