The simplest way to incorporate this hypothesis is to take formula
8 above, but normalise 
for document length (d).
If we assume that the value of 
is appropriate to
documents of average length (
), then this model can be
expressed as
This function is used in the experiments described below (section 7). However, a more detailed analysis of the effect of the Verbosity hypothesis on the 2--Poisson model may reveal a more complex pattern.
