[seek-kr-sms] GrOWL Applet
Serguei Krivov
Serguei.Krivov at uvm.edu
Thu Jan 29 07:10:35 PST 2004
Hi all,
Here are some points/questions:
1. I believe that we have been mapping the axioms, not the model. Be
precise- we use only those semantic relations (isa arrows, restrictions,
etc) that owlapi can track after reading the owl file. And I do not
believe that it does any inference whatsoever to build the model.
2. I thought that DL has algorithm for deciding subsumption and
algorithm that builds model for just one concept. Is there algorithm
which builds model, given set of axioms? Is it implemented?
3. If we draw model per se ( recall that 'model' is a set of (all)
relations between sets which render the axioms true after each concept
mapped to a set and individual mapped to an element, or if you want to
know details here they are http://www.w3.org/TR/owl-semantics/rdfs.html)
then our graph would be dense:
To draw correctly the axioms like
A--isa-->B---isa-->C---isa-->D
We would have to draw:
A--isa-->B---isa-->C---isa-->D
A--isa-->C
A--isa-->D
B--isa-->D
Or may be we can design some rule for canonical drawing of the model
and avoid the mess?
4. I liked how Bertram pictured intersections in those domain maps (see
the paper). I believe that it is natural for DL and when we wanted to
use DL in IMT I wanted to adapt it. But later on we shift to F-Logic,
and here is a point- we have AND and OR as logical operators and we
need to distinguish logical OR of statements from Union of concepts and
AND from Intersections respectively.
Serguei Krivov
-----Original Message-----
From: Bertram Ludaescher [mailto:ludaesch at sdsc.edu]
Sent: Thursday, January 29, 2004 8:13 AM
To: Rich Williams
Cc: Ferdinando Villa; Serguei Krivov; seek-kr at ecoinformatics.org;
GEON-KR at sdsc.edu
Subject: RE: [seek-kr-sms] GrOWL Applet
Hi everybody:
I'm not sure whether this is pertinent to this discussion, but here is
what it reminded me of:
When graphically visualizing an ontology, we can distinguish two
cases:
1. The graph is a canonical representation
2. The graph is one of a number of alternative representations.
Here I'm not talking about graph *layout*, but the actual graph itself
(i.e., the sets of vertices and edges).
(1) is what one would like to have. For example, think of a graph
database having edges representing binary relations such as parent,
sibling, ancestor, etc. For a given family, there is exactly one graph
representing the family relationships
(2) can happen when you come up with a graph of logic formulas.
For example, consider a set of OWL axioms, i.e., a certain set of
first-order logic formulas. You can "graph" them as we have done for
example here (called "domain maps"; Figure 1 of the first paper):
http://www.sdsc.edu/~ludaesch/Paper/icde01.html
However, there is a problem with this "graphing" of a set of logic
axioms: there are many (in fact infinitely many if you're not careful
;-) ways of doing that.
That's why one should distinguish between the *axioms* (syntax) of an
ontology (there are many equivalent sets for a given set of axioms)
and the *models* (semantics) of an ontology. In the ideal situation,
one would have one canonical model (still a symbolic one) that
represents the intended semantics of the given set of axioms.
Then one would "graph" the model (and not the axioms).
