SXSWi panel summary: beyond folksonomies
Saturday 10:00-11:00 Beyond Folksonomies.
The need to categorize and search online information has led in two directions: top-down taxonomy (e.g. hierarchical structures) and ground-up folksonomy (e.g. tagging). One panelist (I believe it was David Swedlow) described folksonomies as the pidgeon language an adult might learn when immersed in a foreign culture; an ideosyncratic shorthand for the information we produce and consume online.
The primary issue with folksonomies is that each person can produce their own shorthand for slicing and dicing information. While I might tag all SXSW data as "sxswi" (because it all comes from the interactive conference) someone else might tag theirs as "sxsw2006" to differentiate it from previous years. Such variations can have greater meaning to each individual, but hinder aggregate-level value creation and community building. Uncertainty also increases the costs and efforts associated with using folksonomies, so less engaged users (i.e. "the masses") will be unlikely to adopt them. Until these hurdles are addressed, folksonomies will have limited impact on markets and societies beyond the technorati.
The panel struggled to find a solution to this problem, and ended up debating the merits of automated tagging. While pure automation resembles a taxonomy, with all the same weaknesses brought on by rigidity, some form of implicit tagging allows users to leverage work that has already been done while retaining the flexibility of context and interpretation. Google searches represent an example of implicit tagging, in which the user receives thousands of "relevant" items, but the search algorithm has already determined which results are likely to be most meaningful based on the amount of activity (links) attached to each. This type of implicit tagging could serve as the lynchpin between organic folksonomies and more structural (and useful) taxonomies.
Some links discussed during the session:
POPFile
AttentionTrust
The need to categorize and search online information has led in two directions: top-down taxonomy (e.g. hierarchical structures) and ground-up folksonomy (e.g. tagging). One panelist (I believe it was David Swedlow) described folksonomies as the pidgeon language an adult might learn when immersed in a foreign culture; an ideosyncratic shorthand for the information we produce and consume online.
The primary issue with folksonomies is that each person can produce their own shorthand for slicing and dicing information. While I might tag all SXSW data as "sxswi" (because it all comes from the interactive conference) someone else might tag theirs as "sxsw2006" to differentiate it from previous years. Such variations can have greater meaning to each individual, but hinder aggregate-level value creation and community building. Uncertainty also increases the costs and efforts associated with using folksonomies, so less engaged users (i.e. "the masses") will be unlikely to adopt them. Until these hurdles are addressed, folksonomies will have limited impact on markets and societies beyond the technorati.
The panel struggled to find a solution to this problem, and ended up debating the merits of automated tagging. While pure automation resembles a taxonomy, with all the same weaknesses brought on by rigidity, some form of implicit tagging allows users to leverage work that has already been done while retaining the flexibility of context and interpretation. Google searches represent an example of implicit tagging, in which the user receives thousands of "relevant" items, but the search algorithm has already determined which results are likely to be most meaningful based on the amount of activity (links) attached to each. This type of implicit tagging could serve as the lynchpin between organic folksonomies and more structural (and useful) taxonomies.
Some links discussed during the session:
POPFile
AttentionTrust
Sunday, March 12, 2006
posted by wae at 10:35 AM
3 Comments:
Do you have time to explain a little more how POPFile was mentioned? I wrote POPFile and am currently working on code to do automatic classification of web pages, and I'm very interested in hearing about the ideas at this conference.
John.
Hi John,
POPFile was mentioned by one of the panelists (J Wynia?) as a successful implementation of automated filtering. In fact, I believe the example was that POPFile was able to catch especially clever spam that the user even initially thought might have been an actual message. Overall the mention was very positive, and intended (I believe) to help people overcome an aversion to automated classification as overly rigid. It sounds like your extension into webpage classification is in line with what the panel was steering towards.
Interesting, thanks. Looks like you are correct about it being J Wynia. There's an interesting article on his blog about how he reads RSS which mentions POPFile:
http://www.wynia.org/wordpress/2005/10/04/why-im-loving-imap-and-thunderbird-as-my-rss/#more-336
John.
Post a Comment
<< Home