Re: Ontology for EDI (was Frames...)

"Peter Clark" <pclark@cs.utexas.edu>
Message-id: <9409232144.AA03667@sage.cs.utexas.edu>
Subject: Re: Ontology for EDI (was Frames...)
To: phayes@cs.uiuc.edu (Pat Hayes), srkb@cs.umbc.edu
Date: Fri, 23 Sep 1994 16:44:38 -0500 (CDT)
From: "Peter Clark" <pclark@cs.utexas.edu>
In-reply-to: <199409232050.AA10532@dante.cs.uiuc.edu> from "Pat Hayes" at Sep 23, 94 03:54:47 pm
X-Mailer: ELM [version 2.4 PL24alpha3]
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 3081      
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

>>> Pat Hayes writes...
>>> Is a fitted carpet IN a room or PART OF the room? How can I write 
>>> axioms (or something) which say, "I don't car about this issue", without 
>>> either saying nothing, or using impossibly large disjunctions?

>> It seems that you are wanting to work at a more abstract level, using 
>> (and axiomitizing) more abstract relations eg. InOrPartOf ("FixtureOf" say)
>> to describe the relation between a room and a carpet. The large 
>> disjunctions only come in when you try to push the level of detail, 
>> and expand the abstract description into a more detailed version.
>> I don't see why you feel "forced" to work at the IN and PART-OF level 
>> if you don't care about it. 

> Well,maybe. But I find again and again that I get caught between two hard
> places. In order to get some conclusions I want - for example, the
> transitivity of IN, or that going IN involves somehow passing through the
> boundary - I need to write enough about the concept that I am then forced
> to take a stance on issues like whether the carpet is IN or not. If this
> were just an occassional difficulty, then thats life; but it seems to be
> pervasive, and its this fact - that this seems to happen again and again -
> that makes me think it maybe has something to do with an inadequacy in the
> tools I am using.
> 
> Of course it is possible to make vaguer concepts, but the tension then lies
> in not making them so vague that its hard to say anything useful with them.

You seem to be assuming that different levels of description can't
(happily) co-exist in an axiomitization, so you're forced to choose one
or the other. And I guess (if you want complete axiomatisation and 
complete inference) this is true, the tension comes from relating the 
levels together:
  [1] If I state in logic IN-OR-PART-OF implies either IN or PART-OF (etc.), 
	then my theorem prover may die a death of disjuncts. 
  [2] But if I don't state this, then I've somehow (?) not fully axiomitised 
	IN-OR-PART-OF.

Something, then, may have to go; Maybe completeness of inference (or
axiomitization) is a candidate -- after all, people die a death of disjuncts 
too, given a few disjuncts to think about together (eg. trying to do 
constraint satisfaction tasks in your head). But this seems to be a better 
option than trying to force a choice between one level of detail or another.
I guess the issue then is: whether the requirement for completeness can be 
relaxed without loosing "important" ("common-sense", even) inferences. 
I would hope that this is the case. Maybe the IN/PART-OF snag is not a 
problem of cutting up the space, but of constraining axiomisation/reasoning 
with the pieces without losing the `common-sense' conclusions. I wonder, what
`common-sense' conclusions would I miss if I followed option [2] above?

Best wishes,

Pete
-- 
Peter Clark                             Dept of Computer Science
email:  pclark@cs.utexas.edu            University of Texas at Austin
phone:  (512) 471-9574                  Austin, Texas, 78712
fax:    (512) 471-8885			USA