Here's a further experiment in mapping the most frequent "words" in the Voynich manuscript to text strings in medieval Persian.
So far, the permutation that yields the most real words in Persian is as follows:
- glyph frequencies from my v101.L0.4 transliteration of the Voynich manuscript
- letter frequencies from the collected works of the poet Kamal Khojandi (1321-1400).
In Khojandi's works, among the common letters two have very similar frequencies: namely م (seventh ranked, 5.5 percent) and ه (eighth ranked, 5.3 percent).
If the Voynich scribes worked from precursor documents in Persian, even contemporary with those of Khojandi, the letter frequencies probably would be slightly different. In particular, the letter م might be eighth ranked, and the letter ه seventh ranked. That would change my mappings.
Applying this interchange of frequencies, I mapped the most common Voynich "word" {8am} to the Persian word سه (in English, "three"). That seems a more plausible mapping than my first attempt.
Above are my revised trial mappings of the top ten Voynich "words".