- Options
- Login
- Desktop Style
Options
Hebrew grammar - what word classes the statistics find in Voynichese
- Hebrew grammar - what word classes the statistics find in Voynichese
-
Hebrew grammar - what word classes the statistics find in Voynichese
lunakreuz > Yesterday, 04:42 PM
Hi everyone, I'm new here!
My background is in statistics, I speak several languages (some better than others!) and have a passion for linguistics.
I've been running statistical tests on the Voynich text (ZL transliteration) for some weeks, always with controls (shuffled text, fake dictionaries, random alternatives).
I'm not claiming a translation, and I've assigned no sound values.
What I've found is grammar, and the grammar behaves like Hebrew. I am not saying this IS Hebrew, just that the language behaves remarkably like it.
It also means the text is not gibberish: the rules below are about as far from random as this kind of analysis can measure.
I know "Hebrew-like" isn't a new claim, but I haven't seen these specific results published anywhere, and if someone has, please point me to them!
Examples:
- It has a "the". In Hebrew, "king" is melech and "the king" is ha-melech: the article glues to the front of the noun. In Voynich, the letter "o" glues to the front of the tall-letter (gallows) words and passes every statistical test an article should: never doubled, switches off where there's no preceding word.
- It has words "the" can never touch. In English you can say "the boy" but never "the ran": articles don't go on verbs. Voynich has a whole class of words (the gallows letters drawn through the bench shape) that refuse the article almost absolutely (z = -21), refuse the plural ending, and avoid starting lines, in every section of the book. That's a verb class behaving like a verb class. This exclusion law is, I believe, unreported, and anyone with a transliteration can check it. Please try to break it.
- It has adjectives that agree, Hebrew-style. Hebrew says ha-ish ha-tov, "the-man the-good": the adjective repeats the article. In Voynich, "the"-marked words follow other "the"-marked words more than chance allows, even after controlling for the known letter-to-letter effects. And the agreement extends to number: when the first word of such a pair carries the plural ending, the second copies it well above chance (z = +3.5 and +4.4 in the two hands).
- It may have gender. There are two plural-like endings (-iin and -dy), and word-stems almost never take both (4% overlap where 14% is expected). Two plurals that refuse to share stems is exactly how a masculine/feminine system looks, Hebrew -im versus -ot. Early days on this one.
- It has final letter forms, like Hebrew sofits. Hebrew nun is נ inside a word but ן at the end. Two Voynich letters work exactly that way, statistically proven pairs.
- It has a plural that takes the article. The -iin ending swaps with other endings on the same word-stems, and "the" attaches to it freely, like Hebrew ha-shomrin.
- It has one-letter grammar words, two of them. Hebrew writes "and" as a single letter (vav) and "to/for" as a single letter (lamed). Voynich has standalone single-letter words doing linker jobs: the letter "d" sits between nouns far above chance (pre-registered test, z = +2.85), an "and/of" word, and the standalone letter "l" shows the same pattern more weakly, a "for/to" candidate.
- The famous word daiin has a job now. It introduces verb-class words at 3.5x baseline and loves opening lines: a step-word like Hebrew az, "then", or possibly a negator like lo. (daiin and dain test as the same word in two spellings.) A second common word, saiin, turns out to do the opposite job: it opens lines but introduces nouns, like a preposition.
- And there are no numerals in 200+ pages, strange for Europe, normal for a tradition that counts with letters, as Hebrew does.
- It has a "the". In Hebrew, "king" is melech and "the king" is ha-melech: the article glues to the front of the noun. In Voynich, the letter "o" glues to the front of the tall-letter (gallows) words and passes every statistical test an article should: never doubled, switches off where there's no preceding word.
-
RE: Hebrew grammar - what word classes the statistics find in Voynichese
oshfdk > Yesterday, 10:13 PM
(Yesterday, 04:42 PM)lunakreuz Wrote: You are not allowed to view links. Register or Login to view.It has a "the". In Hebrew, "king" is melech and "the king" is ha-melech: the article glues to the front of the noun. In Voynich, the letter "o" glues to the front of the tall-letter (gallows) words and passes every statistical test an article should: never doubled, switches off where there's no preceding word.
There are 71 instances of double o in the manuscript, not counting the long oooo sequence from f70r. Also, o is the most frequent first character of the labels in the MS, is this consistent with how ha is used in Hebrew for standalone words?
I'm not sure I understand the rest of the claims, would be nice to have some examples. -
RE: Hebrew grammar - what word classes the statistics find in Voynichese
lunakreuz > Yesterday, 11:46 PM
(Yesterday, 10:13 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.
(Yesterday, 04:42 PM)lunakreuz Wrote: You are not allowed to view links. Register or Login to view.It has a "the". In Hebrew, "king" is melech and "the king" is ha-melech: the article glues to the front of the noun. In Voynich, the letter "o" glues to the front of the tall-letter (gallows) words and passes every statistical test an article should: never doubled, switches off where there's no preceding word.
There are 71 instances of double o in the manuscript, not counting the long oooo sequence from f70r. Also, o is the most frequent first character of the labels in the MS, is this consistent with how ha is used in Hebrew for standalone words?
I'm not sure I understand the rest of the claims, would be nice to have some examples.
So what we call in EVA-o, in Hebrew is the character הַ (ha). It is used in 3 ways. Meaning the equivalent to "the" in English, as a letter by itself, starting a word, and also a number (5). In fact, the reason you don't see any numbers anywhere (others than the pages and on inner corners which were likely added later) is because in hebrew, each letter represents a number too.
