Dear Jim,
Many thanks.
This is all wonderful the hear.
If the person who's lab is responsible for developing and supporting
one of the most widely used DIG reasoners says "these things are not
mutually exclusive in any way", I've got much more confidence that
all will be OK. :-)
Does this mean these different formalisms will remain semantically
interoperable? That's really too general and vague a question to
answer. Let me try that again - Can we expect the assertions made in
one of the more "lite" formalisms will be mappable into one of the
more "heavy" formalisms in a lossless manner. I assume going in the
other direction, the translators must needs be lossy - at least for
those ontological graphs making assertions or defining constraints
not expressible in the more "lite" formalisms. As I believe was
stated on the public-owl-dev list, the hope would be any mappings
would be able to at least avoid logical inconsistencies or
contradictions.
Actually, all of that sounds very indefinite, now that I read it
through again. It's just going to take implementation of specific
Use Cases to get a better understanding of what's possible - and
practical.
It's also great to hear there will be support all around for
ObjectProperties, given the critical role they play when trying to
leverage the power of a reasoner to generate the inferred portion of
your ontology. As these ontologies of biomedical reality (e.g.,
those associated with the OBO Foundry and other efforts across the
globe) continue to grow in the attempt to provide more solid formal
footing for detailed semantic descriptions of very complex biological
data repositories, we absolutely must lean on the reasoners as much
as is feasible as a tool to maintain ontological hygiene and to
extend the leaves of the graphs we are creating to a level of
granularity that keeps them relevant to researchers on the bleeding
edge of their field.
The point you make regarding scalability in the integration space is
also critical. As you say, there has been discussion of this issue
before on this list as a follow-up to last year's OWL ED 2005
discussion/vetting of OWL Extensions, for instance. There was
another such discussion just a few weeks back following the HCLSIG
F2F in Amsterdam, where again Scott Marshall and others made the
point both within and across large-scale repositories, it can be more
effective to consider a "lite" ontological approach, though the
"heavier" more complete OWL-DL (and OWL 1.1) ontologies will
definitely be a critical enabler, when one seeks to support more
complex semantic queries in a complex space such as cross-species,
multi-modality, multi-resolution studies of neurodegenerative disease
as we are charged with in the BIRN project.
Thanks again for providing this very useful feedback.
Cheers,
Bill
On Nov 16, 2006, at 7:51 PM, Jim Hendler wrote:
Bill - I think you see the key point - these things are not
mutually exclusive in any way - high end ontology use needs some of
the OWL 1.1 functionality, but wider adoption is likely to come in
the data integration space, where a lot can be done with the
simpler "OWL Mini" (or OWL Fast, or OWL Prime, or OWLET, or
whatever it might end up being called).
btw, the "?" marks in the graph didn't mean things were being left
out, it meant we weren't sure if including them would cause scaling
problems - I expect we will keep data and object type properties
because a lot of current tools support that distinction (for
example tools which turn OWL classes into web forms generally have
a different handling of each of these)
Work on this continues apace
JH
p.s. There are some resource contention issues which I think are
relatively easily worked out, but there's also some difference in
the philosophy between the current OWL 1.1 people and the people I
mention working on the simpler OWL - but that's mostly because
we've been working in different spaces.
At 7:31 PM -0500 11/16/06, William Bug wrote:
Many thanks, Jim.
I saw posts by you and others - as well as links to more detailed -
and very recent - discussions.
These are all very helpful.
I was particularly interested in the proposal you, Ora Lassilla,
and others have worked on to "absorb" much of the OWL Lite
constructs into RDF++ (I believe that is the name being used - two
proposals - the summary of which you give here - http://
www.nabble.com/perspectives-on-OWL-v.next-and-RDF-tf2624829.html) -
minus many of the class axioms and Datatype Properties, but
including owl:disjointWith.
I recall reading some on this in the past, but given these issues
related to OWL 1.1, those proposals take on a whole new semantic
value now - pun[ning] intended. ;-)
I also can definitely see the sense in the argument when trying to
determine where to go next, there is need to achieve a balance
between simplicity to catalyze wider adoption and the need to
provide new functionality to a subset of users. I think its
actually a four way balance:
simplicity - robustness - performance/scalability -
expressivity
These characteristics are not necessarily exclusive of - or the
inverse of - one another, though usually performance scales with
the inverse of expressivity, at least when comparing across very
large differences in expressivity.
