Layering the Semantic Web: Problems and Directions

I've just taken a first pass look through this [1], and need to think some 
to come to any firm views or deeper understanding.  Meanwhile, a few 
comments and queries:


First, I'll say that very high on my list of desiderata would be the 
ability for an OWL layering to accept base RDF(S) and treat it in a way 
that is consistent with the RDF(S) specifications.  (It would be good to 
understand what aspects of RDF(S) mitigate against such treatment.)  I see 
an important part of the RDF/Semantic Web approach to machine processable 
information is the idea that many diverse applications can use data created 
by other applications.

I would expect to see a kind of variation of Metcalfe's law apply:  the 
value of Semantic Web data being exponentially proportional to the number 
of applications that can create and/or use it.  In my view, generic 
reasoning systems represent a small (not unimportant) proportion of the 
totality of applications:  most applications will be written to perform 
some specific function and will have their "inference" processes 
built-in;  for these, the simple structures of RDF(S) should prove 
sufficient, and, more importantly, simplicity is an important factor in 
promoting its widespread use.  We are already seeing the emergence of 
applications of this kind (RSS and CC/PP spring to mind, I have worked with 
an RDF-compatible form for email data, etc.).  I submit that having direct 
access to data from these applications will be of great value to the 
OWL-driven inference engines of the Semantic Web -- these applications will 
provide the volume and diversity of raw information upon which the 
inference magic can be wrought.


And of the document itself...

Section 1, last sentence:
- should be "section 6"?

Section 2, 2nd para:
- isn't modularized syntax via XML schema also a syntactic underpinning?

Section 2, "Proof and trust ..."
- Yes, I strongly agree with the view that these should be applications, 
not language levels.

Section 2, final paragraph:
- the http: reference is not hyperlinked.

Section 3, 5th para:
- I'm puzzled by this.  I don't regard XML as having any (widely 
understood) semantics to be honoured.  What is meant by this?  (This idea 
of XML semantics reappears later in the document.)

Section 3, 7th para:
- typo: "porposed" -> "proposed"

Section 5, intro:
- I think it might be instructional to explore the minimum parts of RDF(S) 
that need to be removed to avoid the paradoxes.

Section 5.1, 7th para:
- "and would provide more meaning for constructs from lower languages" 
doesn't make sense to me.  I can imagine being able to do more things with 
the meaning that is there, but don't see how more meaning can be created 
for the lower constructs.

Section 5.1, general:
- What are the features of RDF(S) that make the extension approach 
problematic (or:  what might be removed to make it less so?).  I'm 
wondering if there are any small, otherwise inconsequential, changes that 
would make this approach more attractive.

Section 5.2:
- I'm afraid I really didn't understand what was being proposed here.  E.g. 
the section is headed "same syntax, diverging semantics" but in its 
penultimate paragraph talks about "define a frame syntax (sic) for OWL ... 
behaves the same as the non-frame version ... different at the syntactical 
level".  That sounds like divergent syntax rather than semantics.

Section 5.3:
- I find it unclear to what extent this approach differs from section 5.1, 
other than that some element of RDF(S) would not be recognized in OWL.  If 
that element were suitably insignificant, the practical effect might be 
similar to 5.1.  Especially if there were some way to automatically 
recognize unsupported elements of RDF(S) and replace them with equivalent 
supported elements (possible using the extended OWL syntax).

#g
--

[1] 
http://www-db.research.bell-labs.com/user/pfps/webont/layering/layering.html


------------
Graham Klyne
(GK@ACM.ORG)

Received on Thursday, 21 February 2002 11:05:49 UTC