[Article] Decoding Anagrammed Texts Written in an Unknown Language and Script

[Koen G]

The only reason I see to invoke anagramming is to explain the rigid word patterns of Voynichese. But this cannot be the result of arbitrary scrambling. Additionally, it would still be a one way cipher, since all information on initial letter order is lost

[ReneZ]

I would argue that anagramming is only invoked because, without it, nothing seems to work, and anagramming introduces an enormous additional level of freedom.

[Koen G]

I certainly agree that's the actual reason 

[Anton]

I don't think that the authors use the word "arbitrary" in the sense of "random" or "one way". This is precluded by the word "scheme" earlier in the same quoted sentence.

By "arbitrary" they mean either that their method is invariant to the exact scheme of anagramming employed in the VMS (see the context in which the word "arbitrary" is used on p.75 above), or that the rule of anagramming may shift from vord to vord (i.e. be different for different vords). It's difficult to say what of the two things they actually mean, but for sure they are not arguing one-way cipher.

Some combination of abjad and anagramming is a tempting path, of course, because it potentially could explain both for low entropy and strange patterns of glyph positioning.

[ThomasCoon]

When I started researching the VMS 2 years ago, a programmer friend (who works for a very big company) told me that artificial intelligence could solve the VMS "in seconds" if we input the text into an AI system.

That being said, Kondrak hasn't cracked it and doesn't claim to (as JKP and Anton pointed out). He doesn't even claim to decode full sentences.

(pages 84-85, emphasis mine):

Nevertheless, it is interesting to take a closer look at specific examples of the system output. The first line of the VMS (VAS92 9FAE AR APAM ZOE ZOR9 QOR92 9 FOR ZOE89) is deciphered into Hebrew as ועשה לה הכה איש אליו לביחו ו עלי אנשיו .המצות  According to a native speaker of the language, this is not quite a coherent sentence. However, after making a couple of spelling corrections, Google Translate is able to convert it into passable English: “She made recommendations to the priest, man of the house and me and people.” Even though the input ciphertext is certainly too noisy to result in a fluent output, the system might still manage to correctly decrypt individual words in a longer passage. In order to limit the influence of context in the decipherment, we restrict the word language model to unigrams, and apply our system to the first 72 words (241 characters) from the “Herbal” section of the VMS, which contains drawings of plants. An inspection of the output reveals several words that would not be out of place in a medieval herbal, such as הצר ‘narrow’, איכר ‘farmer’, אור ‘light’, אויר ‘air’, אשׁ ‘fire’. The results presented in this section could be interpreted either as tantalizing clues for Hebrew as the source language of the VMS, or simply as artifacts of the combinatorial power of anagramming and language models. We note that the VMS decipherment claims in the past have typically been limited to short passages, without ever producing a full solution. In any case, the output of an algorithmic decipherment of a noisy input can only be a starting point for scholars that are well-versed in the given language and historical period.

[bi3mw]

It would be nice to know if the manipulation of the output was actually limited to just a few spelling corrections. After all, there is a total of 24 possible combinations for the list "priest - man (of the house) - me - people". In my opinion, the hierarchical enumeration is the most likely. So the question is whether the sentence structure was changed to fit this "criteria" or whether it is the unchanged output.

[MarcoP]

The paper is certainly interesting. I was not aware of Knight's results with the Copiale Cipher: that is a great confirmation of the usefulness of software tools, when used with competence, intelligence and hard "manual" work.

I am sorry I am so completely ignorant of Hebrew that I cannot properly evaluate the results of Hauer and Kondrak. My naive impression is that anagram+abjad would be mostly impossible to decipher, but I may very well be wrong. Also, it's a pity that they didn't comment on phenomena like EVA:p/f only occurring at the beginning of paragraphs (see Davidsch's You are not allowed to view links. Register or Login to view.), or the behaviour of the EVA:o- prefix after EVA:-in (as discussed by You are not allowed to view links. Register or Login to view.) or EVA:o- in labels vs plain-text.

A minor error in the paper is that they used the "Bibbia di Gerusalemme" as a source for ancient Italian, while this text was written in the XX Century.

[Koen G]

Nick discusses the article in his latest blog post: You are not allowed to view links. Register or Login to view.

[Koen G]

It's really all over the place again, just got sent the link to the Independent claiming the MS has been "decoded": You are not allowed to view links. Register or Login to view.

Or the VM as click bait...

[VViews]

Ha!
My favorite piece of "reporting" on this so far has to be the one from Fox News, which kicks off its article like this:

"Discovered in the 19th century, the Voynich manuscript uses “alien” characters that have long puzzled cryptographers and historians."

You are not allowed to view links. Register or Login to view.