Lists Home |
Date Index |
On Sat, 13 Nov 2004 13:54:47 +0530, Garfield <email@example.com> wrote:
> Hi All,
> Thanks for the overflowing information on the Semantic Web, and its
> mysteries. Seems to be demystified for the most part of it.
> Seems to be demystified for the most part of it. Well, I
> had another thought in my mind, which I thought worthwhile sharing
> with you folks.
> Are there mappings defined between the Semantic Web and Programming Models ?
<snip>interesting discussion on what this might mean</snip>
it seems to me you're circling the questions of "even if you have
(global) knowledge discovery mechanisms (the enabling pieces of the
Semantic Web) how are you going to exploit them"? In particular, you
> This looks similar to a Javadoc that describes what a class can do,
> what operations it supports, and what are its fields. What I am asking
> for something more than a documentation, but a more formal definition.
Which I think is the issue of the exchange of common semantics. I
can't offer any specific answers for these issues. (The usual, go
look at expert/rules systems, RDF, and of course the whole gamut of WS
stuff, applies, but I think you already know that.) However, since
some of your points are posed as Java specific you may want to look at
JSR 94 and it's kin.
As I've suggested before, rules can be treated as exchangeable
metadata. Basically, couple rules to terms and Ontology traversal can
be treated as a a rules evaluation process: if your rules and my rules
both find the same object then the terms the rules map to must be
describing the same thing (and here's an instance of an object that is
a representation of that thing). What's interesting with this
approach is that the object can be local only; run both the foreign
rules and the local rules locally to see if they are in agreement, if
they are you don't really care what the foreign object might be,
you've got a local object that works for you (there's a heavy weight
way to isolate viruses!).
I don't know if there is any work being done on this approach. We're
sort of indirectly attacking things this way, but more from a rules
discovery of metadata (which then describes abstract objects) than
finding any objects that are directly usable. What might be
interesting was if I could turn my abstract objects into something
like a Castor description and subsequently de-marshall a real object
representation of it. However, I can't see any practical use for this
at the moment...