(12-08-2025, 04:57 AM)anyasophira Wrote: You are not allowed to view links. Register or Login to view.I love this clever way of reverse engineering . It’s nice to show a cypher can look a voynichese and still have meaning behind it regardless if this is the real mechanism. I do have a couple questions and I did listen to your video. I read your paper and I read this thread and I’m really hoping I didn’t miss the answer somewhere. This is pretty hard stuff to follow and I thought I checked before I’m asking . If these have already been Addressed , just let me know and I will go answer my own questions by rereading ? .
Clusters and repetition: are you saying that this is a feature of an encrypted plaintext letter/bigram just so happen to repeat in the plaintext word and therefore looks like words and suffixes being repeated , when it’s just a few repeat plain letters? Or is it because the voynichese encryption “chunks”are similar making all plain text words become repetitive?
Thank you! Before I provide an answer, I want to stress that the procedure of the Naibbe cipher—most notably the use of playing cards to enforce an approximate 5:2:2:2:1:1 set of ratios on a letter-by-letter basis—is meant to reliably achieve the
average statistical behavior across
all of Voynich B. It averages across several subsections of Voynich B that are meaningfully different from each other, such as the "biological" section versus the "stars" section. Locally, a specific tract of text could theoretically veer from the global average. Within a Naibbe ciphertext, there might be a portion of a page where there's more unigrams than bigrams, or a specific tract of text in which the secondary table Beta1 is used more often than Alpha, for example.
I also want to emphasize that the ratios might themselves represent the
average encipherment choices/preferences of the Voynich B scribes (mostly Scribes 2 and 3 in Lisa Fagin Davis's analysis), perhaps looking up choices from tables or something like that without a formal, Naibbe-like card-drawing mechanism. For example, if multiple tables were in a small pamphlet, a scribe might happen to use the table on the front cover more often than a table on an internal page, without any process that formally and systematically skewed the probabilities of certain tables appearing on a letter-by-letter basis.
With that preamble out of the way: Within the Naibbe cipher, clusters of same-prefix tokens are reliably generated if one of the Naibbe cipher's 6 tables happens to be used frequently when encrypting a series of successive or close plaintext unigrams. That's because within a given table, most of the unigram word types start the same way. In principle, bigram word types could also contribute to clustering, in the sense that the prefix
qok can appear in bigram word types, where it takes on a distinct meaning than it does in the specific word types reserved for unigrams.
Word repetition occurs most easily within the Naibbe cipher when there is a string of identical plaintext n-grams all in a row, and they are all encrypted the same way. Again, using the Naibbe cipher's default card-drawing procedure, this is probabilistically disfavored, but it's more likely to occur when it's a string of successive, identical unigrams. But the Naibbe cipher's general framework can accommodate these sorts of repetitive strings appearing through non-random scribal preference (which could happen as a time-saving shortcut; having hand-written text using this cipher, I definitely see the appeal of shortcuts). For example, within the Naibbe cipher, the additive version of the Roman numeral 4, IIII, could be written as
qokedy qokedy qokedy qokedy. Similarly, the plaintext word "benevolentia" could be respaced as "b|
en|
ev|o|l|
en|t|ia," and during encryption, the same bigram prefix could be recycled across all three "e_" bigrams (underlined).
(12-08-2025, 04:57 AM)anyasophira Wrote: You are not allowed to view links. Register or Login to view.Does your system show that the same vords, glyphs and even bigrams like to cluster around line breaks? And places where the line ends?
Does your system start having certain bigrams trending more left or right (Like Patrick Feaster has demonstrated)
At present, the Naibbe cipher lacks mechanisms that reliably bias word/glyph/bigram placements within a line, page, or paragraph—other than the plaintext itself exhibiting those kinds of biases at the unigram and bigram level. This is a known limitation of the Naibbe cipher; see Section 4 in the paper.
One of the things I want to explore is the extent to which the structure of the plaintext can create these biases within Naibbe ciphertext. For example, if the Naibbe cipher were used to encrypt a poem such as Dante's Divina Commedia, the poem's line-by-line structure would have rhyming, repeated phrases, etc. that would theoretically impose greater line-by-line positional biases in the frequencies of plaintext unigrams and bigrams relative to prose such as Pliny's Natural History. Is that sufficient to explain the full extent of the VMS's "line as a functional unit" properties? Maybe, maybe not. But maybe it becomes much easier to achieve "line as a functional unit" properties within a Naibbe-like ciphertext if the plaintext is a poem or poem-like in its structure.
(12-08-2025, 04:57 AM)anyasophira Wrote: You are not allowed to view links. Register or Login to view.Your tables- now
I know you did not focus on actual translation but rather showing something that mimics parts of the Voynich but you do shows charts of plain text letter ratios that match the voynich Glyphs so I am asking how much of your tables are just placeholders for any Latin letter in order to model this approach and how much of this is your actual best guess of what the suffixes, prefixes and Unigrams could actually be in plaintext Latin or Italian. For example did you somewhat choose to map Eva M to a Latin letters that are not as common or is it literally all arbitrary? Or does your cypher fix that? Like it doesn’t matter that certain glyphs and bigrams outnumber others?
Correct: I tried to design the cipher in such a way that it would reliably replicate the Voynichese glyph and glyph pair frequencies. But in this work, I didn't focus on an actual decryption/translation attempt because such an attempt would assume, among other things, that: (a) the Naibbe cipher or something very close to it is the literal VMS cipher; (b) that I have parsed Voynichese tokens correctly or mostly correctly into the affixes that correspond with plaintext letters; © that I have identified the correct plaintext language or close relatives of the plaintext language; and (d) that the tables I have constructed are at least partially correct in their letter assignments. An audacious set of assumptions!
I spent a lot of time manually fitting the tables' unigram word types to plaintext letters by performing Naibbe encryptions of Latin and Italian with placeholder codes (e.g., "unigram_alpha_a") and then comparing these placeholder words' proportional frequencies within the generated ciphertexts to the observed proportional frequencies of words within Voynich B. I did something similar with the prefixes and suffixes, but by comparison, I spent very little time optimizing them. That's because there's so many more of them, and the words they generate can be much rarer and are therefore much more subject to random noise / hypothetical scribal preference. Within this framework, optimizing the prefixes and suffixes essentially means assigning plaintext letters to them such that mock decryptions of Voynich B yield plaintext n-gram frequencies that optimally match those of a given plaintext language—aka a decryption attempt.
(12-08-2025, 04:57 AM)anyasophira Wrote: You are not allowed to view links. Register or Login to view.Anyways incredible work. Really interested how this continues as you fine tune it. I wish I head for math at times like this.
Thank you so much!
