The Voynich Ninja

Quote:d) Did you run statistics on all (most common) groups on the whole text ?

In my app I show the words with the highest number of similar words for every page (see You are not allowed to view links. Register or Login to view.). 
For this purpose it was necessary to calculate the number of similarities for each word on each page. Pages linked together by the same bifolio share indeed similar core words.
(for the bifolios see You are not allowed to view links. Register or Login to view.).

    // Quire 13
    "f75r"  : ["shedy", "ol"],      // CURRIER B
    "f75v"  : ["ol", "chedy"],      // CURRIER B
    "f76r"  : ["chedy", "ol"],      // CURRIER B
    "f76v"  : ["chedy", "qokeedy"], // CURRIER B
    "f77r"  : ["qokeedy", "chedy"], // CURRIER B
    "f77v"  : ["chedy", "qokedy"],  // CURRIER B
    "f78r"  : ["okedy", "ol"],      // CURRIER B
    "f78v"  : ["ol", "shedy"],      // CURRIER B
    "f79r"  : ["ol", "shey"],       // CURRIER B
    "f79v"  : ["qokeedy", "ol"],    // CURRIER B
    "f80r"  : ["qokain", "ol"],     // CURRIER B
    "f80v"  : ["ol", "chedy"],      // CURRIER B
    "f81r"  : ["ol", "chedy"],      // CURRIER B
    "f81v"  : ["ol", "chedy"],      // CURRIER B
    "f82r"  : ["chedy", "qokeedy"], // CURRIER B
    "f82v"  : ["chedy", "qokal"],   // CURRIER B
    "f83r"  : ["chedy", "qokedy"],  // CURRIER B
    "f83v"  : ["shedy", "qokal"],   // CURRIER B
    "f84r"  : ["shedy", "qokedy"],  // CURRIER B
    "f84v"  : ["okedy", "shedy"],   // CURRIER B

Quote:e) You made a graph with word-glyph repetitions compared with the 1 to 20- previous lines. Is that based on your own research on are you referring solely on the work of somebody else?

The graph with word-glyph repetitions compared with the 1 to 20- previous lines is my own research.

Quote:Where is the data that you used for the VMS?

As data for the VMS i have used the transcription auf Takahashi (see You are not allowed to view links. Register or Login to view.)

Quote:f) Further more, i do not understand why you choose to compare that with a poem. 

Poems contain mostly contain similar words. Also the VMS contains many similar words. If there is some similarity then with poems.
In my actual paper I calculate the proportion of identical words appearing near to the word in writing position. (see You are not allowed to view links. Register or Login to view.)

Hi all,

I've just found this forum and, of course, want to contribute

Thanks Torsten for your great work. I'm supporting your opinion of a constructed text as you've described.

The following thoughts are not new. But I want to put it into this discussion because there are relations between your discoveries and the following theory:

To make a text construction as you've described, one could use a disc as shown at f57v. To have a working tool, each of the four circles needs to be drawn at a separate sheet to have each one turnable at a separate layer. If you put the discs together and add another 5th layer with some notches (or just a pointer), you have a simple word generator. Well, some rules need to be added as: How is the word read, what's the start point, and so on. Another way could be a moveable set of four pointers (like a clock) mounted at one end and placed in the middle of the disc drawings.

Third way could be to leave the 4 discs as one sheet and to have one or more turnable mask(s) with some fixed notches to place over the discs (limiting the number of words that can be created). Depending on the position of the notches and the way you turn the overlay you will get a lot of words. It's like a turnable cardan grille.

The words created from such a device would look like the "words" in the VM. Depending on how the discs are turned (or how the fixed notches are placed) all the strange word / paragraph / letter relations would appear as you described.

One effect would be the high frequency words like "daiin" or "aiin" appears as single word and as part of other words. As you can see at You are not allowed to view links. Register or Login to view. these words do appear more often than other words at the discs. It also would explain impossible combinations of words / letters due to the limitations of the discs.

Of course, the disc at You are not allowed to view links. Register or Login to view. could be a draft for the real disc. But anyway, the results would be words as you've described.

If fixed notches in an overlay have been used, it should be possible to recover the position by using the rules you've discovered. But I guess in this case there have been used more overlays (maybe one per page or chapter).

In case the scribe(s) used turnable discs, it could be complicated to discover the algorithm of how the discs have been turned. But not impossible.

Of course, this discs could also be used to encrypt data. But it's rather inefficient. I guess it's used just to generate large amounts of text that looks like an unknown language. My assumption is this: As you have so many ambiguities, it's very complicated to decrypt, even when you have the right discs. So the conclusion would be: It's junk. There is no information hidden, it's just text garbage.

Quote:Of course, the disc at You are not allowed to view links. Register or Login to view. could be a draft for the real disc. But anyway, the results would be words as you've described.

Making such a disc is obvious and i tried such a disc, but unfortunately it does not give those results.

You can easily see that, for example take from 57v :

outer ring
defo
rkedam
ofol
ddal
etc.

or from the third ring

dkedar
aros
echty
tedas
etc.

those do not appear anywhere else in the VMS. 

Not as word, not as part of a word. 

So please provide at least 10 different samples.

