Today my paper Cracking the Voynich Cipher was released. 

A download can be found at sivbuggevatne.com

The paper proposes a full decoding of the Voynich manuscript´s cipher. The decoding is tested in the plant section of the manuscript. It reveals 104 plant names which are all found as one of the first words in each herbal text. The names found lead to plants with the same characteristics as the Voynich illustrations, and the names are closely related to vernacular plant names used in medieval England.

Best regards, 
Siv

Hi Siv, and welcome to the forum!

Please feel free to attach the paper to your post or to post a direct link to your paper.

As a side note, it's not a good idea to use your email address as your username in the forum, since it's visible to indefinite audience including spambots!

so i skimmed it.

whenever someone claims to have a full decoding, i just go to read it, and then am mostly always disappointed.
either, here are 10 words, or 1 paragraph, or even 2 pages, often not convincing, or here are 104 plant names--

those are not full decodings.
a full decoding would allow you to read the whole text.

your plant analysis might bear some fruits, and some of your other commentary is interesting,
but there is no evidence of full decoding in your text.

sorry to sound so negative though, i do congratulate you on your paper, finishing a paper is always a good achievement

(Here is a direct link to the pdf for those who will not find it otherwise: You are not allowed to view links. Register or Login to view. )

Hi Siv

I read your paper since my first impression was that your approach was thorough and intelligent. For instance, you address several oddities of Voynichese early on, instead of just ignoring them. This is a lot more than what we can say of similar attempts.

I especially like your courageous attempt at explaining the positional rigidity of Voynichese glyphs by assuming positional variation. So for example "e" in the plaintext becomes [a] in the middle of words, [y] at the beginning and end. 

I attempted something like this myself a few times, but always found myself running into the same problem, and I'm afraid this problem is impossible to overcome. The thing is, almost all Voynichese glyphs are severely constrained not only by their possible positions in the word, but also by the other glyphs they frequently combine with. This is simply not compatible with regular languages, even if we allow for positional variation. An additional problem is that positional variation will effectively reduce our phoneme inventory.

You solve this by allowing some VM glyphs to represent various sounds, which in itself is not a problem. Again, compared to previous attempts, I find your system reasonable. For example, EVA [p] corresponds to p and b (which only differ in voicedness). You also allow vowels to be dropped in some cases during enciphering. For example (this is just the first case I found while skimming), EVA [chdy] becomes "ldé" which, by assuming the "i" could have been dropped, becomes "lithe". Intuitively, I would say this is problematic: why drop the stressed vowel "i", which is part of the identity of the word, but write the unstressed "e"? 

Both the mechanisms described above (one Voynichese glyph can be various plaintext glyphs and some vowels can be dropped) lead to a cipher where information is lost. This is not a problem in itself, as long as the meaning can still be retrieved. Basically, the steps you take after applying your transliteration system are "interpretative" steps, where you work towards something that makes sense. This is something we see in almost all VM decryption attempts, but yours is one of the most sensible I have seen so far, it is clear that a lot of thought and effort went into this. 

What we often see in Voynichese solutions like yours which propse a lossy cipher is that they work towards a meaning that is supported by the illustrations. But huge sections of the VM text are not illustrated at all, yet they were still written using this cipher. So the question is, as always with this kind of solution, whether or not you accidentally ended up developing a system that gave you enough freedom to always match a Voynichese word to a possible name for the plant.

There are a number of tests that can reveal issues with proposed translations. One is, could you translate a paragraph without illustrations, for example from f108v? Judging by what you write in your paper, I think you already realize that the answer is no, you need to work towards an illustration. Given the freedoms permitted in the interpretative steps, the number of possible plaintexts is simply too great even for a "native speaker" of the original language.

The second test, which Marco mentioned again recently elsewhere, is to take a page of medieval English and encode it using your system. Does the result look like Voynichese, or do you run into issues? My prediction is that you will encounter many sequences that cannot be expressed in proper Voynichese, even allowing for the additional freedoms granted by your system.

I did a quick test on the opening lines of the Canterbury Tales found here You are not allowed to view links. Register or Login to view. , using the yellow table on p23 of your paper. This is the result:

Quote:lary peqeeeeead dy pook of dy kochas of koieekarpire iilon dok oprechchy iied les sloiras sooky dy droqlky of eeorg lod parsad ko dy shooky oeed podad yiare iaeeey en siieg chekoir of iileg iarki yeeqaeedrad es dy cfhoir iilon sapleris yak iied les siiaky praad eeesperad lod en yiare lochk oeed laad dy kaeedry kroppas oeed dy eoeeqy soeeeey lod en dy shon les lochfy koirs eroeeeey oeed seeochy foiiachas eeokan eeachodey dok schapan och dy eeeqlk iied opan ey so prekad lan eeokúry en ler koroqas doeeeey choeeqan fochk ko qoon on pechqreeeoqas oeed pocheearas for ko sakan skroieeqy skroeedas ko fareey lochiias koiidy en soeedre choeedas oeed spaseochche fron yiare sleras yeedy of yeeqachoeed ko koieekarpire dae iiaeedy dy looche pchesfich eeorker for ko saky dok lan lod lochpan iilon dok dae iiary saaky

Or in Voynichese:

