The Voynich Ninja

Moved comment to it own thread.

Barbrey, that's a neat suggestion. Effectively some glyphs would have multiple "registers" and the following glyph "selects" which register is to be read. So rather than 14 glyphs having 14 values, they could have 10 * 4 = 40 values. Some ancient writing systems did have polysemic glyphs, with signs being able to be read as whole words, sounds, or alternative words in different languages. I'm not sure the exact system you suggest is known, but is something which could have been devised at the time.

(02-08-2021, 12:37 AM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.Barbrey, that's a neat suggestion. Effectively some glyphs would have multiple "registers" and the following glyph "selects" which register is to be read. So rather than 14 glyphs having 14 values, they could have 10 * 4 = 40 values. Some ancient writing systems did have polysemic glyphs, with signs being able to be read as whole words, sounds, or alternative words in different languages. I'm not sure the exact system you suggest is known, but is something which could have been devised at the time.

Thanks Emma May, I've just moved it to its own thread.  I think it's an adaptation of Llull, actually.  That's where i got the idea!

Hi Julian

We know that EVA likely inflates glyph counts to some extent. This is almost certain for benches, and arguments can be made for other frequent n-grams in EVA, like [qo], [iin] or even [aiin]. I think this might be an issue because you compare various context: after K, after P, after benched gallows... 

Here is an exaggerated scenario:
- Let's imagine that [iin] is actually one glyph. This adds two characters to the count whenever it occurs.
- Imagine that [iin] is more common following unbenched gallows, with on average one extra [iin] per gallow.

In this case, the counts for unbenched gallows are on average two higher because of EVA, not because of Voynichese.

You write in your conclusion: "The number of glyphs that follow a benched Gallows is typically 3, before the next Gallows is encountered. For un-benched Gallows, this number is typically 5 to 7." If the glyphs following benched gallows tend to be different glyphs than those following unbenched gallows, then it is possible that the different numbers are caused in part by EVA-inflation. It is also possible though that you will get the same proportions with different parsing, but I cannot rule out EVA-inflation just by looking at your data.

(02-08-2021, 02:34 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.Hi Julian

We know that EVA likely inflates glyph counts to some extent. This is almost certain for benches, and arguments can be made for other frequent n-grams in EVA, like [qo], [iin] or even [aiin]. I think this might be an issue because you compare various context: after K, after P, after benched gallows... 

Here is an exaggerated scenario:
- Let's imagine that [iin] is actually one glyph. This adds two characters to the count whenever it occurs.
- Imagine that [iin] is more common following unbenched gallows, with on average one extra [iin] per gallow.

In this case, the counts for unbenched gallows are on average two higher because of EVA, not because of Voynichese.

You write in your conclusion: "The number of glyphs that follow a benched Gallows is typically 3, before the next Gallows is encountered. For un-benched Gallows, this number is typically 5 to 7." If the glyphs following benched gallows tend to be different glyphs than those following unbenched gallows, then it is possible that the different numbers are caused in part by EVA-inflation. It is also possible though that you will get the same proportions with different parsing, but I cannot rule out EVA-inflation just by looking at your data.

Hi Koen,

OK, that's what I suspected you meant :-) I don't think the absolute values of the counts matter too much, only that there is a preferred value, and the distribution shape of the values. For EVA, the count of glyphs following a benched gallows are usually 3 and for un-benched between 5 and 7. For Voyn_101 the number will be N for benched and M for un-benched. Will M-N be different or the same for all transcriptions, and will the shapes of the distributions look the same as for EVA?

EVA example benched gallows pair f1r, line 9:


Same pair Voyn_101:

Here is a set of three cipher wheels inspired by Rene's recent arXiv paper, and based on Stolfi's core-crust-mantle work.



These wheels can account for ~93% of all VMS words if you follow these rules:

• Select none or more glyphs from the outer ring
  
• Append none or one glyph from the middle Gallows ring
  
• Append none or more glyphs from the inner ring
  

How does this relate to the counts between Gallows? Good question!

(The outer ring is Stolfi Crust + Mantle + EVA aoye (17 glyphs), the middle ring is Stolfi Core (8 glyphs), the inner ring is Stolfi Crust + Mantle + EVA aoye (17 glyphs).) 

I do like the number 17 as it jives with f57v.

[attachment=5718]
Basically, I have to determine what should be considered a gallows sign in the first place.
Many people assume 4 characters. But I see 6 characters, without counting the special characters.

You write under Code: (cphesaiin.ol.s.cphey)
Now we have 2x a "P". At first glance they look the same.
At second glance, one "P" has a sharp corner, while the other has a round corner.

One could say that this is a writing difference.

However, if you follow the latest work by LisaFaginDavis, which says that there are 5 different writers. So this has to be taken into account.
If you now look at the sample pages and compare the texts, you will find the same difference.
One has to ask oneself the question, "can 5 people have the same difference in writing," or have I run into a deception?

It's not the first time something like this has happened. Even the Templars broke up the alphabet by using a dot. The same and yet not the same.

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

(02-08-2021, 06:11 PM)julian Wrote: You are not allowed to view links. Register or Login to view.For Voyn_101 the number will be N for benched and M for un-benched. Will M-N be different or the same for all transcriptions, and will the shapes of the distributions look the same as for EVA?

Yeah, that's basically what I mean. I don't know how easy or time consuming it is for you to repeat the experiment with a different text file? If it's not too much trouble, there is something you could do to remove my worries   And I think the result may be interesting either way.

Just make a copy of the EVA file you used and turn a number of n-grams into single characters with find and replace. Since we are only looking at counts, this may even be the same character, like "b".

[ch] -> find and replace with b
[sh] -> b
[ain] -> b
[aiin] -> b
[qo] -> b

Like this, you will have replaced the most annoying n-grams with a single character. If this only changes the absolute values but nothing else, parsing is likely not an issue.

(02-08-2021, 10:37 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.

(02-08-2021, 06:11 PM)julian Wrote: You are not allowed to view links. Register or Login to view.For Voyn_101 the number will be N for benched and M for un-benched. Will M-N be different or the same for all transcriptions, and will the shapes of the distributions look the same as for EVA?

Yeah, that's basically what I mean. I don't know how easy or time consuming it is for you to repeat the experiment with a different text file? If it's not too much trouble, there is something you could do to remove my worries   And I think the result may be interesting either way.

Just make a copy of the EVA file you used and turn a number of n-grams into single characters with find and replace. Since we are only looking at counts, this may even be the same character, like "b".

[ch] -> find and replace with b
[sh] -> b
[ain] -> b
[aiin] -> b
[qo] -> b

Like this, you will have replaced the most annoying n-grams with a single character. If this only changes the absolute values but nothing else, parsing is likely not an issue.

Hi Koen, here are the results without/with your substitutions:

Un-benched pure EVA:

Un-benched modified EVA:


Benched, pure EVA:

Benched, mod EVA:



In summary, the modes change as follows:

p goes from 7 -> 5
f goes from 6 -> 4
k goes from 5 -> 4
t goes from 5 -> 4

cph 3 -> 2
cfh 3 -> 3
ckh 3 -> 2
cth 3 -> 2

Thanks Julian! The difference between benched and unbenched clearly remains.

Am I right to interpret that the stats for unbenched gallows become slightly more similar to each other than in standard Eva? Or is this simply a result of the lower numbers overall?