The Voynich Ninja - Hypervector Analysis of the Voynich Manuscript

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13

(11-08-2020, 08:11 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.
...
The perceived problem with Eva-s vs. Eva-sh is not a real problem, especially when one looks at trigrams.
It is exactly the same as in English, where the s in 'she' is a different s than the one in 'son'.
However, English also has words like 'mishap', where the s is like in 'son'. This problem does not happen in Voynichese.

The s in "she" is pronounced differently from the "s" in "son", and there are numerous examples like this in English because it is a Germanic language with a huge infusion of Romance vocabulary (French). Around here, it's often called a "mongrel" language. There are numerous pronunciations for combinations like "ough" (tough, trough, though, through, and plough).

But how is that exactly the same as English? The shapes are the same in English and the sounds are different. In the VMS the shapes are different (quite a bit different) and whether it is pronounceable is not yet demonstrated.

But assuming Voynichese can be pronounced (which I doubt, at least in an untransformed form), the VMS text is not clustering near English. There are some languages where pronunciation is more closely tied to actual spelling. EVA-s and EVA-S might represent very different things. I don't know whether Caucasian spellings diverge significantly from pronunciation but whether it matters might depend on the language group.

But more importantly, there's no evidence that EVA-S and EVA-s are related in sound or function. EVA-s is an individual character. EVA-S is a combination of EVA-c and the curved-macron-shape that is written above it, which is somewhat presumptuous since that macron-shape is sometimes attached to EVA-h rather than EVA-c. One might be "a", the other might be "cer" (this is just an example). EVA-s and EVA-S do not behave the same in the VMS.

Here are some examples of EVA-S (the left portion of the shape plus the "cap", the character that resembles a curved macron) and EVA-h, the right portion of the VMS glyph (EVA-Sh):

[attachment=4659]

[Image: Examples-of-EVASh.png]

Here are some examples of EVA-s (lowercase). It frequently stands alone:

[Image: SampleEVAs.png]

I don't see anything in these pics that indicates similarity between EVA-S and EVA-s, and it may be problematic for certain forms of computational attacks that the left portion of EVA-Sh (in the first pic) is combined with the macron-like shape above it.

The fact that the macron-like-shape sometimes touches the shape below it might be relevant and might not. In languages that use Latin letters/abbreviations, it does not matter whether it touches or not. Here are some examples:

[Image: ExamplesOfCurvedMacron.png]

In Voynichese, I don't know whether the shape or touching matters, but sometimes it touches the left top, sometimes the left bottom, sometimes the right top (very distinctly so) and sometimes it doesn't touch at all. The "messy" scribe (the one Lisa Fagin Davis has classified as Scribe 2) sometimes connects the macron-like-shape to the bottom left side of the first EVA-Sh curve:

[attachment=4661]

I don't think EVA-S is directly related to EVA-s and since it is frequent, it should probably be given its own slot if text is collapsed down to lowercase.

The s in the first half of Eva-sh is not related to the stand-alone Eva-s.

This is just the same as that the 't' in English 'the' is not related to the 't' in 'tea'.

It is just a writing convention. Icelandic is 'cleaner' in that respect.

English 'sh' and German 'sch' represent the same thing. This cannot be properly interpreted by the hypervector algorithm, but that is part of the uncertainty in the result that one has to live with.

In Voynichese we just do not know at all. Every experiment and every statistic is based on a list of assumptions, and we don't know how good they are.
The best one can do is to try different alphabets and see how much that affects the result.

I did this for entropy calculations, but there we have the advantage that entropy is insensitive to single-character substitutions. Then, the results were not too far apart.

(11-08-2020, 10:37 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.
The s in the first half of Eva-sh is not related to the stand-alone Eva-s.

...

There is no EVA-sh if you are talking about the character that looks like c^c. This is incorrect: sh. (sh)

According to the basic EVA Character chart, it is EVA-Sh. This is correct: Sh. (Sh) and it is, of course, not related to stand-alone EVA-s as I emphasized in my You are not allowed to view links. Register or Login to view..

Are you saying the Takahashi transcript has two versions of c^c (Sh)?

Takahashi perferred to use the upper case notation, but the lower case version was created when merging the interlinear file.

If one uses Sh, one gets the correct representation in the Eva Hand 1 font, because this font does not have the logic that would be needed to interpret sh correctly.

(11-08-2020, 11:03 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.
Takahashi perferred to use the upper case notation, but the lower case version was created when merging the interlinear file.

If one uses Sh, one gets the correct representation in the Eva Hand 1 font, because this font does not have the logic that would be needed to interpret sh correctly.

Are you saying that in the interlinear file there is no distinction between EVA-s and EVA-S? I'm confused. Are you talking about Darrin's file or Takahashi's file or something else? Which interlinear file?

I'm sorry, JKP, but you seem to be worried about a complete non-issue, while in fact there are many real issues in all transliterations.

One can always convert back and forth between sh and Sh using a script (and ivtt does it too), without loss of information.

Real issues are:
- none of the other alphabets can represent ligatures
- we have no reliable way of judging and recording 'word' spaces
- there seem to be several types of d (and other letters) but these are not distinguished by most alphabets

Renez Wrote:[font=Tahoma, Verdana, Arial, sans-serif]I'm sorry, JKP, but you seem to be worried about a complete non-issue, while in fact there are many real issues in all transliterations.[/font]

[font=Tahoma, Verdana, Arial, sans-serif]One can always convert back and forth between sh and Sh using a script (and ivtt does it too), without loss of information.[/font]

I know you can convert back and forth. I'm a computer programmer. I convert all the time. I created my VMS concordance with conversions built into the software app. via a suite of scripts that I coded.

Collapsing EVA-S S (the first part of Sh) down to EVA-s s (which is a separate character that is not part of Sh) is not a non-issue for people doing computational attacks. They are different characters that are attached to different other glyphs in the VMS and which behave in different ways. It has nothing to do with variations in pronunciation.

Darrin was using an already-collapsed transcript so he may not even have known that EVA-Sh and EVA-sh are not the same thing (and that EVA-sh essentially doesn't exist in the VMS). He may wish to do other attacks after working through this one, so the sooner he has a good transcript that makes distinctions between benched gallows and unbenched gallows and between EVA-s and EVA-S, the better. How he does it is entirely up to him, but I'm sure that GIGO is not what he wants.

Eva should not be used for statistical analyses, but interestingly, in this case, since Darren is looking at tri-grams, this is only a relatively minor problem.

I already recommended several posts ago to use the Cuva alphabet (and provided the sed script to do the conversion).

(11-08-2020, 11:26 AM)-JKP- Wrote: You are not allowed to view links. Register or Login to view.Darrin was using an already-collapsed transcript so he may not even have known that EVA-Sh and EVA-sh are not the same thing (and that EVA-sh essentially doesn't exist in the VMS).

I did not at the time, LoL.

Will make the substitutions JKP recommended on EVA and run that as one test. Will also run it against Cuva with ReneZ sed script as another. The results should be interesting.

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13