qokeedy qokeedy qokedy qokedy qokeedy

[RadioFM]

Labels could be abbreviated words, syllables or small n-grams.Like 'PLEI' instead of 'Pleiades'. They could be numbers or index entries to a nomenclator or sth of the sort.

But what about when labels have the same word attached to distinct and quite different images and also appear as very common words elsewhere in the sentence/paragraph continous text?

If Voynichese vords and labels both encode chunks instead of whole words, it's not entirely unlikely that some labelese will match with a particular chunk somewhere in the text. The 'plei' for 'Pleiades' example I gave could pop up someplace else encoding 'mett ere multi plei nsie mee mes col are', for instnace.

I think that it's also likely under these hypotheses that a couple of labels will repeat or at least look similar. For example, taking a selected few of Agrippa's mansions of the moon (D'Imperio '78 ) and dropping some letters,
Alchalh → chal
Achala → chala
Alchil → chil
Alcharph → charph

Ofc there could be filler/null words in the main text, null characters and possibly many ways in choosing how to chunk or encode a certain piece of text, that would add more chaos to the mix. It'd be interesting to try to estimate how that could impact the stats.