Hybrid Vigor

A week ago I had the happy experience of going to a local pub with three graduate students in the iSchool – Michael Braly, Jason Parker, and Geoff Froh.  I am in their debt for the good times and stimulating conversations we enjoyed!  In addition to learning about session beers, sharing notes about travel, cameras and careers, they shared with me some of their enthusiasms from their classes, and folksonomies emerged as a prominent topic.

The term folksonomy is one of those linguistic constructs that suggests its own definition. When I first heard the term, I wondered what would happen if you hybridized a folksonomy and a formal taxonomy (I can’t imagine that I was the first to have this thought, and I wouldn't claim it as my own), but also imagined that the result might be too diffuse to be of much use.

Michael, Geoff, and Jason pointed me to the del.icio.us site for their class on Web 2.0 trends. [aside: is this a great way to share ideas in a class, or what?] Most germane to folksonomies is the Golder and Huberman article, The Structure of Collaborative Tagging Systems, in which the authors cite evidence that the proportion of tags assigned to a given resource stabilizes after about 100 tags. If this result generalizes, then the notion of hybridizing a folksonomy and a traditional taxonomy might be both tractable and useful.

A formal taxonomy provides an established hierarchy, but suffers from rigidity (arguably both a feature and a bug). Collaborative tagging affords an electronic warrant of sorts, bringing to the fore new terms and relationships that will be absent from a static taxonomy.

Whether additional value emerges from this hybridization is unclear and would benefit from actual experimentation (insert hand-waving here). The technical challenges of automatically linking these clouds to the taxonomy are also uncertain (we do agree that, in the words of Erik Duval, librarians don’t scale, don’t we?).

The top level syllogism goes something like:

• There is value in formal knowledge hierarchies that have emerged over time and which are established frameworks for existing knowledge organization
• There is value in the ad hoc semantic clouds of collaborative tagging and      linking
• The hybridization of these values should be complementary

Is anyone doing the experiment?

-----

Image: The Rotunda ceiling of the National Gallery of the Smithsonian Institution.  Photo by the author, February 26, 2006

Comments

[Agromysz]

It's interesting.
------------------------
http://www.jastrzebiagora.com.pl

[Dave Pattern]

Hi!

We've been running a small scale experiment at Huddersfield where we've been using tags to enrich the formal subject headings and visa versa:

http://www.daveyp.com/blog/index.php/archives/57/

Our test OPAC is currently offline for an update, but it should be back online later on today.

I really need to get around to summarising our findings, but my gut feeling is that there's an enormous amount of potential in this kind of dovetailing.

For example, the tagging of just a single book (about the supermarket Tesco's) has created a nice selection of relevant tags on the "Customer loyalty programs -- Great Britain -- Case studies" subject heading:

clubcards, supermarkets, home shopping, data mining

http://webcat.hud.ac.uk:4128/perl/taginfo.pl?tag=270

...not everyone will be familiar with "Customer loyalty programs", but supplement that with "clubcards" and "supermarkets" and suddenly the subject heading is put into a very familiar context.

From the other angle, someone coming across the "clubcards" tag could then be easily pointed to other items with the "Customer loyalty programs" subject heading.

With multiple items tagged, the subject headings become even more enriched:

http://webcat.hud.ac.uk:4128/perl/taginfo.pl?tag=210

I hope this is of interest!

regards
Dave

[Joe]

Your post reminds me that a group of us WebJunction folks had a conversation about with you about this exact topic (vis a vis WebJunction) over dinner a few weeks ago.

We (well at least a lot of us) are very enthusiastic about WJ as a test bed for this concept; it's the perfect context to explore the taxonomy/folksonomy nexus. And since our community is made up library folks, it should be great fun to investigate!

The theory that tags stabilize over time sounds right intuitively, and it's good to see research bearing that out. But it's those pesky corner cases, odd bits of information that don't quite fit most people's ways of categorizing but which must be managed somehow, that will really test the theory. Should be interesting to see what happens with this, on our project and elsewhere.