Mitch Harris wrote: (01)
> I think there is a mismatch between how people are using implicit
> quantifiers here.
>
> True, for any particular problem, it is in a particular complexity
> class or not. Its complexity (computational) does not change.
>
> But for given a set of language operators with specific meanings, the
> complexity of a the set of -all- problems expressible with larger
> selection of operators is at least as difficult (takes as long or
> longer) for a smaller set.
>
> For a particular problem, if you know ahead of time which complexity
> class it lives in, you can gear the algorithm to that. But if you
> don't know that about the input, a simple problem may be expressed
> (equivalently of course) using the larger set of operators and you may
> still be bound by the algorithm -needed- to deal with the larger set
> to do all sorts of expensive, time consuming things. (02)
Mitch's point is well taken. Because the OWL/DL language is less
expressive than a language like KIF or CLIF, there are two important
consequences:
- the complexity of the ontologies that can be phrased in, and thus
the complexity of the problems that can be solved with, OWL/DL is lesser.
- the OWL/DL language forces a discipline on the knowledge engineer
that the more powerful and less complex KIF/CLIF language doesn't. The
knowledge engineer is forced to cast the problem at hand into the
carefully honed elements of the OWL/DL language. (03)
In Mitch's terms, this means that the knowledge engineer is forced to
determine the category of problem and express an OWL ontology that can
deal with that category of problem. (04)
In my experience, people who are constrained to use a language like
OWL/DL to express their ontology can, in many cases, find a way to
express all or nearly all of their intent. By comparison, given KIF (or
CLIF), they will be able to express all of their intent, but they will
do it in a different way. That is, they won't produce a CLIF version of
the OWL model plus the two axioms they couldn't express in OWL. They
will instead produce fewer relations and more, and more complex, axioms. (05)
John is correct that, if you take a CLIF formalization of the OWL
ontology, a good FOL reasoner may perform as well as or better than the
DL reasoner. But if you take the CLIF ontology that the KIF knowledge
engineer prefers to build, it will take substantially longer. Because
"thinking in KIF" will often result in axioms that force the reasoner
"to do all sorts of expensive, time consuming things." (06)
But I agree with John in taking issue with this statement of Martin Hepp's:
> Also, it is more difficult for users to understand an expressive
> ontology, because it requires a better education in logic and more time. (07)
As John says, the problem here is not expressiveness; it is simplicity
or complexity of the language and the artifacts written in it. (08)
CLIF is a simpler language than OWL, and it is easier to read. A
well-written CLIF ontology is often easier to follow than the
corresponding OWL/DL ontology, because the OWL ontology must be
carefully crafted to express the intent. And that often means that the
knowledge engineer uses complex work-arounds in OWL, to make the problem
accessible to the reasoner, rather than to the human. (09)
This is the flip side of the coin above. Humans have an easier time
following a straightforward presentation of the problem, even if the
axioms become more complex. The simpler but less constrained language
is easier for humans to follow, but harder for the reasoner to work
with. The more engineering that must be done to express the problem,
the easier it is for the reasoner, and the harder it is for the human to
follow. (010)
-Ed (011)
--
Edward J. Barkmeyer Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263 Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263 FAX: +1 301-975-4694 (012)
"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority." (013)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (014)
|