[Ontology] UCDs vs ontologies?

Thu Jun 2 02:33:46 PDT 2005

Hello all

I've been more or less silently lurking at IVOA work since 2002 or such, and will be happy
to bring some input, bandwidth permitting, to IVOA "ontological effort" - so to speak :)

I had a look at the first cut at
http://wiki.eurovotech.org/pub/VOTech/VoEventOntology/VOEventOntology_v0.1.owl

First remarks, more to come certainly. Please take those as "thinking aloud" at that
point.

- I would suggest the choice of a "proper" IVOA namespace, not the default Protege one
http://www.owl-ontologies.com/unnamed.owl# - which kind of sucks.

Suggestion : http://www.ivoa.net/voevent#
(go under Protege "Metadata" tab to change that). I will use vo: as a shortcut for this
namespace hereafter

- Concerning vo:ucd. Since UCD vocabulary, as I understand, is going to be a controlled
one, should be good to have a class vo:UCD of which instances will be those UCD
descriptors, whatever those will look like at the end of the day. BTW have UCD folks
consider the SKOS [1] format for RDF expression of UCD vocabulary?
Then the current vo:ucd property would shift from owl:DatatypeProperty to
owl:ObjectProperty, and renamed certainly something like vo:hasUCD, with range vo:UCD

- Same remarks for vo:unit. There are a few ontologies of units around that might be worth
considering, like [2]

- The other way round, I'm puzzled by Classes like vo:PublisherID and vo:EventID. Why not
simply have some identifying DatatypeProperty on vo:Publisher and vo:VOEvent? Actually in
RDF, you do not need in theory any specific identifying property, since the URI is *the*
identifier for all classes and instances ... but things may be trickier (see [3] for more
about identity issues).

- I'm not sure about semantics of classes like vo:What vo:WhenWhere and vo:How. I guess
those look like abstract classes needing specific subclasses. What about adding vo:Who?
Maybe they could be gathered under a common superclass like vo:W5 (the 5 W's).

- I agree that declaration of disjoint classes is worth thinking about. Those declarations
are needed when the ontological general framework allows multiple instantiations. In a
scientific framework like IVOA, I'm not sure this is likely to happen. There is in fact
often an implicit assumption in many ontology-driven systems that direct subclasses of
"Thing" are disjoint even if not declared so ...

- Bottom line, more "Big Picture" : Should be good to think about what is VO-specific and
what could be generic to any scientific framework. I'm currently working for a big project
in the framework of Bio-Medical Science [4], I'm aware of similar efforts in Environmental
and Earth Sciences too [5] and there are certainly more ongoing. I think
cross-pollinization of those projects, avoiding overlap and redundancy of efforts, would
be a great thing. Maybe it's time to overcome the trend towards over-specialization
leading to multiplication of domain languages, and de facto as many ontologies as various
science domains. There might be hopefully still some common ontological consensus across
the scientific community at large, on notions such as System, Object, Observation, Event,
Phenomenum, Experiment, Protocol, Model, Hypothesis, Space-Time, Units ... I would be very
happy to participate in any effort to develop such a generic "science ontology" with
experts from various domains. Volunteers, step forward :))

Thanks for your attention.

Bernard

[1] http://www.w3.org/2004/02/skos/
[2] http://sweet.jpl.nasa.gov/ontology/units.owl
[3] http://universimmedia.blogspot.com
[4] http://europa.eu.int/comm/research/fp6/p1/innovative-medicines/index_en.html
[5] http://esto.nasa.gov/conferences/estc2004/papers/a5p1.pdf

**********************************************************************************

Bernard Vatant
Senior Consultant
Knowledge Engineering
bernard.vatant at mondeca.com

"Making Sense of Content" :  http://www.mondeca.com
"Everything is a Subject" :  http://universimmedia.blogspot.com

**********************************************************************************

> -----Message d'origine-----
> De : owner-semantics at eso.org [mailto:owner-semantics at eso.org]De la part
> de Tony Linde
> Envoye : mercredi 1 juin 2005 21:21
> A : Semantics
> Objet : FW: [Ontology] UCDs vs ontologies?
>
>
> ... (double-posted)
>
> Hi Ed,
>
> Count me in too. In fact, you can count a whole sub-group of the VOTech
> project, DS5. Check out:
> http://wiki.eurovotech.org/bin/view/VOTech/ResourceDiscovery and associated.
>
> And you don't need a new group, there already is one, the semantics
> workgroup:
>
>   semantics at ivoa.net
>   archive at: http://ivoa.net/forum/semantics/
>
> It was set up after discussions between variouspeople a couple of years ago
> but died out.
>
> I'll repost this message there (since I *HATE* cross-posting) and suggest we
> all adjourn to the semantics forum for ontology-like discussions.
>
> Cheers,
> Tony.
>
> > -----Original Message-----
> > From: owner-dm at eso.org [mailto:owner-dm at eso.org] On Behalf Of Ed Shaya
> > Sent: 01 June 2005 19:42
> > To: Elizabeth Auden
> > Cc: Rob Seaman; ucd-sci at ivoa.net; ucd at ivoa.net; ucd-tech at ivoa.net;
> > Data Model IVOA List
> > Subject: Re: [Ontology] UCDs vs ontologies?
> >
> > Elizabeth,
> >
> >     Hurray.  Another ontology fan.  I have taken the liberty of using
> > OwlDoc in Protege to create a browser readable version of your
> > ontology.  It is at
> >
> > http://archive.astro.umd.edu/ont/Documentation/VOEVENT/index.html
> >
> > Would you mind if I integrate what you have into the more general
> > ontology that I have been working on?
> > http://archive.astro.umd.edu/ont/Documentation/index.html
> > (And yes, Rob it does tend to get big, but one can trim it at any
> > level.  I rather see UCDs as a form of topic maps which is quite a
> > close relative of Ontology.  I prefer ontology because I believe one
> > can do more rigorous reasoning and query with it, but many prefer
> > topic maps because they are more loose and easy. )
> >
> > Now that there are two of us interested in Ontology we can form a
> > group.  It has been suggested to me that Ontology discussions should
> > reside in the data model group with "[Ontology]" in the subject Re:.
> > So, I am ccing there too and will take the rest of this discussion
> > there.  I hope you are tuned in Elizabeth.
> >
> > Ed
> >
> >
> >
> >
> > Elizabeth Auden wrote:
> >
> > >
> > > Incidentally, I've posted a first go at a VOEvent ontology (OWL-DL
> > > format) on the VOTech wiki at
> > > http://wiki.eurovotech.org/bin/view/VOTech/VoEventOntology. Any
> > > comments on the structure, concepts, and coverage of this
> > v0.000000001
> > > ontology would be appreciated.
> > >
> > > cheers,
> > > Elizabeth
> > >
> >
> >
>