John, (01)
It seems to me that this discussion has become a matter of religious
tenet, and I see no reason to further defend what I thought was an
ecumenical view. (02)
I would make two minor observations: (03)
John F Sowa wrote:
> But a reasoner, by itself, is not an application program.
> (04)
I would say that depends on whether you think the input queries and
axioms are data or not. There are several valuable application programs
that connect a reasoner to a knowledge base and wrap it with a query
interface. What John is saying may be only that the reasoner becomes an
"application" when you attach it to a knowledge base, but that is true
of a great many "application programs". (05)
> FOL is actually easier for humans to read and write than OWL DL.
> (06)
I think that is debatable, if not outright dicto simpliciter. If you
consider the formulation of "Every dog has exactly one master", the OWL
Manchester syntax is only a bit ugly, while the FOL formulation is
surprising, if not counterintuitive, to a non-logician. (07)
> And it can be stated in controlled NLs. A knowledge compiler can
> translate FOL to whatever format is needed for whatever reasoning
> engine and database is used.
> (08)
This may be theoretically true, but in practice it can be very
difficult. Other logic languages have different model theories, and
related restrictions and conveniences that you have to use to express
the knowledge, and translating a knowledge base from a FOL language
(like KIF or Cyc) into such a language is itself a significant knowledge
engineering task. Automating that process is likely to beget a few
Ph.D. theses. (09)
Our solution to presenting an ontology in OWL form and in CLIF form (to
different reasoning tools) is to maintain both forms, or more
accurately, to maintain a "CLIF macro" form that has a rote OWL
rendition and a well-defined CLIF expansion (thanks to Fabian).
Otherwise, it takes quite some compiler, for example, to convert the FOL
existential and uniqueness axioms above into the OWL: dog is a subclass
of the class defined by "thing that has the property 'thingHasMaster'
exactly once". (010)
And if the FOL phrasing involves negation and the target engine needs
RDF, you get to do a lot of equivalence transforms looking for a Horn
clause (assuming there is a Horn clause formulation). And then you
might consider the complexity of the algorithm for stratifying a FOL
axiom set into a production rules script a la RIF. "It's not easy, Mr.
Fox, sir." (011)
-Ed (012)
The difference between theory and practice in practice is greater
than the difference between theory and practice in theory.
-- unattributed (via John Dilley at HP) (013)
--
Edward J. Barkmeyer Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Systems Integration Division, Engineering Laboratory
100 Bureau Drive, Stop 8263 Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263 Cel: +1 240-672-5800 (014)
"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority." (015)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (016)
|