On Jan 12, 2009, at 10:28 AM, Francis McCabe wrote:
If there are no primitive symbols, then all you have a network of relationships.
Even if there are primitive symbols in PatC's sense, still
all one has is a network of relations :-)
For that to be useful, there has to be a way of relating the network to the Real World™. One might imagine that it is straightforward to 'link' a symbol to a feature of the RW
No, I think this is very tricky. What is it that links a name such as "Cymri" to a country such as Wales? Is it just a kind of mass illusion/convention that Wenglish speakers all implicitly adhere to? (If not, what is it?) And if so, how do we get computers to participate in this natural-language-speaker's convention, when they don't speak our language?
; but relating a sub-graph to a feature sounds more challenging.
Agreed. The analogy I like is that the network of relationships is like a loose cloth which one is trying to keep fitted to a solid reality. It isn't necessary to nail down every single point, but it is important to keep enough of them attached to stop the rest of the network from floating freely. Areas which have no firmly attached points can drift away from the solid shape and take on other shapes. The art is in finding just enough binding points, and a certain amount of slack is inevitable.
On Jan 12, 2009, at 7:26 AM, Pat Hayes wrote:
On Jan 11, 2009, at 11:00 AM, Francis McCabe wrote:
May I assume, based on your responses to this somewhat repetitive thread, that you might recognize that a major issue for Ontology engineering is the symbol binding problem?
Hi Frank. Well, I do indeed think that symbol binding is a major issue, yes. For the semantic web it is the elephant in the living-room right now. However, I am rather at a loss to see how you got there from reading this particular thread. Maybe there's some obvious link I havn't noticed.
IHMC (850)434 8903 or (650)494 3973
40 South Alcaniz St. (850)202 4416 office
Pensacola (850)202 4440 fax
FL 32502 (850)291 0667 mobile