Hi, (01)
In the IDEAS Group work, we decided to use OWL/RDFS/RDF and UML as syntaxes
for our ontology. We found that we could argue forever about how expressive
one language is compared to another, but really it's the quality of the
ontology you produce that matters. We were never going to be able to
influence these languages, so the best bet was to just use them as
pragmatically as possible. (02)
We profiled UML for IDEAS - i.e. created a set of standard UML stereotypes.
We also did effectively the same thing with RDFS - created a new class for
each ontic category. IDEAS is a layered ontology - the foundation layer
describes the ontic categories that we work with. It was against these ontic
categories that we bound the UML profile and the RDFS classes. So, IDEAS
defines Individual, Type and Tuple as its three top categories, and we have
corresponding UML stereotypes and RDFS classes. This close-coupling allows
us to "draw" our ontology in UML and ensure that we can generate RDFS
specifications from the UML diagrams in a repeatable fashion. You can see
some examples of these "drawings" on http://www.ideasgroup.org/3Foundation/.
We can now also generate relational database implementations, with REST APIs
all from the same models. (03)
IDEAS is extensional and 4D, so our initial reaction was that OWL and RDFS
were clearly not expressive enough. OWL seems (at least to my simple
engineer's brain) to be intensional - it allows predicates such as "sameAs"
and "equivalentClass". The W3C documentation also talks about an individual
also being a class (see
http://www.w3.org/TR/2004/REC-owl-guide-20040210/#OwlVarieties) which is
very odd...but rather than whine endlessly about it (though we did do a bit
of that), we decided to make the languages work for us. We found that if you
just treat them as simple syntaxes, and keep the real intellectual work in
the ontology itself, life gets that bit simpler. (04)
The only snag we hit with this approach was that we didn't re-use some of
the key ontic relationships in RDFS - such as "type" and "subClassOf". This
means that we can't get maximum leverage on existing (e.g. open source)
toolsets. We therefore plan to change the RDFS binding appropriately. (05)
The second snag (we've just discovered) is in implementing programmatic
access to the ontology - call them APIs, web services, whatever. If the
service returns RDFS, it can be quite a verbose packet of data (which slows
down implementations). In addition, IDEAS places strong emphasis on
separating names from the things they describe (Chris Partridge is the
expert on this, but I believe this is based on Quine's ideas about
reference). There's no tidy way to do this using the RDFS/OWL syntaxes. We
are therefore considering a simplified XML format for programmatic access to
the ontology. So...there may well also be a vanilla XML binding to IDEAS as
well as the RDFS and UML bindings. The UML model is always the master for
the ontology (as I said, we like to draw our ontologies), and all the other
stuff is auto-generated. (06)
BTW - the other thing we found was that if you treat the RDFS/OWL as just a
syntax, you can have a first order language (e.g. OWL lite/DL) describing a
higher-order ontology (I think Bill Andersen has also talked about a similar
approach). This means you can have a reasonably useful model of the real
world, yet still keep the inference/reasoning folks happy. (07)
Cheers (08)
Ian (09)
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
Sent: 25 October 2008 15:07
To: [ontolog-forum]
Subject: Re: [ontolog-forum] Difference between XML and OWL (010)
Mike, (011)
I'm glad that you found this discussion useful. (012)
MB> I see ontology (OWL or otherwise) as fitting into one specific
> box in the Zachman Framework. (013)
Actually, all 30 boxes address different aspects of ontology.
Each of the six columns focuses on the entities found in response
to one of the six types of questions: What? How? Where? Who?
When? Why? The answers to those questions determine what has
to be represented. (014)
Each of the five rows focuses on the perspective from a person
that plays a different role: planner, owner, designer, builder,
or subcontractor. The product 6 x 5 = 30 boxes, each of which
shows one perspective by one type of person on one aspect of
an information system. (015)
Ed criticized the Zachman framework because many people have hyped
it far beyond its usefulness. And I agree with that criticism.
But I endorse the basic ideas behind the methodology, which can
be generalized much further: (016)
1. The idea of asking questions to elicit information about
aspects of a subject domain. This is a very old idea that
started with Aristotle: each of his 10 categories is an
answer to one type of question. For any give subject domain,
each answer represents one significant type of entity. (017)
2. The idea of considering an ontology from the perspectives of
different people who play different roles. For example,
consider a medical ontology with a focus on the perspectives
of a patient, a general practitioner, a nurse, specialists in
different medical fields, a pharmacist, an administrator... (018)
What I like most about the Zachman framework is its emphasis on
the many different ways of viewing a system. But my major
criticism is that it's only a beginning. The 30 different ways
are just an inadequate approximation to infinity. (019)
MB> What is unfortunate is that UML is so wedded to OO development
> that it does not extend naturally into the semantic space and so
> does not cover the whole of the first two columns. (020)
Yes. But that is also my criticism of the OWL approach. It is
also very limited, and the people who promote it don't admit (or
are not aware) that it is a very highly specialized methodology. (021)
One reason why I recommend that people consider UML, SQL, and
various approaches such as Zachman is that they shows that semantics
is involved with all aspects of meaning. OWL is a one-size-fits-all
notation that can be useful for what it can do, but the subject is
much, much broader. (022)
People have apologized for OWL by saying that it is just a beginning.
That's OK *provided that* somebody tells the OWL users that they are
only scratching the surface of semantics. (023)
MB> However, [UML] is extensible. (024)
Yes. And the idea of bringing together multiple approaches that
had already proved their usefulness separately was its greatest
strength. (025)
My major complaint about RDF and OWL is that they had not proved
their usefulness as de facto standards or just widely used tools
before they became integrated into the SemWeb. That's what I
meant by saying "May God protect us from proactive standards." (026)
John (027)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (028)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (029)
|