[Top] [All Lists]

Re: [ontolog-forum] Quote for the day

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Mon, 03 Jan 2011 03:22:47 -0500
Message-id: <4D218757.2060405@xxxxxxxxxxx>
Pat,    (01)

Let me start with your ending paragraph, which raises critical
issues about the nature of language, logic, and their relationships
to each other and to the world:    (02)

> We create ontologies in part to avoid the ambiguity of human language,
> and discussions in this forum sometimes suffer from that ambiguity.    (03)

First of all, the syntax of natural languages can be used in a way
that is just as clear and unambiguous as any formal language.
Note that controlled NLs, which can have a formal mapping to some
formal logic, can be read as if they were an ordinary NL.    (04)

The fundamental problem is the mismatch between a *discrete*
vocabulary (words, predicates, or other symbols) and a *continuous*
world.  It's irrelevant whether the discrete vocabulary consists
of words like 'horse' and 'give' or predicates like 'horse(x)'
and 'give(x,y,z)'.  And it's irrelevant whether the definitions
are stored in a book or at the target of some URL like
'http://www.infinitevocabulary.org/horse39286.xml'.    (05)

For further discussion of these issues, please see    (06)

    The role of logic and ontology in language and reasoning    (07)

> It is not clear from comments about "no single ontology" whether this is
> intended to deny the possibility of a common ontology that can describe all
> models, or whether it merely means that there are conflicting **models**
> that are logically incompatible.    (08)

I agree.  It is conceivable that science might someday progress to
the point where we have a unified theory of everything, from which
everything else is deducible.  But so far, we don't have anything
that remotely resembles such a theory.    (09)

In fact, modern physics is the closest to having such a unified
theory, but every engineering application of physics uses
special-purpose approximations, since the general theories are
too difficult (or impractical) to apply to specific problems.
See the references in rolelog.pdf to Alan Bundy's work on
ontology repair and evolution for problem solving in physics.    (010)

If physicists don't use a universal set of axioms for everything,
you can be sure that every other field is in far worse shape.    (011)

> It strikes me as equivalent to saying that there is no one human
> language that can describe all models.  Yet we do describe them
> in English, as well as in other human languages (sometimes
> supplemented by graphical or mathematical notation).    (012)

I agree.  That point is consistent with my claim that NLs (or
at least controlled NLs) are capable of stating anything that
can be said in any formal language with equivalent precision.    (013)

But that depends on either (a) having infinitely many words
(or predicates) in the language, or (b) using and reusing
a finite vocabulary in a way that makes the words ambiguous
with a potentially infinite number of word senses for each.    (014)

In fact, the linguist Alan Cruse made the claim that any
word in any NL can have an open-ended number of what he
called *microsenses*.  I believe that he's right.  See
the rolelog.pdf article for references.    (015)

> The basis for general interoperability is to find and
> agree on that neutral ontology that does not **assume**
> any of the contradictory models, but can **describe**
> conflicting models.    (016)

That is false.  Interoperability requires a compromise that
uses the barest minimum of axioms in a highly underspecified
theory.  Note Doug F's example of different ontologies for
renting, selling, or repairing cars.  Each of those businesses
would require very different axioms at the detailed levels, but
they can interoperate efficiently by ignoring the details and
using an underspecified taxonomy of basic terms for talking
about cars and VINs (Vehicle Identification Numbers).    (017)

Please read the rolelog.pdf article.    (018)

John    (019)

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    (020)

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