It wasn't that long ago that folksonomy was supposed to save the universe.  Traditional information structures - taxonomies and controlled vocabularies, among others - were considered hopelessly outmoded, slow to adapt, challenging to apply, even by implication anti-democratic.  The Semantic Web?  Puh-leeze.  Another rigid, inorganic, top-down information model.  Power to the people!

We've been living with social tagging and its advantages - openness, flexibility, collaborative potential - for five years now, long enough to make some of its champions wistful for the heyday of controlled vocabularies with all their advantages - standardization, cross-referencing, browsability.  As tagging has grown in popularity and has made more and more inroads into enterprise and research environments, where structured metadata is critical to meeting institutional goals (and where there are often already extensive repositories of structured metadata to work with and build on), the question of how to make taxonomy and social tagging play nice has come up again and again.

A taxonomy services client of ours uses cheeseburgers as a metaphor to talk about structured tagging, a metaphor we rather like.  The idea is that you get a limited range of possibilities encompassing only the most popular cheeseburger attributes -- say, ketchup, mustard, mayo, lettuce, and tomato.  But if you're anything like the foodies on the Ascentium taxonomy team, you won't want just a fast-food burger.  You'll want a gourmet burger with a greater range of custom ingredients to choose from.  To maintain the integrity of the structured tagging vocabulary, the pickled beets and pineapple rings are separately maintained as user-generated keywords.  Our client's platform accommodates both types of metadata side by side, and permits building both structured tag IDs and free text into the same knowledge base queries.

This is a fine strategy for our client's purposes, but it doesn't involve building relationships between structured and unstructured elements.  In this mixed environment, structured terms inevitably trump free keywords in terms of authority, so creating explicit relationships between different labels that point to basically the same thing ("ketchup" and "catsup", "mayo" and "mayonnaise") isn't as critical.  In an environment where none of the terminology is inherently authoritative, however, spelling out equivalences becomes more important.  Without them, you might search on half a dozen permutations of a term in order to capture all of the content related to a concept.

LibraryThing is a social cataloging application which allows users to add and tag books from their personal library, and based on their tag collections identify new books and connect with like-minded readers.   The site allows any user to designate tags as equivalent; the preferred term is chosen from these equivalencies based on the frequency of its use across the site.  Likewise, users can dissociate tags, making tag relationships subject to the same sort of community negotiation that characterizes the editing of Wikipedia entries.

Strict adherence to a collaborative model like LibraryThing's probably isn't the best solution for every enterprise.  But a more controlled collaborative environment holds out the possibility of maintaining vocabulary standards while at the same time inviting user participation and growing the vocabulary.  After all, structured metadata can't afford to be static; it needs to adapt and change, and user input in the form of tags can be leveraged to help that evolution along.  Just like a chef's artful touch can transform a cheeseburger in a can from something merely edible into something palatable.  (Okay, maybe I've taken the cheeseburger metaphor a little too far.)