[Top] [All Lists]

Re: [ontolog-forum] Ontology vs OWL implementation

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ed Barkmeyer <edbark@xxxxxxxx>
Date: Fri, 02 May 2008 11:40:35 -0400
Message-id: <481B35F3.5070602@xxxxxxxx>
Cati Martínez wrote:    (01)

> I'm new in the Ontology world, and maybe it has been already
> discussed, I'm asking me the question if everything implemented in the
> OWL language can be considered an Ontology. I guess that it's not so,
> but it is difficult for me to say when we can say that it is or not.    (02)

I have yet to see a definition of "ontology" that everyone who thinks 
s/he is developing one, or tools for one, agrees with.  If we take the 
loose definition: "a set of terms and definitions using a 
machine-interpretable language with a formally defined semantics" or the 
even looser definition: "any formal structure of terms and definitions", 
one can argue that any OWL model is an "ontology" because it is a 
representation of a set of related terms using "a machine-interpretable 
language with a formally defined semantics".    (03)

My personal preference is a much stronger definition:  A formal language 
that provides a means of definition of a term using a syntax with a 
well-defined semantics that supports (automated) reasoning is an 
"ontology language".  A model captured in that language is an "ontology" 
if and only if most of the terms it introduces are defined in the model 
using the definition mechanisms of the formal language (possibly 
including references to terms that are formally defined elsewhere).    (04)

That definition is fairly strong, and would tend to eliminate most OWL 
models, and many research projects that use the "ontology" buzzword. 
OWL provides two mechanisms for introducing a term/concept, as a 
"primitive" term that has only a natural language definition, or as a 
"defined" term that has an OWL definition.  As Natasha Noy recently 
observed, there are probably 5 people in the world who have actually 
published "real ontologies" expressed in OWL. All the others are just 
information models -- all, or almost all, the classes they define are 
primitive.  And with a model like that, a DL reasoner can do almost nothing.    (05)

> I'm modelling with OWL some information structure, so OWL is used to
> define the components and relations to these components that compose
> this concrete information structure. Could it be considered an
> Ontology, or only a set of constraints on a data structure?    (06)

I personally don't see a problem with an ontology whose domain of 
discourse is data structures.  That, to me, is not the concern.  The 
issue is whether your model actually contains formal definitions.    (07)

I have argued in this forum that OWL is the next-generation information 
modeling language, and I applaud its use for that purpose.  But I 
distinguish an "information model" from an "ontology" by the language in 
which the definition of the terms is written.  In information models, 
most or all of the definitions are written in natural language.  At best 
an information model expresses some "necessary characteristics" of a 
term, but is silent on the "sufficient characteristics".    (08)

Now, given a small set of fundamental "undefined terms", a first-order 
language like CLIF is almost always able to express the necessary and 
sufficient characteristics that would define a term.  And RDF is also 
relatively competent in that regard.  But OWL/DL is very limited in its 
ability to construct such expressions.  Ultimately it depends on 
definitions that are intersections or unions or depend on simple values 
of simple properties.  Sometimes that can't be done at all, and 
sometimes it can only be done with a significant amount of ingenuity 
(i.e., knowledge engineering).  The latter often obscures the intent to 
the human reader, but it enables the DL engine to do meaningful 
reasoning with it.  So the plain fact is that, in many cases, OWL/DL is 
not an easy language to use for a strong ontology.  And the purpose of a 
"strong" ontology expressed in OWL/DL is *primarily* to enable reasoning 
with a DL engine.  (In a sense, that makes it what OMG calls a 
"platform-class-specific model" -- a model made for implementation by a 
particular kind of software technology.)    (09)

So, if your purpose is to use OWL as an information modeling language, 
please do.  OWL has almost everything you need for that purpose, and it 
is better grounded than any other information modeling language (except 
possibly for ORM).  Whether you can call the result an "ontology" 
depends on your circle of friends, and on the necessity of using the 
buzzword to obtain management approval or funding for otherwise useful work.    (010)

[This is only my view, and it is worth exactly what you paid for it. 
You will doubtless get a lot of other opinions, some of them rather 
better educated. ;-)]    (011)

-Ed Barkmeyer    (012)

P.S. For those of you involved in the summit, consider this a stake in 
the ground:  Use of "formal definition" is one of the primary factors 
(and perhaps the most important) in determining the "quality" of an 
ontology.    (013)

Edward J. Barkmeyer                        Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263                FAX: +1 301-975-4694    (014)

"The opinions expressed above do not reflect consensus of NIST,
  and have not been reviewed by any Government authority."    (015)

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    (016)

<Prev in Thread] Current Thread [Next in Thread>