I’ve been reading through the wiki pages for Brian Lamb’s presentation on wikis etc. at David Wiley’s recent ITI conference. Excellent excellent reading. Brian has done an awesome job as usual in gathering supporting resources. Man, if I could just borrow his brain for a while, I’d have my Master’s thesis done 
(but that’s another story).
Anyway, Brian linked to a post by Jon Udell: Collaborative Knowledge Gardening – basically on how the role of metadata is effectively shifting, and less formally described strategies are showing the promise of being more effective. He compares tools like Flickr.com, which uses “bags of keywords” to the more conventional strategies of complex taxonomies.
I’ve become a huge fan of Flickr, and have to admit that using it has really opened my eyes to the power and flexibility of the “bags of keywords” approach. Just look at the “Favorite Tags” page on Flickr – these are keywords that people have voluntarily added to their photo uploads, and they’ve been found by other users. That sounds like a controlled vocabulary, without the centralized control… It’s also quite cool to be able to enter any tag you want and (usually) find some matching photos. You don’t have to browse a predefined vocabulary, you can just use your own…
I really think that strategies that enable both centralized top-down as well as decentralized bottom-up approaches will be the most successful (easy for The People to use, easy for The Institution to adopt).
{ 2 comments… read them below or add one }
Uncontrolled keywords would probably work really well with latent semantic indexing so that correlations between terms can be made. That way, different people wouldn’t need to know the exact terms that others use in order to find related items.
Yeah. something like that would be extremely useful. Even something as simple as pulling the “bag of keywords” directly from the content – sort of like what the “Related Posts” feature of this weblog does – it’s doing a fulltext match and relevancy ranking of all content on the fly. Something like that takes away the need to predefine taxonomies and relationships…