Having said that, here is my question:
What are you guys graphing: OWL axioms or OWL models? (or sth else)
Bertram
>>>>> "RW" == Rich Williams <rwilliams at nceas.ucsb.edu> writes:
RW>
RW> I like the idea of labeling union nodes as 'is one of' and
intersection
RW> nodes as 'is all of'. In some cases (perhaps most?), it is
possible to
RW> entirely eliminate the intersection node, but this might involve
more
RW> processing of the OWL ontology than is practical. Consider that the
RW> following two OWL fragments are semantically identical, and so it
might be
RW> nice to display them both the same. Otherwise this example would
lead to
RW> displays like:
RW>
RW> Test3--is-a--Test1
RW> |
RW> --is-a--Test2
RW>
RW> And:
RW>
RW> Test3--is-a--[is all of]----Test1
RW> |
RW> ----Test2
RW>
RW> Rich
RW>
RW>
RW> <owl:Class rdf:ID="test1"/>
RW> <owl:Class rdf:ID="test2"/>
RW> <owl:Class rdf:ID="test3">
RW> <rdfs:subClassOf rdf:resource="#test1"/>
RW> <rdfs:subClassOf rdf:resource="#test2"/>
RW> </owl:Class>
RW>
RW>
RW> <owl:Class rdf:ID="test1"/>
RW> <owl:Class rdf:ID="test2"/>
RW> <owl:Class rdf:ID="test3">
RW> <rdfs:subClassOf>
RW> <owl:Class>
RW> <owl:intersectionOf rdf:parseType="Collection">
RW> <owl:Class rdf:about="#test1"/>
RW> <owl:Class rdf:about="#test2"/>
RW> </owl:intersectionOf>
RW> </owl:Class>
RW> </rdfs:subClassOf>
RW> </owl:Class>
RW>
RW>
>> -----Original Message-----
>> From: seek-kr-sms-admin at ecoinformatics.org
>> [mailto:seek-kr-sms-admin at ecoinformatics.org]On Behalf Of Ferdinando
>> Villa
>> Sent: Wednesday, January 28, 2004 8:59 AM
>> To: Serguei Krivov
>> Cc: seek-kr at ecoinformatics.org
>> Subject: Re: [seek-kr-sms] GrOWL Applet
>>
>>
>> Congrats to Serguei and Rich for these encouraging results! Now what
we
>> need is to make the representation easy to understand - of course we
are
>> not asking our "average user" to wade through intersection nodes and
>> restrictions, but at least in browsing mode, we want a smart (and
>> hopefully configurable) graph transformation step to turn this into
the
>> simplest and clearest concept map imaginable. I think we need a
wider
>> group consensus on this, but here are a few points to start with:
>>
>> - recognize Intersection/Union nodes that link two classes and just
>> label them something like "is both" or "is one of", consistently with
>> "is a";
>>
>> - recognize those that point to restrictions and use some variation
of
>> the "is a" theme with a "where" link attached
>>
>>
>> Coming up with a full set of these simplifications won't be easy but
I
>> think the success of the approach will rest on it. Ideas, anyone? If
we
>> could define these transformations parametrically and load them into
the
>> editor as a style sheet, that would be the best. We may even want to
use
>> some smart graphical symbols in the future.
>>
>> Ciao
>> ferdinando
>>
>> On Wed, 2004-01-28 at 11:11, Serguei Krivov wrote:
>> > Here is the first version of GrOWL applet:
>> >
>> > http://ecoinformatics.uvm.edu/dmaps/growl/GrowlApplet.html
>> >
>> >
>> >
>> > It shows camera ontology:
>> > http://ecoinformatics.uvm.edu/dmaps/growl/camera.owl ,from where I
>> > deleted one import statements. Apparently displaying owl with
import
>> > statements in an applet is not going to be straightforward.
Applets
>> > are not allowed to read from urls other then the server they came
>> > from, so import of the files from other host would require some
>> > workaround . It is possible to create a java servlet which would
>> > fetch the files from "foreign" sites, but this would also require
>> > extensive modifications to OWLAPI
>> >
>> >
>> >
>> > I replaced "anonymous" label by "restriction", as Rich said. Please
>> > advice if this is better.
>> >
>> >
>> >
>> >
>> >
>> > Serguei
>> >
>> >
>> --
>>
>> _______________________________________________
>> seek-kr-sms mailing list
>> seek-kr-sms at ecoinformatics.org
>> http://www.ecoinformatics.org/mailman/listinfo/seek-kr-sms
RW>
RW> _______________________________________________
RW> seek-kr-sms mailing list
RW> seek-kr-sms at ecoinformatics.org
RW> http://www.ecoinformatics.org/mailman/listinfo/seek-kr-sms
More information about the Seek-kr-sms
mailing list