[SCL] naming ontologies

[pat hayes]

>Hi Pat...
>
>I'm glad you're thinking about this because it's an important 
>subject.  However, Chris wrote a paper for ECAI 2002 (attached)

Thanks for sending this (which I confess I hadnt read and didnt even 
know about, sigh.) However, after a quick read, I think Chris is 
hoeing a different kind of row here. This is, as the title implies, 
an ontology *theory* where that last word is being used in a 
technical sense, ie a formal ontology. This presupposes a formal 
language to write the theory in, which for us begs the central issue. 
(Chris mentions this in footnote 13, but relegates it to 
'metatheoretic housekeeping'. But Im not suggesting it in that way at 
all: Im suggesting it as part of the language.)

>Content-Type: application/pdf;
>	x-mac-type=50444620;
>	x-unix-mode=0644;
>	x-mac-creator=4341524F;
>	name="ecai2002.pdf"
>Content-Disposition: inline;
>	filename=ecai2002.pdf
>
>Attachment converted: Betelgeuse:ecai2002 1.pdf (PDF /CARO) (000AAA91)
>on Ontology Theory, which covers what you're trying to provide a 
>syntax for below and more.

Lots more; but also less, since Im suggesting this as a syntactic 
device in the language. So, different.

>  There's a conflict between the view there and the view you present 
>- namely that ontologies do not consist of sentences but rather 
>propositions, which seems to be what you're hinting at here:

I really don't care about this issue, or the philosophical tarpit 
about propositional identity and 'content' that it uncovers. For our 
purposes we can say that they are sets of sentences, or even just 
strings, provided only that we can interpret those strings as SCL. 
One thing it does show, by the way, is that simple entailment 
relation described as 'Subsumes' in Chris' framework is likely to be 
too crude.

>>The meaning of this notation is that the name denotes this set in 
>>all satisfying interpretations: this requires that the universe 
>>admits things called 'sets of sentences' which as far as the 
>>semantics are concerned can be thought of simply as functions from 
>>interpretations to truthvalues, i.e. as propositions, which we can 
>>treat as a primitive semantic category (like strings and integers). 
>>So if 'ex:ont' is such a URI then it makes sense to write something 
>>like I(J('ex:ont')) = true where I and J are SCL interpretations.
>
>In general, I think it's getting hard enough just to get the MT and 
>syntax straight for *one* set of sentences that would denote an 
>'ontology' (quotes because Chris has a formal definition of what 
>constitutes one).  Now, you're introducing a whole new set of 
>machinery.

My point was that hardly any machinery is necessary. We just need to 
have things in the universe that can themselves be interpreted. We 
aren't going to say very much about them other than that they are 
denoted and they are true or false. We don't need a full theory of 
them.

>Not that I object in principle.  Someone has got to try eventually 
>to do this and I applaud the effort, but isn't this draining energy 
>away from hitting the 25-meter target - getting the basic SCL stuff 
>right and settled upon quickly?

It would seem to be, except that what I found was that this actually 
simplifies a lot of issues that we need to deal with now in any case; 
in particular, it provides a natural (?) way to accommodate the 
awkward fact that identical SCL sentences might be written in 
different 'languages' depending on how you interpret the lexicon. 
This is a definite *problem*  for a useable interchange langauge, and 
so far we havnt got any way to deal with it . So, two birds with one 
stone, etc. . And I think that providing a way to name ontologies is 
part of the basic stuff in any case.

>Once that's done, I'll be 110% behind doing what you suggest and let 
>me also suggest - although I'll let Chris speak for himself on this 
>- that we use Chris' paper as a starting point to get the ideas 
>right and then figure out where the ontological hooks are to hang 
>the syntax off of.

Well, lets see about that. Im not sure I agree with Chris' way of 
doing these things. For example, why bother to axiomatize the notion 
of being a syntactic part of? Being a syntactic part of is well 
understood already and we have precise, machine-readable ways to 
express it. I think this idea that its not real unless its 
axiomatized is kind of beside the point for most practical purposes 
(though it fits well into a certain tradition in philosophical 
logic). Suppose there were a formal ontology of ontologies: what 
language would it we written in, and what use would it be? I want to 
USE ontologies, not reason ABOUT them. Ontologies, for me, are 
essentially syntactic entities; and I disagree with Chris' view that 
this way of thinking is somehow awkward or unsatisfactory. I also do 
not share his philosophical scruples about using set theory in a 
metalanguage, do not find model theory 'austere and formal' compared 
to a theory of propositions (which I guess Chris feels are made from 
some kind of soft velvet material, as opposed to the wire-mother of 
set theory) and I do not think that it is 'far removed' from 
'ordinary semantic notions' (which Im not sure what those are, but 
whatever they are, axiomatizing them in first-order logic is hardly a 
warm fuzzy way to get at them.)

Pat

>   .bill
>
>--
>Bill Andersen (andersen at ontologyworks.com)
>Chief Scientist
>Ontology Works, Inc. (www.ontologyworks.com)
>1132 Annapolis Road, Suite 104
>Odenton, Maryland 21113
>United States
>Office: 410-674-7600
>Mobile: 443-858-6444


-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes at ihmc.us       http://www.ihmc.us/users/phayes