[Top] [All Lists]

RE: [ontac-forum] A suggestion for ontological discussions at ONTACmeeti

To: "ONTAC-WG General Discussion" <ontac-forum@xxxxxxxxxxxxxx>
Cc: Tim Berners-Lee <timbl@xxxxxx>, Robert O'Harrow <oharrowr@xxxxxxxxxxxx>
From: "psp" <psp@xxxxxxxxxxxxxxxxxx>
Date: Mon, 21 Nov 2005 09:18:05 -0700
Message-id: <CBEELNOPAHIKDGBGICBGAEBKGOAA.psp@xxxxxxxxxxxxxxxxxx>
Gary, your note, just posted, suggests that convergence is perhaps the most critical aspect of the work of the working group.
My position suggests that the working group should move away from a specific notion that "machine intelligence" should be part of (the next round of) core ontological infrastructure.  An alternative is suggested. 
My argument is that controlled vocabularies should be observed using semantic extraction technology and then stood up as the foundation to machine like web services based on the observed use of language (within communities of practice). 
But in doing so, the "system (those who develop the core)" must be aware of its power to control and inhibit innovation and diversity of viewpoint.   (This seems like a simple objective and is within the mission statement of this working group.)  The Roadmap for semantic technology adoption that I produced for US Customs goes into great detail on how ten technologies available in the market today could be integrated , within three months, at a cost of less than 1 M.  The resulting system would be consistent with what I argue for. 
The issue is complex.    What I mean by "machine intelligence" does not preclude advanced pattern recognition programs from computing with the elements of a controlled vocabulary. 
It does preclude messy inference logics such as what have become attached to RDF when OIL was added to produce OWL (Ontology Web Language).  I suggest that this attachment of inference logic to the RDF statement compounded an error made by RDF in treating the graph as a tree. 
[  **  Let me be clear here.  A graph is a set of triples { < a, r, b > }  where a and b are nodes and r is the connector.  The RDF triple makes the a and b "ontologically distinguished".  The RDF triple requires that the a be a subject, the r be a verb and the b be a predicate.  This is useful for the purpose of introducing predicate logics, AND for writing data into a tree structure in the computer.   However, the more general construction is the graph having the form { < a, r, b > } .   This form leans itself to data encoding using a keyless hash table invented, and patented, by Bjorn Gruenwald in 1996.   The more general construction does not have to be aligned with predicate logic, but can be quite easily aligned to be compatible with RDF/OWL.  ** ]
I claim that the leading (and only) RDF/OWL editor, Protege, is unstable, the developers of Protege are constantly making revisions that do not support systems developed over the past months.  Simple things like import and export of information cannot be done reliably.
I claim that the problem is beyond the ability of computer science to solve.    I suggest that deductive mechanisms cannot meet the requirement of real world needs by US Customs or any other large organization that has to deal with the real world in real time.  The argument is specific, detailed and grounded in a scholarly literature. 
is an example of the use of a controlled vocabulary that is derived from
                             "a set of observations by a specific group of individuals" 
about the words and semantic used in the scientific literatures.  Barry Smith's work on this has many, in my opinion, excellent aspects; including a definition of "methodology"
and a set of relationships (which can be extended by the scientific community)
And BFO defines two small ontology schema (set of "concepts"), one for representing static knowledge, <snap>, and one for process knowledge <span>. 
But.....  there is a significant literature about the response degeneracy that all living processes exhibit in the normal course of _expression_.  I have not seen how this normal and common element of processes can be represented using the BFO.  I would like to test the set of proposed relationships to see if this set allows me to talk about Gerald Edelman's work (a noble Prize winner in Immunology). 
The discussion could move in the direction of extending the set of relationships, and in refining this set of relationships in the present of scientific peer review.  
The set of relationships specified in 
<(within organizational scale <snap,snap> and <span,span> relationships>
  part to whole relationship  (i think there is a class of these relationships)
<snap independent,span>
hindrance, prevention
<snap dependant ,span>
<span, snap> relationships between processes and substances
  sustaining in being
  temporal projection
  spatial projection
I hope I made no mistype.  These set of relationships is an ontology in both the sense of being a set of concepts and a referent to reality.
The point being is whether these concepts / referents are sufficient to express the science of biology without imposing a specific set of basis on the science community using this set of concepts ?  Is "intelligence design" for example inhibited by the requirement that authors publish consistant with the BOF?  (I do not think so, but it is a legitimate question.  Yes?)
Most of the work here is in regard to satisfying some funding source, or in the technical issues related to the computer science.  Thus the work is shaped by funding and computer science more than science. 
My voice is about the restrictions or benefits that will come to the science community when controlled vocabularies and ontological models are required for publication and tenure.
***** On a different matter
There is a position that some views are not considered. 
John Sowa has forwarded his views to me and has included a statement from one of the participants - whose position argues that I should not be allowed to post to the forum.
I am happy to not make anymore posts if the majority of those in the working group who wish to vote will ask this of me.  So please vote - but do it publicly. 
John's notes was posted in my web log at:

Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (01)
<Prev in Thread] Current Thread [Next in Thread>