Dear UOS-ers, (01)
At the risk of sending something that may seem out of context, the
following is an abridged (offline) exchange I had with Mike Uschold
about my take on the subject of maturity of ontology technology in
general and integration in particular, based on our actual experience
at Ontology Works in our work with customers. Mike urged me to send
this and I agreed, so here it is.... Editorial comments in brackets
to separate them from what was said at the time. Comments welcome. (02)
.bill (03)
Begin forwarded message: (04)
> From: "Uschold, Michael F" <michael.f.uschold@xxxxxxxxxx>
> Date: March 2, 2006 2:39:43 PM EST
> To: "Bill Andersen" <andersen@xxxxxxxxxxxxxxxxx>
> Cc: <lobrst@xxxxxxxxx>
> Subject: RE: [uos-convene] Essential or Not
>
> -----Original Message-----
> From: Bill Andersen [mailto:andersen@xxxxxxxxxxxxxxxxx]
> Sent: Wednesday, March 01, 2006 11:14 PM
> To: Uschold, Michael F
> Cc: lobrst@xxxxxxxxx
> Subject: Re: [uos-convene] Essential or Not
>
> Right....
>
> BA: [...] Let me toss out some my admittedly incomplete ramblings
> on the subject of maturity, based on my experience building and
> applying ontology-based products. I would seriously enjoy hearing
> both of your perspectives on this, since you've both been around
> the block a lot.
>
> BA: First, let me say that, being a representative of a company
> that is selling ontology-based products, I can't believe that our
> technology is too immature to sell. I may be a lot of things, but
> I'm not dishonest. We do sell what we do because it works.
>
> MU: I'm talking about semantic interoperability technology
> specifically, not ontology technology more generally. The grand
> vision of seamless semantic interoperability across all
> applications at all times in the way that .html works for the web,
> is way way far away. Todays technology is only adequate for point
> solutions under limited sets of circumstances and assumptions. Like
> small scale, or manual mapping, or incomplete/inaccurate mappings
> and translations, etc.
> --
> BA: [Agreed]
>
> BA: Ok, that said, what I meant with that rather flip comment in
> agreement with Mike [about immaturity] is that I believe that we're
> all still a pretty long way from the visions that many of us have
> for the potential of ontology for integration.
>
> MU: OK so we are on the same track here, more or less.
> --
> BA: Let me stress that last part. I mean for *integration*. This
> is one reason we (OWI) have focused only on building better
> database systems. That is an application area where users have a
> concrete need because conventionally-based database projects with
> complex data and queries often fail under the dual stresses of
> design and implementation complexity and complexity = cost. Thus,
> when I say that ontology (and in our experience, the application of
> our ULO) saves money for customers, this is what I mean.
>
> BA: Now, on to integration. I think we're a long way from doing
> that, or even knowing how to do that, assuming we can first figure
> out what it means or even what "ontology" means. First a sidebar
> about logic and OWL.
>
> BA: Even though I love to work with logic, I do believe that many
> are too bent out of shape over it and religious issues surrounding
> it. We're bombarded daily with the "Why don't you use OWL?"
> question. We use a logic programming semantics at Ontology Works
> because it's the only way we (or anyone else) knows of to do the
> things we do fast enough to satisfy our customers -- database
> product customers with LOTS of data. Period. Dot. No OWL except
> in an import/export mode because it just won't (even with the
> proposed extensions in 1.1) do what we need [in terms of
> expressiveness and query speed]. Maybe SWRL or some other rule
> extension will help, but we're not holding our breath and meanwhile
> we'll just keep on working.
>
> BA: What does all that have to do with integration? Well, I think
> it's going to depend on what notion of integration you're after.
> For our [near-term] purposes, integration will mean being able to
> query 20 databases as if they were one and moreover, being able to
> extend the 20-database installation to a 30-database one without
> paying a non-linear marginal cost for the addition of the extra
> databases. If we do that, it will be a HUGE win for ontology and
> we think that a principled ULO is the only way to get there.
>
> MU: you may be right. We could and maybe shuold pose this as
> something that the group could agree on or not, and what the
> arguments are pro and con. Namely, that a principled ULO is the
> only way to get there. Some may prefer a weakened version: there is
> nothing else out there that is as promising.
> --
> BA: [Agreed. Experience has a way of settling these kinds of
> arguments]
>
> BA: As for loftier goals of effecting integration between
> independently constructed logical theories (ULOs if you will but
> that makes no difference [as Menzel stresses]), sure, there are
> lots of theoretical accounts for how that could be done. From a CL
> perspective, just take theory A and theory B (with disjoint
> languages) and link them via bridge axioms between their
> constituent relations (the simplest being equalities to arbitrary
> complex formulas). Then check the thing for consistency.
> Undecidable, but we may be able to get close enough to, with
> empirical testing to fill the gaps, judge this as a workable
> integration of A and B. Another route is to force A and B into a
> decidable logic, like OWL-DL. That's a non-starter as any notion
> of ULO as I understand it will thumb its nose at being limited to a
> dyadic logic. You can do it, but you've probably tossed the baby
> out with the bath water in doing so. Or we can all chuck our
> favorite ULOs and agree on one über-ULO. Also unlikely.
>
> BA: Where does that leave the UOS effort? I think we need to gain
> *experience* using these things *individually* to see what they do
> well, and where they break, and for what applications they do well
> and break. That will cut through a lot of the philosophical
> arguments [because all that matters at the end of the day are
> results]. Once the field has been narrowed and the candidates
> refined through actual use and actual failure in actual
> applications (as opposed to academic posturing), then we can
> decide what's worth the effort to integrate and the best way to do
> it. Meanwhile, having discussions as to scope, structure, and
> content and doing some experiments with logic and theorem proving
> to see how these things line up [as Menzel and Gruninger suggest in
> recent posts] is valuable work well worth funding. But I certainly
> see no great urgency in this - for this reason the big push for an
> "upper ontology summit" has been a mystery to me. I'm wondering
> who really wants this, why *now*, and for what [real-world] purpose.
>
> MU: I'm pretty much in agreement with all this.
>
> So: the only disagreement we [may] have is you wish to make a
> stronger statement about whether there are any viable non-ULO
> approaches to solving 'the semantic interoperability problem'.
> This is getting more into quibble territory.
> --
> BA: Agrees that other approaches may work. The ULO approach is way
> too far from being adequately explored to rule it out in favor of
> other less well explored approaches (e.g. category theory). Again,
> experience with actual information systems will be the deciding
> factor. (05)
Bill Andersen (andersen@xxxxxxxxxxxxxxxxx)
Chief Scientist
Ontology Works, Inc. (www.ontologyworks.com)
3600 O'Donnell Street, Suite 600
Baltimore, MD 21224
Office: 410-675-1201
Cell: 443-858-6444 (06)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/uos-convene/
To Post: mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal: http://ontolog.cim3.net/
Shared Files: http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit (07)
|