Re: Implementing a graph algebra
Date: 22 Nov 2005 02:24:20 -0800
> Thanks, Jim. To me graph-like meta-models of data with their algebras
> are clearly "database theory" stuff. But I understand this forum albeit
> called that is really more (exclusively?) about relational theory.
> Maybe I'll try a graph theory place, or boost::graph as you suggest.
What I meant with my poor terminology was that graphs are insufficient for knowledge representation, purely on the fact that not all relationships are reducible to dyadic predicates and graphs with their source, target, edge structure are necessarily binary in nature. A good example is by Darwen (If my memory serves me correctly), who highlighted the following predicate: Teacher A uses Book B to teach Course C.
Propositions that fulfill this can't be broken down to a binary (and hence graph) representation. Now you can of course say there is a relationship R, with properties A, B and C, and store that as a graph (and I believe there is a lot to be said for this), but then you have moved to a different conceptual level where you are interpreting an n-ary predicate (the logical level that this group tends to discuss), but _encoding_ it using a graph (the physical level). This is the very source of confusion that has plagued the semantic web, who confuse using RDF as a tool and a model, and so now have a extremely inconsistent and brittle structure.
imo ;) all best, J. Received on Tue Nov 22 2005 - 11:24:20 CET