Tuesday, December 7, 2004

More Social

Further discussion on blog networks: Stephanie is researching the matter, and suggests that private communication is replacing blog conversations. Jon replies, echoing Diane and indulging in some clever wordplay to answer Torill.

Lilia and Stephanie 's paper (thanks Jill and Elin) is useful in establishing the existence of clusters and presenting a useful visualization (although I do wish they'd explained the embedding in 3-space they used, and that I understood the difference between figures 3 and 6) They look at a group of weblogs that link to each other at least 3 times in a span of 7 months; they start with knowledge management but discover interesting secondary clusters in cooking. Is this our old IR friend, topic drift? I bet it's not -- I bet it's a distinct phenomenon. How would we prove that?

What Efimova and Hendrick haven't (yet) done in this paper is to develop metrics that characterize the link graph; with the right metric, we'd be able to measure its change over time. (Cyclometric complexity? Mean or rms out-degree? Mean shortest path between all node pairs? Edge count?) With a metric in hand, we could measure the change over time, and then we'd have an answer. (Old papers of tangential relevance here and there).

I'd also love to see an expansion of their distinction among kinds of links; the decision to study links among posts and to exclude blogrolls and comments is, I think, very useful. Can this be expanded to semantic role without becoming arbitrary? It would be nice to be able to distinguish

A.B. Clump proposes his 23rd Conjecture (link), but fails to observe that the ocelot and the sloth are counter-examples to his hypothesis


A. B. Clump (link) dropped by for lunch, we had a cheese sandwich and discussed his recent paper.

Would the Nanard's old idea, "Should Link Anchors Be Typed Too?" be helpful here?