The Voynich Ninja

(14-04-2017, 08:49 PM)Anton Wrote: You are not allowed to view links. Register or Login to view.

Quote:Despite Davidsch' and Anton's remarks, I maintain that it is (unfortunately!) impossible to prove that the text is meaningless. All one can do is demonstrate that a hypothesized method of text generating is feasible.

This is effectively the same as to say that "it is impossible to prove that the text is meaningful. All one can do is to demonstrate that the text can be unambigously mapped to a certain plain text in some language. But the fact that such mapping fits may be purely accidental."

Ceteris paribus, if one demonstrates a method (of text generating), for me that would present a definite argument (I wonder why it won't for you). The problem is that noone has yet succeded in that, and for sure cannot succeed as long as existing transcriptions are used which simply do not address all peculiarities of the script.

So we are comparing two odds here:
1) the chance that the VM can be perfectly mapped to a single known text, while it has nothing to do with that text
2) the chance that someone can come up with a way to generate Voynichese exactly, even if it wasn't generated that way

I agree that both odds are small. If there is only one known text the VM converts to perfectly, then the chance of that being wrong is infinitely small. But if someone comes up with a procedure to flawlessly produce Voynichese on the fly, (case 1), I'd still consider the chance that this is a coincidence. It may be possible, with enough patience and knowledge, to write a program or procedure that does the same for Latin for example. Only, Voynichese is somewhat simpler, so it "feels"different.

Quote:Imagine a scenario where the VM is the 100th book this person has filled with gibberish language. It has become his second nature to make it look and feel like normal language and the words pour out as his pen flies across the vellum. Basically, the writer is already very well trained in writing Voynichese, and through the years it has become what it is now.

This would mean that the text is utterly meaningless, but no schemes or procedures have been used at all.

Codex Seraphinianus of 15th century?   Why not. But random text generation (and "words pouring out" is just that) also falls under my notion of the "procedure".

Quote:It may be possible, with enough patience and knowledge, to write a program or procedure that does the same for Latin for example.

Here the key limitation is that the procedure should be compatible with 15th century level of productive forces (knowledge, technology, machinery...)

Yes, that's true, though my Latin analogy is somewhat flawed since we can judge it more accurately than Voynichese. Latin would be much harder to generate procedurally since we know the vocabulary pool that's allowed and the ways words are allowed to combine.

For Voynichese we don't know whether the words in the manuscript represent the complete vocabulary or not. And more importantly, we know preciously little about its grammar or lack thereof...

There is of course yet another possibility. This would be (e.g.) that only the first words of each line contain meaning, and the rest is meaningless filler. Thus the text would be 90% meaningless.
A more subtle variation of this would be that the first word is a pointer (a number) that says which word on the line is the one that is meaningful.

Again, there are both arguments for and against these possibilities, but the point was not to go into that here.

With respect to Donalds point about Montemurro:

The problem is that the transition probabilities have been extracted from the end result (the text as we know it). This creates a vicious circle. As Nick already pointed out, this is also not an approach that can be assumed for an early 15th century person.

The same method was already used by Bennett, and implemented in Jacques Guy's Monkey program, though that strictly works with single characters. I think one of the main differences is that in the present method words are split into a start and an end. This helps to avoid frequent long words, which will naturally arise in the more standard 'monkey' procedure.
There are some nice example texts in Bennett's book, for example pseudo-English generated by this procedure, based on transition probabilities of Edgar Allen Poe's writings.

Donald: thinking about the structure of your argument a little more, it seems to me that what you have put forward is a kind of hybrid of traditional cryptologists' letter contact tables and modern cryptologists' state transition models.

The point of compiling letter contact tables is to try to understand what a letter context predicts about immediately following (or preceding) letters.

The point of building state transition models is to try to understand the sequentially threaded implicit grammar that goes just beyond what you get in letter contact tables.

Modelling Voynichese solely as a (large) set of prefixes and suffixes is arguably even more reductive than Gordon Rugg's table (if more empirical). But I don't honestly think { {pick a prefix} x {pick a suffix} } really counts as a valid state transition model in any useful sense of the phrase.

