Re: [OEP] OWL and Semantic interoperability $swbpd

All

What seems to me to be the most important point from all of this is the one that
I think none of us would dispute, that OWL is incomplete and lacks certain
features required by many applications.   Mike is a major user for semantic
interoperability, and I believe we have to take his comments seriously.  Some of
them chime all too well with my own experience.

Mike specifically cites numerical calculations such as coordinate and unit
transformations as a gap.  Our experience would strongly support this along with
the need to incorporate numerical inequalities into the framework so that I can
say, and classify on the basis of, a legal speed is from >=10 & =<30.  Apart
from clinical terminology, where it would be useful but not critical because
normal ranges are typically factored into a different bit of the architecture,
almost all the other non-trivial applications we know of require the ability to
use and process numerical (and string) concrete data types.  Not being able to
do so in OWL is a serious barrier to its use..

There are other constructs which are not present which many applications do
need, leaving aside my pet peve, there is the lack of the ability to negate a
property, so that it is impossible to express notions such as all surgeons of
type X are authorised to perform all operations of type Y. (It is part of the DL
fragment, there are papers on how to do it, but it is in neither the language
standards nor any of the classifiers.)

There are lots of other examples.

I am less concerned whether it is SWRL or some other formalism that fills these
gaps than that they be filled, and urgently.  My  experience at the moment is
that too often I find myself trying to convince people to move from
representations that meets their needs to one which meets needs they weren't
aware of while failing to meet the ones that are met by their current solution.

Identifying what is needed to achieve semantic interoperability and the gaps
between that and the existing OWL (and other) standards is a worthwhile and
urgent exercise.

Regards

Alan


"Uschold, Michael F" wrote:

