[Top] [All Lists]

Re: [ontolog-forum] Using controlled natural languages for ontology

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Mon, 14 Mar 2011 13:08:30 -0500
Message-id: <4D7E599E.3060705@xxxxxxxxxxx>
Ali,    (01)

I agree with some of the points you make, including the observation
about the usefulness of the following report by Mike and Michael:    (02)

> I should mention that I have found this work to be very useful in
> helping clarify for me the role of ontologies in an organization
> http://stl.mie.utoronto.ca/publications/ker.pdf    (03)

But I'd like to add that the report was written in 1995, when the
Semantic Web was still in the early talking stage.  The sad story
is that after 16 years, it is still a better overview of the nature
of ontology and its promise for interoperability than anything that
has been written about the Semantic Web.    (04)

Furthermore, note the word 'promise'.  The discussions of examples
are stated as hypothetical "would be" and "can be" statements *if*
somebody were to follow the methodologies outlined in that report.    (05)

That report does contain some examples of tools that were
integrated with mainstream IT, at least for certain aspects
of what an ontology needs to do:  Express/STEP and IDEF1X.    (06)

But today, the only SW product for Ontology is OWL, which still
isn't integrated with mainstream IT as well as Express/STEP and
IDEF1X were a couple of decades ago.    (07)

> Fundamental principle:  the primary purpose of the CNL is not
> for the SME to verbalize the ontology, but for the SME to read
> and verify the verbalization created by the KE.    (08)

> While I find this to be in principle, a laudable goal, I wonder
> if it is actually necessary and might not possibly be a distraction
> of attention and resources.    (09)

What is the alternative?  Not to have any verification by the SME?    (010)

> the focus is on how much we should expect an SME to follow the
> machine readable representation of the ontology.    (011)

That is why I have recommended CNLs:  they focus on the ontology
while avoiding the machine-oriented details.    (012)

> Why is it beneficial to have an SME be able to read the innards
> of a deployed ontology?    (013)

The SME is the expert who understands the subject.  The KE may
be an expert in the computer side of things, but the KE is just
a novice about the subject.  The SME is the only one who can
tell whether the ontology has any resemblance to the reality.    (014)

> Dave had mentioned how a car driver does not need to know the
> inner workings of an engine to drive a car...
> What benefit do we gain from exposing SME's to a particular
> formalism or machine readable segment of an ontology?    (015)

It's true that most drivers don't know how the car's innards work.
But SMEs who designed the car had better know how the engine works,
they're the only ones who are qualified to check the ontology, and
they almost certainly don't know any formal ontology language.    (016)

> I mentioned how nuclear engineers don't need to know how the
> control room was designed to use the control room    (017)

There are multiple SMEs for any large project, and the people who
designed the control room are also SMEs.  They need to check
the ontology about the control room and its relationships to
the nuclear reactor.    (018)

> in many ontology deployments, any subject matter expert will
> only be able to follow only a fragment of the whole. Meaning,
> that a particular representation choice that takes into account
> other fragments might not superficially strike a particular
> SME as intuitive or "right".    (019)

Yes, of course.  That is why you need a CNL.  It's the only
notation that *all* the SMEs can read in order to verify the
part of the ontology that they are responsible for.    (020)

> a particular choice of an axiomatization might be unwieldy
> and meaningless to an SME.    (021)

A particular theorem proving method might be unintelligible
to the SMEs.  And some Kn. Rep. languages, such as OWL, are
so convoluted that a simple fact has to be rewritten in a
hopelessly opaque way.  But that is another argument for
expressing the ontology in a CNL -- and then compiling it
into some lower-level representation, if necessary.    (022)

CNLs and knowledge compilation reinforce one another.    (023)

> Should an SME follow both the highly expressive and more restricted choices?    (024)

That depends on whether the restricted choices change any of the
assumptions.  If the knowledge compiler faithfully compiles the
general principles into the restricted form without making new
assumptions, then the SME should not have to verify anything more.    (025)

But if the restrictions *change* any assumptions, the SME *must*
review the changes.  As an example, please note the disaster that
occurred at the Hyatt Regency in Kansas City:    (026)

    http://en.wikipedia.org/wiki/Hyatt_Regency_walkway_collapse    (027)

The engineer who designed the walkway wrote a safe specification.
But the contractor who implemented it made a "minor" change that
simplified the construction.  As a result, 114 people were killed
and 216 were injured.    (028)

John    (029)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (030)

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