W3C home > Mailing lists > Public > www-webont-wg@w3.org > November 2002

Re: SEM: Light review of semantics document

From: Deborah McGuinness <dlm@ksl.stanford.edu>
Date: Mon, 11 Nov 2002 10:37:33 -0700
Message-ID: <3DCFEADD.113160E1@ksl.stanford.edu>
To: Christopher Welty <welty@us.ibm.com>
CC: www-webont-wg@w3.org

I have had a lot of discussions with people about what ontologies are in
common usage.  I think the only thing that we all agree on is that there is a
spectrum of descriptions for ontologies.  I dont know that we want to get into
a big description of it since i dont think the discussion is short.  if people
want to refer to something on the topic, i wrote about it in an "ontologies
come of age" paper [1].  that paper includes a spectrum that a number of us
came up with and i refined after a panel at aaai on ontologies.

Thus, i would be careful below with wording such as ontologies
referring "to what we "traditionally" ... call ontology in Computer Science,
and "knowledge base" to refer to a mixture of ontology and instance data."

including class descriptions in our ontologies means including some instances
because of hasValue and oneOf.
Since someone will ask the question - I wanted to head it off here.  We should
NOT say  that owl documents with instances in them are not to be referred to
as ontologies.
I agree with jim's point though that a document with only instances and their
associated class tags  in it is not what i think of as an ontology and also is
not something that i would like to promote as a way of using the term.

i am not opposed to instance data set for this.

i do not usually make a big deal of a distinction between knowledge bases and
ontologies since the overlap is too great and many people consider knowledge
bases ontologies.  I had not thought about this point directly in the guide
writing but it probably would not create as much confusion if we downplay that
distinction in the guide.

deborah

[1]
http://www.ksl.stanford.edu/people/dlm/papers/ontologies-come-of-age-abstract.html

Christopher Welty wrote:

> Jim,
>
> If by, "I believe strongly that this is not a critical issue of language
> design, it's simply a suggestion we develop consistent terms so we
> get our message out," you mean that you just want a term that we will
> agree to employ in our human interchange, and not something in the OWL
> language, then that's fine, make something up and we'll add it to the
> Guide.
>
> It appeared from the discussion that Jeff and you wanted something more
> formal, that required an adjustment to the meaning of the owl:ontology
> tag.
>
> In the Guide, Mike and I chose to employ "ontology" (not the tag, but the
> natural language term) to refer to what we "traditionally" (it's a short
> tradition) call ontology in Computer Science, and "knowledge base"
> to refer to a mixture of ontology and instance data.  We did not draw
> the distinction you want there, but if I understand you
> correctly, the Guide is right place to talk about this.
>
> -Chris
>
> Dr. Christopher A. Welty, Knowledge Structures Group
> IBM Watson Research Center, 19 Skyline Dr.
> Hawthorne, NY  10532     USA
> Voice: +1 914.784.7055,  IBM T/L: 863.7055
> Fax: +1 914.784.6078, Email: welty@us.ibm.com
>
> Jim Hendler <hendler@cs.umd.edu>
> Sent by: www-webont-wg-request@w3.org
> 11/09/2002 06:59 PM
>
>         To:     Christopher Welty/Watson/IBM@IBMUS, www-webont-wg@w3.org
>         cc:
>         Subject:        Re: SEM: Light review of semantics document
>
>
>
> At 3:00 PM -0500 11/8/02, Christopher Welty wrote:
> >I have to agree with Pat here, guys.  The meaning of an "owl:ontology"
> tag
> >inside an RDF document is simply that the document contains OWL syntax,
> >not some hard to pin down notion of a separation between definitions and
> >data.
> >
> >What you, Jeff and Jim, want to accomplish can be done with comments,
> >since the distinction you want does not exist in the language nor in its
> >interpretation - it only exists in the minds of some people.  I think
> what
> >you want is sort of like the distinction C compilers make for ".h" files
> -
> >in point of fact there is no difference between  the contents of a ".h"
> >and ".c" file, just a methodology enforced by some compilers.
> >
> >That using "ontology" to describe a set of instances does not match your
> >definition is not, I think, the point.  Maybe the tag is inappropriately
> >named.  But don't get me started on mis-named tags.  "Property" is
> already
> >a LOT worse.
> >
> >-Chris
>
> Chris/Pat - I think you guys misunderstood me - I believe that all of
> these things are OWL documents, but I'm concerned with a small matter
> of usage.  The way I see it, there are documents which are clearly
> owl ontologies because they define terms and properties and the like.
> There are also owl documents that only use those terms and, in fact,
> there is no reason that there will be any trace of any OWL vocabulary
> in those documents.  For example, if Chris defines an ontology about
> people, I could have a document which contains only the following:
>
> Namespace definitions to RDF and to Chris' document
>
> <rdf:RDF>
>    <chris:person rdf:id="Hendler" />
> </rdf:RDF>
>
> by the definition "uses owl terminology" this is NOT and owl
> document.  By the definition "uses terms from an owl ontology" this
> is an Owl document.
>
> So I am asking for terminology that would
>     i. let me differentiate this document from an arbitrary RDF
> document (and Pat, please note I wasn't being anti-logical, but it
> seems to me we don't need this distinction to have a logical meaning
> in the formal sense -- I'm simply looking for a common term to mean
> RDF documents that are expecting to be linked to owl ontologies) --
> Jeff called this a data document, which Pat didn't like.
>    ii. lets me differentiate this kind of document from an owl
> document which does contain class and property definitions and
> restrictions.  I DO KNOW what to call the ones that have that (an
> ontology), but not what to call the other ones.
>
>   I believe strongly that this is not a critical issue of language
> design, it's simply a suggestion we develop consistent terms so we
> get our message out.
> Technically, it is clear to me the document above is an RDF document
> - it would use the RDF Model Theory and all would be happy.  But
> colloquially, we need to be able to discuss these documents with a
> term that people in the outside world can understand.
>
>   In class, I refer to these as "Owl data sets" and the students get
