Wednesday, January 04, 2006
New
discussion about signal pathways
and complex ontology
This is part of a discussion that will be moved to a Wiki page soon.
Part of a private
discussion, (therefore the other side
is omitted)
You and I share many sensitivities about what the tools are and what the tasks might be.
Your interest in pragmatics, insights from specific use cases, is shared, and this is where I feel the SW tools have strong limitations. However, I also strongly agree that the current approach to communication of data within a world science community is going to be aided by the use of OWL. Specifically this work by the BioPAX project - if the science continues to control the computer science.
The limitation in the W3C standards comes in the clearest form in moving observation (measurement) to be assigned to standard signs within the community agreed to model. Not only are lowest common denominators constraining the community agreements, but reductionism takes a particularly fundamentalist stance - that has to be agreed to or there can not be agreement. The best science is done by 1% of the scientists. The rest are followers. (Opinion)
If this measurement is assigned factually, the "meaning" of the data is left open. The measurement is structural. The meaning is something that is always subject to interpretation. So let us build "Semantic Interpretation Environments", and let the scientists be scientists.
http://www.bcngroup.org/beadgames/generativeMethodology/complexOntology.ppt
The same is true for business and computer science. Computer science must adopt a more servant role, rather that imposing the "knowledge engineer" in the middle of everything. The untapped potential that you speak about awaits mankind.
As with pure mathematics and physics, there is a separation between the formalism and the scientific practice. Those of us who know the foundations of logic and mathematics know that this should be expected. Many, or most, in the SW and AI community do not have this realization. So they over sell the work, and in this enthusiasm they blur the difference between the formal system and the natural systems. This is a dis service that has not been balanced by the service provided,(in many people's opinion.)
In the case of data interoperability between computer programs one wants to be able to exchange data without human intervention.
This is structural and also never involves the full notion of "meaning", a point that is hotly debated by the AI/SW enthusiast. This is where the dis service becomes extreme. It is also unnecessary that meaning be reduced to structure. see Rosen's viewpoint
http://www.ontologystream.com/beads/nationalDebate/314.htm
Protege cannot ever achieve both completeness and consistency ( a result I draw from Godel, and others (Penrose, Rosen)).
Conjecture: The solution is then to "open" up the system to allow "human" inference and the abduction of new categories (signs) in real time, all the time.
An interesting corollary to this conjecture is the possibility that mathematical techniques like latent semantic indexing (or my work on shallow link analysis iteration and parcelation (SLIP) ) can identify when there are new patterns of co-occurrence (or other "relationships"). This is semantic extraction done right, and so far none of these semantic extraction systems has been able to reach a full potential, as you know. Semantic extraction, like so many other things in AI/SW is misnamed and this naming causes us to not understand. It should be merely "pattern measurement". Meaning is not extracted by the mathematical tools, it is only patterns that are extracted. The assignment of a pattern to a category with assigned "meaning" is an induction/abduction that is outside of computational ability. (opinion)
Thus the task, assisting world side scientific discussion, in front of us has two separate sub tasks:
1) structural interoperability of data (which is what the current BioPax activity should continue to focus on)
2) the development of data interpretation interfaces that act like semiotic (sign) systems and allow the human to be in charge of abduction (induction) of signs related to interpretations of meaning from profoundly interesting measurement devices.