Cribs in the Voynich MS

[Mark Knowles]

Let's assume that the 28 labels around the central star on You are not allowed to view links. Register or Login to view. are indeed related to the mansions of the moon. In that case I would be astonished, because some of the labels repeat themselves. The labels, along with their numbers of occurrences are as follows:

oar alys (1)
okcheys (1)
okeody (2)
okodchy (1)
okeod (3)
okeyd (1)
okeeydy (1)
otchy (1)
sarydy (1)
ykeody (1)
okolar (1)
saiir (1)
saral (1)
ykeydy (1)
oteol (1)
oteeys (1)
okeol (1)
otody (1)
okody (1)
ytody (1)
oeesy (1)
ytory (1)
ykeey (1)
ochoyk (1)
okeey sar (1)

It's really odd that okeod occurs 3 times and okeody twice if we consider that each mansion should have its unique name and star. Also, the labels themselves are also very similar to each other. This makes me think that when it came to picking the Voynichese labels the author was either drawing from a bucket of premade labels (and maybe sometimes mistakenly picked a label that was already taken?), or was just picking the next "closest" "free" label, whatever the criteria for those might've been. This is all assuming that a rather big Voynichese dictionary was involved, of course.

It might be interesting for these 28 words to calculate the sum of the Levenshtein distances to every other word in this list. Then sort the 28 words by this number with the greatest first. Then look to split this list into two those with high Levenshtein distance sums and those with low Levenshtein distance sums. Then focus effort on on those with high Levenshtein distance sums.

The interesting question is how easy it is to split the list into two. Is there are continuous descent in the figure for Levenshtein distance sum or is there a jump at some point? If there is a clear jump then it is easy to split the list into two.

(In practice whether one uses the Levenshtein Distance Sum or Levenshtein Distance Mean it makes little or no difference. Though to calculate the mean involves one more computation)

Levenshtein distance measures the similarity of the spelling of one Voynichese word to another Voynichese word. The mean Levenshtein distance measures the similarity of the spelling of one Voynichese word to a group of other Voynichese words. Calculating the mean Levenshtein Distance for each member of a group of Voynichese words to the other members of the group of Voynichese words allows one measure how similar each Voynichese word to the group as a whole. Sorting the group by the mean Levenshtein distance, allows to sort the members of the group in order of similarity to the group as a whole. This allows one to sort the group of Voynichese words by the distinctive down to the most homogeneous. Splitting the group into two by the mean Levenshtein distance allows one to split the group into distinctive Voynichese words and homogeneous monotonous words. Focusing on the distinctive Voynichese words I think is likely to be more profitable than the monotonous Voynichese words. The key difficulty is deciding where to make the split. A guide to that is looking at the graph of descending mean Levenshtein distances and deciding where the most obvious step change in the graph is; it would be preferable to have a formal reliable definition for how to locate that step change point.