[Top] [All Lists]

Re: [ontolog-forum] Levels

To: Chris Menzel <cmenzel@xxxxxxxx>
Cc: "[ontolog-forum] " <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Wed, 14 Feb 2007 13:11:45 -0600
Message-id: <p06230905c1f9073005ae@[]>
>On Monday 12 February 2007 14:46, Pat Hayes wrote:
>>  >On Feb 12, 2007, at 12:12 PM, Pat Hayes wrote:
>>  >>>  You're right that import statements should not be considered part of
>>  >>>  an ontology.  I agree it's the imported axioms that are part of the
>>  >>>  ontology.
>>  >>
>>  >>  Wait. Of course the imports statements are part
>>  >>  of the ontology. What are you guys talking about?
>>  >
>>  >*If* ontologies are logical theories
>>  Well, if "theory" has to mean deductively closed,
>>  then I would say no, they are not. That is, an
>>  ontology to me means a set of sentences (usually
>>  finite, usually not deductively closed) in an
>>  ontology language. Which is, roughly, a formalism
>>  with a model theory.
>I agree it seems to accord better with the way the term "ontology" is used to
>think of an ontology (or at least the purely logical part of an ontology) as
>the *axioms* of a logical theory, rather than as axiom+consequences (though
>I'm not sure it matters much.    (01)

Well, only for clarity in communication, c.f. a 
recent email from Karen which drew strong 
consequences from the closure reading.    (02)

>>  >, then it seems to me that this 
>>  >confuses a mechanism for saying what (some of) the axioms of an 
>>  >ontology are with the ontology.
>>  Why do you call this "confuses"? The assertion of
>>  what some of the axioms are can be part of the
>>  theory.  You seem to be insisting on a 'levels'
>>  distinction here which I fail to see the reason
>>  for (and plenty of reasons to reject.) BTW, Cyc
>>  has been doing this for years.
>Come on, Pat, no one suggested it can't be done.    (03)

My point was that it is BEING done.    (04)

>  Here's why I said the two
>are being confused.  Intuitively, an ontology is about a certain domain, say,
>faculty and admin at TAMU.  The relevant objects, properties and relations in
>that domain are indicated by the names in one's ontology language.  If one of
>the names is now a URL pointing to your ontology at IHMC, then, on this
>intuitive understanding of what an ontology is, your ontology at IHMC becomes
>part of the *subject matter* of my faculty and admin ontology.    (05)

If by 'subject matter' you mean that it is in the 
universe of discourse, yes. But all sorts of 
things are in the universe of discourse of any 
reasonably complicated assertion: your TAMU 
example may well refer to numbers, dates, 
time-intervals, publications, etc.. If you use 
'imports' it also refers to ontologies 
(considered as Web entities). If this bothers 
you, you shouldn't be using 'imports'. You can 
always grab the actual ontology text using a Web 
browser and cut and paste it into your document, 
if you don't want to use an imports statement.    (06)

>  That seems
>wrong;    (07)

I have no idea what you mean by "wrong". It seems 
to me to simply be a fact, and follows from the 
model theory in the specs. And I have absolutely 
no intuitive problem with it (I suspect because I 
don't take the notion of "subject matter" as 
being as salient as you apparently do.)  After 
all, if you are using CLIF then you already have 
all numerals and all Unicode character strings in 
your universe of discourse, before you even write 
an axiom. Are these part of your subject matter? 
If not, how do you define this notion?    (08)

>conceptually, at any rate, I think we *should* distinguish statements
>expressing genuine domain knowledge with statements concerning the location
>of relevant domain knowledge.    (09)

In a Web logic, the "location" (actually this is 
not strictly correct) of domain knowledge *is* 
domain knowledge. That is the whole point of 
'imports', and the motivation for giving it a 
coherent model theory. Part of what you know 
about X is how to find out more about X. In some 
extreme cases, this is ALL you know about X; and 
it can still be useful.    (010)

>  So, while I am most certainly not "insisting"
>on it, I am indeed suggesting that there is an important conceptual
>distinction to be made here between "object-level" statements concerning the
>entities, types,  and relations of interest in the ontology
>and "meta"-statements about the ontology itself, notably statements about
>where certain object-level statement can be found on the web.    (011)

And I want to strongly deny this conceptual 
distinction. I think it reflects a pre-Web way of 
thinking about knowledge. And PLEASE don't say 
that imports is meta-level, even in scare quotes.    (012)

>But really, we're simply involved in a quibble here.  The important (?) thing
>is to have a consistent understanding of what an formal ontology is.    (013)

But I think it is a very important quibble. The 
fact that imports statements are genuine 
assertions, not some kind of 'markup metalevel' 
that can be handily ignored to protect the 
innocent intuitions, is not merely a quibble: it 
is central to the SWeb vision. So, I insist, 
imports statements *are* part of a formal 
ontology. See    (014)

http://www.websemanticsjournal.org/ps/pub/2005-23    (015)

