data blogs
joel, 11:50am 12 April 2005Some SPIRE folks and I have talked about using folksonomies to do for data sharing what they´ve done for picture and bookmark sharing. In particular, one thing that makes tagging sights like flickr and del.icio.us so powerful is that every user is free to add tags to already tagged resources. In other words, the metadata attached to a resource can evolve as the community gains new understandings of what the resource relates to. We´ve been talking for a while (mentioned it in at least one proposal) about developing a system that allows scientists to tag data according to what they find it useful for, and that attaches those tags to the data. But, in a sense, the Technorati-blogging infrastructure already provides such a system.
For example, I blogged a link to an NBII dataset, tagged it both “foodweb” and “habitat”, and now it shows up at technorati.com/tag/foodweb and technorati.com/tag/habitat. If anybody used this data, they could tag it according to what they use it for. For example, the habitat data could be tagged “colorado river toad”. All the tags of a dataset can be found by following the linkbacks. (flickr provides a more straightforward way of viewing all the tags associated with a picture.)
More generally … People have been contrasting folksonomies with the semantic web, as if they were in opposition to each other. In many cases, however, folksonomies can easily slip into the semantic web framework, playing the role currently played by taxonomies. any thoughts on this?
Ebiquity Blog
April 16th, 2005 at 8:16 am
A number of people have picked up on the idea of using the tagging concept of Technorati etc as a way of making semantic mark-up an easier and more natural process. When I first came across the idea of tagging and the way it was being used by Technorati, del.icio.us and Flickr it immediately struck me as the ‘human’ approach to mark-up for web resources.
It seems I wasn’t alone: my project supervisor pointed me in the direction of a blog entry at http://www.thespoke.net/BlogReader/SingleEntry.aspx?id=78637#id78637from one of the people at MIT working on the Haystack project (http://haystack.lcs.mit.edu). If you’re not aware of this, then I suggest you check it out - it’s still very much work in progress but shows a lot of promise for making semantic mark-up of data user-friendly.
My personal viewpoint is that ontologies and folksonomies are complementary: tags have the advantage of being freeform and easy to use, whilst ontologies allow the mark-up to be reasoned about semantically. By tagging data and then grounding those tag phrases in ontological descriptions I believe you could create a happy medium where humans can easily classify data and machines can process and reason about those classifications automatically.