Quote:lary peqeeeeead dy pook of dy kochas of koieekarpire iilon dok oprechchy iied les sloiras sooky dy droqlky of eeorg lod parsad ko dy shooky oeed podad yiare iaeeey en siieg chekoir of iileg iarki yeeqaeedrad es dy cfhoir iilon sapleris yak iied les siiaky praad eeesperad lod en yiare lochk oeed laad dy kaeedry kroppas oeed dy eoeeqy soeeeey lod en dy shon les lochfy koirs eroeeeey oeed seeochy foiiachas eeokan eeachodey dok schapan och dy eeeqlk iied opan ey so prekad lan eeokúry en ler koroqas doeeeey choeeqan fochk ko qoon on pechqreeeoqas oeed pocheearas for ko sakan skroieeqy skroeedas ko fareey lochiias koiidy en soeedre choeedas oeed spaseochche fron yiare sleras yeedy of yeeqachoeed ko koieekarpire dae iiaeedy dy looche pchesfich eeorker for ko saky dok lan lod lochpan iilon dok dae iiary saaky

Now I certainly made some mistakes. And the presence of an interpretative step makes that you still have some leeway to nudge this closer to Voynichese. But I don't quite see how it will get close enough...

Hi Pythagoras and Koen G, 
 
thank you both for the comments and for reading my paper. I appreciate it a lot!
 
Koen G, you are spot on! Your question is very good: 
«So the question is, as always with this kind of solution, whether or not you accidentally ended up developing a system that gave you enough freedom to always match a Voynichese word to a possible name for the plant.”
 
It might be a problem that my proposal for the decoding of the cipher gives many choices when translating. I have wondered about this myself. Does it give too many choices? Since this problem does not in itself lead to the conclusion that the decoding is wrong, I find that there is no point in stopping. It gives results. It is only needed to be done more work to study, limit the choices and find out why it is at it is. It occurs to me: Maybe this problem, which might not be a problem, has made people stop before? Resulting in the puzzle not to be solved. 
 
And I believe you will find, if you have time to study the paper, that the choices are not that many. With the proposed decoding, it is not possible to find other plant names in the herbal texts. If at all, it will not be within the same language, or a name of a plant that fits with the illustration, or within the first words of each herbal text.
 
And, yes, it is possible to get meaningful translations of pages without illustrations. It will come. I believe that the choices for translation will be reduced with further studies, and the seemingly looseness of the cipher might be tightened. There are several things do. For example, a study of the plant names is needed to further specify spellings and phonology. A systematic work on how the words are used throughout the manuscript is also needed before one can claim to have good translations. I hope you saw that I did not claim to have correct translations. This will take time. 
 
It is needed to keep a lot of things open. For example, the aim of this manuscript might not be to transfer information in the same way as Chauser did. 
We also need to keep it open whether the language is Middle English or not. The Middle English dictionary gives results, and the plant names found are closely related to vernacular plant names used in medieval England, but it is still needed to be kept open what kind of language from medieval England (or the area around England) it might be. 

 
Your second test of my decoding uses my yellow table, but it does not use the decoding specified (the following tables). These specifications are needed to be taken into account. The result will be very different. It will be something like the PDF I attach.


Best regards,
Siv

Hi Siv

Thanks for your reply. I want to stress once again that I admire much about your paper. The way you try to keep a logical system while at the same time juggling Voynichese's many oddities is not an easy task. Several of the plants you use seem like the most sensible identifications (water lily, viola...) though as always with the VM others are much less certain. 

And it is especially refreshing that you offer a relatively clear way for others to test your system. Often there is a lot of hocus pocus hidden under the hood, which makes falsification impossible.

My two points of criticism remain though:

1) I think that purely statistically, you would be surprised at how many sentences your system could produce from one Voynichese sentence. "P" that could also be "B", vowels that can be inserted... Those things all multiply. More fundamentally, I think the combination of vowel dropping and fuzzy consonants is a big problem. Some writing systems like Hebrew drop vowels, but those use a very rigid consonant system. If both your vowels and consonants are uncertain, I'm afraid the cipher is too lossy.

In your attachment, how is it you get from "whan" to [aiin]? This already feels like a big stretch, since you tend to transliterate [aiin] as "ewn" which becomes something like "hewen" for "heaven". Having [aiin] stand for "hewen" or whatever spelling of the word is perfectly fine, but if "whan" is also [aiin], possibilities are getting out of hand. Further on we have these translations:
[daiin] - dóun, dóuen, theuen, thung
[aiin] - even, wen
[kaiin] - twein
[taiin] - keome
[oaiin] - houven
[chaiin] - leove
[saiin] - swene

This means that [aiin] can correspond to ME óun, óuen, euen, ung, even, wen, wein, eome, uven, eove, wene. And these are not spelling variations, they are different root forms with different pronunciations and different meaning in Middle English.

If every word in a Voynichese sentence has even only two different translations, then a sentence of ten words will give you over a thousand options. Note that this does not take into account the usual ambiguity like homonyms; the system conflates unexpected, unrelated words.

2) I don't think dialect matters much when it comes to testing a cipher. Agreed, things like omission of initial "h" might influence certain statistics, but overall the letter combinations that are possible in any English dialect far outweigh those that can be achieved in Voynichese. I think you will notice this if you try to convert more English texts to Voynichese: no matter which dialect or type of source text, you will keep running into the same problems.