The Voynich Ninja

(05-02-2017, 01:55 PM)nickpelling Wrote: You are not allowed to view links. Register or Login to view.If you take the three (EVA) letters 'e', 'o', 'l', and count the occurrences of their six permutations (in the Takahashi transcription), I think you get the following:

eol - 961
ole - 39
oel - 0
leo - 2
loe - 0
elo - 0

These numbers strongly imply that Voynichese has very strong letter adjacency rules in play: and it is surely far from coincidental that both eol and ole are the only two permutations where l immediately follows o. (For reference, 'ol' occurs 5507 times in the same transcription.)

Without any doubt, these numbers are inconsistent with auto-copying if you are trying to argue that errors introduced during autocopying are the systematic source of the random variation within Voynichese.

Dear Nick,

 I didn't argue with errors and random variations. Since I assume that the VMS was created by a human mind I expect systematic variations. Moreover the autocopying hypotheses means that every word is at the same time the result of the copying process and it is also a possible source for further copying steps. In this way I suggest a self-referential system for the VMS.

The error in your example is that you suggest that the letters are used independent from each other. But this is not the case. The groups 'ol', 'al', 'ar' and 'or' are replaced with each other in a systematic way:  "Another example is that 'o' is interchangeable with 'a' before 'l' and 'r' ..." [You are not allowed to view links. Register or Login to view.]. This means that instead of 6 possibilities only two exists. This two possibilities are 'e' + 'ol' and 'ol' + 'e'. Your two words with 'leo' are 'oleoeder' and 'doleodaiin' and belong in the category 'ol' + 'e'. 

The matrix of possible permutations is therefore:
ole  39  [font=Courier New]ore 11   ale  10  are   7[/font]
eol 961  [font=Courier New]eor 464  eal 140  ear 171[/font]

The most frequent word containing 'eol' is 'cheol' (172 times). For 'eor' this most frequent word is 'cheor' (100 times). In the case of 'eal[font=Trebuchet MS]' it is  '[/font]cheal[font=Trebuchet MS]' (30 times) and in the case of '[/font]ear[font=Trebuchet MS]' it is[/font] 'chear' (51 times). As more frequent a word is as more similar words exist for it. Moreover since 'cheol' (172 times) is frequent also the word 'sheol' (114 times) is frequently used. In this way the result fits with the autocopy hypotheses.

(05-02-2017, 03:08 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.Thanks Nick, but don't hold your breath that Torsten will reply constructively.

Dear Emma,

please don't comment replies I will write in the future.

Quote:I pointed out the problem of word structure to to him in 2014 after his first paper. Over two years later he still hasn't adequately responded to it, saying that the writer was free to write as he pleased within his 'concept of language'.

Please don't change the facts. You did this in February 2016 [see You are not allowed to view links. Register or Login to view.]. I responded to you on your blog two times but until now your answer to my second post is missing. 

Quote:It's sad, but he's another Rugg with a neat answer which is all wrong.

Please use arguments even if you didn't like my hypotheses.

Torsten,

Please in a few words define what you mean in lay mans terms what, "auto-copying" is.  Are you implying that the VMS is not any known alphabet meaning the letters are just scrambled?

(05-02-2017, 03:08 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.I pointed out the problem of word structure to to him in 2014 after his first paper. Over two years later he still hasn't adequately responded to it, saying that the writer was free to write as he pleased within his 'concept of language'. I don't think he has an answer because it would spoil his theory.

As I recall, you recently mentioned cheol vs chole vs echol, which reminded me of the eol permutation thing I'd noticed a while back but had forgotten. Voynichese has plenty of similar strongly asymmetrical permutation sets, but I'm pretty sure that eol was one of the most extreme.

It's also interesting that we see all four gallows inserted in the middle of ch, but never anything else in the middle of ch. That's a rule all of its own as well, etc etc.

Torsten: so what you are saying is that even though Voynichese is autocopied from itself, the "itself" from which it is autocopied (i.e. that seeds the copy) contains near-universal adjacency rules which are then preserved in the autocopying?

And that these near-universal adjacency rules that are in the 'seeding' part of the text (and then preserved in the autocopied part of the text) may or may not be language but you're not interested?


And that all the word variations arise not from a conscious attempt to produce randomness, but instead from the autocopyist assembling many small blocks that retain those near-universal adjacency rules?

So... why is it that eol is 25x more frequent than ole, again?

I'd point out that you may both be right - we could well be seeing genuine text which is hidden in between null text generated via a mechanism like Torstein's auto-generating hypothesis. Nick / Emma are picking up on the message, Torstein on the gibberish.

Either way, this is fascinating on both sides of the argument so let's not bicker.

(05-02-2017, 05:23 PM)nickpelling Wrote: You are not allowed to view links. Register or Login to view.

(05-02-2017, 03:08 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.I pointed out the problem of word structure to to him in 2014 after his first paper. Over two years later he still hasn't adequately responded to it, saying that the writer was free to write as he pleased within his 'concept of language'. I don't think he has an answer because it would spoil his theory.

