(9 hours ago)ReneZ Wrote: You are not allowed to view links. Register or Login to view.Those entropy calculations are still very depending upon the type of transliteration;
some more aggressive "experts" here are favourizing an "h2" of 2.3 for Voynichese, while the change of transliteration file can produce a value of ~2.9.
While the conditional entropy calculation is dependent on transliteration, this is not the only reason for low entropy. With my Slovenian transliteration alphabet, I increased the h2 to 2.3, however the minims still remain ambiguous and required to be read in the context. I noticed that EVA does not have separate equivalent for the Voynich glyph that looks like Latin u. Instead, the
ee the two curved shapes connected with the rounded line at the bottom are considered
ee (or
cc in Currier). This makes the Voynich language look so strange, because three, or even four equal vowels cannot be found in a natural language. Because of the ambiguous writing, some
u-shapes are more clear than others, however, when there are three same vowels, like
eee, they represent either
ue or
eu. In a similar way
ii was used for
ji or
ij before the letter
j came into use in the 16th century.
I am sure that if the Voynich text, which I believe is early Slovenian, would be written down with 44 Glagolitic letters, the h2 would increase, but that was not the option. The option chosen by the author was to invent a few new letterforms. Of course, the alphabet of 100 different letters would also increase the h2, but would such text be readable?
I Used my transliteration alphabet SLO-VA to compare it with ZL transliteration.
My starting comparison of h2
ZL transliteration – all characters
Entropy:
h0 = 5.700439718141093
h1 = 3.9785010591329724
h2 = 2.2069807941051414
SLO-VA transliteration
Entropy:
h0 = 5.882643049361842
h1 = 4.040270713023543
h2 = 2.2971498898978986
Note: I used the same
ch and
sh combination for
č and
š Slovenian sounds. Minims were read as EVA
n as
v, EVA
in as
n and EVA
iin as
m, and EVA
iiin as
im, unless there is a visual indication that
m is
iw, or
n is
iv, etc. I transliterated some EVA
sh glyphs as
be, when initial
b was written clearly and was connected with horizontal line in the middle to the next letter. This still leaves enough room to improve entropy, but that is how far one can get without understanding the language.
Transforming EVA
n,
in,
iin and
iiin into
v,
n,
m,
im increased h2 to 2,28. I imagined that final
n to
n to
w,
iv, final
iin to
jiv,
nv, iw, and final
iiin to
jiw,
nu would further increase the h2.
After converting
q to
p, the h2 does not change, because it is paired with the same letter (PO compared to EVA qo)
Continuing transforming 4870 EVA
ee combinations to
u increased h2 to 2.3
(I assumed that all
ee are
u, because there seem to be no double letters, although there are a few double vowels, but too few to make statistical difference).
Transforming 22388
ch to
č brings the h2 to
2.4.
Converting
ch in SLO-VA transliteration to
č increases h2 to
2.39
Changing
sh to
š increases h2 to 2.41, and changing
y to
i, j, ij, ji would also increase the h2, while inserting
e or
a for the missing vowels may decrease the h2.
While the trend shows increase of h2 with some changes, this only affects the transliteration.
ZL transliteration cannot be fully converted to SLO-VA transliteration with find and replace, however even the approximate conversion indicates what increases or what decreased the h2.
I consider SLO-VA also as transcription alphabet that converts Voynich glyphs to Latin letters, not to Slovenian medieval alphabet. Only the words that can be unambiguously transliterated and are comprised from the Latin letters that have stable phonemic value in Slovenian and Latin, could be converted with simple substitution (with one-to-one or two to one (
ch, sh to
č, š), or one to two ( EVA
p, f to
sv, zv, cv), like EVA rod, dar, dal, kar, char.
The more complicated problem is how to normalize the text to be comparable to medieval Slovenian writing, because neither spelling nor pronuciantion was standardized. Different writers used different Latin letters for different sounds (
c for
k, h, z, č., s for
s, z. ) Late in the 16th century
j and long
s were also used, but
r, w, y were dropped.
There is no suitable medieval Slovenian text for comparison, and the Wikipedia articles are absolutely no match to the medieval Slovenian language. Because the Voynich text is too dificult to normalized for comparison with any medieval Slovenian text, because alphabet, spelling and vocabulary has changed drastically, even the simple 'most frequent word test' does not work.
The language is based on the peasant language that was very simple. Words were created in a similar way as Timm explained, there was very little variety of prefixes and prepositions, a lot of repeated suffixes which are in VM also misread, like
n and
iv,
am,
ed,
il, a lot of words that are spelled the same, but pronounced differently, due to different value of vowel or different accent.
Because of all the above, the Voynich manuscript is so resilient to computational analysis. Unless the text is properly normalized for the language peculiarities, the tests would generate incorrect statistical data and the Voynich grammar based on these data would be incorrect. However, most peculiarities Voynich researchers pointed out can be explained by way of Slovenian language.
The main indications that the language is Slavic is are the word structure and morphology. While Slovenian is close to Czech and Croatian, the differences were noticable even in the middle ages.