I recently ran into Aggie Donkar, a former student of mine, now graduated. We got to talking about her Masters paper, which sounded interesting and which I subsequently read. As an aside, I’d just like to say that a very large percentage of our Masters papers sound incredibly interesting, and I only wish I had the time to read them all! Anyway, Aggie’s Masters paper was indeed very interesting, though I had some questions regarding her methodology. But I’m not going to critique the paper here. My point is, one of these methodological questions led me to wonder about something.
In her Masters paper, Aggie looked at the tags used by a small community of del.icio.us users. She started with the community, or actually with a set of URLs, though the URLs were for sites that were pretty much only going to be of interest to a predefined community.
So I wondered, is it possible to identify communities within delicious by looking at who is using overlapping sets of tags? In other words, a community is defined as the set of users who use a specified set of tags. Would there be some “centroid” tag, like Aggie says sga is for her data? Are some tags more prototypical to the community, and some further and further out on the periphery, as Rosch says is the case for the way people actually create categories? Are there boundary object tags, that link communities?
This doesn’t sound too difficult to me, honestly… though I admit I don’t have to tools (intellectual or technological) to do it myself. It sounds to me like pretty straightforward network analysis. Indeed, it wouldn’t surprise me if someone had done this already. Anyone know if this has been done, or anything like it? If not, will someone please do it?
1 Comment
Fred
Jeff, you might be interested in the work of Alla Zollers, she is looking at this problem. One of her papers was at the WWW tagging workshop: http://www.ibiblio.org/www_tagging/
I must say I’ve also thought about this question, and struggled with a methodology. The early Huberman tagging paper dealt with genre, but I haven’t seen anyone narrow down on community from an ontological perspective. How do clusters tag alike, how are tagging behaviors adopted or normalized. In my community, we’ve got one set of people that use the social-networks tag, another uses social_networks, and a third uses socialnetworks. In the end I follow them all.
Another useful mechanism might be delicious networks. I imagine these are adopted and used by only a small fraction of the userbase, but they’re the best example of community that’s tangible. It would be interesting to compare the nature of the clusters to the blogosphere…since ego is pushed to the margins in delicious, people are actually making their community decisions based on content (as opposed to findability). Interesting stuff.