xml-dev - RE: Success factors for the Web and Semantic Web

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Yes, and that is a matter of degree.  At 
every step of this, we commit to trusting 
the system we are using, in other words, 
reliability of source and means of inference. 

But that's ok.  We test it until we trust 
it.  Free form text is tough because all 
we have is text analysis (something for 
which, if the domain of discourse is 
identified, the ontology while not 
ideal, is fit).  Given an XML for an 
RFP, things are better because we know 
where to look for what.  Remember, this 
isn't a universal; it is a local politic 
of communication, but perhaps we should 
just say, it is a protocol.   

In the oldeCALSDaze, we talked about 
Product/Process/Organization as interrelated 
domains to define a project. 

So I need assertions for products, processes 
to develop and manage the development of 
products, and organizational models for 
policies to control the processes.  A service 
model provides a registry of keys by which 
I discover that some site services a domain. 

I gave examples of how the RFP process has 
to at the outset, analyze the communication 
and determine characteristics related to 
a request for features of a product such that 
by this one can respond with one of four basic 
responses, plus detect any "universals" that 
must be filtered to prevent complications.

You wanted usable examples.  This one has 
the virtue of having been explored in detail 
with lots and lots of government and industry 
capital.  PDES, CALS, all of these initiatives 
had similar overlapping goals.  Like HyTime, 
they try to tell the whole megillah, and so 
are a little complex.  On the other hand, 
subsets (just as HTML as GenCoding works), 
work and can be readily grokked.  

Take a layer at at time.   Look at UDDI, for example, 
as the schema by which the first layer 
of communication is defined, then successively 
peel off each layer below that.  Remember, 
once you start the communication, it isn't a 
serial process.  It is highly parallel and 
has many aspects of real time control.  One 
issue is rollback can't be as granular and 
that leads to how coarse can a transaction 
be.  I am drifting away from ontology, but 
to me, again, ontology is just another 
record of authority, a document type, to 
be processed as a means to make communications 
coherent over some time/event distance.

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: Sean B. Palmer [mailto:sean@mysterylights.com]


> We can't entirely put away natural language. In fact, we'd be
> hard pressed to give a closed definition for "natural language".
> If you can create an RDF from the web pages, you can do it
> from other texts.  The formality and structure of the texts make
> a difference of degree.

Well, yes. But you can only extract RDF from Web pages if they are really
well marked up. It relies on the semantics of XHTML, not the semantics of
the natural language contained within XHTML. I've felt the sting of this
problem myself [1], i.e. [2].
<sigh/> it would be so much easier if we could just add RDF to XHTML...

[1] http://www.mysterylights.com/xhtmltordf/
[2] http://www.mysterylights.com/xhtmltordf/xhtmltordf.xsl