The Voynich Ninja

Have you thought about taking only the substantial values and then combine those with the weight of occurrences (=counts)

Then try to figure out how to put it in 1 overview or graph.  

Finally, as you know, i have a list of those letters based on word position and weight of occ.(=counts).
I do not see how to combine that as well, but if that would work, we would have a nice total view.

(22-10-2016, 08:48 AM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.If finally put together something that I hope might help in the analysis of the previously posted data. The attached image presents "radar charts" computed on the basis of the clustering input (the csv you can find in You are not allowed to view links. Register or Login to view.).

I have manually added in red the cluster boundaries produced by the two algorithms discussed You are not allowed to view links. Register or Login to view.. Brighter red lines correspond to the "macro cluster" the were identified combining the outputs of the two algorithms. Fainter red lines correspond to some of the dubious cases (eg EVA:k k could belong with F K P T or with EVA:t t).

[edit: I had to redraw the d chart using a different Y-axis scale... it looked weird because of some LibreOffice bug]

It would seem that a and y are most similar because they both appear often after d. I find that outcome acceptable as I consider them dependent on the following characters.

(23-10-2016, 04:02 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.It would seem that a and y are most similar because they both appear often after d. I find that outcome acceptable as I consider them dependent on the following characters.

Yes, that's basically the only thing they have in common, but it's a quite relevant feature.
a / a and y / y are the only two characters that often follow d / d, and they follow d with similar frequencies (da occurs in 30.2% if a-words, dy in 31,5% y-words).

Marco and Emma May, it's very interesting you've been talking about the functions of a and y. Earlier today (without seeing your posts), I was trying to compare where <a>-groups and <y> appear.

I've been intrigued by word endings like -dy and -daiin because many vords are valid with both endings (ex. choldy, choldaiin). But equally interesting are patterns in strings like this:



I don't know how to interpret it, but I've been wondering if there is a relation between three distinct groups of combinations:

Group 1: ch, Sh, d, s,
Group 2: or, ol, ar, al, aiin, ain, an, am
Group 3: Gallows k, t, p, f and <o+gallows> and <y+gallows>

Hi Thomas, do you mean that ch, sh, d, and s are interchangeable? ch and sh certainly seem to be but all four are not.

(25-10-2016, 06:05 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.Hi Thomas, do you mean that ch, sh, d, and s are interchangeable? ch and sh certainly seem to be but all four are not.

Hi Emma May: I definitely don't think they are completely interchangeable, but I wonder if there is some relation between them (perhaps in terms of a cipher rather than a linguistic system, but I've been open to both ideas).

I would say that d is more closely related to k and t than ch and sh. Which is what Marco's stats say too.

True, that's a good point....

Marco, is there a database of characters and frequencies I can use? I am thinking of developing an app that enables you to globally substitute letters for characters. I would tag each character with a "VOYNII" number (as in ASCII) convert the text into VOYNII and then start looking at transforms. It would also be easy to "class" the characters by assigning them to number clusters, or assigning VOYNII numbers to letter combinations.

(02-11-2016, 06:48 AM)normstone Wrote: You are not allowed to view links. Register or Login to view.Marco, is there a database of characters and frequencies I can use? I am thinking of developing an app that enables you to globally substitute letters for characters. I would tag each character with a "VOYNII" number (as in ASCII) convert the text into VOYNII and then start looking at transforms. It would also be easy to "class" the characters by assigning them to number clusters, or assigning VOYNII numbers to letter combinations.

Hi Normstone,
I don't know if official frequencies data-sets exist. 

I always compute the numbers from scratch using Takeshi Takahashi's transcription (and I did incur in errors, with this approach).

Maybe you can try Robin's Restful API?
You are not allowed to view links. Register or Login to view.