[Article] Can Evolutionary Computation Help us to Crib the Voynich Manuscript ?

[bi3mw]

Daniel Devatman Hromada, 7 Jul 2021

Abstract:

Voynich Manuscript is a corpus of unknown origin written down in unique graphemic system and potentially representing phonic values of unknown or potentially even extinct language. Departing from the postulate that the manuscript is not a hoax but rather encodes authentic contents, our article presents an evolutionary algorithm which aims to find the most optimal mapping between voynichian glyphs and candidate phonemic values.


You are not allowed to view links. Register or Login to view.

[Scarecrow]

The first experiment led to identification of multiple substitution alphabets which allow to map 240 EVA letters, contained in 40 distinct words present in the Calendar, onto 35 feminine-name-ressembling sequences enumerated among 13815 items of ??????????. 

Results of second experiment indicate that if ever the Calendar contains lists of hebrew names, then these names would be more probably feminine rather than masculine.

This is, as far as we can currently say, all that could be potentially offered as an answer to the question Can Evolutionary Computation Help us to Crib the Voynich Manuscript?. Everything else is - without help coming from experts in other disciplines - just a speculation.

Another interesting approach, quite usual Voynich results.

[MarcoP]

Everything else is - without help coming from experts in other disciplines - just a speculation.

Experts already pointed out that entropy excludes a simple substitution cipher of an ordinary language. E.g. Lindemann and Bowern, Character Entropy in Modern and Historical Texts, 2021 (but this is known since Bennett's 1976 book):

At the character level, Voynichese is strikingly different from any other text in the Wikipedia and Historical corpora. The character set size and frequency of characters is conventional, but the characters are combined in an extremely predictable way, as indicated by an unusual conditional character entropy that is distinctly lower than any of the 316 comparison texts.

...

If Voynichese is encoded text, it must be more complex than a simple substitution cipher. A simple monoalphabetic substitution cipher will have absolutely no effect on conditional character entropy, because the same characters simply shift places with one another.

[nickpelling]

I believe Elisebeth Friedman described this a decade or more prior to 1976.