Re: What's the point in using OWL-S?

> [Gerhard Austaller]

> I still have not understood what exactly an agent can achieve with the owl-s
> description of a service.
> 
> Let's take the Bravo Air example. For an agent to book a ticket it has to
> "understand" the concepts used there. But where's the big advantage in
> having an interface (in WSDL) defined or an annotated WSDL like WSDL-S? I
> admit having the exact order of steps defined to achieve something is a step
> in the right direction...
> 
> Now I want my agent also to book a ticket at another airline. So if this
> airline uses another ontology that ontology also as to be understandable to
> the agent made by an programmer. So do we assume that all airlines agree on
> one ontology and further also on the process how to do something?
> 
> Do I miss something? Can please somebody explain or give me pointers what's
> a programmer can gain from an owl-s description?

Your question is phrased in a confusing (and I think confused) manner.

When you say, "But where's the big advantage in having an interface
(in WSDL) defined or an annotated WSDL like WSDL-S?", it sounds like
you're questioning WSDL.  So it's not Owl-S per se that you're having
trouble with, but the whole semantic-web-services idea, right?

[from your second message]
> One example that can be often found in semantic web examples is the e.g.
> travel agency agent that can negotiate with several source like airlines
> agents, hotel agents about prices, travel times, room reservations and so
> on.
> 
> But if even the hotels can't agree on standardized ontologies and/or
> interfaces how should this work. I don't think that a travel agency wants to
> implement an agent that support dozens of interfaces to be able to talk to
> the most important airlines and hotel booking systems... 
> 
> Given this example I am wondering what is needed at least to make this
> scenario working in near future.
> 
> I know that there is research on matching different ontologies and finding
> "somehow" corresponding concepts in the ontologies but I don't think that
> this will work in near future.

It seems to me that hotels will be strongly motivated to agree on
standard notations for reservation services.  Furthermore, if the
problem is simply to book a hotel room (or 97 hotel rooms for 97
different clients) then there will be little need for "semantic" web
services (compared to web services without the "S" word).

In my opinion, the place where Owl-S may be useful is when several web
services are needed to solve a problem stated in terms that don't
mention web services (as such) at all.  The problem might be to
reserve a room and an airplane flight for a conference at a certain
location and time.  In order to solve this problem, a reasoning system
needs to find web services that can achieve subgoals that the problem
comprises.  Owl-S provides a framework for connecting two things:

 - At the abstract level, expressing what subgoals are achieved by
   message exchange with a web service.
 
 - At the concrete level, expressing how messages are encoded using
   (e.g.) SOAP

The ontology-matching problem enters into this picture, but not as the
central problem.

[Note the qualification: "in my opinion."]


                                         -- Drew McDermott
                                            Yale University
                                            Computer Science Department

Received on Saturday, 3 June 2006 16:16:16 UTC