I’m of the opinion that you tweak the appearance and use of both taxonomies and folksonomies depending on a particular application and audience, but that you should retain the ability to store and harvest both types of metadata. For a curatorially-specific application you may push the taxonomy to the fore and maybe even not display the tags; for something more public facing the formal taxonomy may only be in evidence during searches and it’s the tagging which is the obvious form of metadata on the page.

Some sites manage to do both quite successfully, but usually – as in Flickr – it is one form of metadata which is obviously dominant, in this case, folksonomy.

Where it gets interesting is when you start to look at the cross-overs between the two. A virtuous circle would obviously be a scenario where user-centred folksonomy begins to help inform formal taxonomy – not with the aim of ever replacing it, but as a means of helping curatorial staff appreciate what users understand by a particular object or context. Cory’s point about how collections of tags themselves begin to say interesting things about the data being tagged is also vitally important. Flickr have the concept of a “cluster” (“other types of polo: mint/car”) but I see this going much, much further. Imagine, for example, that tag data itself was available via a harvesting mechanism or machine search across a range of sites – you’d then be able to draw these inferences across a much wider spectrum of content. Not forgetting, of course, when you put formal things in place around tags that they lose their immediacy…

Folktaxonomy anyone?