Re: about representing persons (and other things) in our ontology

Le 18/11/2009 14:29, Evain, Jean-Pierre a écrit :
> Pierre Antoine, all,
> 
> Good summary of the situation but if we consider the current scope
> of MAWG which is to map between different metadata metadata
> representations, and as I tried to explain during the call. I cannot
> see how a URI/URL will point to e.g. a complex(type) description
> within an instance based on a structured schema like tva or mpeg7
> personTypes or organisationTypes.  These instances might as a whole be
> seen a linked data via the uri/url of the file location but not chunks.
> Or are you proposing to have the Xpath added to the uri/url? Is this
> feasible?

I'm not suggesting anything of the sort. I'm ony saying : *if* we have a
URI instead of, or in addition to, a more or less structured value, we
should not lose this URI in the process, to let linked-data-aware
clients use this URI the way they like. After all, this is an API for
the Web.

In many of the in-scope format, we won't have a URI; that's ok, we will
stick to the embeded description. But some of the in-scope formats are
RDF-based (and I guess some others may allow to embed a URI), so this
consideration is in the scope of the WG.

> Furthermore, does this really make sense?  This would certainly lead
> to a degree of human intervention. Is it the purpose of our action?

I agree on that, this is why I *don't* suggest we should produce a URI
and/or linked-data when it is not provided by the underlying format.

> If all the metadata to which we want to map were RDF, it would be
> simpler (also because complex structures wouldn't be expressed like in
> xml.  But this is not the case.

No, but some are. Those will provide a URI. Why lose this URI?

> I think that what we want to do is to provide simple attributes and if
> data from complex structures have to be concatenated into something
> meaningful, then we have to define the method in the API.

I entirely agree on that. The API will have to flatten structured data
into a single label. The only additional structure I propose to keep is
a URI, if given by the underlying format.

  pa