[Top] [All Lists]

Re: [ontology-summit] First Model Bench Challenge

To: ontology-summit@xxxxxxxxxxxxxxxx
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Mon, 07 May 2012 09:14:29 -0400
Message-id: <4FA7CAB5.8010501@xxxxxxxxxxx>
Leo and Doug,    (01)

> “Type” and “Class” are pretty much used interchangeably in ontology 
>engineering.    (02)

I agree that there is a huge amount of confusion, and both of them are
also confused with the word 'set'.  My recommendation is to use the
following definitions:    (03)

  1. A class is a set of all x of a given type.    (04)

  2. Every type can be specified by a monadic predicate that is
     true of every instance of that type.    (05)

These two definitions are sufficiently short and general that they can
be specialized to all the major notations and schools of thought.    (06)

> Semantic analysis is NOT the same as ontological analysis.    (07)

I agree, but I would avoid the word 'semantics'.  It has acquired too
many diverse meanings in too many different contexts.  In logic, it
has become synonymous with model theory.  In linguistics, every school
of thought has a different definition.  In common usage, it is vaguely
synonymous with 'meaning', which is another word that I never use
in any technical sense.    (08)

I would also avoid the term 'semantic analysis' because it inherits
all the vagueness of the word 'semantics'.  It sounds technical,
but it is impossible to state a definition that is compatible with
each of the many ways of using the word 'semantics'.    (09)

>> Natural languages are the ultimate knowledge representation
>> languages -- in generality, flexibility, *and* precision.    (010)

> i am surprised at the claim of precision.  You have often
> argued that natural languages lack precision, but that such can
> be achieved through the use of controlled natural languages in
> which the meanings of the terms is precise and the syntactic forms
> are limited.    (011)

I have never said that NLs lack precision.  That's what Lotfi Zadeh
says, and that's a point on which I strongly disagree with him.    (012)

What I have said is more complex:    (013)

  1. Every artificial language or notation is an abstraction from
     some natural language.    (014)

  2. Every textbook that teaches the formal notations of mathematics,
     logic, or programming languages uses some NL as the metalanguage.    (015)

  3. It is possible to state anything in any of those artificial
     languages in the metalevel NL with *exactly* the same level
     of precision.    (016)

  4. But the NL restatement of the artificial language would require
     much more careful writing and qualification to state as precisely
     as the artificial language.    (017)

  5. Therefore, the artificial language provides an excellent
     methodology for forcing precision in the NL.    (018)

  6. *However* the statement in an artificial language might *look*
     precise, but that precision is very often an illusion because
     the author who wrote it wasn't using the notation correctly.    (019)

  7. Therefore, I recommend controlled NLs, which can be read
     as easily as the NL, but they can be tested automically
     by the same kinds of tools used for the artificial languages.    (020)

> If NL were precise, the task of writing programs to encode something
> stated in NL or to translate from one language to another would not
> be nearly as difficult as it currently is.    (021)

The really hard part is the task of getting a clear understanding
of the subject matter.   For example, suppose you want to hire
a contractor to build a house.  The initial discussion will be
extremely vague.  There is no way that anybody could build a
house to your satisfaction from that discussion.    (022)

But suppose you were trying to describe a house you have been
living in for many years.  For that discussion, you can give
crisp, precise answers to a huge number of questions.    (023)

But there are still some questions on which you cannot be
precise:  What is the color of the house?  Do you mean the
original paint that was used or the way it has faded or been
stained by the weather?  Etc.    (024)

> The precision issue arises from the multitude of aspects of NL.    (025)

What I was trying to say is that they are *not* aspects of the NL.
They are aspects of the *subject matter*,  Every question about
your house that you can state precisely in logic, you can state
just as precisely in English.    (026)

But any question about your house that you can only answer vaguely
in English *cannot* be stated precisely in logic.    (027)

If you say "My house is white" in English, you can translate
that to "white(MyHouse)".  But that precision is an illusion.    (028)

>> Re formal logic:  What it says so precisely may be totally
>> different from what the author had intended.    (029)

> Agreed.    (030)

That's the main point I was trying to make.    (031)

John    (032)

Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/   
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/  
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2012/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2012  
Community Portal: http://ontolog.cim3.net/wiki/     (033)
<Prev in Thread] Current Thread [Next in Thread>