As I recall, you recently mentioned cheol vs chole vs echol, which reminded me of the eol permutation thing I'd noticed a while back but had forgotten. Voynichese has plenty of similar strongly asymmetrical permutation sets, but I'm pretty sure that eol was one of the most extreme.

It's also interesting that we see all four gallows inserted in the middle of ch, but never anything else in the middle of ch. That's a rule all of its own as well, etc etc.

Indeed, it is easy to point out that so many of the possible permutations that a writer could make either don't exist or occur so few times that they might be errors.

Let's take the word [chedy], which occurs about 501 times (I'm using voynichese.com for the numbers) and should be the centre of a big network of permutations. Let us agree that it has four characters [ch], [e], [d], and [y], which means there are five possible places to insert a new character. Let us further agree that the following 22 characters are part of the Voynich script: [o, y, a, e, ch, sh, k, t, f, p, ckh, cth, cfh, cph, d, s, l, r, m, n, i, q].

Here are the stats for attempting to insert each of these characters into the word [chedy] in the five different positions:

[o]: [ochedy] 8; [choedy] 0; [cheody] 89; [chedoy] 0; [chedyo] 0
[y]: [ychedy] 13; [chyedy] 0; [cheydy] 0; [chedyy] 0; [chedyy] 0
[a]: [achedy] 0; [chaedy] 0; [cheady] 1; [cheday] 0; [chedya] 0
[e]: [echedy] 1; [cheedy] 59; [cheedy] 59; [chedey] 1; [chedye] 0
[ch]: [chchedy] 0; [chchedy] 0; [chechdy] 0; [chedchy] 1; [chedych] 0
[sh]: [shchedy] 0; [chshedy] 0; [cheshdy] 0; [chedshy] 0; [chedysh] 0
[k]: [kchedy] 22; [chkedy] 5; [chekdy] 0; [chedky] 1; [chedyk] 0
[t]: [tchedy] 33; [chtedy] 2; [chetdy] 1; [chedty] 0; [chedyt] 0
[f]: [fchedy] 11; [chfedy] 0; [chefdy] 0; [chedfy] 0; [chedyf] 0
[p]: [pchedy] 34; [chpedy] 0; [chepdy] 0; [chedpy] 0; [chedyp] 0
[ckh]: [ckhchedy] 0; [chckhedy] 11; [checkhdy] 2; [chedckhy] 0; [chedyckh] 0
[cth]: [cthchedy] 0; [chcthedy] 7; [checthdy] 1; [chedcthy] 0; [chedycth] 0
[cfh]: [cfhchedy] 0; [chcfhedy] 0; [checfhdy] 0; [chedcfhy] 0; [chedycfh] 0
[cph]: [cfhchedy] 0; [chcphedy] 3; [checphdy] 0; [chedcphy] 0; [chedycph] 0
[d]: [dchedy] 27; [chdedy] 2; [cheddy] 0; [cheddy] 0; [chedyd] 0
[s]: [schedy] 7: [chsedy] 0; [chesdy] 0; [chedsy] 0; [chedys] 0
[l]: [lchedy] 119; [chledy] 0; [cheldy] 0; [chedly] 0; [chedyl] 1
[r]: [rchedy] 11; [chredy] 0; [cherdy] 0; [chedry] 0 [chedyr] 1
[m]: [mchedy] 0; [chmedy] 0; [chemdy] 0; [chedmy] 0; [chedym] 0
[n]: [nchedy] 0; [chnedy] 0; [chendy] 0; [chedny] 0; [chedyn] 0
[i]: [ichedy] 0; [chiedy] 0; [cheidy] 0; [chediy] 0; [chedyi] 0
[q]: [qchedy] 1; [chqedy] 0; [cheqdy] 0; [chedqy] 0; [chedyq] 0

Now, Torsten only considers words with at least 4 occurrences to be sec[font=Verdana]urely valid ("For words occurring le[font=Verdana]ss than four times transcription errors become important." Timm, 2016, p2), so we will use this to sort each permutation into one of three classes: does not occur, does not validly occur, and validly occurs.

[/font][/font]Here are the stats: of the 110 possible permutations for inserting a character into [chedy] only 16 (14.5%) occur validly, a further 14 (12.7%) occur invalidly, and 80 (72.7%) do not occur at all. Nine of the characters, [a, ch, sh, cfh, cph, m, n, i, q], cannot be inserted anywhere in the word validly. Also, no character can be inserted validly into the slots after [d] or [y] of [chedy].

Remember, [chedy] is the third most common word in the text, so most words will show even fewer permutations!

Almost every theory I have ever comes across—linguistic, cryptological, and even Rugg—account for the very obvious word structure. Torsten's theory is the only one which handwaves away some of the best evidence we have.

(05-02-2017, 05:35 PM)nickpelling Wrote: You are not allowed to view links. Register or Login to view.Torsten: so what you are saying is that even though Voynichese is autocopied from itself, the "itself" from which it is autocopied (i.e. that seeds the copy) contains near-universal adjacency rules which are then preserved in the autocopying?