> it, I'd be happy with that term.
>
> So, I ask Pat/Chris and anyone else inclined to help out:
>
>   what name shall we use for documents that are in the class with the
> following properties:
>
> Document a rdf:RDF document AND
> Document uses terms from a owl ontology document AND
> Document NOT a owl ontology document.
>
> IMHO, If we call such a document an "ontology," we're going to
> confuse  a lot of people.
>
> Finally, such documents not only will, but DO exist (in case someone
> is going to argue that this is specious) -- there's a number of
> examples in [1], for example [2] which has no hint of the daml
> namespace in it, but is linked to an ontology which is clearly
> defined in DAML [3].
>
> [1] http://www.daml.org/data
> [2] http://www.daml.org/2002/02/chiefs/af.daml
> [3] http://www.daml.org/2002/02/chiefs/chiefs-ont
>
> >Dr. Christopher A. Welty, Knowledge Structures Group
> >IBM Watson Research Center, 19 Skyline Dr.
> >Hawthorne, NY  10532     USA
> >Voice: +1 914.784.7055,  IBM T/L: 863.7055
> >Fax: +1 914.784.6078, Email: welty@us.ibm.com
> >
> >
> >
> >
> >pat hayes <phayes@ai.uwf.edu>
> >Sent by: www-webont-wg-request@w3.org
> >11/08/2002 11:34 AM
> >
> >         To:     Jim Hendler <hendler@cs.umd.edu>
> >         cc:     www-webont-wg@w3.org
> >         Subject:        Re: SEM: Light review of semantics document
> >
> >
> >
> >
> >>Sorry Pat, but I gotta agree with Jeff on this one- you signed onto
> >>this group which had "ontology" in the title, fully knowing by
> >>charter we would create something called an ontology language.
> >
> >Well, sure, but apparently what you and Jeff meant by 'ontology'
> >wasn't what I meant. Thats the trouble with English, right? Which is
> >one reason why we are doing all this in the first place...
> >
> >><owl:ontology>
> >>    statements
> >></owl:ontology>
> >>
> >>was in DAML+OIL, has been in OWL from day 1, a prereq for WG members
> >>was familiarity w/D+O, so you should have been aware that was there.
> >
> >I have no problem with that, but I have always understood this to
> >simply be an XML marker for the presence of OWL syntax in the RDF
> >graph. Why do we need to say anything more than that? "owl:ontology"
> >isn't in the graph, right? So its not in the namespace, and it has no
> >semantics. If "owl:ontology" is in the OWL namespace, then we ought
> >to say what an RDF triple which includes that name means.
> >
> >>The issue we need to address is that IN ENGLISH USAGE (not formal logic)
> >
> >I do wish you would make at least an effort to disguise your built-in
> >anti-logical knee-jerk, Jim. It just gets in the damn way. What we
> >are all doing here, whether you like it or not, is using ENGLISH to
> >talk ABOUT a FORMAL logical language. When using ENGLISH it is often
> >a good idea to use words which refer to concepts that actually make
> >some sense in the context being talked about.
> >
> >>there is a need for us to say whether there is a difference between
> >>documents that look like
> >>
> >><rdf:RDF>
> >>
> >><owl:ontology ... />
> >>...
> >>
> >><owl:class rdf:about="Moose">
> >>    <owl:restriction>
> >>      etc
> >>    </owl:restriction>
> >></owl:class>
> >>
> >></rdf:RDF>
> >>
> >>and documents that look like
> >>
> >><rdf:RDF>
> >>
> >><a:Moose rdf:id="MyMoose />
> >>
> >></rdf:RDF>
> >>
> >>(and those which mix some of both).
> >
> >The difference is that one of them contains OWL syntax and the other
> >does not. End of story.
> >
> >What about the many other cases, such as  <rdf:RDF>....</rdf:RDF>
> >which contains non-ground RDFS, say? What about one of those that
> >contains RDFS which would break fast-OWL? What do we call those?  If
> >some RDF uses rdf:bag, is it instance data? What about an rdf:List,
> >described using bnodes? You (and Jeff) are confusing two different
> >distinctions: ground versus non-ground, and RDF vs OWL. That is a
> >dangerous confusion to incorporate into an official terminology.
> >
> >>Jeff's usage is consistent with the outside world's usage, and I
> >>suggest if we don't use it we will confuse everyone in the world
> >>except for logicians
> >
> >I'm tempted to respond that anyone who you would classify as a
> >non-logician is already confused in any case. But I won't.
> >
> >>-- given that, I'd suggest we use it -- i.e.
> >>
> >>Ontology documents are those that define classes and properties.
> >>Instance documents are defined by using RDF to produce instances (or
