[Top] [All Lists]

Re: [ontac-forum] Problems of ontology

To: mitioke@xxxxxxxxxxxx, ONTAC-WG General Discussion <ontac-forum@xxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Sat, 20 May 2006 23:53:11 -0400
Message-id: <446FE427.5070606@xxxxxxxxxxx>
Ken and Leo,    (01)

KE> I hope you and the others do not mind if I add a short comment.    (02)

Everybody's welcome.  I just listed the names of the people
I was responding to.    (03)

KE> I learned from my own lengthy experience with automatic
 > indexing systems, concept databases and the lexical accounts
 > of several thousand root concepts found in most languages,
 > that it is usually sufficient to look up definitions as needed.
 > We determined to keep them short and distinctive in the
 > architecture of our methods and measure.    (04)

It sounds as if you have been working with terminologies,
rather than formal axiomatizations.  But I believe that the
points you are making are very well taken for formal systems
as well as less formal indexing systems.  In fact, I would
recommend that we use terminologies for the purpose of
indexing and selecting whatever detailed microtheories
are needed for any given task.    (05)

KE> I have specific examples and computational performance
 > data to substantiate this claim if needed.    (06)

This discussion could definitely benefit from more data and
practical experience about what works and what doesn't work.
Please contribute whatever examples you feel are relevant.    (07)

LO> I think that many if not most of us who have been in this
 > ontology business long enough know the difference between
 > ontologies and terminologies.    (08)

Yes, we all know that.  But just look at what people are
actually doing:  They are claiming to be aligning ontologies,
but they are treating them as if they were terminologies.    (09)

Look at the work on database alignments and at most of the
work on natural language processing.  They're ignoring almost
all of the detailed axioms (except for type-subtype).  Even
when they have access to the full axiomatization of Cyc,
they're just using the hierarchy.  (I won't mention actual
systems because I don't know whether the people who are
doing this work want to be dragged into the discussion.)    (010)

So far, there is an enormous amount of evidence for the value
of large terminologies, both for human and computer consumption.
There is also a great deal of evidence for the need for detailed
axioms at the lower levels.  But so far, there is ZERO evidence
for the value of detailed axioms at the upper levels of the
ontology.   Just ask Lenat -- the one with the most experience
in working with large ontologies on planet earth.  He says that
the upper levels are not important.  I didn't completely believe
him in 1999 when I finished my KR book, but now I do.    (011)

LO> I should state that in principle the specific salient
 > interpretations of a given natural language utterance can
 > be arbitrarily approximated, given a rendition in a logic
 > or logics.    (012)

You're preaching to the choir.  You can *always* achieve such
precision on narrow topics, as Peirce, Whitehead, Bundy, Lenat,
I, and many, many others have said.  But you cannot achieve
such precision in a global theory of *everything*.    (013)

LO> WordNet is NOT an ontology, but primarily a thesaurus
 > organized by psycholinguistic principles (at least according
 > to George Miller, Christiane Fellbaum). People link their
 > ontologies to WordNet in order to have a term to ontology
 > "concept" mapping, and get the benefit of synonmy and weak
 > term semantics (hypernymy and hyponymy).    (014)

Yes, that is exactly the point I was making:  if ontology A
and ontology B are aligned via WordNet in the middle, the only
relevant axioms that both A and B can depend on is what is
shared with WordNet -- the subtype/supertype relations.
In other words, any detailed axioms associated with the
upper levels of A and B are irrelevant.    (015)

Summary:    (016)

  1. Terminologies are valuable.    (017)

  2. Microtheories with detailed axioms are valuable for
     reasoning about specific problems or tasks.    (018)

  3. There is little or no evidence that detailed axioms at
     the upper levels (i.e., anything more than a cleaned-up
     WordNet) are of much, if any value.  In fact, they are
     more likely to be of negative value, since they can
     cause unexpected contradictions.    (019)

Therefore, I suggest that we concentrate on putting detailed
axioms in the microtheories and use terminologies for indexing
and selecting whatever microtheories are appropriate for
any given task.    (020)

John    (021)

Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (022)
<Prev in Thread] Current Thread [Next in Thread>