I trimmed the ccs since I get scared if I have to scroll a cc list.
On 13 Sep 2007, at 15:21, Chimezie Ogbuji wrote:
On Wed, 2007-09-12 at 14:42 +0100, Xiaoshu Wang wrote:
Chimezie Ogbuji wrote:
[snip]
But please also see how dangerous such practice will be: "Ian
Horrocks1,
Bijan Parsia, Peter Patel-Schneider, and James Hendler, Semantic Web
Architecture: Stack or Two Towers"
at "http://www.cs.man.ac.uk/~horrocks/Publications/download/2005/
HPPH05.pdf"
Yes, I'm *very* aware of those two schools of thought :).
Please note that that paper explains that the syntactic intersection
approach does not yield the same semantics when extended in a variety
of ways. That is, query languages, for example, can detect the
difference between a DLP OWL ontology understood in LP vs. the first
order way.
I think the
domain of discourse is an additional criteria that isn't often
considered, however.
Eh. Not really. First of all, "domain of discourse" has a couple of
specific technical meanings so we should be a bit wary about it.
My quote from [1] was only to ensure we have a
proper definition of "closed world assumptions" - I often find threads
on this topic begin with a poor definition and people end up talking
past each other.
This is a good practice as there are several meanings and CWA doesn't
really cover all the possible nonmonotonic formalisms and is rather
biased toward databases.
Those two papers address a different (non-trivial)
question: whether the semantic web stack is best built on top of
Description Logic or Logic Programming
Well, the two tower's paper doesn't say that. It just says that an
syntatic intersection doesn't ensure semantic interop, so it's wrong
to blithely assert that it does.
(or at least on a framework which
includes a mapping between the two [1]).
[1]http://www.cs.man.ac.uk/~horrocks/Publications/download/2003/
p117-grosof.pdf
DLP isn't a mapping. For mapping from, e.g., LP to FOL, look at
things like clark's completion. This cannot be done in the most
general settings since it would require second order features. But,
e.g., with nominals, you can restrict the models of an OWL KB in a
number of ways. One obvious thing you can do is domain closure, i.e.,:
owl:Thing subClassOf {A, B, C}
where A B and C are individuals and {} is the "oneOf" operator.
Note that DLP-LP and DLP-FOL do coincide for certain classes of
answer for certain classes of query (since their ground entailments
coincide).
This has gotten a bit too deeply technical perhaps. From a language/
infrastructure design perspective, it seems clear that having some
sorts of non-monotonic features are quite useful in a number of
circumstances. For example, having integrity constraints (i.e., some
sort of data validation) can help with input and cleansing. The
questions are *which* features (and which variants), how to
incorporate those features in a usable way, and how to implement them
in a useful way. The DatabasEsque Features OWLED task force is
putting together a report on some of this:
http://code.google.com/p/owl1-1/wiki/DatabasEsque
Cheers,
Bijan.