> >>individuals) that are members of those classes with those properties.
> >
> >Sorry, I object to this, because it doesn't make sense. You are
> >presuming something that is false: that RDF can only be used to
> >describe ground facts. We ought to use a naming convention that warns
> >the world not to get this confused, rather than casting the confusion
> >in stone.
> >
> >I know that RDF is often used for instance data, but it can be used
> >for other things. I frequently meet people in the DAML community who
> >are surprised to hear that RDF allows bnodes; there is deployed
> >software with serious bugs arising from the misunderstanding that RDF
> >is used only for ground facts.  Also, this convention makes nonsense
> >of the fundamental presumption that one can merge RDF graphs. If
> >OWL/RDF really is RDF then this terminological usage becomes
> >nonsensical even when applied to actual documents. It already doesn't
> >make any semantic or operational sense.
> >
> >Pat
> >
> >>
> >>OWL documents include each or documents which combine both.
> >  >
> >>I'm happy if someone wants a different term for "instance" documents.
> >>
> >>I think the above is consistent with our current documentation.  I'm
> >>happy to see someone suggest rewording the above (Written quickly
> >>and not formally) in a way that is more technically correct -- but
> >>this is how most of the rest of the world will refer to what we
> >>have, so we should make it easy for them....
> >>   -JH
> >>
> >>
> >>
> >>
> >>
> >>
> >>   1:42 PM -0600 11/7/02, pat hayes wrote:
> >>>>Pat, there is a terminology problem here. What you and Peter call
> >>>>ontologies are different from what I call ontologies.
> >>>
> >>>Yeh, I had that impression. My problem is that I don't really know
> >>>what you are talking about.  I have never come across any useful
> >>>definition of "ontology" in our non-philosophical sense other than
> >>>something like "set of sentences" or maybe a document containing a
> >>>a set of sentences, etc.. If there is a real difference in your
> >>>mind between ontologies and other OWL thingies, then we ought to
> >>>get this clear and incorporate it into the language in some way.
> >>>
> >>>>My practical
> >>>>definition is that OWL ontologies are only those OWL documents that
> >>>>include the <owl:Ontology> tag.
> >>>
> >>>Hmm, I have to confess that I wasn't aware that 'owl:Ontology' was
> >>>in the OWL namespace. What is it supposed to mean?? Does it appear
> >>>in the RDF graph anywhere?
> >>>
> >>>But OK, an ontology is a *document*. In what language? I'm guessing
> >>>it has to be in  OWL/RDF/XML, right? So an OWL/RDF graph is not an
> >>>ontology(?)
> >>>
> >>>>All other OWL documents are not OWL
> >>>>ontologies. Now, you are correct that a document with <owl:Ontology>
> >>>>could consist of nothing but ground facts, and as such you don't
> >>>>technically need to have a separate class of document for data.
> >However,
> >>>>the fact is, people only use the <Ontology> tag when they are defining
> >>>>vocabularies (this statement is based on common usage in DAML). Are
> you
> >>>>suggesting that these people should include <Ontology> tags is all of
> >>>>their documents (see daml.org's list of data sets for a number of
> >>>>examples of DAML documents without these tags)?
> >>>
> >>>I really don't give a rats about this tag, to tell you the truth,
> >>>but certainly people should somehow mark their OWL as being OWL; if
> >>>they don't, then they can't complain if an OWL engine misses it
> >>>entirely. We might want to follow RDF's lead and register an OWL
> >>>media type, though I think that idea is wrong-headed, myself. All I
> >  >>care about is that we have some way to detect well-formed OWL which
> >>>is being asserted. Well-formed OWL means what it means as defined
> >>>by the OWL specs. The distinction between ground and non-ground OWL
> >>>is unimportant, seems to me, and there is no need to even refer to
> >>>it. If some piece of OWL has 10|6 ground facts and one non-ground
> >>>fact, I'm cool with that. What would you call it? Data with a dash
> >>>of ontology?
> >>>
> >>>>Or are you suggesting
> >>>>that we should call these ontologies too?
> >>>
> >>>If we use the term at all, then yes, they are ontologies, in much
> >>>the same sense that a gazetteer is a book.
> >>>
> >>>>   I think the later would really
> >>>>confuse users to call every document an ontology, but only some
