Friday, October 28, 2011

A bit of ontology maximization

Let me deal with a set of keywords K containing a finite collection, k1,k2,k3,..
These keywords equivalently point to the same terminal, a URL resource, as in:

X.Y.(k1,k2,k3,..).URL

where I use my simple semantic syntax.  In that search order, all the kn are equivalent. But we want to keep the kn pruned in the sense that seldom used kn are removed to a separate longer semantic string:

A.X.Y.(k4,k5..)

Hence preserving the -iLog(i) rule, where i is the probability of occurrence.  Under that rule, I think we can prove that the optimum ontology would have each of the linked sets: K1,K2,K3.. contain keywords with uniform probability of occurrence.

There, a simple rule Imagisoft donates to the open source folks. So we come up with the simpler of ontology formats, yields a smaller set of prune and branch rules.  Each prune and branch rule comes with adjustments, like knobs on a radio, letting the CEO dial in a precision to his company semantic graph.

The Imagisoft proposal imposes little burden on the employees, except they occasionally blog something interesting to the company, complete with any web links.  They blog with straightforward business writing, going from general to specific.

No comments: