Saturday, November 20, 2004
Center of Excellence Proposal
à
White Paper on Incident Information Orb
Architecture (IIOA) à
Types of Ontology for Crisis
Management à
Adi Structural Ontology Part I
à
Cubicon language description
à
Topic map standard, the RDF and Readware
constructs
.
. .
and the Mark 3
A
National Project is possible, and this National Project should be based on
diversity and objective and publicly transparent peer review.
BCNGroup
Founding Committee, 2004
It is very clear to us that there are, in fact, several knowledge processing systems that have merit in specific ways. The founding committee has raised a number of issues about how one satisfies, on principled grounds, concerns related to application of the notions underlying five or six approaches.
For example, how does Richard Ballard’s Mark 3 technology become deployed within a society that now uses natural language? For Ballard’s theory of information, one sets aside natural language in favor of directly developing a model of each thing and each aspect of each thing in the world, as computer data encoded in the form of an n-ary. Other approaches, by Readware or Applied Technical Systems, have specific concerns that can be addressed in an objective community based evaluation. The short white paper begins to shape this evaluation.
Three Types of Ontology for
Crisis Management
Many approaches to a formative knowledge science have been developed. All have various types of merit, but an objective evaluation of the generality of potential deployments is needed.
Is this correct?
It is not merely the question of transitioning industry and massive government communities to something that must in a real sense be taken on faith. Unless knowledge technologies are deployed without cultural or proprietary encumbrance, the community that is receiving the technology will not make a fair judgment. Cultural encumbrances are different, but sometimes related to proprietary encumbrance. Both types have roots in the non-accessibility of knowledge about the nature of human knowledge and the nature of the technology.
For example, does one have to simply buy into Richard's belief that there is no real knowledge system but the Mark 3 or Mark X, and future evolutions of the Mark 2, Mark 1. How would this be different from the Cyc Corp polemics that took, along with the TREC competitions, almost all funding away from things like Hendler's work and anything else ontology-like. Government funding has pored into what most now see as a mythology invented by Doug Lenat. The Cyc Corp approach is pursued as if business always trumps science, and has long ago left open objective peer review behind. It is now only marketing and political contacts that fund Cyc Corp projects. But is this much different from the many other partial solutions that the IT business sector uses to generate revenue?
How is one marketing activity different, or similar, from Lenat's belief in a machine encoding of human common sense? This must be a question of objective and fair peer review. In the information sciences, objective and fair peer review has not occurred over the past three decades. During this time, the sense has been that the markets can make the optimal choices, and the markets have optimized value for the IT business sector.
It is also a question of accepting, or not, that one commercial solution is simply far better that reasonably developed and deployed topic maps, or even reasonably well developed and deployed RDF or OWL ontologies. If we accept one solution based on marketing or political contacts, what do other people do with their work? How can we evolve a community rather than a business consortium?
What the information sciences has NOT ever had is diversity.
There are a number of good choices that can be made in regards to the DHS requirements to have in place some type of information system that will allow information to flow that is reasonably complete, transparent and secure.
We treat this reasonably complete, transparent and secure requirement, as we see it, in the
White Paper Architecture for Managing Incident Information
We suggest that Orbs with cA and eC could fulfill the reasonably complete, transparent and secure requirement, but we do not say that topic maps could not, or that the Mark 3 could not. We see also that RDF solves a class of problems, but we also suggest with great humility that RDF and OIL do not solve many of the most critical problems that will arise in a natural world where diversity will generate things that one did not know about.
As Michel Biezunski stated in Chapter 2 of the book “XML Topic Maps, Creating and Using Topic Maps for the Web” (Jack Park. Ed, 2003):
“The relationship between RDF and topic maps is currently being
studied. Research has progressed far
enough to show that there is a distinction between types of high-level models
for information, in order, on the one hand, to provide to information owners a
neutral language, for knowledge representation and finding aids and, one the
other hand, to provide a way for computers to run applications.”
The key here is that information owners are humans, who own knowledge in a sense that is distinct from the notion of ownership of physical property. Computers require a synthetic construction so that the computer processor can execute precise mechanisms. Humans are not “imprecise”, which is bad; in contrast to computers that are precise, which is good. The language used is often misleading.
It is the sense of six or seven people that the approach based on categorical abstraction and event chemistry is a good choice, and one in which misleading language can be corrected.
We make the conjecture that ontologies based on a sub-structural ontology that is itself based on a framework is a reasonable way to both model human perceptual and cognitive abilities - and to use the scalability of computer process running within a two sided Semantic Web, one side being the human communities and the other side being the computer processes.
Recently, we have found and are exposing to a wider community the work by Adi. This new effort is not because the Adi structural ontology is the work of a friend, but because it has interesting theoretical principles and has a commercial software package that demonstrates very high precision-recall - as well as conceptual role-up properties that are not exploited as yet.
We also make the argument that topic maps, done in the way we saw in the IRS help desk deployment, would be a reasonable approach to creating the information system that the DHS needs for any one of a hundred different deployments.
We make the case that the group at Top Quadrant have exceeding fine techniques for the design of complex information systems that need to be deployed in a social environment with transition barriers.
The monitoring of port security is the one that is up before us now, by some accident of nature. A project at NARA is also within reach.
We ask for comments on the three types of ontologies, we conjecture, are needed to address the needs of this DHS task.
Our work on this is at:
http://www.ontologystream.com/beads/nationalDebate/165-1.htm
A National Project is possible, and this National Project should be based on diversity and objective and publicly transparent peer review.
Does anyone disagree?