Lists Home |
Date Index |
On Jun 12, 2004, at 11:19 AM, Miles Sabin wrote:
> Bill de hÓra wrote,
>> Miles Sabin wrote:
>>> I've read all of those documents and there's nothing there that's
>>> new since Tarski on the MT front, and the rest is KIF, say, with
>>> angle brackets.
>> Funny you say that. I've read the C# specification and there's
>> nothing new in there since Turing, and the rest is C++, say, with
>> garbage collection. I don't think either of us should be overly
>> disappointed however.
> Not comparable. Model-theoretic semantics for formal langauges are two
> penny: pick up any halfway decent textbook on mathematical logic and
> you'll find at least one example which in all likelihood will look
> quite similar in to the RDF MT in general outline (well, more similar
> to the RDF MT than a TM configuration would look to C# code). The only
> novelty in the RDF MT is the hoops it has to jump through to align
> itself with the awkward informal semantics of pre-MT RDF.
Well sort of, but the RDF MT does provide the mathematical foundations
that had been lacking and are rather helpful if we are actually talking
about "machine understanding".
What is unique, and what was actually a surprisingly difficult task
even for MT experts was the ability to express a description logic MT
in RDF -- that is to unify the OWL and RDF model theories. As is well
described in the WebOnt WG archives, for the longest time our logic
experts thought it couldn't be done.
You are free to offhandedly dismiss this work as trivial and nothing
new, however you should realize that some of this work has been done by
people who are *very* well respected in their communities -- this is
the sort of work that should grow on people once some of the technical
underbrush becomes understood. To me it is very analogous to how the
decades of work on SGML had received little attention until XML came
out. Was XML at all new? No. One of the huge improvements, at least to
me, with XML was the specification by EBNF productions. Perhaps that is
a technical change, but perhaps that is part of why it has such a huge
adoption -- writers of parsers can be very precise about what is legal
and what isn't.
>>> So where's the beef?
From a technical point of view, one of the principles of RDF is that
there is no negation -- no ability to state NOT . This is because of
the Web and the requirement to have a completely monotonic logic. If
you can't say NOT, then there is no risk of finding out later that some
document negated something you had inferenced.
It was thought not possible to write the OWL MT using a *syntax* i.e.
triples, in which all statements had to be asserted as true.
Technical? very. New? it appears so. Meaningful? remains to be seen.