Experimental replica of VMS properties with a given corpus

[bi3mw]

For some time I have been dealing with a fundamental question. Have you ever tried to use a ( medieval ) corpus to generate text that has similar characteristics to the VMS using cryptographic methods? The horse is saddled here, so to speak, from the other side. I mean no random text, but text that can be decrypted again (meaningful). For example, is it possible to replicate the steep curve of word lengths ? Are immediate,multiple word repetitions possible ?

It is therefore a matter of experimentally replicating the known properties of the VMS with a given corpus ( modeling ). Does that make sense in your opinion or is that too experimental ?