W3C home > Mailing lists > Public > public-swbp-wg@w3.org > January 2005

Review of ODM Chapter 16 - UML Profile for RDFS and OWL

From: Holger Knublauch <holger@smi.stanford.edu>
Date: Fri, 28 Jan 2005 15:34:10 +1000
To: "'SWBPD'" <public-swbp-wg@w3.org>
Message-ID: <E1CuOmU-0007Xz-0s@frink.w3.org>

Thanks for the invitation to review chapter 16 of the ODM
draft.  Sorry for the long delay with my response.

I don't have any technical objections to this draft.  The
mapping would allow to represent many OWL DL models using
UML notation, and the chosen mapping is straight forward.
This approach fills the gap of visual modeling standards
for OWL and should hopefully be compatible to existing tools.

The only item that needs to be clarified is the purpose of
the mapping.  In its current form, rdf:Properties are mapped
into OWL classes, so that they can be used in multiple places
in the diagram.  This reflects the syntax of RDF, where properties
are stand-alone entities, independent from specific classes.
The obvious mismatch is that in UML, attributes and associations
are typically assigned to a single class.  I am sure the working
group has thought a lot about this fundamental mismatch, and
the current proposal is probably the most feasible way of achieving
at least syntactic interoperability.

However, the question remains where do we go from here.
The resulting "UML" diagrams will be of little use for people who
are approaching them with an object oriented background, because
properties are not modeled as they would be in UML.  So while the
mapping is syntactically correct, a much more important contribution
would have been to provide a semantic mapping between UML diagrams
from the real world, and OWL models from the Semantic Web, so that
developers can map between existing (legacy) UML diagrams or source
code, and OWL/RDF models.

I did not follow the state of the work in the ODM group closely enough
but I would be interested to hear about the plans for such a second
kind of bidirectional mapping between UML and OWL, because this is what
we as tool developers are particularly interested in.

On this occasion I would also like to hear of any real implementations
of UML-OWL mappings so that we could include them into Protege-OWL.

Holger
Received on Friday, 28 January 2005 05:34:54 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:09:41 UTC