The position you choose to target in this space really depends on
what you are trying to achieve, obviously.
Cheers,
Bill
On Nov 16, 2006, at 3:01 PM, Jim Hendler wrote:
actually, we're trying to move the discussion to [EMAIL PROTECTED]
there's a thread there that expresses some of my concerns about
moving away from the OWL syntax - given that the primary tools out
there right now still assume the OWL is integrated with the RDF
graph...
At 2:52 PM -0500 11/16/06, William Bug wrote:
As I expected, the experts are listening. :-)
Many thanks, Holger. That's extremely important to know.
I will dig into the thread for more detail. One main concern would
be whether that was just a token gesture to stay compatible for
now, as opposed to a commitment to remain compatible, until or
unless an effective alternative is provided to representing very
large knowledgebases in RDF.
I think more than anything I was a bit overwhelmed by the
collective picture given by those half-dozen or so presentations
from last week's meeting. The meeting seemed "fresh" enough, so
that it could be expected to be reflective of the status quo. I
assume there was much heated discussion during the meeting, that
would have filled out such detail - or such has been carried out on
the owl-dev list.
I suppose it's also a good idea to dig into the OWL Extensions list
hosted by Jim Hendler's lab:
http://lists.mindswap.org/mailman/listinfo/owl
Cheers,
Bill
On Nov 16, 2006, at 2:33 PM, Holger Knublauch wrote:
However, I'd point out, in the last document, where they describe a
mapping of OWL 1.1 to RDF, they make the following caveat:
Not every OWL 1.1 ontology can be serialized in RDF. In particular,
ontologies using the following features of OWL 1.1 cannot be
serialized:
1. punning and
2. annotations on axioms.
Please see a recent thread on the public-owl-dev mailing list about
this: http://www.nabble.com/Limitations-of-OWL-1.1-to-RDF-mapping-
tf2639224.html
Bijan states:
"The RDF mapping has lagged behind the others, but the plan is to
extend the mapping to cover these cases."
Holger
TopQuadrant, Inc.
http://www.topbraidcomposer.com
http://composing-the-semantic-web.blogspot.com/
Bill Bug
Senior Research Analyst/Ontological Engineer
Laboratory for Bioimaging & Anatomical Informatics
www.neuroterrain.org
Department of Neurobiology & Anatomy
Drexel University College of Medicine
2900 Queen Lane
Philadelphia, PA 19129
215 991 8430 (ph)
610 457 0443 (mobile)
215 843 9367 (fax)
Please Note: I now have a new email - [EMAIL PROTECTED]
--
Prof James
Hendler
[EMAIL PROTECTED]
Dept of Computer Science
http://www.cs.umd.edu/~hendler
AV Williams Bldg
301-405-2696 (work)
Univ of
Maryland
301-405-6707 (Fax)
College Park, MD 20853 USA
Bill Bug
Senior Research Analyst/Ontological Engineer
Laboratory for Bioimaging & Anatomical Informatics
www.neuroterrain.org
Department of Neurobiology & Anatomy
Drexel University College of Medicine
2900 Queen Lane
Philadelphia, PA 19129
215 991 8430 (ph)
610 457 0443 (mobile)
215 843 9367 (fax)
Please Note: I now have a new email - [EMAIL PROTECTED]
--
Prof James Hendler [EMAIL PROTECTED]
Dept of Computer Science http://www.cs.umd.edu/
~hendler
AV Williams Bldg 301-405-2696 (work)
Univ of Maryland 301-405-6707 (Fax)
College Park, MD 20853 USA
Bill Bug
Senior Research Analyst/Ontological Engineer
Laboratory for Bioimaging & Anatomical Informatics
www.neuroterrain.org
Department of Neurobiology & Anatomy
Drexel University College of Medicine
2900 Queen Lane
Philadelphia, PA 19129
215 991 8430 (ph)
610 457 0443 (mobile)
215 843 9367 (fax)
Please Note: I now have a new email - [EMAIL PROTECTED]
