(24-04-2024, 12:22 AM)pjburkshire Wrote: You are not allowed to view links. Register or Login to view.[quote="asteckley" pid='59113' dateline='1713912207']
... so two groups of training materials need to be identified.
I think your direction of thinking is a good one. But one has to determine what the framing of the problem should be. And I'm not sure what you have in mind there.
Not all machine learning involves the following type of training data, but its still helpful to describe my point in this way:
- One has many "examples" (training cases) that include an input set of data and an output "answer".
- Once trained, one can then give it the data representing the Voynich, and it will predict the answer for the Voynich case. ("predict" can be a misleading word, but just means it converting the input data. into the output answer).
You seem to be suggesting approaches to segmenting the kinds of data - and doing so with more and better data than Kondrak used -- but what are you thinking that the input sets of data and the answer produced actually represent?
The goal is to answer the question; What does this text say?
It's going to involve working with the operators of the application to define what they need as input beyond the transcribed text data from the manuscript.
I recommend starting with pages You are not allowed to view links.
Register or
Login to view. to You are not allowed to view links.
Register or
Login to view. . The first page, You are not allowed to view links.
Register or
Login to view. , is a mystery and Quire 8 is weird. After that things get even more complicated. There are a lot of Zodiac pages but they have a lot of illustrations and only a small amount of text.
I recommend starting with only transcribed text data and not images but it might be useful to include a brief description of the plant on each page. For example, the plant on You are not allowed to view links.
Register or
Login to view. has white flowers and the plant on You are not allowed to view links.
Register or
Login to view. has red flowers. This information may be more useful in determining which word is the color white and which word is the color red than in doing plant identification.
Many investigations have gone into the VMs jungle and not all have come out. Previous attempts to match visual identifications with medieval herbal images and then compare the two written texts have not found positive results.
If the goal is to correctly interpret the text, then beyond statistical analysis, there is the need to interpret specific examples of VMs text. So, the question is, which ones? Particularly if the botanical investigations have not been successful so far. Which pages of text should be targeted?
Perhaps, if deception is involved in the VMs, the author is a step ahead here and has separated the hidden text into several segments - marked by variations of Stolfi's markers in the VMs zodiac sequence, it the cosmos and in the central rosette, etc.
In order to lend potential validity to these markers, VMs White Aries clearly shows the connection of these markers with the actual religious history of the Fieschi popes and the origins of the tradition of the cardinals' red galero.
There is a difference between 'proof' in a modern sense and in a sort of medieval determination. Modern proof is beyond a reasonable doubt. Medieval proof is if the ideological association works correctly, it is valid. Paired heraldic, azure, bendy patterns and red galeros are two elements that have a specific, verified existence in the religious history of the medieval church. The interpretation then turns on the investigator's knowledge of these specific historical details.
In the end, if it can be solved, it will be solved by a guess.
In other problems, we might guess that the hierogylyphs with a circle around them are important and guess that they might be the name of a Pharoah. And be right.
Or we might guess that the double glyph in a cipher is the double letter LL because that is common in English. And be right.
Isn't that what AI does, after all? An LLM guesses what word comes next. It just gets very good at guessing - and that is what we call intelligence.
Human beings are very good (and creative) at guessing, and at ways of improving their guesses. (We're not talking about baseless guesses, but educated guesses, the guesses of the duly prepared.)
I honestly think that one day - if it can be solved at all - someone (most likely after long study and reflection and frustration) will make the right guess and the solution will become clear. They will find the key and the door will be unlocked. But it will involve a leap, an insight, a hunch, that only a human being - the best guessing of creatures - can make, and it will seem almost like an accident.
(25-04-2024, 05:56 AM)HermesRevived Wrote: You are not allowed to view links. Register or Login to view.They will find the key and the door will be unlocked.
I think that is the problem. People are looking for a key. What if there isn't a key? What if I am right and it is not an encryption but a lost language?
Why are so many people so sure that it is an encryption?
I watched this old video of an interview with Greg Kondrak. He thought it was an encryption. Maybe that is why he failed. Maybe it wasn't wrong training data. Maybe he was asking the wrong questions.
Solving The Voynich Manuscript With Ai | Greg Kondrak
February 3, 2023
You are not allowed to view links.
Register or
Login to view.
(25-04-2024, 12:22 PM)pjburkshire Wrote: You are not allowed to view links. Register or Login to view.I think that is the problem. People are looking for a key. What if there isn't a key? What if I am right and it is not an encryption but a lost language?
Why are so many people so sure that it is an encryption?
I answered something similar here: You are not allowed to view links.
Register or
Login to view.
Something has been done to the plain text, and it seems this has the effect of encryption.
There has been research looking at properties of a wide variety of natural languages. None of them were consistent with Voynichese's properties. Even those like Hawaiian or Mandarin that might match one weird property fall down with regard to others.
Your lost language would presumably be related to an existing language studied, and so how likely is it that it differs enough from its surviving siblings to match all of Voynichese's bizarre properties?
If you are imagining a language isolate that could have been spoken enough to make it into the manuscript by 1438 but is utterly unknown with no trace now, that language isolate would also have to have very unusual properties from other language families. This seems low probability.
(25-04-2024, 12:45 PM)tavie Wrote: You are not allowed to view links. Register or Login to view.If you are imagining a language isolate that could have been spoken enough to make it into the manuscript by 1438 but is utterly unknown with no trace now, that language isolate would also have to have very unusual properties from other language families. This seems low probability.
Unless the entire community and all their creations were intentionally obliterated by the Spanish Inquisition or some similarly motivated group intent on total extermination of the heresy of anti-Christian Zodiacs and naked women.
(25-04-2024, 01:58 PM)pjburkshire Wrote: You are not allowed to view links. Register or Login to view.Unless the entire community and all their creations were intentionally obliterated by the Spanish Inquisition or some similarly motivated group intent on total extermination of the heresy of anti-Christian Zodiacs and naked women.
The You are not allowed to view links.
Register or
Login to view.s, nearly exterminated by the Hussites in 1421?
(29-04-2024, 12:45 PM)nablator Wrote: You are not allowed to view links. Register or Login to view. (25-04-2024, 01:58 PM)pjburkshire Wrote: You are not allowed to view links. Register or Login to view.Unless the entire community and all their creations were intentionally obliterated by the Spanish Inquisition or some similarly motivated group intent on total extermination of the heresy of anti-Christian Zodiacs and naked women.
The You are not allowed to view links. Register or Login to view.s, nearly exterminated by the Hussites in 1421?
I never heard of them before but maybe. Did they have a written language?