RE: Where are the semantics in the semantic Web?

Of course, the hotel service request description is not a standard but just
a sample for discussion. We need to add some other elements, such as
location, type of hotel, etc. in to this service description file. People in
this domain need to negotiate to reach an agreement. Otherwise, we have to
guess what we are talking about by logic even if the topic is exactly the
same.

But that's not the point for this discussion. Professors, please do not
ignore my focus in the discussion as, we should separate service description
from any technology to develop the service. Let's see another example:

An airline company has a Web service for selling ticket, thus its Web
service may have an operation named as "sellTicket".

A airline ticket agency provides a Web service to help its customer to buy
ticket, thus its Web service may have an operation named as "buyTicket".

A travel agency provides Web services to help its customer to buy airline
ticket, train ticket, etc. thus its Web services may have one operation
named as "buyAirlineTicket".

And we can have many more samples for this instance.

Since the semantics of such Web services are the same, what can be expected
by adding semantic annotations onto WSDL elements? By logic, since the
meaning of such service and functions are the same, can we say, the
semantics of such operations will be the same, which then means:

sellTicket = buyTicket
buyTicket = buyAirlineTicket
sellTicket = buyAirlineTicket
... ... ... ...

Is the result a semantic chaos? However, if we separate service description
from any concrete WSDL file, there is no chaos since service description is
described without relation with any operations, objects, etc. inside WSDL
document.

<ServiceRequest>
....
ALL requirements are described here ...
....
</ServiceRequest>

Actually, we can use either OWL-S or WSMO to describe such service
semantics. That's my suggestion to OWL-S and WSMO: remove any relationship
with WSDL grounding and merge diffferent parts into one single document
since people may wish to search the requirement of the input/output
variables. In this way, such service description can be used by either WSDL
Web service or REST Web service.

Again, I can demonstrate such statement by real project development:

1. the following two Web services perform exactly the same function, thus
they should have the same service semantics.

http://157.182.136.51/agswsprojs/geoWebService/Service1.asmx?WSDL 
http://157.182.136.51/agswsprojs/geoWebService/Service5.asmx?WSDL

Service1 has one function, two input variables, and one output varible.
Service2 has one function, one input variable, and one output varible. In
such a simple example, how can you add semantic annotations into such
different WSDL documents and tell people they are the same?  

2. the following WSDL (by SOAP) and REST (by HTTP/POST) Web services share
the same service request document and perform exactly the same function and
generate exactly the same result.

http://157.182.136.51/agswsprojs/geoWebService/Service5.asmx?WSDL 
http://157.182.136.51/agswsprojs/HttpService/getService.aspx (HttpServer)

Please remember W3C defined Web service as "A Web service is a software
system designed to support interoperable machine-to-machine interaction over
a network". So we should develop something that can be used by either
WSDL/SOAP or any other ways for the interoperable machine-to-machine
interaction over a network. Adding semantic annotation onto WSDL cannot
match such requirement but may lead to semantic chaos and we should avoid
this potential problem.

Regarding XML, RDF, OWL, etc. I don't really care about which way is better
for service description. RDF of course has advantages to give a more
accurate definition and I tried to use it in my paper "Removing Syntactic
Barriers for Semantic Geospatial Web Services" (URL:
http://www.ucgis.org/summer2005/studentpapers/shi.pdf ) presented at UCGIS
Summit and your critical comments and advice to it will be greatly
appreciated.


Happy Thanksgivings and have a nice holidy!

-----Original Message-----
From: Jim Hendler
To: drew.mcdermott@yale.edu; 'public-sws-ig@w3.org '
Sent: 11/22/05 6:08 PM
Subject: RE: Where are the semantics in the semantic Web?


