More on Vector Models

Rehashing the merely obvious:

If a schema declares a vocabulary, the set of all schemas 
used by an enterprise is the set of the enterprise vocabulary. 
Each instance of a vocabulary is a vector against that 
vocabulary, and each vocabulary is a vector against the 
enterprise set.   The enterprise set is a vector against 
its ecosystem of actively communicating entities.  Messages 
are the terms of enterprise vectors.

If the value of indexing is expressed as the function of the 
density of objects in addressable space so that performance 
is inversely proportional to the space density (actually, the 
address space itself), XML vocabularies increase the 
density of the space as well as introducing ambiguity 
and uncertainty through semantic loading and can actually 
hurt the performance of the system. (yes|no ?)

That's why Bosworth's presentation has merit.  The problem 
however, is that it simply moves the calculation of the similarity 
metric away from the apriori schema declaration into raw 
microparsed vector results. A schema is the declaration of a 
space where occurrence indicators are a determinant of frequency 
and therefore, similarity given a rule that frequent terms are 
less important than rare terms within a document (term vectors), 
and more important across documents (document vectors).

len