This idea is embryonic at the moment, so I would welcome
comments - and also pointers to prior art if it exists.
In this method, to encrypt a word, we first separate out the vowels from the
consonants, and order each group of letters. (This is similar to Philip Neal's
alphabetic anagrams idea.)
For example:
aldebaran => aaae bdlnr
pleiades => aeei dlps
algol => ao gll
Then we label each group of vowels with a shorthand. For example:
aaae => 4o
aeei => 1o
ao => 8a
For the consonants, we perhaps use a simple monoalphabetic cipher:
bdlnr => cemos
dlps => emqt
gll => hmm
The encrypted word examples are then:
aldebaran => 4ocemos
pleiades => 1oemqt
algol => 8ahmm
The appeal (at first sight) of this scheme is that it might explain the strange
prefix/stem/suffix or crust/mantle features of the VMs words, and also the
repeated words.
To lend some support to it, we can look at the frequency distribution in the VMs
of the first pair of glyphs in each word, and compare the distribution with the
ordered vowel lists from words in a plaintext language.
For the Herbal, the distribution looks like this (top 10 only shown):
(order, glyph pair in Voyn101 encoding, number of occurrences, normalised
frequency)
1: 4o 303 0.08426029
2: 1o 218 0.060622916
3: oh 159 0.044215795
4: 1c 153 0.042547274
5: ok 143 0.03976641
6: oe 103 0.028642936
7: 8a 94 0.026140155
8: 9h 93 0.02586207
9: 2o 84 0.023359288
10: 9k 75 0.020856507
Words in an English dictionary:
1: ae 779 0.08168187
2: ei 528 0.055363324
3: eo 512 0.053685647
4: ee 475 0.049806017
5: e 343 0.03596519
6: a 320 0.03355353
7: aei 309 0.032400128
8: aee 299 0.031351577
9: i 286 0.029988466
10: o 279 0.029254483
The normalised frequency distributions look comparable.
Frankly, I'm not sure how to proceed further with this, so would welcome ideas
:-)