Thursday, July 28, 2005

Reversed folksonomy

Firstly, some background. A colleague/friend discussed their phd with me a while back, which was around images and how to catalogue them. The conclusion - as I got it - was the importance of determining the goals of the user accessing the image (at that instance in time) driving the cataloguers thought process when describing the image. Consider a picture of a bridge - the metadata about that bridge that is required, and the terms used to access it, vary greatly between an architect studying tensions, to a historian interested in who built it, to a child who wants "any old bridge" for a school project. I've thought a lot about this over the years, mainly from the perspective of how to profile the user searching in more innovative ways so the right images (or any type of information) are retrieved to meet the users current goal.

Folksonomies begin to solve one half of this problem - we have a range of users with different goals/views of some "content" each providing their own description from their viewpoint. In theory, if their viewpoint can be mapped to a user class, and other users are also known to be within that class, then we should be able to improve relevance of hits greatly. This falls down when you have a user who's doing some searching for their kids homework, but its still something I want to keep picking at as its an interesting diversion on these wet summer days.

But this ain't what I'm supposed to be posting about - this is!!! Go on, take a click!

Now, whats interesting about this site is that the folksonomy concept is back-to-front. "We provide the term, you provide the content" seems to be the philosophy. I like that! And it got me thinking - its something we can apply to the content in our OPACs just as easily. Put a word up each day, let users link works/records to it. It may take a long time, but eventually an interesting folksonomy would appear (its a bit like asking users to flag "books like this one").

But would it take a long time? I start thinking Bigfoot and Silkworm and the whole ethos of leveraging the network effect of libraries. Say each Talis library is provided a different word each day on their OPAC from a central store of, hmmmm, 50,000 words. Users could link whatever book or dvd or cd to that word they felt "matched". Next day, different word. Talis harvest this centrally. So, thats 100ish libraries, 364 days a year (go on, you can have xmas off!). Each word would have 7 different libraries, of many different user types, linking mutliple works/records to it. In a year, we have a complex folksonomy of all Talis library holdings. And, to a limited degree, these terms are "controlled".

Hmmm - think I'm now moving into "pie in the sky" stuff so I'll stop, and have a lie down. And maybe a beer...

No comments: