The Voynich Ninja

(30-10-2016, 10:34 PM)Koen Gh. Wrote: You are not allowed to view links. Register or Login to view.I love this stuff, Thomas, especially the way you visualized it. 

This makes me wonder about two things.

1. The words that don't fit, what causes them to not fit? What is different about them? Can they be divided in groups?
   
2. What would it look like if something like this were made for a language like Spanish or Chinese or..?  How big is the difference with something we could observe in normal language?
   

Thanks Koen - to answer your questions:
1) Things that don't fit: mainly words that have two gallows, because the Verifier can't generate more than one. There are also some words with groups before the gallows, such as <chek> or <cheek> that I have trouble with...
2) I am sure that a similar grid would not work for German, Latin, or maybe any Indo-European language. Chinese would likely work, maybe Hebrew and Arabic?

(30-10-2016, 10:38 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.Can you explain why this tool would be useful? Or what fundamentals of word structure it reveals?

Good question. There are so many peculiar features of Voynichese word structure, and I was theorizing an underlying system that could explain them all. For example, my small table would account for:

1) Why many vords spelled with <t> can also appear with <k> - and what the overall relation between the gallows might be.
2) Why many vords spelled with 2 <e>'s can also appear with 1 or 0.
3) Why many vords spelled with <sh> can also appear with <ch>
4) Why many vords spelled with <yk> or <yt> can also appear with <ok> or <ot>
5) Why most words beginning with <q> can also be valid without it
6) Why many words ending in <daiin> can also end in <dy>
7) Why many words spelled with <ar> and <al> can also appear with <or> and <ol>

(30-10-2016, 10:53 PM)stellar Wrote: You are not allowed to view links. Register or Login to view.It does reveal a system of some sort.  Wow Mr. Coon if that many vords can be made from left to right maybe I can adjust the voynich numerology table to work but it would have to be nine columns.

Thanks stellar

(31-10-2016, 12:41 AM)-JKP- Wrote: You are not allowed to view links. Register or Login to view.Tom's observation in this post isn't the first time he was on the right track. I don't know if he realized it at the time, but a couple of months ago, maybe three months ago, he posted something that showed he was headed in this direction, had the right idea and I supported his comments then. I don't have time at the moment to hunt up the post but I'm sure some members will remember it (at least I hope they remember it because it was a substantive observation).

Thanks -JKP-, I appreciate the confirmation and kind words. I also believe that the text is "structured in a way that is different from natural language" - but I only have a small idea of where to go from here. I think the whole forum would be very eager to see your ideas, when you decide to post them  - if it's any encouragement, I am sure they would be fine without total perfection. But we absolutely respect your decision either way of course. As for diglyphs and monoglyphs, do you perhaps mean something to do with <y> and <d>?

It's similar to the regex created by Philip Neal:

You are not allowed to view links. Register or Login to view.

Quote:[… The transcription is a variant of EVA..]
The regular expression is something like
^[qd_][aoy_][lr_][ktpfKTPF_][CS_][eE_][d][ao_][lrmn_][y_]
(C =3D, ch S =3D, sh E =3D, ee KTPF =3D, the complex gallows)
correction: the last line, above, is how the passage appears online. I think it is meant to read(C =ch; S =sh; E =ee; KTPF = the complex gallows)
In other words, you chose any one character from each set in square brackets and rewrite the_ as zero, for instance qo_k_Ed__y -> qokeedy. The null character _ can occur anywhere: I would not interpret it as an unwritten vowel. I should admit a serious difficulty with the regex, namely that it generates most Voynichese words but also a great many words which do not occur, but I still think this is the basic pattern.

Stolfi developed his "core mantle crust" theory based on these same observations (You are not allowed to view links. Register or Login to view.) as did Gordon Rugg for his Cardian Grille theory.
So yes, it seems fairly obvious that vords have a very strong internal alignment, which IMHO points towards an artificial language. I've always had a sneaking suspicion that Don of T had the right idea but applied it in the wrong way... Of course, if it is an artificial language, it's probably lost as we'll never know what the nouns and verbs represent.

(30-10-2016, 10:22 PM)ThomasCoon Wrote: You are not allowed to view links. Register or Login to view.

Hi Thomas, what is the EVA equivalent of uppercase S in your table (columns 2 and 9)?

Hi Marco - that is an error (thank you for catching it). It should be lowercase s (s) alone. Column 2 should have <s> - <ch> - <Sh>, and column 9 should  have <d> - <s> <ch>. I was trying to generalize why there is often <saiin> and <daiin>, words ending <oldy> and <olsy>, etc. (But it's still not a perfect table)

Thanks. Although it would be quite easy to check this system, I honestly have to say I applaud to the effort, not to the system itself.

