What I like most in this new architecture is that the related links are now driven by del.icio.us. Our edit team is tagging content in del.icio.us. The engineers are pulling down the del.icio.us RSS feeds. And then we create matching logic based on the common tags. We also link back out to del.icio.us pages via the tags for the article on display.
This is a first step with several more ideas for leveraging tags coming soon. We need a more densely tagged data set behind us before some of the other plans can become real. The accuracy of the related links will also be a little shady, I'm sure, until we get more sophisticated with our tagging. But we're all excited about the possibilities for the site now that we have these tags. New ideas seem to crop up daily.
The downside is that we're probably going to phase out or at least simplify the robust taxonomy that we spent so much time and energy building and refining over the years. It's hard to look back on that effort and to think of walking away from all that valuable code and history of manhours.
Of course, there's still a need for structured tagging, and we will continue to tag in ways that enable us to create new sections of the site and to help advertisers optimize their marketing campaigns. We built a lot of functionality into the site that is dependent on tagging in a normalized way that would evaporate if we moved completely to freeform tags.
For example, we have advertisers who want to reach people interested in storage products. There are probably 10 different ways to target storage on the site with different kinds of marketing including contextual targeting, behavioral targeting and lead generation programs. Eliminating a high level structure to our tagging would mean that our freeform tags would have to be incredibly precise at all times.
Our lead engineer Derek Butcher defined this problem in an interesting way. "Structured tags give you low precision but high volume. Freeform tags give you high precision but low volume." David Weinberger explained it in terms of Trees and Leaves. I won't make any bets on the future of folksonomies yet, but I'm certain we've done a good thing here. And I'm starting to see how del.icio.us can become a business with a revenue stream instead of just a good idea worth betting on.
UPDATE: David Weinberger asks twice, "I’m not sure what it means that Infoworld is applying matching logic to del.icio.us feeds. Does that mean they’re looking at tags from non-Infoworlders?"
Not yet. At the moment, we're simply matching articles that use the same InfoWorld-specific tags. If Article A and Article B share Tag Z, then they are related. If Article A and Article C share Tag Z and Tag Y, then Article C is more related than Article B. Then we rank chronologically.
Well, this is the intent, anyway. Give us a few days to get it all working right.