> Thanks Deborah.
>
> Can you please give some good examples of the richness that OWL provides for
> supporting semantic interoperability. Then I will include them.
>
> It was not my intention to be negative or controversial.
>
> I want to start off the note with the things that OWL does well, and then
> note the limitations.
>
> Thanks,
>
> Mike
>
> -----Original Message-----
> From: Deborah L. McGuinness [mailto:dlm@ksl.stanford.edu]
> Sent: Monday, February 07, 2005 10:00 AM
> To: Uschold, Michael F
> Cc: Jim Hendler; public-swbp-wg@w3.org; Jones, David H
> Subject: Re: [OEP] OWL and Semantic interoperability $swbpd
>
> a couple of points that come out for me from the thread so far:
>
> 1 - i am also reactive to the feeling that is left from the note that owl is
> inferior.
> i find its expressive power to be superior to other things that people have
> tried to use for semantic interoperability in at least the database world.
> 2 - we might try to keep subjective words like inferior out of the note
> 3 - i think it is clear that we need to be precise with what definition(s) we
> are considering for semantic interoperability
> 4 - if we are using a kind of mapping as a definition, i like think one of the
> strong claims of owl (and all description logics)
> is that they can be used to express  necessary conditions for membership in a
> class, and
> necessary and sufficient conditions for membership in a class,
> and then provide subsumption and recognition reasoners to identify when
> classes are subclasses of other and to identify when
> instances are members of a class.
> Additionally, OWL provides convenient ways to make equivalence statements that
> can be used to facilitate mappings.
> (additionally it provides convenient ways to make different from statements.)
>
> one can look at history and go back to the prose family of configurators that
> used an early DL - CLASSIC - and could have been viewed to be solving semantic
> interoperability problems since it interfaced with dozens of databases.
>
> Deborah
>
> Uschold, Michael F wrote:
>
> Thanks Jim for your input.
>
>
>
> More concrete dicussion can take place when the note has some meat on it
>
> for you and others to assess. However, I can respond to some of your
>
> points.
>
> See [MFU] below.
>
>
>
>
>
> -----Original Message-----
>
> From: Jim Hendler [mailto:hendler@cs.umd.edu]
>
> Sent: Sunday, February 06, 2005 2:29 PM
>
> To: Uschold, Michael F
>
> Cc: public-swbp-wg@w3.org
>
> Subject: Re: [OEP] OWL and Semantic interoperability
>
>
>
> for the note is:
>
>
>
>> This note addresses the role of OWL in overcoming problems of semantic
>>
>> heterogeneity.  We briefly characterize what we mean by semantic
>>
>> interoperability, and what the challenges are. We describe some OWL
>>
>> constructs that are designed to support semantic interoperability and
>>
>> illustrate them with examples. We highlight their strengths and
>>
>> limitations.  The main message is that OWL is no silver bullet for the
>>
>> general problem of achieving semantic interoperability.  The support
>>
>> provided is very limited.  Many of these limitations will be overcome
>>
>>
>>
> by
>
>
>
>> the Semantic Web Rule Language (SWRL) that is currently under
>>
>> development.
>>
>>
>>
>
>
> Mike - let me put this in a friendly way -- I don't like this at all.
>
> First of all, when you compare OWL to any of the alternatives it is
>
> way better -
>
> [MFU]
>
> Which alternatives to OWL is OWL superior to? In what way? Perhaps you
>
> have something different in mind by 'semantic interoperability' than I
>
> do.
>
>
>
> When I compare OWL to alternatives (such as CWM or SWRL, and perhaps
>
> FLogic) for providing semantic interoperability , OWL comes up short
>
> (see below).
>
> ==
>
>
>
> If you stopped after the sentence "OWL is no silver
>
> bullet ..." I'd be fine.  However, you go on to say "the support
>
> provided is very limited" which I don't agree with, and which also is
>
> a subjective statement at best.
>
>
>
> [MFU] By limited, I meant it cannot support some extremely simple and
>
> minor things that are critical to many real world interoperability
>
> scenarios. e.g.  doing arithmetic to convert between different units
>
> (e.g. feet/meters). One can do this very easily using CWM, and SWRL.
>
> That assertion is not subjective. Going from this fact to an
>
> interpretation that OWL is "very limited" is open to discussion, and
>
> perhaps should be re-worded for different emphasis. For example: "OWL
>
> provides x,y and z to support interoperability. a,b,and c are not
>
> supported by OWL. The extent to which this is a limitation depends on
>
> the nature of one's application." would probably be better. In the
>
> experience of myself and some of my colleagues, OWL's support for
>
> interoperability is indeed very limited. Without simple arithmetic and
>
> rules, most of what we need is not supported.
>
>
>
> I would be very grateful if you [Jim or anyone] could give some examples
>
> of non-toy real-world problems of interoperability that can be solved
>
> with ONLY the support that OWL provides. That would be a good way to
>
> start off on a postive note; and then move on to some of the
>
> limitations, concluding that OWL is no silver bullet, and that some kind
>
> of rule language is needed to perform interoperability beyond what OWL
>
> currently gives.
>
>
>
> ===
>
>
>
> I would even further object to the
>
> last sentence - first, SWRL has no official status and may not even
>
> be the eventual choice for a standard, second SWRL has not been
>
> tested in interoperability issues, and it is unclear to me why more
>
> expressivity should increase the ability to do reuse -- if anything I
>
> would actually expect the reverse in practice...
>
> [MFU] I mean to say that more expressivity increases the ability to
>
> express and execute mappings between one ontology and another for the
>
> purpose of interoperability. I'm not making any claims about reuse.
>
>
>
> ===
>
>   So let's keep this factual and focused, and leave the hypotheticals
>
> out of WG notes
>
> [MFU] That is my intent. I will rely on you and others to identify
>
> hypothethicals that may unintentionally find their way into the note.
>
> ===
>
>   -JH
>
> p.s. And in case anyone is interested, I would like to deny in the
>
> strongest possible terms that "The support provided (for
>
> interoperability) in OWL is very limited",  the URI/RDF basis of OWL
>
> makes the sharing of terms possible in a way that no KR langauge in
>
> the past has ever had, and is a good part of the reason that OWL is
>
> now the most used KR langauge in the history of AI.
>
>
>
> [MFU] Again, I'm not making claims about reuse and sharing, per se. What
>
> do you see as the relationship/connection between reuse/sharing and
>
> interoperability?  This note does not focus on reuse/sharing per se. It
>
> is no doubt part of the bigger picture of semantic interoperability that
>
> I should perhaps explore more.
>
>
>
> As I read and reread your comments, it seems likely that we have two
>
> fairly different things in mind by 'semantic interoperability'. What do
>
> YOU have in mind?  Perhaps when you think of semantic interoperability,
>
> you are mainly thinking of the reuse and sharing of ontologies and
>
> ontology terms using the URI/RDF basis?  Perhaps THIS is what you see as
>
> being superior to all the alternatives? On that we can agree.
>
>
>
> The way that OWL is inferior, is in its abilty to express mappings
>
> between ontologies to that applications using different ontologies can
>
> interoperate. This is what I mean by semantic interoperability.
>
> ===
>
>
>
>
>
>
>
>
>
> Professor James Hendler                   Director, Semantic Web and
>
> Agent Technologies
>
> Maryland Information and Network Dynamics Lab.   301-405-2696
>
> 8400 Baltimore Ave, Suite 200                    301-314-9734 (Fax)
>
> College Park, MD 20742
>
> http://www.cs.umd.edu/users/hendler
>
>
>
>
>
> --
>
>  Deborah L. McGuinness
>
>  Co-Director 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
>
--
Alan L Rector
Professor of Medical Informatics
Department of Computer Science
University of Manchester
Manchester M13 9PL, UK
TEL: +44-161-275-6188/6149/7183
FAX: +44-161-275-6236/6204
Room: 2.88a, Kilburn Building
email: rector@cs.man.ac.uk
web: www.cs.man.ac.uk/mig
        www.opengalen.org
        www.clinical-escience.org
        www.co-ode.org