Independent Analysis Identifying Languages Beyond Currier A/ B, compared to René Zand

[npcompl33t]

Several AI red flags in this thread, but maybe be research is sound?

I definitely understand the risks of accepting AI work uncritically.

I've used AI with help assembling historical reference works, as well as executing some of the tests, but AI did not generate the methods or conclusions. I'm happy to discuss those in more detail, i know it can come across like a lot of technical jargon.

The TLDR is i wanted to use a different method than PCA over bigrams, for the reason i said in my post. 

JS divergence, in my optinion, is the appropriate tool for this type of analysis, regardless of whether or not AI is used, and i think using the transition matrix rather than bigrams captures most of the bigram information while introducing additional context. Hellinger distance is a useful crosscheck as well.


I was able to back out some descriptive bigram-like rules since i submitted this post originally, using the above classifications provide.

you can plug these into voynichese yourself and pretty easily see the impact. Some of them I haven't seen utilized as discriminatory features, because they aren't as useful in the classic A/B split.


In particular, using these splits, 'cho' is actually the most useful bigram for discriminating (yes i know currier identified chol and chor as discriminatory words, but i haven't seen 'cho' specifically discussed, and definitly not as a better discriminator than ED.  After CHO, it is ED, DY, EO, then AL.

ED , EO, and very minimal AO are used in RZ's splits, but not the others. 

Happy to answer any questions!