W3C home > Mailing lists > Public > public-media-annotation@w3.org > September 2010

Re: Using the Ontology for Media Resources in the Semantic Web

From: Yves Raimond <yves.raimond@bbc.co.uk>
Date: Tue, 14 Sep 2010 12:37:43 +0100
Message-ID: <AANLkTim83j6kYced9b5NfBFGT5Cv7WqRcukOaWU9+iCM@mail.gmail.com>
To: RaphaŽl Troncy <raphael.troncy@eurecom.fr>
Cc: Media Annotation <public-media-annotation@w3.org>
Hello Raphael!

>
> [glad you jump on this thread ;-)]
>
>> Sorry to jump on that particular example, but I don't think I've seen
>> it mentioned before. I can understand why such a modelling can be
>> appealing, but it is also potentially harmful. In your particular
>> example, if I loaded all the quad data in any triple store and asked
>> whether the audience is applauding (ASK WHERE
>> {<http://dbpedia.org/resource/Audience>
>> ex:humanActivity<http://dbpedia.org/resource/Applause>}), I would get
>> "true" as a result, which can get quite confusing! However, it is only
>> true in the context of that media fragment.
>
> I agree and I voice the same concerns.
>
>> I am a bit concerned by
>> the fact that this relies too much on some Named Graphs semantics that
>> hasn't been properly defined yet. Moreover, as you mention, there is
>> no way of serialising all this information in a single document...
>
> Yes, and we would not suggest to use Named Graphs before there is stable and
> consensual document that explains their syntax and semantics, something we
> don't have yet.
>
>> Did you consider event-based annotations (which are properly
>> understood and already deployed, e.g. by the BBC) instead?
>
> Yes, naturally, but this is not the issue we are discussing here.
>
>> If it was
>> dropped in favor of a named graph approach, is there a document or an
>> email thread detailing the reasons why?
>
> Hum, I think you misunderstood the purpose of my example. Forget about the
> named graphs. The current Ontology for Media resources allows to attach only
> *simple* annotations to media or part of media. By simple, we meant either a
> plain literal (see this a free text annotation) or a URI (see this a single
> concept from a controlled vocabulary). The issue is how could we attach
> *complex* annotations to media or part of media? By complex, we meant a RDF
> graph of arbitrary complexity that could describe precisely a scene.
>

I am not sure I understand, indeed - what is the status of that the
examples you mentioned above, then? Is that something that is likely
to make it in the final document? If you want to tackle the "complex
media annotation" scenario, and want to stick to your ma:relation
framework, then you will have to use Named Graphs in the way you
described it above, which we are apparently both concerned about?
Pointing to a SKOS concept doesn't cause any issues, but pointing to a
Named Graphs relies on some semantics that isn't quite there yet.

Therefore, I am guessing there are only two possible outcomes 1)
Dropping the "complex" annotations from the scope of the final
document or 2) Move to another scheme than the ma:relation one for
complex annotations, which was what I pointed at in my previous email?

Kind regards,
Yves

> †RaphaŽl
>
> --
> RaphaŽl Troncy
> EURECOM, Multimedia Communications Department
> 2229, route des CrÍtes, 06560 Sophia Antipolis, France.
> e-mail: raphael.troncy@eurecom.fr & raphael.troncy@gmail.com
> Tel: +33 (0)4 - 9300 8242
> Fax: +33 (0)4 - 9000 8200
> Web: http://www.eurecom.fr/~troncy/
>
>
>
Received on Tuesday, 14 September 2010 11:38:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 14 September 2010 11:38:19 GMT