---

Torsten

I have many more questions but now we have thread polution, i will try to ask only the important ones and 1 question at a time:

Quote:For this purpose it was necessary to calculate the number of similarities for each word on each page. Pages linked together by the same bifolio share indeed similar core words.

Is it possible that you describe or disclose me the way you check for similaraties?


By asking about  "word-glyph repetitions compared with the 1 to 20- previous lines."  i was hoping you tell me exactly what you did and what you did not test.

I would like to duplicate your test and in order to do that i need to make a good basis-compare mechanism.

I understand from the app that you simply take a word, or line and compare that 1-on-1 to other lines.

What i want to do, is take 1 page.
 compare the first word of a page with all other words and count the similar letters per word
 compare the second word etc. etc. until the last one.

Of course then do the same and show the letters that deviate from that word and show that count.

Thirdly then do the same and show exactly which letters are similar and which deviate per word.

That could be done on a page -page basis as well, but that would give a load of data if i compare each word to each of the other pages as well.

Quote:Is it possible that you describe or disclose me the way you check for similarities?

Hello Davidsch,

sorry for not replying sooner.


I have published the code for my app via GitHub: You are not allowed to view links. Register or Login to view.

In SimilarResults.swift you can find the code for determining the top two words with the most similarities for each page. Starting point is the function calcTopWords(voynichLoader: OriginalVonyichLoader) at the end of the file. (see You are not allowed to view links. Register or Login to view.)

The results of this function are stored as compareMap in line 507. (see You are not allowed to view links. Register or Login to view.)

To compare two words in Line 844 the function Levenshtein.getDistanceOptimized() is called.
(see You are not allowed to view links. Register or Login to view.)

The class "AutoCopyTextGenerator" is used to generate the pseudo text.
(see You are not allowed to view links. Register or Login to view.)

The class "Glyph" stores the dictionary "similarElements". This dictionary determines which glyphs or ligatures can replace each other.
(see You are not allowed to view links. Register or Login to view.)

Thanks, but going through the code will take a lot of time,
could you provide 1 example and describe it in layman's terms?

Quote:could you provide 1 example and describe it in layman's terms?

Did you ask for an explanation of generating a new word? Or for the explanation of comparing two words?

In life there is always a place for heroism. You just need to be away from this place.

No not a new word.

If you compare 2 words  ABCmXYZ AND 123M456
what do you compare and what is the process ?

For example.

ABCmXYZ 
123M456

i can compare 
A with 1
A with 2
A with 3
A with M
etc.

then
A with 2 
etc.

or i can only compare positions

A with 1
B with 2
etc..

or i can only compare 2-grams or only 3-grams
etc...

For comparing two words a modified implemenation of the Damerau-Levenshtein distance is used (see You are not allowed to view links. Register or Login to view. and You are not allowed to view links. Register or Login to view.). The idea behind the Levenshtein-algorithm is to count the number of steps needed to transform one word into another one (see also Torsten Timm, 2014, How the Voynich Manuscript was created, p. 6 - You are not allowed to view links. Register or Login to view.).

To consider the peculiarities of the VMS script, a function 'isSimilar()' in line 41 in Levenshtein.swift was added. This function determines if two glyphs are similar or not (see 
You are not allowed to view links. Register or Login to view.). The map 'similarGlyphsMap' defines which glyphs are similar to each other (see You are not allowed to view links. Register or Login to view.).
If a glyph is deleted, added or replaced by a similar glyph, this is counted as one change. If a glyph is replaced by a non-similar glyph, this is treated as deleting one glyph and adding another glyph. For instance, since 'l' is similar to 'r' the distance between 'chol' and 'chor' is one. And since 'l' is not similar to 'k' the distance between 'chol' and 'chok' is two.

Another example is the calculation of the distance between 'chol' and 'chlor'. The Damerau-Levenshtein distance handels the transposition of two characters as one change. Therefore the replacement of 'ol' with 'lo' counts as one change. Secondly there is an additional glyph 'r'. This also counts as one change. Therefore at least two changes are needed to transform 'chol' into 'chlor' ('chol' -> 'chlo' -> 'chlor' or 'chol' -> 'chor' -> 'chlor'). The edit distance for 'chol' and 'chlor' is therefore two.

For perfomance reasons the distances for all similar words to 'chol' are stored as 'chol_distanceMap' in SimilarResults.swift. This map shows that for 'chol' 33 similar words with an edit distance of one and 218 words with an edit distance of two exists. 
See You are not allowed to view links. Register or Login to view.:
let chol_distanceMap: Dictionary<String, Int> = [
    "shol" : 1,
    "shod" : 2,
    "otchol" : 2,
    "chor" : 1,
    "cfhol" : 1,
    "ol" : 2,
    "chol" : 0,
    "chal" : 1,

(18-05-2016, 08:45 AM)Andrey Wrote: You are not allowed to view links. Register or Login to view.In life there is always a place for heroism. You just need to be away from this place.

Andrey, I don't see what's the relevance of this to the subject matter of the thread and to the ongoing discussion, and what place do we need to be away from.

Maybe, a language barrier problem? In any case, please keep the discussion on topic and avoid spamming.