[XESAM] Ontology sketch. Feedback needed. This time with attachment.

[Mikkel Kamstrup Erlandsen]

2007/6/1, Antoni Mylka <antoni.mylka at dfki.uni-kl.de>:
>
> Mikkel Kamstrup Erlandsen pisze:
> >
> >  * contributor (DC)
> >  * creator (DC)
> >  * description (DC)
> >  * language (DC)
> >  * publisher (DC)
> >  * subject (DC)
> >  * title (DC)
> >  * license (an extensible vocabulary with predefined values GPL, LPGL,
> > MIT etc)
> >  * uri
> >  * category (a controled vocabulary that maps to the cats in the
> > extended onto)
>
> This particular one may be tricky. This "controlled" vocabulary should
> be extensible. I think it's not a good idea to have to agree on all
> categories at the very beginning. I would rather go for the notion of
> type. A resource may be a file, email whatever. This field would point
> at a type of a resource. (In NIE it would be expressed with rdf:type
> pointing to an URI of a class (like Document, File or whatever), in the
> simplified XESAM "language" it could point at some type identifier).


What you call "type" is what we have called "Source" so far I think. The
category of an object is fx Contact, Video, Image or the likes. The source
specifies how we obtained this object fx EmailAttachment, File, Archive...

I agree that the vocabulary should be extensible, but it should still have
predefined values.

Evgeny pointed out on IRC that the above proposal was to simple to be
usable. Things like Contacts and Email does not really fit into it in a
natural way. He had a proposal coming up with a bunch of extra fields that
should accomodate this if I'm not mistaken. Evgeny?

The question if you want to have a simple list of types, or an
> inheritance tree of types remains open.


It will be a tree. At least that's what we've settled on at the 2007-05-15
meeting.

>  * mime
> >  * creationDate
> >  * modificationDate
> >
> >
> > With this simple onto you can actually do quite a bit of nifty stuff.
> > With this in place it might also be easier to agree on extended ontology
> > as Antoni already suggested.
> >
>
> I'm all for. With this set (or something similar) it should be possible
> to agree on a 1 to 1 mapping between NIE-core and XESAM-core.
>
> The properties from the extended part of the ontology could be mapped
> via the subproperty relations to the ones from core (as much as
> possible). A simple mechanism could realise the basic inference rule.
>
> IF a prop b AND prop subPropertyOf prop2 THEN a prop2 b
>
> This shouldn't be too difficult and time-consuming, even in systems
> where performance is a priority.


Right, I think most indexers can (and does) support this already. Anyway it
has always been an unspoken premise of the field ontology that we used
inheritance.

Is it possible that you can change you vocabulary when you talk xesam? :-)
We use "field"[1] by convention and you call them "properties" (which
ofcourse is the correct rdf lingo). It could reduce the possibilities of
confusion a bit :-)

Cheers,
Mikkel

[1]: This seems to be the standard term used in most indexers where a field
value is always a literal (as it is in xesam)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freedesktop.org/archives/xdg/attachments/20070601/483e54ec/attachment.htm