How to recombine glyphs to increase character entropy?

[MichelleL11]

Usually it would increase entropy. However, in the VM this is never certain, because its predictability works on various levels. So of you introduce this new "u", chances are that this "u" itself becomes very predictable (often preceded by "ch", followed by "d" and so on).

This makes sense and is the reason why subbing out bigrams for single plaintext letters is the way to push the entropy toward a more “language-like” number.  You can think of the use of bigrams as the artificial “cause” of the low entropy (high predictability).  What Koen’s work shows is that it is possible to address the entropy in this manner.  Of course it doesn’t explain precisely how it was done just shows it is possible to move the numbers in the right direction.  

I have used the list of bigrams as suggested combinations in testing out cipher approaches without success - but as common with the VM - my tests indicate it can’t be the only effect in use.  But l greatly appreciate having some direction to pick from all the myriad of choices once you try to test out verbose ciphers.

[Ruby Novacna]

Usually it would increase entropy. However, in the VM this is never certain

Even if it is not certain, it is just a proof of common sense: double e is a good candidate to transcribe u, which does not exist in EVA. Maybe it won't be right everywhere if the combinations ee and ei sometimes exist.

[Ruby Novacna]

I have used the list of bigrams as suggested combinations

I don't know how this list looks like, but if there is a ch bigram, when it is only one glyph, or the pc bigram, when it is p+half of the ch glyph, how do you do it?

[Searcher]

I must admit that I don’t like math and calculations, it seems to me quite boring, so I usually don’t delve too much into statistics, into various tests.

Nevertheless, to find the problem, I ran some experiments with the text, and the results were disappointing. I am sure that you have done all the same manipulations before me, but I hope that I will add something new. I need a little more time to note all my observations, but in general, I can conclude that it is impossible to increase entropy of the VMS text to the necessary level just reconstructing glyphs or ligatures, either excluding spaces and changing borders of the tokens.

[Koen G]

It is a difficult task for sure, which shows just how far removed Voynichese is from plainly written language. It s easy to "translate" isolated words, but nobody manages to translate a paragraph without interpretative steps and other hocus-pocus.

You may be a bit quick to conclude that reaching language-like entropy is impossible though. I did not manage to increase entropy to the required levels just by recombining glyphs, but I came somewhat close. Rene's results look even more promising, and I am very curious to see what he did.

[Aga Tentakulus]

I must vigorously disagree.
Voynich is not far from a natural language. You are only trying to solve encryption like a crossword puzzle, but disregard the diversity and inventiveness of cryptology.
That is why the results are so far away.

Why do you not understand the simple.
If 5 people have the same writing tolerance, then something is inherently wrong.
Here, deception comes before encryption.

[Koen G]

I don't quite understand your reply, Mr Tentakulus. I say Voynichese is far removed from plainly written language, and then you vigorously disagree and accuse me of ignoring cryptography? Cryptography is the opposite of plainly written language, and what we are investigating here is to what extent one form of cryptography (the verbose cypher) could solve certain issues.

[Ruby Novacna]

It s easy to "translate" isolated words, but nobody manages to translate a paragraph without interpretative steps and other hocus-pocus.

It's true that some words jump out at you by their great resemblance to existing words. 
It's also true that this does not concern entire paragraphs. 
The third truth is that to ''translate'' separate words requires us to pay attention to the glyphs that make them up.
And I keep asking myself why the glyph "ch" is considered to be the joining of two letters?

[Aga Tentakulus]

It is not the Voynichtext where is far away, it is the solution path.
It's the same experiments with the same path that gives the same result.
When I say 4o is not equal to qo, they are already possibilities. But if I use EVA for the programme, it only gives one possibility.
But it could contain the prefix ex- and qu- in the same, whereby ex- can also stand alone.
With daiiin, one wonders how to evaluate the three iii. Where is the difference in eva with i, ii,iii.
Three times the same sign most likely gives a wrong result.
The whole thing is full of so many errors that a justifiable result will never appear.
This is not a criticism of you Koen, but of the attempted solution.

Translated with You are not allowed to view links. Register or Login to view. (free version)

[Koen G]

And I keep asking myself why the glyph "ch" is considered to be the joining of two letters?

That is just the way it is written in EVA, but nobody says how it should be considered. For experiments like these, representing the bench glyph as a single character is a obvious choice.