Writing a fake Voynich by rules: copies, habits and the link between words

[Mauro]

So I changed one thing: the scribe now learns its letter habits seperately for each hand, and for each section within a hand, instead of one table for language A and one for B. With that the curve comes out almost right: ....

I think this observation supports a view of 'punctuated evolution' of the VMS text, which I think is also supported by the recent studies of Layfield & @LisaFaginDavis and @npcompl33t. Some sections are more uniform (ie. Q13 Balneo/Bio), while some others are more varied (and with a recoverable internal 'evolutionary' gradient according to @npcompl33t) but in-between sections the changes are greater.

One more question please: what's the percentage of hapax legomena your model generates? I'm curious because a structure prefix+midfix+suffix does not seem enough to me to account for many of the hapaxes.