[Top] [All Lists]

Re: [ontolog-forum] Ontology-building vs Data Modelling (was Two ontolog

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ed Barkmeyer <edbark@xxxxxxxx>
Date: Mon, 18 Jun 2007 18:59:12 -0400
Message-id: <46770E40.4060300@xxxxxxxx>
Peter F Brown wrote:    (01)

> Further, I don't understand why ontology building is an exercise in
> application-specific modelling: I can build an ontology without giving a
> damn about how it's used, where or by who    (02)

In so many words, I doubt that.
If you don't know whether, how or why it will be used, why would you bother?
(I realize that like the Red Queen, you didn't say you would, only that you 
could.  And I can't deny that, even if I try with both hands.)    (03)

> - it's my encapsulated take on
> reality and I can then *use* it to build something, if I choose    (04)

The problem with a "take on reality" is that a "take" has a viewpoint, which 
engenders a degree of accuracy, a scope of interest, a degree of depth, etc. 
You can only use an ontology to build something that is consistent with that 
viewpoint.    (05)

IEEE 1471: Every model has a viewpoint.  And an ontology is a model.    (06)

> Not so many moons ago, we had an extensive and intensive discussion
> around the differences between data modelling and information modelling
> - and now we have ontology modelling and, why not, concept modelling.
> Would someone venture to distinguish, group, eliminate and add, as
> appropriate?    (07)

"What is in a name?"  People who distinguish among these ideas have particular 
concepts, scopes and (to them) important characteristics in mind.  At one time 
the term 'data modeling' meant SQL, and was distinguished from 'information 
modeling' in that the latter didn't commit directly to a database schema or to 
the database concepts 'table and row' versus 'class, property and instance'. 
(Among other things, that meant you modeled relationships the same way, 
regardless of whether they were 1-to-1, 1-to-many, or many-to-many.) In the 
same way right now, some want to use 'data modeling' to mean capturing the 
model in XML Schema.    (08)

I think it is useful to consider the OMG ideas of the "platform-independent 
model" and the "platform-specific model", BUT I think most 
"platform-indepedent models" (PIMs) are actually targeted to a specific 
implementation technology class in some way.  (The original purpose of the OMG 
idea was not to throw away everything they learned from CORBA and start over 
for J2EE and Webservices.  A PIM built from CORBA probably translates quite 
well to Java and WSDL/SOAP.)  So an 'enterprise object model' PIM that has an 
obvious rendering in Java is not likely to have an obvious rendering in SQL.    (09)

So, is an information model a PIM?  I don't know.  But it is some kind of 
abstraction from an XML schema or an SQL schema.    (010)

Whether a 'data model' means the same thing as 'information model' or means 
something like an XML schema is up to the speaker.  The ISO 10303 modeling 
language EXPRESS says it is a 'data modeling language', and it turns out to be 
halfway between the two.  "entity" ("class") and "relationship" ("object 
property") are not distinguished in the language; so all the "product data 
models" in ISO 10303 are thinly disguised SQL models in which "entity" means 
'table'.    (011)

In a similar way, IMHO, the folks who think their 'ontologies' represent 
Platonic truth are deluded.  They are models made from a particular viewpoint, 
and they are at best accurate from that viewpoint in their time of creation. 
Very few truths outside of pure mathematics are timeless.  Even accepted 
scientific truth undergoes slow refinement.    (012)

So I personally would say that an 'ontology' is an 'information model' with 
one important distinguishing feature: the language in which the model itself 
is captured is well-defined.  (NIAM was never formally documented, IDEF1-X had 
several fuzzy notions, EXPRESS was badly defined, and UML is deliberately 
ambiguous.  And so on.  ORM may be an exception.  Terry Halpin has a formal 
logic interpretation for ORM, but it relies on alethic semantics, plus 
reification and relational 'projection'.)    (013)

That the language is well-defined doesn't make the ontology a better model. 
It should make it possible to determine whether a given ontology really is 
"correct" or "complete" in some particular viewpoint.  And on a few occasions, 
it may make it possible to know when two concepts really are the same, in 
spite of different authors, different viewpoints in construction, and 
different terminology.  (And no information model was sufficient for that.)    (014)

-Ed    (015)

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    (016)

"The opinions expressed above do not reflect consensus of NIST,
  and have not been reviewed by any Government authority."    (017)

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    (018)

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