[Top] [All Lists]

Re: [ontolog-forum] OntoNotes and the Omega ontology

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Thu, 30 Sep 2010 22:45:53 -0400
Message-id: <4CA54B61.4090901@xxxxxxxxxxx>
David, Ferenc, and Ronald,    (01)

Several different threads have sprouted, but they are all
addressing related topics.  Rather than comment on each point
in a different thread, I'd like to make some points that
relate to all of them:    (02)

  1. There is a major difference between the metaphysical motivation
     that has led to a huge number of words for related topics.
     Although I have a high regard for *good* metaphysics, there
     is so much dubious metaphysics that is being kicked around
     in this forum, that I'd prefer to avoid it, if possible.    (03)

  2. The simplest way to avoid metaphysics is to focus on the
     mapping to some version of logic and use only the terminology
     of that logic.  For simplicity, I'll just use Common Logic.    (04)

  3. When we use CL semantics, there is a very bare minimum of
     terms:  quantifiers, Boolean operators, functions,
     relations, and names.    (05)

  4. It's often useful to have a few more terms, which can be
     defined as special cases of CL terms:    (06)

     A variable is a name that is governed by a quantifier.    (07)

     A constant is a name that is not governed by a quantifier.    (08)

     A proposition is a relation with 0 arguments.    (09)

  5. If anybody feels the need for a word like 'entity' or 'thingy'
     that is true of everything, it is possible to state an axiom
     that is true of the monadic relations Entity or Thingy:    (010)

     CLIF:  (forall (x) (Entity x))
            (forall (x) (Thingy x))    (011)

     CGIF:  [@every *x] (Entity ?x)
            [@every *x] (Thingy ?x)    (012)

     The CGIF syntax actually makes it possible to express the
     previous two statements without using the name x:    (013)

     [Entity @every]  [Thingy @every]    (014)

This terminology and notation is definitely *not* adequate
for discussing metaphysical issues.  On the contrary, it is
so sparse that it makes it *impossible* to express them.    (015)

For that reason, it enables us to see through the veil
(or smokescreen) of terminology and focus on exactly what
any particular ontology will mean from the point of view
of an implementation in some computable form.    (016)

JFS>> Cyc has formal definitions for over 600,000 concepts.    (017)

DE> In my mind I use "concept" thusly...   social security number
 > = {SSN, TIN, taxpayer identification number, EIN, SIN, empl_id,
 > soc_sec_no, SOC-SEC-NBR, XJ37H, ...} to count as a single concept.
 > While I certainly acknowledge that things such as social security
 > number & TIN are in fact different things, because they tend to be
 > represented as 9 digit numbers, they are often treated as the same.
 > How would Cyc count this/these thingys?    (018)

I don't know how Cyc counts them, but if I were using CL,
I'd make the following decisions:    (019)

  1. Assign monadic relations for SSN, TIN, EIN, SIN, etc.    (020)

  2. Since any SSN can be used as a TIN, I'd write    (021)

     CLIF:  (forall (x)
               (if (SSN x) (TIN x))    (022)

     CGIF:  [@every *x]
            [If (SSN x) [Then (TIN x)]]    (023)

FK>  Unless you accept mental operations at work, we never
 > get a steady picture in focus.    (024)

When I talk about psychological issues, I definitely accept
the existence of mental entities and mental models.    (025)

But when I talk about a statement in logic, I don't talk
about such things.  They do not appear in the notation.    (026)

FK> I do not need more primitives than object, property and
 > relation to build an ontology for any domain.    (027)

CL is even simpler.  It just uses relations.  You can use
a monadic relation named 'Object', a monadic relation
named 'Property', and call n-adic relations relations.    (028)

FK> a verb is required to represent a predicate    (029)

The word 'verb' is a linguistic term that is used to talk
about a syntactic category in natural languages.  The
problem of how to map NLs to Common Logic (or any other
computable notation) is extremely complex.  It is best
not to mix NL terms with the terms of any formal notation.    (030)

RS> In the ontology (metaphysical flavour) we have adopted 
 > a version of actualism  everything has an actual concrete,
 > physical existence in the present.  Things commonly regarded
 > as abstractions exist as social norms shared by the members
 > of a community.  Signs employed in speech acts play a central
 > role in creating abstractions and also in providing our
 > knowledge of everything not in the here-and-now.    (031)

You have a well thought out metaphysics that can be very
helpful for analyzing aspects of the world and ways of
talking about the world.  But none of those fine points
will have any influence on what a computer does with
anything you represent in a computable form -- unless
you specify them by axioms in some version of logic.    (032)

I don't want to belittle anybody's metaphysics, but I do
want to emphasize that the computer has no commitment to
any kind of metaphysics.  It blindly pushes around whatever
symbols we type.    (033)

RM> Wordnet defines entity as "that which is perceived or known
 > to have its own distinct existence."    (034)

Actually, the WordNet documents call those sentences 'glosses'.
They are guidelines to the humans who read the documentation
and use the WN synsets in some computerized application.
But the glosses have no effect on any computer processing.    (035)

RM> I'll go with the term entity as originally suggested by John Sowa
 > with the implied connotation of being when considering metaphysics.    (036)

That's fine with me.  But I want to emphasize that from the computer's
point of view, the only thing that matters is some axiom like    (037)

    (forall (x) (Entity x))    (038)

John    (039)

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    (040)

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