Semtation GmbH
cfillies@semtalk.com
SemTalk is
a MS-Visio based graphical modelling tool, which is used for a broad range of
applications as there are
The native
modelling language supported by the SemTalk [1] consistency engine is settled somewhere
in the middle between RDFS and OWL. It supports multiple inheritance, instances, object- and data type properties. Graphically we
follow the approach of the UML tools with boxes for classes and labelled arcs
for object type properties. Data type properties are being displayed inside the
rectangle of the class. We prefer this relatively compact notation in contrast
to the graphical DAML notation used in VisioDAML.
We
experienced the SemTalk language constructs as being as complex and powerful
enough to express most of the
In
collaboration with Ontoprise GmbH we have created an
F-Logic export interface to communicate with Ontobroker™
and OntoEdit™ [2].
In
collaboration with Network Inference Ltd. SemTalk has been customized to cover
full OWL graphically [3]. Because the SemTalk engine has been left unchanged,
it can not be used to check complex expressions, disjointness
or equivalence. The Network Inference product “Construct™” is designed for
reasoning on the graphically created OWL model with the Cerebra™ engine.
SemTalk has
export- import interfaces to RDFS and DAML. The main goal of these interfaces
is to make use of existing ontologies in various SemTalk modelling scenarios.
These
interfaces are limited to the language subset of the SemTalk engine. E.g. a
DAML disjointness axiom is being ignored by the DAML parser, DAML lists etc. are not being recognized. This
limitation applies to all DAML and RDFS imports described in the following
chapter. For the OWL implementation these restrictions do not apply
anymore: Full OWL can be parsed and
generated. We expect significantly higher quality of the import once all tools
will support OWL. It is currently not planned to complete the
Screenshots
of the resulting models are in the appendix. The resulting models will be made
available on http://www.semtalk.com
Loom |
We did
not try to convert the Lisp files |
OilEd |
After fixing
some issues on the On the
other hand this model is quite close to OWL. We tried to rename some XML
elements to OWL, but finally failed to import it mainly because of the
combination of “cons”- |
OntoEdit |
|
OpenKnoME |
We did
not try to convert the Smalltalk files |
Protégé |
Using
RDFS import. Ignored
by SemTalk RDFS Import even if the SemTalk engine could represent them:
|
Terminae |
We did
not try to convert the text / Oil files |
WebODE |
Failed to
import classes as rdf:description with rdf:type Class |
KAON |
Successful
import after manually removing the XML-namespace “a:” |
The overall
impression from a SemTalk standpoint is, that SemTalk
failed to import DAML models with complex expressions. This issue has already
been fixed for OWL, which is in turn not supported by the current versions of
the other tools. SemTalk succeeded in importing taxonomies from all tools,
which support DAML or RDFS.
From a
Being a
graphical OWL editor has not been the major goal of SemTalk in the past. The
first solution for OWL is the Construct version of SemTalk developed with
Network Inference early 2003. The intension of “Construct” is to replace the
non-graphical OilEd by an easy-to use graphical tool.
The
problems we found using the more sophisticated features of OWL in practice are,
that:
1. Fillies,
C., Wood-Albrecht, G., Weichhardt, F.: A Pragmatic
Application of the Semantic Web Using SemTalk, WWW2002,
2. Fillies, C.; Sure,Y: On Visualizing the
Semantic Web in MS Office, IV02
3. Fillies,
C., Ng, G., Thunell, A.: Cerebra Construct:
Inferences for End Users, WWW2003,
The
following screenshots show results of imports from OilEd,
OntoEdit, Protégé and KAON. The last screenshot shows
a part of the OilEd model
rebuild with the OWL shapeset made for Construct.
Screenshot 1 Oiled DAML Import
Screenshot 2 Instances of the
Screenshot 3 Instances of the Protege
RDFS Import
Screenshot 4 KAON DAML Import
Screenshot 5 Subset of the OilEd
Model redone with the OWL Shapeset