My take on the question: (01)
Every computer program performs reasoning of some form; any "if-then"
is reasoning, broadly construed. The problem of creating a precise
definition for "reasoning" is similar to the problem of creating a
definition for "agent". There is a gradation of complexity, and for
the purpose of useful categorization (or even discussion) one has to
choose a cutoff point which may appear somewhat arbitrary. So one
tries to use terms in the most common sense. A "reasoning system"
should designate a computer system that performs reasoning beyond the
typical reasoning performed by procedural code, as its *distinctive*
characteristic. That means that if the system does not generate
implied inferences from information, it cannot perform its primary
function. That definition would exclude RDBSs, as they can store and
retrieve data just fine without any attached declarative rules. More
specific than that, a "knowledge-based reasoning system", would be a
systems that uses declarative knowledge according to explicit rules
that are *also* in declarative form, i.e. the reasoning rules are
themselves part of the knowledge base. This excludes neural-net type
reasoning and reasoning which is primarily embedded in procedural code,
as well as database systems that can as an optional feature perform
reasoning with data input transformations, views, triggers, other
attached procedures, or SQL statements. (02)
The point of using ontologies for me is not that it does something that
is *impossible* to do some other way but that it provides a way to
specify knowledge that is more easily communicable to other systems,
more easily comprehended by people, and more easily maintainable.
Perhaps this is the difference between "essential" and "most
cost-effective", as we debated recently. The numerical advantage of
ontologies over alternative ways of achieving reasoning over
declarative knowledge may make an interesting and perhaps significant
academic study, but at the moment the preponderance of opinion seems to
be that they do have substantial benefits beyond what other paradigms
provide. If one really wants to pursue this line of discussion, I
would suggest that one first try creating a definition that describes
what ontology-based systems do that other systems don't usually do (not
"can't do") outside their procedural code. (03)
If you don't think that any of this can be "proven", recall that the
joint communique is a position statement that expresses the beliefs of
the panelists. Projections of anticipated relative effectiveness can
only be proven by multiple examples over time. (04)
If anyone thinks that clarifications or caveats are required for
statements in the Joint Communique, please formulate your concerns and
put them in the Extended remarks. If you think that one of the
statements is factually incorrect according to any normal
interpretation of the words, then the statement may need modification. (05)
Of course, suggestions for modifications or additions to the joint
communique are always in order, especially from the panelists. (06)
Pat (07)
Patrick Cassidy
MITRE Corporation
260 Industrial Way
Eatontown, NJ 07724
Mail Stop: MNJE
Phone: 732-578-6340
Cell: 908-565-4053
Fax: 732-578-6012
Email: pcassidy@xxxxxxxxx (08)
-----Original Message-----
From: uos-convene-bounces@xxxxxxxxxxxxxxxx
[mailto:uos-convene-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Bill
Andersen
Sent: Monday, March 06, 2006 1:10 PM
To: Upper Ontology Summit convention
Subject: Re: [uos-convene] KBR vs SQL (09)
On Mar 5, 2006, at 17:06 , Obrst, Leo J. wrote: (010)
> Oh, come on, Bill. Since when are relational databases reasoning
> systems? Since programming was invented, i.e., a program represents
> some programmer's thoughts about the data and what he/she means about
> that data? So programs that use databases are reasoning systems?
> Because programs encapsulate human reasoning? (011)
Simple existence proof: (012)
[1]
(forall (?x ?y) (=> (or (brother ?x ?y) (sister ?x ?y)) (sibling ?x ?
y))) (013)
[2]
create view sibling as
SELECT *
FROM brother
UNION
SELECT *
FROM sister (014)
Which piece of syntax, under evaluation, is "logical"? Which is
reasoning? (015)
I have no doubt that some sufficiently clever programmer could embed
a tableau or resolution prover into a complete RDB. The connections
between databases and theorem proving are well elucidated in:
"Principles of Database & Knowledge-Base Systems Vol. 1 & II", by
Jeff Ullmann. (016)
> This is silly, sorry.
>
> If ontology is about organization, you should be very happy with OWL,
> since as a description logic (also called terminological logic,
> classification logic), it is focused on classification and nearly
> nothing else. Deductive reasoning doesn't really even figure until
you
> add SWRL. (017)
OWL is precisely a deductive system, so it's a candidate for the
organizational task. There are many other candidates with other
properties, however. (018)
> I think you need to re-evaluate your rhetoric about this stuff! ;)
> (Hey, I'm am undergoing rhetorical revamping myself; I need
company!). (019)
Well, perhaps, but it gets people thinking.... :-D (020)
.bill (021)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/uos-convene/
To Post: mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal: http://ontolog.cim3.net/
Shared Files:
http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki:
http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/uos-convene/
To Post: mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal: http://ontolog.cim3.net/
Shared Files: http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit (022)
|