The Voynich Ninja

Pages: 1 2 3 4 5 6 7 8 9

Excellent. There are no surprises, because the distribution always remains intact.

>>Here's a display made that way of the distribution of [daiin] on Scribe 3 pages.
Strange image there. What is going on?

There is a possible problem that arises in a part (I described it somewhere on this forum..) where the pages are foldouts.
The horizontal lines are there twice as long, and it causes statistical issues.
I suggest you manually break those up into the average vord-length first.

(09-12-2021, 09:39 PM)Davidsch Wrote: You are not allowed to view links. Register or Login to view.There is a possible problem that arises in a part (I described it somewhere on this forum..) where the pages are foldouts.
The horizontal lines are there twice as long, and it causes statistical issues.
I suggest you manually break those up into the average vord-length first.

Data for unusually long lines should naturally sort to the "long" end of the line-length axis -- in this case, the top of the image (where you can see just a few faint squares in a sea of black). So I think this situation may already be accommodated, unless I'm misunderstanding your suggestion.

Thank you. It seems you read my suggestion correctly and have taken care of any problem there.

For the Daiin image: I do not understand why the red bar is there so in the middle and not more evenly distributed,
more something like the Ok-ot image. Could I ask you once more (sorry) to explain what we are looking at here in the daiin image, what explains the pixels filled and what if they are black?

Again sorry, if the question is stupid, I try to understand, because I think your extensive work is very interesting.

(14-12-2021, 01:20 AM)Davidsch Wrote: You are not allowed to view links. Register or Login to view.For the Daiin image: I do not understand why the red bar is there so in the middle and not more evenly distributed,
more something like the Ok-ot image. Could I ask you once more (sorry) to explain what we are looking at here in the daiin image, what explains the pixels filled and what if they are black?

The red lines represent overall distribution, as opposed to the distribution of whatever specific phenomenon we're trying to examine. In the [daiin] image, red shows the overall distribution of vords, so that the brightest bands represent the most common line lengths and the darkest bands represent the least common ones. Ten- and eleven-vord lines show up brightest. The blue and green channels then show the distribution of [daiin] specifically. The brightness scales are different -- all channels are maximized, which is to say that the highest value for each one, whatever it comes out to be, is scaled to the maximum pixel value of 255. But in general a redder tint should reveal an area where [daiin] is relatively less common, while a more blue-green tint should reveal an area where it's relatively more common. Black means [daiin] is uncommon, but vords in lines of that length are also uncommon in general. White means [daiin] is very common, but so are lines of that length in general.

With the [ok] and [ot] display, the red shows the overall distribution of glyphs by line length as *counted* in glyphs, and the spread of values is larger than the spread of values for line lengths counted in vords, so that values are a bit more evenly distributed. (My impression is that the bulk of blue-green values "sloshes" from right to left as the images alternate, with possibly some additional top-bottom "sloshing" at the far left. I *think* this provides decent evidence of a positional imbalance extending throughout lines of all lengths, but I hope someone will give me a wake-up call if I'm being a cheshire about this.)

(01-12-2021, 04:11 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.Another side of the subject is the possible impact of page layout. In particular, You are not allowed to view links. Register or Login to view. I noticed that image-breaks (frequent in Herbal pages) appear to have an effect on nearby words. The extra-long lines in the Pharma section might also be of interest.

That data you gathered about behavior around image breaks is really intriguing -- reminiscent of line-break behavior to a point, but also distinct from it.

I'm unsure how best to integrate breaks around images into the kind of rightward/downward display I've been experimenting with, but I did try one thing that was relatively easy to do: making plots of glyphs appearing immediately adjacent to intruding images, either to the left or right of them. In my other plots, I'd been treating intruding images as interchangeable with "ordinary" vord breaks.

Here's a plot of [<->o] in blue and [o<->] in green for all folios combined, where <-> represents an image intrusion.

[attachment=6098]

I'm not entirely sure whether there's enough data here for apparent patterns to mean anything. But it looks as though [<->o] appears later on average in lines than [o<->] does. If so, that would seem to suggest that patterning within lines transcends image intrusions to at least some extent, such that behavior around intrusions is partly constrained by where in a line the intrusions occur.

For whatever it's worth, here's a similarly structured plot of *all* image intrusions.

[attachment=6099]

Strangely enough, I remember that Marco P's above-linked discussion of glyph occurrences at image breaks was actually on my mind when I left Patrick Feaster the comment on his blog encouraging him to join Voynich.ninja, and I wondered what he'd have to say about it. So reading this thread feels a little uncanny. Ain't the internet grand?

