The Voynich Ninja

My point is by looking at all the gallows characters and by aggregating we can say what the shape looks like. This is why I referred to image recognition. What the shape actually looks like is important for my research. There is natural variation in how a shape is drawn.

(16-05-2019, 09:54 PM)Mark Knowles Wrote: You are not allowed to view links. Register or Login to view.JKP: I am not focusing on how it looks in modern scripts. I am focusing on how it looks independent of any scripts.
...

WHAT????

That's impossible. You can't focus on how a shape looks independent of any script references. Comparison cannot occur in a void.

JKP: Of course it can. Do need a script to compare a shape with a circle?

---

If I see a drawing of what I think looks like a triangle on a piece of paper to answer the question as to whether I think it is intended to be a triangle does not require any reference to scripts.

Even if you call it a figure-8 instead of "8" (as in the number or letter), you are still using references.

If you want visual references instead of script references, call it a figure-8 so we know what you are talking about.

As it is, by using vague references, you keep asking us to read your mind. Sorry, but we can't do that.

I don't have a big problem with understanding this.


You can take images of a bunch of q's and a bunch of 4's and line them up in order from more q - like to more 4 - like.
These images could come from a variety of sources or even be mathematically drawn.

You could mark the point where q's end and 4's begin.
For any given source you could count how many are more q like and how many are more 4 like.
Mark seems to think most are more 4 - like in the VMS.

My guess would be that intermediate forms would dominate.
Also, that cutoff point will be arbitrary.

I imagine that theoretically a mathematical function could be designed to draw the shapes, by adjusting a coefficient of "pointy" and/or a coefficient of "round".

In that sense, the shapes would be independent of a manuscript.

Meaningful results though would only come from comparison of shapes from actual manuscripts to the mathematical ones.

(17-05-2019, 12:28 AM)DONJCH Wrote: You are not allowed to view links. Register or Login to view....

You can take images of a bunch of q's and a bunch of 4's and line them up in order from more q - like to more 4 - like.
These images could come from a variety of sources or even be mathematically drawn.

You could mark the point where q's end and 4's begin.
For any given source you could count how many are more q like and how many are more 4 like.
Mark seems to think most are more 4 - like in the VMS.

My guess would be that intermediate forms would dominate.
Also, that cutoff point will be arbitrary.

... [deleted for brevity]

Don, I did suggest very early in the thread something along those lines by saying, "By all means come up with an algorithm that classifies them. As long as it's the software doing the work, there's no reason not to, and such an algorithm could be generalized to other letters."

But that suggestion got no response.

So it's like... we take the number 4 but forget about its meaning and origin until it becomes a meaningless form and then see whether it looks like the VM shape?

(17-05-2019, 02:50 AM)Koen G Wrote: You are not allowed to view links. Register or Login to view.So it's like... we take the number 4 but forget about its meaning and origin until it becomes a meaningless form and then see whether it looks like the VM shape?

It's like, the ancient Greeks introduced the concept of the perfect circle. While no circles in the real world were perfect, the perfect circle was amenable to geometrical and mathematical manipulations and analysis. Results were meaningful and useful when transferred to the physical world.

It's also like, ANY curve can be described by a sum of exponentials and Rene has made good use of that in your support in the other thread. Sure, those exponentials may be "meaningless" but are still useful to plot a smooth curve reflecting the real data.

In the same way, create a (theoretically infinite) series of shapes between q and 4. Where on the series does the VM Eva-t shape lie? To calculate mean and sd I think you need the general formula describing the shapes.

Of course, I may be completely wrong, as I always say, I am a novice compared to some of you!

Hello Don,

what you write is not wrong. However, the thread title doesn't say:
"The top left hand (etc) is a triangle".

Furthermore, Mark says that the purpose of all this is to be able to find the origin of these shapes. At the same time, we are not allowed to look in documents contemporary to the Voynich MS to find examples and counter-examples.

So, just looking at the Voynich MS, in most t and p characters, the leftmost tip is mostly rounded.
In the q characters, it tends to be a bit more pointed, but is also often rounded.

In principle, I suppose one could measure and compute stats of the slopes of the oblique lines, but I think that should be left to someone who thinks that there is a useful outcome of such a non-trivial exercise.
(Implication: it won't be me).

In the opening post, Mark wrote:

"I am completely convinced that the top left hand loop of the gallows characters is a "4" rather than a "q", but I am intrigued as to how much disagreement there is in this."

Unfortunately, he didn't let us know whether he meant medieval 4, modern 4, or any old 4 (which could include quite a bit of variation).

Also, he didn't say "shape of 4" in the first post, he said, is a "4" rather than a "q", implying that he's talking about letters.


In his second post, he wrote:

"For me this is an important question as it has implications in terms of the script used, some scripts may be more likely to have the kind of angular "4" shape in their script and other might not."


Thus, implying that he was interested in scripts. However, later, he said he was not evaluating the shape in the context of scripts.


He also wrote:

"If one can transform the problem from the context of the Voynich manuscript to just images containing circles or triangles then one transform it into an image recognition problem."

This is familiar territory for me. I'm keen on computer graphics, and I'm keen on automating this kind of process so the computer does most of the work. This comment is what prompted me to mention that an algorithm could be used to evaluate the shape and be generalized to other shapes. My comment received no response.

However, to do image recognition you need references, so you can establish the extents for the amount of variation. Each shape will have its own profile because in script (which Mark says he doesn't want to use as a reference), some variations are meaningful and some are not, and the tolerances for each character and for each style of script will be different.

Whether it's worth the time depends on how many shapes will be evaluated and the end-goal. Personally I don't think it would substantially assist in the decryption of the VMS but that doesn't mean I would discourage someone else from doing it, as they may have some other goal in mind.