RE: Where are the semantics in the semantic Web?

Dear Jacek,

Thanks very for your kind advice. I agree with your explanation as that the
focus of current efforts is on "machine" other than "semantics". Thus as
Uschold inferred, the "machine processible semantics" is just the
Description Logic or whatever logics. Many people asked me off-the-list
what's the difference between what I am proposing from what other people are
proposing. By now, I have to say, the difference is I focus on the semantics
inside the message exchanged between the service providers and service
requesters, while others may deal with the semantics of the framework they
built (WSDL/SOAP, etc.) for exchanging the message but not the message
itself.

I also discussed with many people how can we bridge SW and WS. I think the
solution is just to focus on the message we exchanged over the Internet.
This is the entrance to bridge the two together as I think. The difference
of the message exchanged in SW and WS is that in SW, we exchanged HTML
document, while in WS, we transfer the command to invoke functions on remote
machines to do something. I suggested that such a command be an XML
document. In this way, we can ignore the framework (WSDL, SOAP, REST) but
focus on the semantics of the command document. I think this is the
semantics of the Semantic Web Services. When both SW and WS can exchange the
documents over the Internet, we can bridge them together.

Such a solution has been demonstrated by my pilot projects as I showed to
some people in this group. By exchanging the same XML document as the
command message (service request), I invoked the same functions for
geospatial data processing/analysis and got the same result by either WSDL
or REST Web services. That's why I suggested to Dr. Haas that W3C have a
long term goal to unite both WSDL and REST Web services under the same
semantic Web service framework (I tried to search such information as
semantics for REST Web services but could not find any valuable
information).

As for AI applications in SWS, I think the first step is to describe the
service semantics (the command in XML document). The second step is then to
use RDF/OWL to make the description more accurate and specific for service
discovery and matchmaking. Since AI is an "artificial" intelligence,
machine's intelligence depends on the human design otherwise machine cannot
colone itself. However, human intelligence is also limited since when we
design and write a program, given the example of a if...elseif...else loop,
we understand that we cannot handle every possible case. If we don't know
what may happen, then machine cannot process any extra condition which is
classified into "else" cases.

I strongly suggested that W3C formulate standardized domain-specific service
semantics. If Web service is for sharing the computing resources
interoperablly, then semantic Web services should share service semantics.
In this way, the same semantic description can be deployed by any service
providers to perform the same kind of tasks and every participating party
will understand the meaning of the command to process the service. AI does
not necessarily depend on desciption logics but can be implemented based on
standards and agreements. This has been demonstrated by many researches on
service chaining in geospatial domain, although at that time, the
terminology of Web services was not coined.

It seems W3C needs to build a SWS encyclopedia of standardized service
semantics as the guideline for developing SWS. Otherwise we have to continue
to guess the meaning of data and service semantics by logic which as you
said, could only cover small scopes of the service semantics, thus the logic
approach is of course not suitable for all.

Best wishes,

Xuan

-----Original Message-----
From: Jacek Kopecky
To: Shi, Xuan
Cc: public-sws-ig@w3.org
Sent: 12/7/05 8:41 PM
Subject: RE: Where are the semantics in the semantic Web?

Dear Shi Xuan, 

there are levels of machine processability. The current Web is very much
machine-processable - my browser processes and displays HTML on a daily
basis, and it even processes the more powerful flash programs. And my
SOAP stack processes WSDLs with ease.

With this in mind, you have to see that TimBL's definition of "semantic"
as "machine-processable" does not demand full "machine-processability",
just a higher level of it. The Semantic Web people and the SWS people
have different use cases, they want the machines to do different tasks,
therefore they will represent different information in a
machine-processable way. For SemWeb, it's probably about search by
inference; for SWS it's about discovery and composition of processes.

Whoever criticizes WSDL as "not semantic", really means "not semantic
enough", and that's what WSDL-S wants to mitigate (at least partially),
for a limited set of use cases (again). Nobody here wants the big
solution, as that would be strong AI. What we want is to solve small
well-defined problems, and we try to use the same tools so that we
benefit from unexpected correlations that might arise, among others*.

So what I'm saying is - with SemWeb and SWS we are trying to scope the
AI discussions in smaller compartments, and I'm afraid the recent
threads went well above any of these small scopes.

Best regards,

Jacek

*other benefits of using the same tools is that they get hardened
(debugged) and better known when more people use them, so they get
easier to use. And I'm not talking about WSMO vs. OWL-S - who don't
share much - but I'm talking about W3C trying to standardize something
akin to WSDL-S, building on RDF and XML/WSDL, those are the common
tools.


On Fri, 2005-11-25 at 14:18 -0500, Shi, Xuan wrote:
> When Tim Berners-Lee defined the semantic Web, the word "semantic"
meant
> "machine processable".  Now that Web services are designed for
> "machine-processable", WSDL is criticized as not "semantic". The word
> "semantic" in Semantic Web Services seems different from that in
Semantic
> Web?
> 
> If Tim Berners-Lee's definition is still effective, we can understand
both
> XML and RDF/OWL are machine-processible. Which way we should go? Still
it's
> an issue of agreement and standardization, otherwise, we have to
continue
> our debate. Especially according to Tim Berners-Lee's definition, WSDL
is
> machine-processible then why should we again add "semantics" onto such
> machine-processible (thus "semantic") WSDL document? Or we are talking
about
> something different in the domains of SW and SWS?
> 
> -----Original Message-----
> From: Joachim Peer
> To: jeff@inf.ed.ac.uk
> Cc: Harry Halpin; public-sws-ig@w3.org
> Sent: 11/25/05 9:00 AM
> Subject: Re: Where are the semantics in the semantic Web?
> 
> dear all,
> 
> i've followed this thread with great interest. i have tried to
summarize
> some technical (pro XML) arguments in a little paper which is attached
> to this mail
> 
> kind regards!
> Joachim
> 
>  <<rdfxml.pdf>> 
>

Received on Thursday, 8 December 2005 14:48:07 UTC