David Weinberger led the BOF on tagging, folksonomies, usage, usability, scalability and spam.
The discussion was interesting, though nothing new came out of it per se.
David summarized well the situation by saying that the size of the tagging ecosystem, either counted as numbers of users, tags or documents is tiny looking in an Internet scale perspective. And the day tags become really deployed, tagspam will be unleashed to scale as well.
As I commented during the BOF, tags worked great to "rally" people or documents around well identified "meeting" places. All ETech-related pictures on Flickr or Buzznet will be tagged "etech", and so will most of blog posts. Tags also work great in the personal bookmarking context, thanks to the sheer simplicity of del.icio.us.
Beyond these two cases, tag consumers will need to start incorporating functionality like simple stemming of tags, essentially creating a more organized framework where semantic ambiguity is avoided as much as possible: New-York=NY=NYC, venture capital=vc=venture financing, etc. The vc tag is typically a good example of a "bad" tag: too short and confusing. Applications allowing users to tag things will also need to help desambiguify tags (like vc or ny) through lightweight UI tricks (lightweight is key, otherwise nobody will bother tagging at all).
I was chatting to a fellow ETech-er this morning, who is developing a service featuring tags in a resonably well defined semantic context. I recommended to offer a set of defined keywords/tags that people could choose from, and then allow the addition of free form tags. This might work in his (semantically defined) context.
Comments