Monday, January 23, 2006
[361] ß parallel discussion in “national debate” bead thread
The Second School of Semantic Science, on
founding
The post-OWL paradigm
non-hierarchically defined ontology
Communication as part of a Protégé eforum (local copy of relevant part of the discussion)
Samson,
First, I noted that your phrase
"set/subset semantics of class/subclass relationships" ß (link to google on this phrase)
is interesting language and language that does in fact capture the essence of some technical restrictions that are coded as enforced Protege-frames behavior. As seems the case, "semantics" has a restricted technical meaning ... imagine talking about the semantics of arithmetic !!
But those who know me know that I have made this type of observation before. The "meaning" of an abstraction needs to be restricted to instances, on the one hand; and abstractions (or categories) on the other hand... and yet we have only one word.
Some interesting google comes up with "set/subset semantics of class/subclass relationships"
The first retrieved element is
http://www.mondeca.com/BernardVatantKT2001paper.htm
Bernard Valant is someone whose position is (somewhat) close to mine, about complexity and limitations of the class-subclass ontologies, and is a member of the Topic Maps community.
In any case.... the line of research has to do with non-hierarchically defined ontology.
Those workers whom develop ontologies in the class-subclass tradition will know of this "set/subset semantics of class/subclass relationships" distinction; and should see this as an interesting phrase. But the phrase is (according to google) loosely and more strongly associated by usage in the Topic Maps community ...
In "complex systems" research (as defined at Los Alamos and other places), each "natural system" has an private part (an endophysics in this community's language) and an interface (autopoiesis) between the private part and the environment (the exophysics). This complex systems theory, and language, implies that class - subclass relationships are not "real" but only "abstract".... and yet one can see how useful taxonomies generally are.
Taking this point of view, for a moment..
So there is "in reality", ie in "natural ontology" no nested class, sub-class relationships. I agree that we are not talking about angels and archangels, where perhaps real nested ontological realities are assumed (I imagine) to actually exist independent of our abstractions (our way of thinking about reality).
Again, according to this viewpoint....
These set-subset and class-subclass "relationships" exist in the formalism, in the abstract notions that humans sometimes thinks about (or with) - but NOT IN THE REALITY of natural systems.
Set-subset would appear to be not precisely the same as class-subclass in this regard. Of course.
So there are instances where someone doing "ontology" with Protege-frames or Protege OWL would have various problems expressing what one wants to express, even if one is talking about everyday things - like e-business services. Practical problems in web service ontology (OASIS SOA-IM standard) arise due to the natural complexity of social expression, and when one relies only on Protege frames and/or OWL then one simply cannot reach solutions that are available if one uses an n-ary approach, or that may be available if other kinds of ontological modeling standards were commonly known and had free software supporting it. The set of limitations of Topic Maps is a completely different discussion (of course).
If one is talking about structure the set-subset, Protege-frames and or OWL, paradigm makes a lot of sense. One might have a set of things and a subset of those things. But the class-subclass distinctions seem to be more completely NOT ontological in nature. The human class with subclass female and male is not a structural distinction, like a bag of four marbles and a subset of two of those four. The set-subset distinction for the class subclass relationship is in the abstract, not the concrete. At issue here is the function of things that are categorized into classes. (I hope I am not stepping on any toes here, but when one moves to orchestrated composition of responses to web service requests, then the function of things comes strongly into consideration (in ways that I have not got clarity on, but this is where this line of research seems pointed)). We are looking for helpful insights.
The nature of class-subclass distinctions is more related to how we know, how human knowledge might be organized and this is why OWL and Protege frames is being used, or is being attempted to be used, to express human scientific knowledge of gene and cellular expression.... as in the BioPAX work (www.biopax.org ) .. I use the BioPAX work only as an example of what I consider to be excellent work on communicating human knowledge about how biological science categorizes knowledge into categories where class – subclass taxonomy is useful.
I do not know Steven Wartik, but I have seen this specific problem with taxonomy based ontology models.
One needs the class-subclass relationship to define transitive properties and to thus enable the OWL retrieval of information - what is called OWL reasoning with one of the "reasoner".
So we come back around to the real limitations and strengths of the Protégé ontology paradigm, whether expressed as a frames or as RDF with OIL ie as OWL.
Paul
Prueitt PhD (mathematics and quantum cognitive neuroscience)
The Taos Research Institute
Taos New Mexico