> >>>>ontologies are <Ontology> ontologies. In any case, all of our
> documents
> >>>>need to be a lot more clear about terminology (e.g., which definition
> >of
> >>>>ontology does our WG use) and about how people should use ontologies
> to
> >>>>describe real content.
> >>>
> >>>As to the last point, the distinction between ontology and data
> >>>just seems to make things more confusing, suggesting a distinction
> >>>in meaning that isn't there.
> >>>
> >>>Pat
> >>>
> >>>>
> >>>>Jeff
> >>>>
> >>>>pat hayes wrote:
> >>>>>
> >>>>>   >pat hayes wrote:
> >>>>>   >>
> >>>>>   >>  >Here's some initial comments on the Semantics document dated
> >Nov. 3:
> >>>>>   >>  >
> >>>>>   >>  >1) Sect. 2.2. The syntax needs the ability to represent
> >>>>>documents that
> >>>>>   >>  >consist soley of facts (that is, something other than
> >ontologies).
> >>>>>   >>
> >>>>>   >>  ? Can you explain what you mean by "other than ontologies" ?Do
> >you
> >>>>>   >>  mean, not in OWL?
> >>>>>   >>
> >>>>>   >
> >>>>>   >Part of this depends on what you consider OWL. From your
> response,
> >I
> >>>>>   >assume that you think of OWL as just a language for defining
> >ontologies,
> >>>>>   >and that you must use it with RDF in order to describe data
> >>>>>
> >>>>>   No. I fail to see the distinction you are drawing between
> 'ontology'
> >>>>>   and 'data'. I don't know what you mean by this, or what importance
> >it
> >>>>>   has. One can have valid OWL documents which consist of nothing but
> >>>>>   ground RDF facts. So?
> >>>>>
> >>>>>   >(e.g., a
> >>>>>   >product catalog, a univeristy's course offerings, etc.). I tend
> to
> >think
> >>>>   > >of OWL as an extension to RDF, so this data is still part of
> OWL,
> >it
> >>>>>   >just has the standard RDF syntax.
> >>>>>   >
> >>>>>   >In any case, our model theory must talk about data to the same
> >extent
> >>>>>   >that it talks about ontologies.
> >>>>>
> >>>>>   It does. It always has done. What is the problem?
> >>>>>
> >>>>>   Pat
> >>>>>
> >>>>>   --
> >>>>>
> ---------------------------------------------------------------------
> >>>>>   IHMC                                    (850)434 8903   home
> >>>>>   40 South Alcaniz St.                    (850)202 4416   office
> >>>>>   Pensacola                               (850)202 4440   fax
> >>>>>   FL 32501                                        (850)291 0667 cell
> >>>>>   phayes@ai.uwf.edu
> http://www.coginst.uwf.edu/~phayes
> >>>>>   s.pam@ai.uwf.edu   for spam
> >>>
> >>>
> >>>--
> >>>---------------------------------------------------------------------
> >>>IHMC  (850)434 8903   home
> >>>40 South Alcaniz St.                                           (850)202
> >4416   office
> >>>Pensacola                                                      (850)202
> >4440   fax
> >>>FL 32501  (850)291 0667    cell
> >>>phayes@ai.uwf.edu
> >http://www.coginst.uwf.edu/~phayes
> >>>s.pam@ai.uwf.edu   for spam
> >>
> >>
> >>--
> >>Professor James Hendler           hendler@cs.umd.edu
> >>Director, Semantic Web and Agent Technologies             301-405-2696
> >>Maryland Information and Network Dynamics Lab.            301-405-6707
> >(Fax)
> >>Univ of Maryland, College Park, MD 20742                  240-731-3822
> >(Cell)
> >>http://www.cs.umd.edu/users/hendler
> >
> >
> >--
> >---------------------------------------------------------------------
> >IHMC  (850)434 8903   home
> >40 South Alcaniz St.                                             (850)202
> >4416   office
> >Pensacola                                                        (850)202
> >4440   fax
> >FL 32501  (850)291 0667    cell
> >phayes@ai.uwf.edu
> >http://www.coginst.uwf.edu/~phayes
> >s.pam@ai.uwf.edu   for spam
>
> --
> Professor James Hendler    hendler@cs.umd.edu
> Director, Semantic Web and Agent Technologies              301-405-2696
> Maryland Information and Network Dynamics Lab.             301-405-6707
> (Fax)
> Univ of Maryland, College Park, MD 20742                   240-731-3822
> (Cell)
> http://www.cs.umd.edu/users/hendler

--
 Deborah L. McGuinness
 Knowledge Systems Laboratory
 Gates Computer Science Building, 2A Room 241
 Stanford University, Stanford, CA 94305-9020
 email: dlm@ksl.stanford.edu
 URL: http://ksl.stanford.edu/people/dlm
 (voice) 650 723 9770    (stanford fax) 650 725 5850   (computer fax)  801 705
0941
Received on Monday, 11 November 2002 12:34:03 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:55 GMT