Re: A certain difficulty

I guess I'm finding this whole thread a little confusing. The original
poster quoted someone else as saying.

> 
> "We (a working group of 7 technicians from the WAP FORUM Telematics Expert
> Group) tried it (RDF).  We tried like hell for over a week's time and we
> never got it. Sure we could put some things together with nodes and arcs,
> but after that we had no idea where to go.  We downloaded every thing we
> could find, only to become more confused."
> 

And when this, I thought "This is a modeling problem." It's not the
XML syntax (though that may be an issue, see below), it's that they 
didn't know how to build RDF/ RDFS models, or what to do with the 
models once they had been built.

But, later on in this discussion, the complaint has been subtly changed. 

> > That this group of engineers made a sincere effort to implement
> > RDF and failed, is saddening
> 
> The basic problem with RDF (actually the specs) is the XML part.
> This was always a how-do-we-get-there-from-here problem.  RDF has a
> pretty detailed data model.  It might have sufficed - as I believe at
> one point it did - to consider XML as just one possible serialization
> syntax....  
>

There are two things wrong with this:

	(1) They didn't try to "implement" RDF. They tried to use
	RDF to model something (I think).

	(2) XML is a fine, if somewhat verbose, serialization spec. 

Amplifying the second: the basic problem with modeling languages like
RDF is that they are fairly abstract. Years and years of experience
in knowledge modeling has taught the AI community that most people,
be they programmers or no, have difficulty building robust and reusable
conceptual models. What's more, forcing people to focus on syntax 
(*any* serialization syntax, not just XML) makes modeling orders of 
magnitude more difficult (it forces people to think at the conceptual 
level and pay attention to things like "did I get the tags right" at 
the same time).

This implies that RDF needs an accessible user interface.

<beginplug>

Which leads into a plug for Protege-2000. Protege-2000 is an open-source 
(written in Java) knowledge-modeling application that we at Stanford 
Medical Informatics have produced. The user interface is based on 
years of helping medical personal (e.g. physicians and their ilk)
build complex conceptual models of their domains. 

What's more, Protege-2000 is highly modular and extensible, with 
components that can be reused in applications. So, for example, 
someone building an RDF schema can:

	(1) Use Protege-2000's user interface to define the schema
	and create RDF resources. 

	(2) Use the KnowledgeBase object (from the package 
	edu.stanford.smi.protege.Model)in a program to read the 
	RDF in and make queries against it. Note that this isn't
	very difficult-- we've got a lot of examples and the source
	code for the entire application is in the download. 

	(3) Ignore the serialization syntax entirely. 

There are two caveats:
	
	(a) Our RDF support isn't perfect (though we've been working 
	like hell to remove the imperfections).

	(b) Modeling is still difficult. We've just removed some of the 
	unnecessary difficulties.

Even in spite of these caveats, we think Protege really will help a 
lot of people with learning the basics of, and using RDF.

Try Protege-2000. And, please, give us feedback (and, please, feel
free to tinker with the source code and point out errors). 

	www-smi.stanford.edu/projects/protege

</beginplug>


William Grosso

-- 
William Grosso 			grosso@smi.stanford.edu
Phone 650-498-4255 		http://www.smi.stanford.edu/people/grosso/

Received on Tuesday, 22 February 2000 12:54:43 UTC