>  > >Suppose I'm writing my ontology for 
>>  >TAMU faculty and admin again, and you've got a nice higher-level 
>>  >ontology for universities over there at IHMC.  My statement "import 
>>  >(reiterate, endorse, whatever) Pat's university ontology" is not part
>>  >of my ontology
>>  YES IT IS. Check out the specs. In both OWL and
>>  CL , the imports is a *sentence*, and the
>>  truth-recursions apply to it like any other
>>  sentence.
>Well hang on now; the issue isn't whether imports has a respectable model
>theoretic semantics.    (016)

But I think they are closely related. Unless you 
wish to distinguish between the universe of 
discourse and a somewhat fuzzier notion of 
'subject matter'? What does the latter mean, if 
not the former?    (017)

>  *No one* has questioned that, least of all me.  The
>issue is what a formal ontology is -- and "ontology" is not to be found in
>the CL spec (except somewhat extraneously in Appendix C), which IIRC was a
>conscious decision on our part.    (018)

It used to be there, but other (:-) people 
preferred 'module' as being a less contentious 
terminology. OWL ontologies are called 
"ontologies" quite robustly.    (019)

>  Granted, OWL has an explicit definition of
>what counts as an *OWL* ontology, but that won't do across the board,    (020)

Obviously one has to remove the 'OWL' part, but 
the general idea of a set of assertions in a 
language with a semantics, perhaps encoded in a 
document which can be transmitted by a transfer 
protocol, is good enough for me. In fact, the 
usage "OWL ontology", "CL ontology" "RDF 
ontology", etc. seems a useful and indeed widely 
used form of words which is based on this 
generalization.    (021)

>there is some question as to whether it is worth trying to identify a general
>notion that works for any representational context.
>>  >I'm not dogmatically wedded to the idea that formal ontologies are 
>>  >logical theories of some ilk, but if you're right, and my import 
>>  >statement is literally part of my ontology, then formal ontologies 
>>  >are not (in general) logical theories,
>>  Well, it depends on what you count as a logic, I
>>  guess. Is CL a logic? Because CL has importing
>>  with a full model theory,
>Well, I certainly don't want to quibble about that either.  FWIW, in general,
>I do tend to think language+model theory is sufficient for calling something
>a logic (CL in particular I think of as a Web logic, a logic for a new
>generation, if you will ;-), but I have to admit I was thinking of logic in a
>traditional first-order way in my comments in this thread.    (022)

But that begs the question, surely, since 
traditional logic doesn't have imports statements.    (023)

>>  cf. section 6.3 in the ISO final draft
>>  http://cl.tamu.edu/docs/cl/24707-31-Dec-2006.pdf
>>  See also lines E17 and E20 in the semantics table
>>  in 6.2.  As I put quite a lot of work into
>>  getting this (I hope) right, I guess I am rather
>>  unwilling to admit that it is broken or
>>  irrelevant.
>None of this is to the point, Pat, as "ontology" is not defined in CL.    (024)

It is all based on my identifying "subject 
matter" with "in the universe of discourse". If 
you didn't intend that identification, what is 
the "subject matter" of an ontology?    (025)

>  (For
>the record, though, I think you got the stuff on importation exactly
>right. ;-)
>>  >and we'd better get clear 
>  > >about the connection between the former and the latter.  You seem to 
>>  >be favoring the idea that ontologies are rather more concrete than 
>>  >I'd been thinking.  Do you think it would be better to say that a 
>>  >logical theory is only one of several components of an ontology?
>>  See above. Whatever we decide should be the
>>  definition of "ontology" (God alone knows why we
>>  need a definition, but...), it had better be the
>>  case that the 10,000 or so existing OWL
>>  ontologies turn out to be ontologies by our
>>  definition.
>I would have to agree with that.
>>  Now, *they* are collections of
>>  assertions in OWL (or maybe the XML documents
>>  which encode those assertions, or the RDF graphs
>>  which encode..., etc.; choose your favorite
>  > encoding level). These are finite entities and
>>  their sentences satisfy semantic constraints
>>  which fully determine their meanings. And the
>>  'imports' sentences that they contain are part of
>>  them, just as much as any other sentences.
>Well, I still have the concerns I first expressed above, that this is mixing
>things together conceptually that should be distinguished.    (026)

Why "should be"? I suggest they should NOT be, at 
least in the very strong sense that you are 
implying.    (027)

IHMC            (850)434 8903 or (650)494 3973   home
40 South Alcaniz St.    (850)202 4416   office
Pensacola                       (850)202 4440   fax
FL 32502                        (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (028)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (029)

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