Wednesday, January 26, 2005

Enterprise Distributed Categorisation: I Get on the Folksonomy Bandwagon

I haven't posted for a while, but like many out there, I've been thinking about services like delicious which can offer the benefit of developing a folksonomy (a.k.a. free-tagging, ethno-classification, distributed tagging) within a corporate environment. On the wicky wacky world wide web, the benefits seem obvious - flat metadata categories which help you to easily identify useful information (power laws usually meaning that only a restricted set of tags become wholeheartedly utilitarian) from the vast quantity (and quality) that is available. Tags let you let others do your spadework for you.
I have been wondering whether in an enterprise environment, a similar folksonomy approach could be used for the rapid (and flexible) development of iterations of a classification system. Taxonomy projects always seem to be tagged as inflexible, long-winded and time-consuming - couldn't folksonomies get rid of some of these disbenefits?
Owing to my own priorities, I've been thinking about this for intranet metadata and search integration. It strikes me that the flat categorisation of a folksonomy could be clustered and analysed using the same techniques as are used for a card-sort, and that key, popular categories coming out of a distributed categorisation exercise could be used to form the structure of an intitial taxonomy.
I suppose my suggestion would be to allow a "tag this page" section in our CMS (which at present has very limited metadata functionality) which allows for multiple space delimited tags in the same fashion as delicious. This functionality would be explained to current content authors, who could tag new and existing content in an unrestricted manner for a set period of time. At the cut-off point, the tags would be analysed for popularity and similarity (synonyms, misspellings etc.), and a reduced set then used for an open (card) sort. The results of this could be analysed and transferred into a first iteration of a taxonomy, along set rules.
The taxonomy could then be exposed to the search engine and CMS - authors would now tag their content using the taxonomy, and, for ongoing maintenance would still have the option for creating a new tag. New tags could be processed by the taxonmomist and added/amended to the overall classification as necessary. I'm not sure whether this would work in practice, but it might be the outline of using rapid, distributed categorisation to move to a stable classification, and it would certainly reflect a consensus opinion...
[NB: Clay Shirky (I think) mentioned that delicious tags can be moved into hierarchies (xml>xslt, for instance or mammal>dog), although I can't find out how to do this, or any reference on delicious as to how to do so - otherwise, I might possibly suggest allowing users to set their own "corporate" hierarchies and then clustering them in this way rather than just going for the clustered approach.]
Has anyone implemented anything similar or looked at the pros and cons of such an approach? I know Adam Mathes Folksonomies - Cooperative Classification and Communication Through Shared Metadata touches on some of the issues, as does Ulises Ali Mejias A del.icio.us study -
Bookmark, Classify and Share: A mini-ethnography of social practices in a distributed classification community
, and the always helpful Lee Bryant at Headshift with Can social tagging overcome barriers to content classification?
[NB(2): check out extisp.icio.us while you are at it - if you haven't already - a nice visualisation of your own delicious personal information cloud.]
UPDATE:
Should have checked Louis Rosenfeld's bloug out first - some of this is discussed in a post there, makes for interesting reading.

No comments: