A One-Page Ledger Method for Generating Voynich-Like Text

[Dunsel]

I would say the biggest problem in the script is on line 60.

Code:

"".join(random.choice(VOYNICH_ALPHABET) for _ in range(length))

Each character is generated independently there. However, Voynich words exhibit strong positional dependencies. Certain characters only appear at the beginning or end of a word, such as qo-, -dy, and -in for example. Statistically, the result is therefore immediately recognizable as random noise.

You are ABSOLUTELY correct.  But, that's not my generator. User You are not allowed to view links. Register or Login to view. gave instructions on how to create a fake language and I followed those instructions to create that generator.  The output of mine is on page one of this post.  He believes that the copy and mutate system I've described with rules is too cumbersome and not a good way to make a pseudo language and that just inserting random words periodically is.

One possible solution might be to assign weights to the individual parts of the word.

PRECISELY... which is what my ledger does.  Weighted letters.

The point I hope he gets is that making a pseudo language, if that's what the Voynich is, requires some rules.  Reducing those rules to a minimal set that still produces something statistically like a language is what I'm trying to prove. Random words just don't work.  In order to make that generator work, yes, that random word generator has got to go. But, what rules do you apply when creating a new word or modifying a word? That's where my ledger works and works well, as long as it's used with human intuition and constraints.

But, if the Voynich really is a pseudo language, no generator will be able to exactly reproduce it without cheating.  Statisticallyl?  Yes, I think so.  Word for word?  Absolutely not.  Could his idea for a generator work?  Very possibly.  It has potential.  Is it going to require a good set of rules to do so?  Of course.