Password
Title
Content
Elias Torres has a good example of where folksonomies break down, with the tag " queso" at del.icio.us being used for both " things involving cheese" and " a Semantic Web-enhanced Atom Atom Protocol server ". Best I had so far was the difference between "chat"@lang=en and "chat"@lang=fr, but without any real-world data.
But I'm pretty sure the problem isn't inferring semantics from tags per se, more like inferring the wrong semantics. There's a set of docs on the topic of cheese, there's a set of docs on the topic of the Queso app. The del.icio.us tag covers the union of these sets of resources (they probably aren't quite disjoint, Elias' post mentions both for a start).
I think I got the semantics messed up myself in a different way recently, mixing up classes and instances. It's probably best to do the semantics with a bit of indirection, SKOS seems in the frame. I wonder if this works: there is a skos:Concept associated with cheese, there is a skos:Concept associated with the Queso app. Each of these has a bunch of skos:altLabels, both bunches including "queso". A subject identifier for each might be Queso's home page, the Wikipedia entry for cheese/foodstuff. Both of these concepts are narrower than a third concept, one with a subject identifier of http://del.icio.us/tag/queso.
Dunno, not sure about the exact terms, might still have got wrong. Bringing in the Tag Ontology would probably help, also it'd be cool to use the new WordNet in RDF/OWL. Whatever, although there are obvious uses for subject-oriented tagging, I suspect the fruit hangs lower around personal tags & ontologies, along the lines of http://del.icio.us/danja/queso etc.
Â
Tags
Date