Following on my recent post outlining a tag-oriented approach to collaborative filtering, some reactions and further ideas:
Kate Ehrlich at IBM Research, where they've developed the Dogear enterprise social bookmarking application, offered these thoughts: "[re:] some of the issues you raise about tag overlap... There are a couple of issues with it as I see, One is that there is a high degree of variance in tag use. So, for instance, you and I may bookmark the same pages but use completely different tags. And vice versa. We may use similar tags but on completely different (by some measure) webpages. The other is the power law problem. There are a few people who contribute a large number of tags. Their data tends to overwhelm the system when it comes to looking for matches so one has to do some kind of normalization first..."
What Philip and Kate are working on are more advanced than what I described, to be sure. But Misiek Piskorski at HBS wrote me a very thoughtful note that said essentially, "great idea, but first we have to get more people tagging!"
Also, Bill Ives posted recently on an interesting roundup of social bookmarking sites.