Re: Where are the semantics in the semantic Web services?

hi Xuan,

> I specially appreciate the "law of the semantic web" as he said by the 
end
> of the paper "The more agreement there is, the less it is necessary to 
have
> machine-processible semantics". I think it's the same to semantic Web
> services.
> 

My thinking of using "semantic Web services" is that we
- need to write down a sufficiently large part of the semantics of a 
service using some kind of well defined 'building blocks',
- that can be assembled using some form of 'semantic glue' to capture 
service semantics.

The building blocks are domain-specific terms defined in some kind of 
shared ontology (just ANY form of agreed data type definition, IMHO that 
could even be plain XML DTDs),

The 'semantic glue' is a well defined (logic-based) language that allows 
to take these terms to form sentences like:

"if the world state statisfies an instance of formula F1 before service 
execution, then the state will satisfy an instance of formula F2 after 
successful execution." 

- the formulas F1 and F2 refer to the agreed domain terms and eventually 
combine them using some kind of logical connectors like AND, OR, NOT etc. 
- by "instance of a formula" i refer to the fact that formulas may be 
partially or fully grounded by the input data of the client or the output 
of the server (see [1] for details)

While I also agree with the "law" (or paradox? ;-) of the semantic web as 
expressed above, i think that semantic Web services can provide added 
value if they provide a nice and usable form of 'semantic glue' that 
allows people to form statements using their agreed vocabulary. 
 
> Question 2 still is whether semantic Web technology (RDF/OWL) is 
suitable
> for semantic Web services? RDF/OWL are good at defining the "nouns" not
> "verbs". We can define one object is a subclass of another object. But 
how
> can we define one function is a subclass of another function? Such as in
> OWL-S approach, "BookSelling" is a subclass of "SellingService" and
> "SellingService" is a subclass of the root ontology class "e_Service". 
Do we
> need to find some other ways to define the relationship of "verbs"?

i am not sure if one really needs *verbs*. Personally,  i like the idea of 
characterizing operations by describing the precondition state and the 
resulting effect state. I think the description (characterization) of 
states does not require verbs, because states are just static snapshots. 

But of course, sometimes one would like to use some form of "past passive 
participles" of verbs to describe effects, e.g. for an email sender 
service, one could be tempted to state that the post-execution state 
entails "sent_mail(m, from, to)" (which of course has its roots in the 
verb 'sending').

BTW - for the practical implementation of semantic Web services i have 
designed an XML based markup language called SESMA [1], which might be 
seen as a subset of OWL-S that is less generic than the latter (all a 
question of using the right tool for the job, personal preferences etc.)

i have implemeneted several running (protoype) systems using this language 
(e.g. [2]), and feel that SESMA is at least very useful if one needs to 
come up with a service description quickly

cheers!
Joachim


[1] http://www.ai.sri.com/WSS2005/final-versions/WSS2005-Peer-Final.pdf
[2] http://elektra.mcm.unisg.ch/pbwsc/docs/eswc05jpeer.pdf

Received on Saturday, 17 September 2005 09:25:58 UTC