Xuan Shi -
   Let me once again point you to the slides from my XML talk [1] - I 
contrasted Xlink (which is essentially identical to what you propose 
to do for services) to RDF, showing why they are crucially different. 
There's a big difference between what you have below (because I am 
unsure what "roomtype" is and what values are allowed) and 
http://ex.org/hotel#roomtype which could dereference to an RDFS or 
OWL document which would exactly answer that question.  I could also 
then tell if Holiday Inn's "roomtype" and one at some Inn in Japan or 
China was the same thing, or something that might be different -- 
these are not minor differences - the focus on links is crucial to 
understanding the Semantic Web as I said in that talk
  -Jim Hendler

[1] http://www.cs.umd.edu/~hendler/presentations

At 15:32 -0500 11/22/05, Shi, Xuan wrote:
>My suggestion is that service description can be separated from service
>development. Let's describe the service first. Supposed we have such a
>service description:
>
><ServiceRequest>
>    <Service Name="SearchHotelInformation">
>       <Function Name="WebService.Hotel.SearchInformation">
>          <InputVariables>
>             <CheckInDate></CheckInDate>
>             <CheckOutDate></CheckOutDate>
>             <NumberOfCustomer></NumberOfCustomer>
>             <RoomType></RoomType>
>          </InputVariables>
>       </Function>
>    </Service>
></ServiceRequest>
>
>it's then easy to develop such a Web service.
>
>If such service description can be a domain standard, then ALL hotel
service
>providers have to follow such standard to develop their Web services in
any
>way they want by reading this request document as their starting point
to
>process the request (the format of response should be standardized
also).
>
>The problem to your "interesting question" is if such developers do not
>follow the standards since we can develop Web services in anyway we
want,
>then there may be no automated agent to communiate each other. This was
>discussed as the most difficult level for interoperability in GIS
community
>since such people/organization just do not want to share.
>
>If even we cannot reach such a simple domain specific agreement on
service
>description, how can we guess the semantics we generated in varied ways
by
>logic? Actually, the logic way may be just another standard and
agreement
>that enforces developers to follow on. If their actions are formulated
>within the logical inference scope, then you can get an answer.
However, how
>can we process any sort of extra actions not within that scope?
>
>Considering that multiple Web services can perform exactly the same
function
>by different interfaces/approaches, since they should have the same
service
>semantics, adding semantic annotation onto WSDL may not be the right
way
>since the objects/elements in WSDL interfaces/documents are different.
This
>means the same service semantics will be described in different
>terminologies. Is this the result we want to see? Or, service
description
>should be separated from any technology for service development?
>
>
>-----Original Message-----
>From: Drew McDermott
>To: public-sws-ig@w3.org
>Sent: 11/22/05 2:52 PM
>Subject: Re: Where are the semantics in the semantic Web?
>
>
>
>>  [Shi, Xuan]
>
>>  However, if service semantics is developed based on standards and
>>  agreements, then everything is clear and we do not need logic for
>>  matchmaking.
>
>Well, yes, but that takes all the fun out of it.  You seem to be
>saying that human developers, given enough clear information about web
>services, can write any desired program for interacting with web
>services.  That's certainly true.  The more interesting question (to
>me, anyway) is whether there is a point in "generality space" where it
>pays people to describe web services formally enough that automated
>agents can write the programs, or at least play a role in writing
>them.  The descriptions would have to be written without detailed
>knowledge of what program was going to be required, which seems to
>indicate that the notation should be neutral and general-purpose.
>Such notations tend to look like logic of some kind.
>
>Of course, the answer to the "interesting question" may well be No.
>
>                                              -- Drew
>
>--
>
>                                          -- Drew McDermott
>                                             Yale University
>                                             Computer Science
Department

-- 
Professor James Hendler			  Director
Joint Institute for Knowledge Discovery	  	  301-405-2696
UMIACS, Univ of Maryland			  301-314-9734 (Fax)
College Park, MD 20742	 		  http://www.cs.umd.edu/~hendler
(New course: http://www.cs.umd.edu/~hendler/CMSC498w/)

Received on Wednesday, 23 November 2005 16:44:34 UTC