W3C home > Mailing lists > Public > public-xg-mmsem@w3.org > March 2007

Re: Comments / Suggestions for the Interoperability Deliverable

From: RaphaŽl Troncy <Raphael.Troncy@cwi.nl>
Date: Wed, 28 Mar 2007 18:35:08 +0200
Message-ID: <460A993B.D630433@cwi.nl>
To: Tobias BŁrger <tobias.buerger@deri.org>
CC: public-xg-mmsem@w3.org

Dear Tobias,

> to pick up some aspects of the discussion from the list in [1]:
>
> do you think that it would make sense to include a general section on
> how interoperability for metadata standards/vocabularies could be
> achieved using semantic technologies?

Perhaps.

> I can imagine five options:
> (1) Through standards + consensus building among the community
> (2) Building upon a core or foundational ontology like DOLCE, SUMO or
> others and align voacbularies to them
> (3) Through mediation between vocabularies based on a central vocabulary
> (4) Through integration / harmonisation of vocabularies
> (5) Through mapping or matching between concepts

Well, then you definitively need to expand these 5 bullets items, because I
don't think I agree :-)

First, a general remark: your assumption is that we try to integrate various
standards (vocabularies?) that strongly overlap, which is not the case. I
believe that all these standards are made for different purposes. They
contain sometimes overlapping notions, but the point of the various use cases
is to show what is the added values of using together these individual
standards. The music use case is a good example for that: using a FOAF
description, ID3 tags, low level features you can extract from the signal,
external knowledge (wikipedia, amazone, etc.) allow you to make new service !
The research question is then how to you can use these standards together
(not merging/mapping them) and most of the use cases develop the idea
thatgoing through the semantics of OWL/RDF is a nice path ...

Second, I can understand (1). I don't see why (2) makes the situation better.
You will need almost 3 mappings (link your standard A to the upper ontology,
your standard B to the upper ontology ... but still need to establish mapping
between A and B). The group of Frank v. Harmelen has made nice experiments in
the medical domain with such an approach. I can understand (3) only if the
two vocabularies have a strong overlap. What means (4) ? How it is different
from (1) ? For (5), do you think about an automatic approach ?

> I think the one that the XG goes for is (4),

Again, this is not true. In the case of the Music Use Case, you just see how
many music ontologies there are, and you pick the best one, MO. Then you
adapt what you can extract (ID3 tags) to this vocabulary ... so it is rather
(3) in your category than (4)

> But perhaps this part would be to theoretically as we want to
> demonstrate practical things.

It is certainly a good idea, but I encourage you first to develop these ideas
... feel free to create a wiki page if you think it is necessary ...

Best regards.

    RaphaŽl

--
RaphaŽl Troncy
CWI (Centre for Mathematics and Computer Science),
Kruislaan 413, 1098 SJ Amsterdam, The Netherlands
e-mail: raphael.troncy@cwi.nl & raphael.troncy@gmail.com
Tel: +31 (0)20 - 592 4093
Fax: +31 (0)20 - 592 4312
Web: http://www.cwi.nl/~troncy/
Received on Wednesday, 28 March 2007 16:35:34 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:21:21 GMT