Pete,

what would happen if you had "dctag:subject=UTube funk music soul" so  
that the definition of dctag was closely related to but not  
purporting to be dc:subject?  (I am assuming dctag was defined in a  
namespace.)

Liddy

On 03/11/2006, at 12:15 AM, Pete Johnston wrote:

>
> Jonathan said:
>
>> On a different note, do RDF and Topic Maps fit into this
>> discussion anywhere?  Triples are a specialised sort of
>> metadata.  But I don't know if you would think of them as tags.
>
> RDF and Topic Maps are both generic data models/frameworks for
> representing information - information about anything. (Well,  
> that's the
> case for RDF; I'm less familiar with TM but I think that is the  
> case for
> TM, too).
>
> So, using RDF, various aspects of "social tagging" could be  
> represented
> using the RDF data model: the information that is created by a  
> "tagging"
> operation, the process of tagging itself, whatever part of "the social
> tagging domain" you decide you want to represent.
>
> And many of the Web-based social tagging systems do expose the data  
> they
> hold (or some subset of it) using the RDF model (typically using RSS).
> e.g. del.icio.us exposes an RSS/RDF feed of my recent entries
> http://del.icio.us/rss/PeteJ
>
> And pushing that through an RDF parser, I get
>
> http://librdf.org/parse?&uri=http%3A%2F%2Fdel.icio.us%2Frss%2FPeteJ
>
> As an aside, I might quibble about some of the conventions used in  
> that
> del.icio.us RDF representation. The triple
>
> <http://www.youtube.com/watch?v=sIsYDWD1lgA>
> <http://purl.org/dc/elements/1.1/creator> "PeteJ" .
>
> asserts that that YouTube resource was created by "PeteJ" (which -
> leaving aside the question of whether a string can create anything and
> working on the basis that we are talking about an agent with the
> name/label "PeteJ" - I don't think is the case. I didn't create that
> video.)
>
> And perhaps more interestingly, I think the use of the dc:subject
> property
>
> <http://www.youtube.com/watch?v=sIsYDWD1lgA>
> <http://purl.org/dc/elements/1.1/subject> "YouTube funk music soul" .
>
> is arguably something of a stretch. DCMI describes the dc:subject
> property as "The topic of the content of the resource."
>
> Leaving aside what it means to concatenate the several tag literals  
> into
> a single literal, I don't think tagging is limited to indicating
> has-topic/"about-ness" relationships. Certainly, I use tags in
> del.icio.us to indicate all sorts of relationships. Taking that  
> example
> there, my "YouTube" tag is an indicator of the
> publisher/distributor/host of the resource, not the "topic" of the
> resource. It's not about "YouTube". The other tags in that example  
> are,
> arguably, closer to "topics", but still, I'm not sure that video is
> "about" soul or funk - my intent was really to indicate "genres".
> Elsewhere I use tags to indicate the places where events take  
> place, the
> format of a representation, or even some sort of "status" indicator
> (like "current", "to-read" etc), and of course del.icio.us itself
> defines conventions like the "for:user" tag.
>
> So I think approaches which equate tagging to some sort of informal
> subject classification only tell part of the story: it may be true  
> that
> in some systems tagging is limited to indications of
> subject/topic/"about-ness", but in the general case that is not the
> case.
>
> (Oh, and that's a great video, BTW.)
>
> There has been some interesting work on representing various facets of
> social tagging using RDF. See e.g. Richard Newman's tag ontology
>
> http://www.holygoat.co.uk/projects/tags/
>
> and Dave Beckett's XTech 2006 paper
>
> http://xtech06.usefulinc.com/schedule/paper/135
>
> These approaches (I think) seek to model rather more of "domain" than
> most social tagging systems currently expose, at least in
> machine-processable forms - particularly the idea that a "tagging"  
> takes
> place in a context: the existence of a relationship between a resource
> and a tag is asserted by some agent at some point in time.
>
> Pete
> ---
> Pete Johnston
> Technical Researcher, Eduserv Foundation
> Web: http://www.eduserv.org.uk/foundation/people/petejohnston/
> Weblog: http://efoundations.typepad.com/efoundations/
> Email: [log in to unmask]
> Tel: +44 (0)1225 474323