And that these near-universal adjacency rules that are in the 'seeding' part of the text (and then preserved in the autocopied part of the text) may or may not be language but you're not interested?

No, adjacency rules are not preserved from the seeding part they are a result of the autocopying process. If a word is copied and some glyphs are replaced with similar ones the order of the glyphs stays unchanged. If you copy for instance 'chol' and replace 'ch' with 'Sh' and 'ol' with 'ar' you get 'Shar'. On first view this word looks different but the adjancency rules are still in place. If then 'Shar' is used as source for generating a new word 'chol' is still a possible outcome. See for instance 'Sheol chol Shar chol' in line f106v.P.25. 

Even if you change something while copying a word some elements of the source word remain. Therefore it is no surprise that the autocopying process results in adjacency rules.

Recursion can result in very interesting effects:
[attachment=1124]

Quote:And that all the word variations arise not from a conscious attempt to produce randomness, but instead from the autocopyist assembling many small blocks that retain those near-universal adjacency rules?

There is no randomness. I wouldn't expect from someone before 1500 that he would be able to generate something random. In my opinion the VMS was written before 1500. This means that the scribe didn't know what randomness means. Moreover, as far as I know cognitive processes have virtually no way of generating truly random events. It would be like trying not to think of something. Therefore I didn’t find it surprising that something written in medieval times resembles more to human language then to a random text generated by a computer.

If you want to fill a page with text it is much easier to copy text then to generate new text. In this way the text in the VMS is in my eyes a medieval form of the 'Lorem ipsum' text [see You are not allowed to view links. Register or Login to view.].

Quote:So... why is it that eol is 25x more frequent than ole, again?

'ol' is more frequent as suffix then as prefix. The most frequent word with 'ol' after 'e' is 'cheol'. It is used 172 times. The most frequent word using 'e' after 'ol' is 'olkeedy'. It is used only 42 times. Since 'cheol' if more frequent then 'olkeedy' the autocopying process leads to more words related to 'cheol' then to 'olkeedy'.

(05-02-2017, 06:27 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.Indeed, it is easy to point out that so many of the possible permutations that a writer could make either don't exist or occur so few times that they might be errors.

During writing the VMS the scribe generated a network of 6837 similar word types out of 8026 word types. To argue that he has missed to add the words 6838 and 6839 which you would add to the network is not a valid argument in my eyes. Keep in mind, if you want to add something to the VMS it is necessary to remove something else. Otherwise it would be necessary to add some pages to the VMS.

It is not as easy as you think to call rare words an error. Most times a rare word is similar to other rare words. If you call one word within the network of 6837 similar word types an error you have to answer the following question: Why is only this word an error and not the other 6836 connected word types?

Quote:Let's take the word [chedy], which occurs about 501 times (I'm using voynichese.com for the numbers) and should be the centre of a big network of permutations. Let us agree that it has four characters [ch], [e], [d], and [y], which means there are five possible places to insert a new character. Let us further agree that the following 22 characters are part of the Voynich script: [o, y, a, e, ch, sh, k, t, f, p, ckh, cth, cfh, cph, d, s, l, r, m, n, i, q].

I disagree. First 'dy' is used as ligature. In fact the word 'dy' occurs 270 times in the VMS. Therefore 'chedy' has only three elements [ch],[e] and [dy]. 

Secondly, the only possible place to add a new character is at the beginning of a word. Within a word you can only duplicate 'e' and 'i' and you can add a formerly removed glyph.

Thirdly the glyphs are not independent from each other. For instance it is possible to replace similar glyphs [see You are not allowed to view links. Register or Login to view.]. In this way 'Sh' is only another way to write 'ch'. The same is true for the gallows glyphs. Moreover special rules to write certain glyphs at the end or beginning of a word exists. For instance in most cases at the end of a word 'a' is written as 'y'. 

What was the motivation to write glyphs differently at the end or at the beginning of a word? If your intention is to hide that only similar words are copied all the time, it is a mistake to repeat something. The easiest way to remove something repeated is to change it. One feature of the script used for the VMS is that in many cases one additional quill stroke is enough to change a glyph into another one. For instance, it would easily be possible to change 'ch' into 'Sh', 'k' into 't' or 'e' into 's' [You are not allowed to view links. Register or Login to view.].

Torsten,

This is a highly plausible argument for the text representing numbers to words.  Just think about all you have to do is change one or two glyphs in the suffix of a vord and then you have a vast array of new words to use for the vord.  Everyone here is in denial of your research well not everyone I like it, but Torsten the text is not meaningless.

oops if you found that prefixs change at greater occurrences then suffixs my argument still stands for the voynich.  But what is the weight of what we are seeing here maybe you could apply a gephi graph to prove the argument for prefixs out weighing suffixs in a huge %.

What do I have to do to wake up the Voynich Community to this fact well I guess you are.

As you can see most of the vords share the same prefix yet suffix's change little.

Why the heck can't anyone see this but me!!!!!!!!!!!!!!!!!!