Voynichese is a numeric cipher?

[magnesium]

My general point of view on cipher models is that it's healthy to build them out and then test their properties against Voynichese. For example, every cipher model out there makes some kind of assumption as to what a Voynichese token represents on average. This should have immediate implications for how Voynichese tokens relate to each other. In principle, this sort of thing could be modeled and tested through statistics like mutual information and the co-occurrence of ciphertext word types, akin to what Montemurro and Zanette did for the Voynich Manuscript itself: You are not allowed to view links. Register or Login to view.

This would be a fun way of testing the behavior of various cipher proposals, such as my Naibbe cipher, Matlach et al.'s steganographic cipher, this numeric cipher, and JoJo_Jost's "vowel bridge" cipher (on specifically Bavarian).