Re: eric and a picture of eric...

(+cc: www-archive, hoping nobody minds)


s/subclass/subproperty/, btw., since foaf:depiction is a relation.

http://swordfish.rdfweb.org/photos/2002/02/27/001698.JPG depicts eric and
nadia, just as the picture of ian horrocks and jim hendler in
http://www.cs.vu.nl/~dieter/mm/webont/ depicts them both.

One way to think about it is by having a notion of a sub-image, defined as
a region of an image. There is defintely a sub-region of 001698.JPG that
depicts Eric, and since that's the case, any images that include that
sub-region also depict him.

What we're getting into is the difficult of characterising the content of
a scene. Pictures that depict non-existent entities, people etc are a
variant of the problem...

I reckon keep foaf:depiction liberal for now, and then use types to
classify the images themselves more explicitly, or intermediate nodes that
talk explicitly about the mode of presentation within the image. Mostly
that'd be unncessary, but could be useful for the pedantic.

Dan


On Thu, 28 Feb 2002, Eric Prud'hommeaux wrote:

> On Thu, Feb 28, 2002 at 01:16:10PM -0500, Charles McCathieNevile wrote:
> >
> > right, but luckily this is RDF so we can add a new property and say it is a
> > subclass of depiction and deal with the implications later...
>
> I believe it would be dangerous to make it a subtype as it would lead
> people to the wrong conclusion, that depictsProxy implies everything
> that depicts implies. If it implies less, it should be a supertype.
>
> > On Thu, 28 Feb 2002, Libby Miller wrote:
> >
> >
> >   http://swordfish.rdfweb.org/discovery/2001/08/codepict/codepict.jsp?mbox=eric@w3.org
> >
> >   nadia took a photo of ericP; the database now contains both the photo
> >   and a depiction of her with the photo. suspect this is not a true
> >   codepiction ;)
> >
> >   Libby
> >

Received on Thursday, 28 February 2002 18:57:15 UTC