For example, the word "ooy" would be translated as ההי , ha-hi, or in English "that one". Please bear in mind that I am working with a very large dictionary body of biblical medival Hebrew for comparison as that's where the grammar seems to be going, and it makes sense for the time.
I also have the full document in a format that looks like this (more or less): The-noun(pl)-adj -and noun - verb - for... etc. So they grammar looks like it works so far.
- o = "the". One letter glued to the front of nouns, like Hebrew ha. Inside a word it's just an ordinary letter. Also the number 5. Regarding f70 I believe it to be a number but I dont have all the numerals worked out yet and I am not 100% certain on what that semi circle is to be honest. I wont make guesses.
- the gallows (tall letters k, t, and their top-line variants p, f) = the noun letters. These words take "the" and take plurals.
- a gallows drawn through a bench = the verbs. They refuse "the" almost absolutely (z = -21), refuse plurals, avoid starting lines.
- the plain benches (ch, sh) = a second noun type that never takes "the", like Hebrew body-part and possessed nouns ("your hand", never "the hand").
- -iin / -in = a plural that takes "the". Like Hebrew -im, Mishnaic -in (which is why I am using this variant of Hebrew)
- -dy = a second plural. Stems pick one ending or the other, almost never both: the shape of a masculine/feminine split, -im vs -ot.
- n and m at word-end, i inside = final letter-forms, the way Hebrew nun changes shape at the end of a word.
- d on its own = a one-letter joining word between nouns, "and/of". Hebrew writes "and" as one letter too (vav), glued to the next word.
- l on its own = a weaker linker, "to/for", not certain yet. Like Hebrew lamed.
- daiin = a step-word: ushers in verbs, loves opening lines. A "then" (Hebrew az), possibly a "not". dain is the same word in another spelling (that took MANY tests).
- saiin = the mirror image: opens lines but ushers in nouns. Preposition-like.
- y at word-end = a final vowel.
- e = confirmed to be a vowel, riding on the bench and gallows consonants. This means this is hebrew partly vowelised (hebrew usually only writes its consosnants), which is why it is not a 1:1 translation. This part vowelisation was common at the time in certain texts, based on my research.
That's the grammar so far. I can post the numbers and the controls behind any single line if it's useful. -
RE: Hebrew grammar - what word classes the statistics find in Voynichese
Rafal > 8 hours ago
Well, you say your background is statistics.
Then you should realize the problems that Voynichese words make when you try to assign and convert them to any language.
About 40% of words end with -dy if I remember correctly. You won't find such patterns in any language with some many words ending in the same way. Could it be that some plural form, as you say, is so common in the text ???
There is not enough Voynich common characters to cover common letters in 99% of languages. The letter entropy is too low for any language so simple symbol substitution could work, And so on.
But I am curious to see how you could develop your Hebrew theory. So don't give up, maybe you will manage to overcome the obstacles where others failed
-
RE: Hebrew grammar - what word classes the statistics find in Voynichese
lunakreuz > 7 hours ago
(8 hours ago)Rafal Wrote: You are not allowed to view links. Register or Login to view.Well, you say your background is statistics.
Then you should realize the problems that Voynichese words make when you try to assign and convert them to any language.
About 40% of words end with -dy if I remember correctly. You won't find such patterns in any language with some many words ending in the same way. Could it be that some plural form, as you say, is so common in the text ???
There is not enough Voynich common characters to cover common letters in 99% of languages. The letter entropy is too low for any language so simple symbol substitution could work, And so on.
But I am curious to see how you could develop your Hebrew theory. So don't give up, maybe you will manage to overcome the obstacles where others failed
Yes, my major was in econometrics and I have certificates in programming, also work in technology. I am not sure this matters however!
To work on the problem I have built my own programme which I am running.
No, -dy is only 17%(ish) of characters, what you mean is -y which is at 40.6%. This is unremarkable, under Hebrew partial vocalisation (common for the time) that is one letter covering the whole final-vowel class, and final-vowel classes run high everywhere: 90% of Italian words end in a vowel.
-dy itself sits mostly in Currier B (23.5%) and pairs with -iin as complementary endings: stems take one or the other, rarely both (4% overlap against 14% expected), and neighbouring words copy the ending well above chance. That's inflection behaving like inflection.
On entropy we agree completely. 2.3 bits kills simple substitution (which is what you are thinking of, a 1-2-1 comparison), this is why I'm not proposing one (as you can see for my proposed substitutions above), that would be very silly of me considering it has been proven otherwise by other researchers time and time again. If you don't vowelise Hebrew (which only writes consonants) you can see how this cannot be a 1:1 comparison.
Just to clarify, I am also not saying this is Hebrew - I cannot make that claim yet. Just that the grammar behaves just like Hebrew from the time (not modern).
For example, numers do not behave like Hebrew numbers. Numerals in the MS behave like Gobar Arabic numerals. I am far from assigning them yet, but a Hebrew reader from that time in that region would be familiar with Gobar Arabic numeral as they were used about 47% of the time in the 14th Century.
My analysis is based on how Linear B was decyphered: No values assigned, no substitution claimed: grammar first, against permutation nulls, values last.
Curiosity welcome.
(Yesterday, 11:23 PM)RobGea Wrote: You are not allowed to view links. Register or Login to view.For Hebrew-like claims --See tavie's curated solution list - Currently 4 Hebrew, 2 Arabic 1 Aramaic
You are not allowed to view links. Register or Login to view.
Hi, I am not claiming this is Hebrew. Just Hebrew grammar for now and Arabic numerals. When I am further into my research I can say more.