Hey, didn't I just do this on some other networking site Defrag invited me to? O.k., I'll play.

I know you are all terribly excited about flying to Denver to hear a little old lady librarian talk about taxonomies; I might just whip the pencil out of my bun and go wild!

But wait, before you tiptoe away...

I could probably call my talk "It's All Good," but that's been overdone. Anyway, my talk (which as of today is yet evolving!) will focus on what taxonomies can learn from folksonomies and vice versa, and how nice, tidy semantics and general third-order messiness are symbiotic and complementary. (In other words, it IS all good.)

Increasingly, as I ponder the massive quantity of bibliographic data the librarian community has built over the last hundred-plus years and the piles of user-supplied data mounding up on the social Web, I'm taken less with the obvious distinctions between taxonomies and folksonomies ('expert' versus 'user'; expensive versus cheap; slow vs. just-in-time; precise versus messy; flat versus heavily faceted), but with the very different focus of each ontological approach.

The bibliographic record (also known as a surrogate record) is an item-level description focused on "the book in hand" (that's an actual expression in LibraryLand), whereas folksonomies tend to focus on collection-level aggregation: "put this thing with other things that are like it." Another way to think about this is that traditional taxonomies focus on description, while folksonomies focus on discovery (since the implied corollary to the previous statement is "so I can find it again").

Furthermore, taxonomies are record-focused, whereas folksonomies tend to be data-focused. The ultimate unit of expression in a taxonomy is a record representing an intellectual item, whereas the ultimate unit of expression in a folksonomy is a tag representing a group of like items.

Which is better? Like I said: It's All Good. An item-level focus tends to reap precise, high-quality facets, whereas a data-level focus tends to produce discovery-oriented data sets that are highly portable and reusable. Furthermore, both ontological types have other strengths and weaknesses. Not only that, but emerging metadata tools are blurring the line between both groups.

My talk will focus on (at least) three blended collections that demonstrate the strengths of a blended approach: Librarians' Internet Index, LibraryThing for Libraries, a del.icio.us set developed by librarians, and an automated metadata generation project at UC Riverside.