- From: Bullard, Claude L (Len) <clbullar@ingr.com>
- Date: Tue, 22 Jul 2003 13:17:26 -0500
- To: "'Graham Klyne'" <gk@ninebynine.org>
- Cc: www-tag@w3.org
I've read the signproc document. It was illuminating. I believe these pull together some points various contributors have been making. There is another paper on analogical reasoning systems which I think the Semantic Web advocates will benefit from as it explores the processes by which information systems proceed from analogical to logical reasoning (abduction, induction, deduction). http://www.jfsowa.com/pubs/analog.htm Some limitations of automated inferencing systems are made clear. The initial abduction limits the scope of the logical assertions (GIGO) and there are scaling problems. So we punt to the ontological commitment notions (when parties agree that x IS_A Y, that is sufficient for the architecture. If automated systems agree, it is sufficient if the parties using it agree it is acceptable. No Golems.) 1. It is unlikely as others have stated that a single set of upper level ontological assertions will emerge for the Semantic Web. I believe serious AI researchers accepted this long ago. 2. It doesn't matter as long as those that do emerge work acceptably well for accepted cases. 3. Pattern seeking systems will find them. The usefulness of each discovery depends on what one attempts to deduct. It is fun to compare a cat to a car but not useful when packing the cat for a trip to Europe or driving it to the grocery store, but may be for gassing it up. len From: Graham Klyne [mailto:gk@ninebynine.org] I was offline when I drafted my last response, and have now done a little digging. Short verion: [1] http://www.jfsowa.com/pubs/signtalk.htm [2] http://www.jfsowa.com/pubs/autotalk.htm [3] http://www.jfsowa.com/pubs/signproc.htm Longer version: It seems the keywords Sowa and Lattice appear in two different contexts: (a) Sowa's formation of an upper-level Ontology which is based on earlier philosophical works and appears in his book on Knowledge Representation. I don't think that is going to assist in the particular topic that lead to this discussion. (b) "John Sowa's potentially infinite open-ended lattice of theories", which might have some bearing on the topic.... This, too, seems to be related to upper-ontology work, but has more resonance with the what-does-a-URI-identify question. Folling this line, I found some possible jump-off points: http://www.jfsowa.com/pubs/autotalk.htm This slide suggests a link between theories and "what-is-identified": http://www.jfsowa.com/pubs/autotalk.htm#s17 and: http://www.jfsowa.com/pubs/signtalk.htm note reference to "Language Games": [[ Words only have a precise, formalizable meaning with respect to a particular language game. ]] and (this is a paper with much more detail, which I've yet to read): http://www.jfsowa.com/pubs/signproc.htm #g -- At 10:52 22/07/03 -0500, Bullard, Claude L (Len) wrote: >Mine too, but I'm almost terrified to leave this in >the hands of the experts. :-) > >I saw a reference to Scott's work in the Google listing. >I don't know what the originating relationships are. I >know that Sowa writes with unusual clarity on the issues >of concepts and set theory. He derives from Peirce and >clarifies that as well which is no mean feat. It may >be that for the architecture, one has to admit that >the theories about why it works are available but not >as important as capturing the how. In the case >of one URI = one concept, that is easy to do: >assignment. To the case of proving that there is >only one concept to which that assignment can be >made, that isn't doable except insofar as assignment >to the empty set (the theory of all theories) makes sense. >Sowa is clear about the lattice membership. > >len > >-----Original Message----- >From: Graham Klyne [mailto:GK@ninebynine.org] > >Getting out of my depth, but... does this have any relationship to the >work that Dana Scott did back in the 1970s on lattices and a theory of >computation, which in turn provided some basis for denotational semantics >of programming languages? I recall that the notions of approximation and >monotonicity came into that work, with some reference to functions being >ordered according to some notion of "accuracy". > >#g >-- > >At 09:45 22/07/03 -0500, Bullard, Claude L (Len) wrote: > > >You could research John Sowa's lattice theory for > >more precise language to describe this notion. > > > >Apologies but Google returns far too much material > >to provide a precise URI to start the research > >if you aren't already acquainted with it. And > >that tells us something about URIs and precise > >identification. :-) > > > >len > > > >-----Original Message----- > >From: Graham Klyne [mailto:gk@ninebynine.org] > > > >We agree that we, as people, try to use a URI to refer to > >a "single", more or less consistent, concept that is a topic of > >communication. But there is no way to formalize this single concept: I > >think the best we can do is to describe it as a kind of "locus" of > >denotations from interpretations that satisfy some formal statements we can > >make about it. > >------------------- >Graham Klyne ><GK@NineByNine.org> >PGP: 0FAA 69FF C083 000B A2E9 A131 01B9 1C7A DBCA CB5E ------------------- Graham Klyne <GK@NineByNine.org> PGP: 0FAA 69FF C083 000B A2E9 A131 01B9 1C7A DBCA CB5E
Received on Tuesday, 22 July 2003 14:17:33 UTC