Marco, if I recall correctly, the take away lesson from that study of yours is that vords to the immediate left of images tend to be typical line-final vords, and vords to the immediate right of images tend to be typical line-initial vords. Your study supports the idea that an image interruption is a functional line break.

Patrick, I seem to recall that your current understanding of labels in the VMs, is that of functional line units of a single vord. In other words, a label tends to begin like a typical line-initial vord, and end like a typical line-final vord.

I'm reminded of the metadata of telegrams and ham radio communications done in Morse Code, whereby the start and end of a string is indicated inherently in the structure of the communication itself — think of old movies where someone reads a telegram out loud, and pronounces the "stop" at the end of each line. Or protein synthesis, where specific triads of base pairs code for "start here" and "stop here". Each of the information strings in the VMs, whatever their meaning or intended use, seem almost as though they're designed to contain explicit indicators of where one ends and another begins.

Given Patrick's demonstration of downward effects at the paragraph and page level, and the old problems of Grove words and the affinity of [f] and [p] for the first lines of paragraphs, it wouldn't surprise me if Voynichese also contains apparent metadata for "start a new page", "same paragraph new line", "new line new paragraph", and other formatting and line placement instructions.

The question then becomes, what kind of information, recorded and shared and of value to people in the early 1400s, would have benefitted from having this kind of formatting and placement metadata attached to it? For example, might there have been any situations in the olden days whereby someone would orally dictate strings of information to a scribe whose writing surface he couldn't see, to be transported and read out loud elsewhere by a courier, to others taking dictation from the courier? In such a setup, formatting and placement metadata would help ensure that the final recipient received the strings of information in the same order, grouped and laid out in the same way, as the originator. Just a thought.

(14-12-2021, 09:44 PM)RenegadeHealer Wrote: You are not allowed to view links. Register or Login to view.Marco, if I recall correctly, the take away lesson from that study of yours is that vords to the immediate left of images tend to be typical line-final vords, and vords to the immediate right of images tend to be typical line-initial vords.

As I read it, Marco's analysis is more nuanced than that. Some of the patterns are like line-break patterns, while others differ from them. For an example of a difference, gallows are common line-initially but rare immediately after an image-intrusion. Same with [q]. And yet there are also enough similar patterns to suggest some kind of continuity between the two situations, as with line-initial and post-image-intrusion [y] and [s].

(14-12-2021, 09:44 PM)RenegadeHealer Wrote: You are not allowed to view links. Register or Login to view.Patrick, I seem to recall that your current understanding of labels in the VMs, is that of functional line units of a single vord. In other words, a label tends to begin like a typical line-initial vord, and end like a typical line-final vord.

Yes, that's where I am right now, at least with the most common "label" vords. Endings such as [ry] seem particularly telling. But I await counterarguments and am sure there must be some.

(14-12-2021, 09:44 PM)RenegadeHealer Wrote: You are not allowed to view links. Register or Login to view.Given Patrick's demonstration of downward effects at the paragraph and page level, and the old problems of Grove words and the affinity of [f] and [p] for the first lines of paragraphs, it wouldn't surprise me if Voynichese also contains apparent metadata for "start a new page", "same paragraph new line", "new line new paragraph", and other formatting and line placement instructions.

All the fascinating discussion in You are not allowed to view links. Register or Login to view. of "multipass" phenomena could perhaps point to one need for metadata. If meaningful text was in fact written in different orders (left-to-right, top-to-bottom in columns, etc.), it might have been necessary to be told the order to read it correctly without making a lot of false starts.

You mention downward effects at the page level -- I have no evidence myself as yet about those one way or the other, and the idea that they might exist is still only speculation as far as I know. The first lines of paragraphs do seem distinctive all around, though; it's not just gallows that behave oddly there.

(14-12-2021, 09:44 PM)RenegadeHealer Wrote: You are not allowed to view links. Register or Login to view.The question then becomes, what kind of information, recorded and shared and of value to people in the early 1400s, would have benefitted from having this kind of formatting and placement metadata attached to it?

Just about anything -- *if* the metadata were tied to the encoding method itself.

Scaling paragraphs by {rightwardness, downwardness} makes good sense in the abstract. But I would like to have a better sense for how it transforms texts of known structure. To that end, I am worrying the convenient King James Bible* (Kalevala can wait). One pattern is apparent even to a non-linguist: complementary distributions of objective vs subjective pronouns. For example, "thou" and "thee:"