The Voynich Ninja

Here's a recent paper, "On the Directionality of Text in the Voynich Manuscript: An Edge-Based Approach":

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

Quote:Abstract

In this paper, we examine some linguistic properties of the Voynich manuscript. Our approach focuses on the properties of 'word' edges. It is shown that there are clear differences of the symbol distribution between the left edges and the right edges. We argue that these differences indicate the use of capitalization. Since the capital letters are placed at the right-edge, we conclude that the directionality of the text is right-to-left.

Paper is attached.  What do you think?  Are their conclusions sound?

Before I read it, I'll make the comment that capitalization wasn't very important to medieval scribes (except the professional scribes that wrote bookhands). They used it, but often inconsistently. Capitalization was the job of rubricators (in their eyes).

On a related subject... punctuation wasn't very important either (just as spelling wasn't rigidly standardized). In fact, in some cultures punctuation wasn't even used in the 19th century. I helped someone with some genealogy research and numerous family letters, which were well-written (both script and language), had no punctuation whatsoever.

So... we have to keep in mind, while reading analyses of medieval documents, that the priorities of those who were literate (which most people weren't) were different.

I don't quite get how they jump from the fact that group A symbols are never left-edge to the hypothesis that they are capital symbols. In a left-to-right text, typically, a capital symbol would occur exactly on the left edge.

All in all, they miss the possibility that the properties they observe are due to vord construction. E.g, if a vord cannot start with glyph X, you will never see it in the left edge, to begin with. If vords rarely start with X, you will find X in the left edge rarely still.

There are numerous ways of mapping the text to the Latin alphabet and since the earlier transcripts have been mostly abandoned in favor of later ones, the authors should have included a brief transliteration chart rather than making hundreds or thousands of readers hunt up old transcripts in order to understand the paper.


The conclusions are based on many assumptions:

• The text is natural language.
  
• The text includes capital letters.
  
• The spaces are literal.
  

Based on the assumption that the text is natural language and includes capital letters, the author assigned some of the letters to uppercase so that "the number of the uppercase letters is closer to the half". This is problematic because it greatly reduces the number of glyphs available to describe an alphabet, and also affects entropy.

This is one of the conclusions:

"If our capitalization hypothesis is correct, it entails that the symbols in the Voynich manuscript is a
variation of European alphabets. The use of capitalization is limited to European alphabets, such as Latin
script, Greek script, and Cyrillic script. Therefore, the Voynich script is written in either one of these systems
or some variation related to them."

The author does acknowledge that acceptance of numerous glyphs as capitals "significantly" reduces the phonetic values, but gives no explanations as to why readers should accept a conclusion that includes too few values to define an alphabet in any of the languages mentioned in the previous conclusion.


It doesn't look to me like it's well argued or completely thought out.


Personally, I don't care if Voynichese turns out to be right-to-left or left-to-right because I always look at it both ways, but a paper for specific directionality should be argued better than this (it should explain why someone would get rid of a big chunk of the alphabet in order to include capitals).

The author has a 50-50 chance of being right, it's probably one direction or the other (probably not vertical), which means that even a coin-toss (without any research), has a pretty good chance of being right.

Thanks for sharing. 
I must say I don't agree at all with the paper. First of all, I believe there are plenty of indications in the layout that we are dealing with a left-to-right script.

For a You are not allowed to view links. Register or Login to view. I wrote about layout, I made the following image:



Note: 
- The large P-shapes at the beginning of paragraphs, which appear much more like capitals than anything at the end does.
- The ends of paragraphs are at the bottom right (marked in purple). The open space would be on the bottom left in a right-to-left script.
- Left margin often aligned better than right.


Additionally, it's a weird conclusion that capitalization must point towards Europe, while at the same time claiming the script is right-to-left. I'd rather say the use of larger initials on the left and layout suggestive of left-to-right is what points towards European custom.

Koen, the point of the paper is that the scribe intentionally masked a right-to-left script under the left-to-right appearance. In this perspective, one should not judge by the appearance, but rather by the text properties.

(09-04-2019, 02:18 PM)Anton Wrote: You are not allowed to view links. Register or Login to view.Koen, the point of the paper is that the scribe intentionally masked a right-to-left script under the left-to-right appearance. In this perspective, one should not judge by the appearance, but rather by the text properties.

Right - I found this unlikely given the extent to which the layout suggests left-to-right. 
But if I must accept deceitful layout as an option I must follow JKP in that it's more of a 50-50.

I'm not sure the properties are well argued either.

It's true that punctuation is usually at the ends of words (and sometimes the ends of lines), but there might not be any punctuation.

It's true that capitals are usually at the beginnings of words, but first you have to be sure you are assessing the beginnings of words. If the spaces are not literal, and if text wraps from one line to the next without punctuation (which happens often in medieval text), then the statistics are not actually based on words, they are based on VMS tokens, which might be something quite different from words.

It's weird that they single out symbol "Y" (cFh) in group B. "W" (cPh) and "X" (cKh) could be in this group as well, as they don't appear on the right edge in their (very obsolete) data either.