Re: geo example

* Bernard Vatant <bernard.vatant@mondeca.com> [2004-04-27 17:26+0200]
> 
> Dan
> 
> > http://dmoz.org/Regional/Europe/United_Kingdom/England/Bristol/ shows a
> > topic from open directory.
> 
> Would you call this resource a "subject indicator", to use topic maps dialect?

I don't know enough about the detailed mysteries of topicmaps there, but 
certainly this seems close. It is, colloqially, a subject or topic, 
thought of a thing-in-itself, and distinct from both the thing(s) that
topic represents, and from documents and data representing either the
topic or the, er, subject of the topic.

I give up! this terminology is too overloaded to say anything ;)

Put another way... the "etc/England/Bristol/" node in the DMoz topic
graph (and its equiv in SKOS) are things that represent "the concept/idea of the
place Bristol", rather than representing Bristol directly. There is an 
additional level of indirection compared to 'raw' RDF (and OWL).


> 
> > I can imagine how this works in SKOS. And
> > how we might represent Bristol as a place in RDF. For discussion: how to
> > represent (to machines) the connection between these two worlds.
> >
> > We could easily redo the open directory (dmoz, see
> > http://dmoz.org/rdf.html)
> > in SKOS.
> 
> As an on'n off ODP editor [1] I had proposed internally to have a look at integration of
> ODP in the SW framework.
> Not with a great feedback, actually. Remember that RDF dump of dmoz uses an old RDF
> format.

Yep, it is pretty scruffy, was backed by very early RDF code from Guha
and friends when at Netscape/Mozilla. There are scripts around to,
rather minimally, fix it up (namespace URIs, character encoding, etc).
Those scripts could probably be adapted to emit SKOS rather easily. 

> > And we can doubtless find some RDF representation of Bristol as
> > a member of a class "City", with location info, population, etc...
> >
> > My hypothesis is that a new property in SKOS, skos:conceptualizes,
> > could be used to relate bristol-the-skos-concept to
> > bristol-the-thing-in-the-world.
> 
>  ... looks to me indeed very close to the notion of "subjet indicator".

I suspect so too. Topicmap people have often said that RDF confuses
things with their representations. I believe RDF allows such modelling 
errors to be made, but also that it allows perfect clarity. TMs, on my 
understanding, try to have more built-in facilities for preserving 
those distinctions.

cheers,

Dan

Received on Tuesday, 27 April 2004 11:55:30 UTC