What I think you then do is try to draw the kind of conclusions you would draw from a state transition model if that were the only valid state transition model possible, along the lines of "because this is the only valid state transition model, Voynichese is therefore meaningless".

However, there are infinitely many other state transition models that aren't constructed as prefixes x suffixes purely after the fact (as yours is), and the failure of your model to please really can't be taken as a theoretical refutation of all of them, can it?

(15-04-2017, 09:13 AM)nickpelling Wrote: You are not allowed to view links. Register or Login to view.Modelling Voynichese solely as a (large) set of prefixes and suffixes is arguably even more reductive than Gordon Rugg's table (if more empirical). But I don't honestly think { {pick a prefix} x {pick a suffix} } really counts as a valid state transition model in any useful sense of the phrase.

Because I wrote the blog pages in the order I did the analysis, only correcting things if I subsequently discovered them to be wrong (i.e. mistakes), it's easy to misinterpret what I wrote.    I could have avoided that if I simply wrote a paper after doing the research, leaving out any ideas which I pursued along the way that were eventually dropped, such as that words are composed from prefixes and suffixes.   But I also wanted to show the path the research took, not just the final result.   Perhaps I should have advised people to read my blog backwards.

My theory is that words are composed of individual glyphs, not prefixes and suffixes.   At each state, a glyph is output and then a transition is followed.

Quote:@Fisk
My theory is that words are composed of individual glyphs, not prefixes and suffixes. 

Donald do I understand you correctly?  This would leave the vocabulary of the Voynich to be very limited.

The situation can be boiled down slightly - we don't have to assume an entirely open field when wondering whether the text contains a meaning or not.

For a start, we know the text is not random. It obeys certain You are not allowed to view links. Register or Login to view.. Therefore, it follows that a text generation mechanism is in place.

What is the purpose of a text generation mechanism? Either to:

1. encode information
   
2. generate pseudo-meaningful text
   

So we now have only two options - the text is meaningful, or the text is meant to look meaningful. Can we devise a thought experiment to narrow this down further?

If 1), then the mechanism could be a) natural or b) artificial.
If it's a), then it's a natural or artificial language.
If it's b), then it's a code or cipher.

If it's 2), then a deliberate deception is involved.

Or am I missing an option?

Let us examine 2) first.
If we assume a 15th century origin, then it is plausible to assume the mechanism was in place to make the text superficially plausible, and to avoid frequency analysis of letters (using an invented alphabet would again reinforce this point). Such aims could also unintentionally generate the effects we see with modern analysis - the use of a generating mechanism could incidentally cause the text to obey Zipf's and the other Benson laws; there could be an element of intentional re-use of words in different sections and labels; etc etc etc.

If we return to 1).
For many reasons which have been outlined elsewhere, the idea of a "natural" language can probably be discarded. But the highly non-trivial layout of word structure does suggest an artificial language, albeit that such languages didn't start appearing until the mid 16th century.

If we look at codes or ciphers - well, it's not a simply cipher. The whole thing could be a code book, in which case the meaning has probably been lost. Or we could be looking at some sort of steganography. We'd have to examine this a bit further.

(15-04-2017, 03:39 PM)coded Wrote: You are not allowed to view links. Register or Login to view.

Quote:@Fisk
My theory is that words are composed of individual glyphs, not prefixes and suffixes. 

Donald do I understand you correctly?  This would leave the vocabulary of the Voynich to be very limited.

I'm unsure why you think that.   Using Table 1 on You are not allowed to view links. Register or Login to view., just under 90% of the words in the Voynich Manuscript can be generated, along with some others.    Variants of this table (with different probabilities) were used to generate You are not allowed to view links. Register or Login to view. (EVA text:  You are not allowed to view links. Register or Login to view.).

(15-04-2017, 08:30 PM)davidjackson Wrote: You are not allowed to view links. Register or Login to view.Or am I missing an option?

Yes you are, but it's rather a far-fetched one: that it's meaningful but not a language.   For example, it could be a musical score.   I have generated music algorithmically, and like my proposed solution to the Voynich Manuscript, it also used random input to generate apparently meaningful output.