As discussed before, and by others, there is a rather simple Regex formula for the text, which looks quite similar. (as You are not allowed to view links. Register or Login to view.)

I did not test that formula either, yet, but I strongly suspect it does not hold for all the words. 
At the moment two other classifications methods that identify >98% of the text, have been described by me, but that is only 1/3 of the work.

I assume/guess  this ThomasCoon of Regex method will be satisfactory for something like 75% of the total text. But then what ? 
Then we need to find some sort of primary system, device or method which is used as a basis of that method.
I am interested in that primary system, and possibilities to create that.

For your method presented, are there connections, logical cohesion:

• between the letters itself in the system
  
• the position of the letters
  
• the distance of the letters in the Latin Alphabet (why is this often 4, for the first two letters?)
  

(31-10-2016, 08:47 AM)davidjackson Wrote: You are not allowed to view links. Register or Login to view.Stolfi developed his "core mantle crust" theory based on these same observations (You are not allowed to view links. Register or Login to view.) as did Gordon Rugg for his Cardian Grille theory.
So yes, it seems fairly obvious that vords have a very strong internal alignment, which IMHO points towards an artificial language. I've always had a sneaking suspicion that Don of T had the right idea but applied it in the wrong way... Of course, if it is an artificial language, it's probably lost as we'll never know what the nouns and verbs represent.

I'm not yet convinced of artificial language, but I definitely think the text is artificial. I haven't done any work on this theory, but I wonder if each vord / segment of a word (core-mantle-crust) is input for a corresponding chart, where the presence or absence of certain vord components (q, a gallows letter, aiin vs. am, etc.) are all "instructions" on how to produce to plaintext outputs. This is completely a non-tested idea though.

(31-10-2016, 02:07 PM)Davidsch Wrote: You are not allowed to view links. Register or Login to view.I assume/guess  this ThomasCoon of Regex method will be satisfactory for something like 75% of the total text. But then what ? 
Then we need to find some sort of primary system, device or method which is used as a basis of that method.
I am interested in that primary system, and possibilities to create that.

For your method presented, are there connections, logical cohesion:

• between the letters itself in the system
  
• the position of the letters
  
• the distance of the letters in the Latin Alphabet (why is this often 4, for the first two letters?)
  

Thanks Davidsch - you're right, the primary system is necessary. My system doesn't do that (and I don't claim it does), but I hope it might help to figure out the relation between the letters / components making up vords. I'll take a look at your classifications. You're also right that the Regex is very similar.

(31-10-2016, 02:01 PM)ThomasCoon Wrote: You are not allowed to view links. Register or Login to view.Hi Marco - that is an error (thank you for catching it). It should be lowercase s (s) alone. Column 2 should have <s> - <ch> - <Sh>, and column 9 should  have <d> - <s> <ch>. I was trying to generalize why there is often <saiin> and <daiin>, words ending <oldy> and <olsy>, etc. (But it's still not a perfect table)

Thank you for the clarification, Thomas!

As Sam pointed out, your table is equivalent to a regular expression. It is similar to Neal's but there are differences of course.

I tried translating it to a unix (grep) regular expression and it seems to match 57% of the unique words in Takeshi Takahashi's transcription.
Here's a random sample of 30 unmatched words with the command I used:

grep -Ev  '^(q|d)?(s|ch|sh)?(o|e|y|ol)?(k|t|p|f)?(ch|sh)?(e|ee|eee)?(o|a)?(r|l)?(d|s|ch)?(o|a)?(n|in|iin|m|r|l)?y?$' ../allwords_uniq | sort -R | head -30  | sort

aloiiin
ataiin
aty
chdchy
cheeg
cthoepain
dchodees
dcthy
esechor
kotchody
laiiin
lfchy
ltai
ochockhy
oechedy
olekeey
olsai
opaiiral
pcharalor
pokar
polkeedal
qepoepy
qoedeey
qoteode
qoykeey
rodeedy
sheoeky
shoikhy
socthey
todai

It would also be interesting to measure how many "non Voynichese" words are matched by a regex, but this is more difficult to do. For instance, the regex '[a-z]*' matches all alphabetic strings, 100% of Voynichese included, but it also matches any random alphabetic string.

Thanks Marco - well, there's definitely room for improvement then! .

I made some changes to the table, but it is still not totally perfect. I definitely don't think this is the full answer, but I've been looking for a way to generalize similar components of vords.



Just an idea - Back to the drawing board then