Thursday, December 22, 2005
Lattice of ontologies
Function/structure descriptions
Jennifer Vendetti
Stanford University, Protege Team
I thank
you for your note.
I know some about the history of the framed based KIF (Knowledge Interchange Format) and that Protege started as a KIF editor. I am "almost" right about this, yes?
As RDF began to be funded by DARPA and marketed by W3C there was a decision to evolve Protege towards RDF and then OWL. Yes?
As this occurred, did the frame based approach and the OWL approach develop as two different developments? Or were the two approaches mixed together in some way?
For example, does Racer (the inference engine) act on frame based ontology (I guess it is proper to call this OKBC - Open Knowledge Base Connectivity - rather then KIF. ) Or is KIF not related to OKBC?
Somehow it did not occur to me that there are two Protege packages - Protege-Frames and Protege-OWL.
http://protege.Stanford.edu/overview/index.html
I worked with a group who were trying to get Protege-OWL to work at Customs. I was trying to integrate semantic extraction technology with well-specified concept representations, but there was just too little understanding of the big picture. And there was WAY too much Java and programming language stuff and jargon. Too many moving targets.
My RoadMap was the result of the struggle. In the RoadMap we advanced some suggestions that (we felt) were more consistent with how the human brain/perception system works.
We parted ways from the (at least partially incorrect) notion that logic is "how thought follows". The RoadMap was NOT done as a business process but as a collaboration between a number of technologists, cognitive scientists, linguists, and mathematicians.
My background is in applied semiotics, computational linguistics, semantic extraction, artificial neural networks, quantum theory, general systems theory, mathematics, cognitive neuroscience, and theoretical immunology... I say this not to be egotistical but to remind others that I have a great number of scholarship fields to follow - other than IT, Java, and this constantly changing set of "standards". This is why I have called for some type of completion and stabilization of an editor/visualizer that works to represent "sets of concepts".
I feel that frames, as Schank described early on, is a really good way to represent a concept, as having slots and fillers. However, I have been critical of the logical entailment that folks (Hendler, Berners-Lee and others) are attempting to put into the W3C work.
The criticism is based on extensive backgrounds in cognitive neuroscience, and quantum theories of mental event formation (Pribram, Edelman, Penrose, Rosen etc). see:
http://www.ontologystream.com/beads/nationalDebate/314.htm
I know also of four or five "sets of concepts" representation that is build into software, Knowledge Foundations Inc, Applied technical Systems Inc, Readware, WebMind and a few others. I invented my own called Orbs (Ontology referential base)... all of these have some right to be called "ontology" but certainly are not OWL.
I see the OWL
projects fail because what is attempted might not be doable using RDF plus
logic organized as "theories". If any proof that ontology with
theory can work one would expect to see this from Cyc Corp, but we do not see
this - at least not in the general B-2-B case. But OWL seems
mandated by government work, and many of us simply cannot understand why the continued
support.
Sorry to abuse the group so much, but the W3C approach just seems wrong - whereas the frame based approach seems proper and doable.
The question I hope you can answer in the affirmative is"
Is Protege-frames completely
separate from the OWL and RDF standard. Is this true?
This question is very important to me personally, and I am so excited to maybe have stumbled into a way to get leverage on what I see as wrong and what i see as right about the approaches to ontological modeling.
Thank you