[Top] [All Lists]

Re: [ontolog-forum] a skill of definition - "river"

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ali Hashemi <ali.hashemi+ontolog@xxxxxxxxxxx>
Date: Sun, 15 Feb 2009 23:41:22 -0500
Message-id: <5ab1dc970902152041r4ef3bacbib6db986f37742177@xxxxxxxxxxxxxx>
I bristle at tooting my own horn, but some thoughts:

1) If you've given a definition of River, and it is inadequate when you encounter something as the Okavango River, ought it not indicate that you only need to update your definition of river? Isn't the whole point of defining something trying to abstract the generalizable qualities / properties of the object/entity under consideration? You can still have a monotic logic, you just need smart revision policies :P.

2) I'm not sure there is such a difference between the intensional and extensional approaches. Here, I would point to the ontology design algorithm I developed in my thesis. It tries (i think pretty successfully) to bridge the gap between the two approaches, starting first with concrete examples, and "learning" through falsification hypotheses to refine an intensional definition that captures the semantics contained in sets of accepted and rejected concrete (empirical / extensional / examples) models.

From where I stand - extensions correspond to fragments of models (Tarski sense).  Any set of axioms A has associated with it a set of models MA.  The beauty / difficulty of ontologies is finding the right match between the two. Limiting ourselves by defining what exists only to things we have concretely encountered seems too restrictive; we would like to conjecture as to what would count as well. Yet as many have noticed, if the set of acceptable models is too expansive / restrictive, the quality of the resultant ontology suffers.

To me, this makes a strong case for developing a set of referent ontologies sooner than later, (ideally unified under some umbrella).

If our goal is to move towards ontology interoperability, instead of focusing on whether Extensional or Intensional approaches are superior, or whether one particular ontic category hierarchy is appropriate for all, I think efforts would be more fruitful in explicating / generating mappings between what various peoples find useful.  I.E. take the IDEAS hierarchy and compare it to DOLCE or SUMO -- what's being reused? what are people disagreeing on? what are the implications of the different choices? But perhaps I digress here :P.



On Sun, Feb 15, 2009 at 2:20 PM, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
Mike and Mitch,

I would like to comment on the following point:

MB>> According to that definition the Okavango is not a river.

MH> The Okavango surely is a strange kind of river.
 > Do you really expect to hold natural language to the same
 > strictness standards as formal ones?

This question has nothing to do with the differences between
natural languages and formal languages.  It is the result
of trying to map a continuously variable world to a discrete
set of labels (i.e., words, terms, symbols, concepts, signs).

As a continuous fluid (at least to a degree far below human
perception), there is a continuous range of ways that water
can flow across a surface.  For various purposes, people label
those ways of flowing that happen to be significant for their
interests.  The way they group them and label the groupings
depends on what they consider important in their  environment.
The kind of language, natural or artificial, is irrelevant.

This is a commonly discussed issue in philosophy:

Immanuel Kant:

   "Since the synthesis of empirical concepts is not arbitrary
   but based on experience, and as such can never be complete
   (for in experience ever new characteristics of the concept
   can be discovered), empirical concepts cannot be defined.

   "Thus only arbitrarily made concepts can be defined synthetically.
   Such definitions... could also be called declarations, since in
   them one declares one's thoughts or renders account of what one
   understands by a word. This is the case with mathematicians."

Wittgenstein's *family resemblances* :

   Empirical concepts cannot be defined by a fixed set of necessary
   and sufficient conditions. Instead, they can only be taught by
   giving a series of examples and saying "These things and everything
   that resembles them are instances of the concept."

Waismann's *open texture* :

    For any proposed definition of empirical concepts, new instances
    will arise that "obviously" belong to the category but are
    excluded by the definition.

As Kant observed, precision depends on the kind of concept, not
on the kind the language used to define the concept.

As Waismann observed, if you state a precise definition for an
empirical concept (such as 'river'), you will simply exclude
many reasonable examples, such as the Okavango River.

John Sowa

(•`'·.¸(`'·.¸(•)¸.·'´)¸.·'´•) .,.,

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    (01)

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