Re: The Q model

* Lars Marius Garshol
|
| Hmmm. I think probably you're reading more into this statement than
| I intended by it. Clearly, RDF represented in Q is still the same
| RDF, and so the same OWL/RDFS semantics will apply as when RDF is
| just triples, but, as you note below, this will be a semantics that
| ignores the two extra elements.

* Peter F. Patel-Schneider
| 
| This seems to be akin to extending propositional semantics to
| first-order sentences by ignoring the quantifiers and variables.  I
| just don't see how it is obvious that it will work.

Ah, no, this is not what I mean. What I mean is that you can use the
existing RDFS/OWL semantics on Q instances, provided you ignore two of
the elements in each quint (specifically, statement-id and context).
That may sound silly, because it means we're back to RDF, but it does
mean that the RDFS/OWL semantics can be applied to Topic Maps
represented in Q, although with the limitation that reification will
be ignored (as it is in RDF), that variant names will be ignored
(which is not much of a loss), and that the semantics only apply to
statements in the unconstrained scope.

This means that you get RDFS/OWL semantics for Topic Maps provided you
restrict yourself to an RDF-compatible subset of Topic Maps. That's a
gain for Topic Maps users, though unlikely to excite RDF users much.

Of course, some porting work is necessary to describe exactly how this
would work, but I can't really see how it's possible for this to fail.
 
* Lars Marius Garshol
|
| No, I do not. I think I can see a way to take context into account,
| but I don't have this worked out, and it's definitely possible that
| taking the two extra elements into account will break it.
 
* Peter F. Patel-Schneider
|
| Yes, well then why the claim?

I guess it looks like a claim when you read the slides (and ignore the
fact that the slide says "A possible solution is (possibly) to
..."). To people hearing the presentation I think it was quite clear
that this was something I thought was possible, and wanted to try out,
but definitely did not guarantee would work.
 
| This seems to be just like saying that you have a syntax that
| marries propositional temporal logic and higher-order predicate
| logic, but no semantics for the combined syntax. 

That's precisely what I am saying. (Well, if by "temporal" you mean
"time-based", then that's not what I mean; scope in Topic Maps is
generalized context, not just time context.)

| What makes you think that a reasonable combined semantics is even
| possible?

Let me be clear here: I think the model is valuable even if no such
combined semantics is possible, for two reasons:

 (1) It's useful for implementations, for conversion, and for general
     understanding of the relationship between the two models.

 (2) It's useful because it makes it possible to find out if a
     reasonable combined semantics is possible.

I believe it's possible to do a combined semantics that takes into
account context because given a full Q model, you can build from it
one model for each context in the full model, where all quints in each
of these models have the same context. You include all quints which
have that context or a weaker context (and it's easy to know which are
weaker).

For each such model you are back with the case we had at the top of
this email: standard RDFS/OWL semantics apply, and any quints you
infer have the same context as the other quints in that model. The
union of all the enriched submodels would then be the result.

This definitely lets you get RDFS/OWL semantics for one context at a
time. Whether you can efficiently do the inferences for all contexts
at once I don't know, and that really is the big open question. It
wouldn't surprise me that much if someone has already answered this.

Reification I admit I have no ideas on how to handle, but I'd consider
this a success even if it were not covered.

| Note that I'm not saying that this is not possible.  I am instead
| saying that you, as the proposer, need to show that it is indeed
| possible.

I can hardly disagree with that.

-- 
Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >

Received on Thursday, 4 August 2005 23:18:05 UTC