Thursday, February 16, 2006
Generative Methodology Glass Bead Games
n-articulated ontological framework
On using RDF to model biological expression
Link back to part of the “solution” to translations between
RDF / OWL and Models of Information
à- [167]
Concurrent Anticipatory web discussion à [30]
Discussion about the depth of OWL indexes
If I might, I would like to ask those here who have OWL projects to estimate the average depth of their class hierarchies.
So if we could get some ordered pairs
(domain of project, average depth)
this would be useful to some research we are doing on ontological modeling in general, and in information systems mapping specifically.
The study focuses on what constrains the addition (design) of a new concept to an existing ontology. Core questions include:
1) Is it easier to create a high level class or to attach a subclass to an existing class.
2) when one finds and uses an OWL ontology, is this ontology designed in a way that captures the Requirements, Motivation, Context, Goals of those who need to use the ontology?
3) are there blueprints, profiles and related structural work that can be used to run simulations of the ontology (as embedded in a use) ? If so, how are such simulations instrumented, measured and discussed?
A related focus is how are ontologies "best" designed. Does one start with a set of high level concepts that in some way "cover" the "semantic space", and the arrange these into hierarchies, or extend subclasses from the top concepts? In biology, scientists rely on a theory of evolution to justify deep class hierarchies. But in business ontology? Where does the justification for specific class - subclass "assertions" come from? The answer is, of course, from the designer of the ontology. But what goes into the design of an ontology?
It seems to us that there are many constrains on where a concept ends up in an ontology. But a global measure of these constrains taken together might be the average depth.
I will elaborate a bit. We have been studying the SOA IM (Service Oriented Architecture Information Model). The way in which this IM is constructed is similar in many respected to how UML is constructed, and similar to how object oriented programming is structured. There is a notion of "parent" that does not translate (easily at least) to the notion of a super-class). The notion of parent is similar to a previous state of the "system". SOA-IM entity "ActiveInputs" has as a "parent" "Decision". Whereas this is not a class, subclass relationship; this type of "parent" notion is common in UML, and in things like Petri nets.
One might also talk about how relational database schema are constructed (designed). Please correct me if I state something incorrectly, but it seems to me that UML, relational database schema and object oriented programming are ALL quite different than the DL-logic based ontologies.
This difference centers on the class, and subclass definitions. It is interesting that the notion of a class and the notion of a subclass are in fact part of the "ontology" of an OWL ontology. It is these definitions that give power to the OWL indexing (reasoning) that is created.
This power is given to the designer of a particular class, properties of that class, sub and supper classes, and restrictions.
reading material
"A tool for Storing OWL using database technology", by Roldan-Garcia and Aldana-Montes
http://www.mindswap.org/2005/OWLWorkshop/sub1.pdf