An attempt at extracting grammar from vord order statistics.

[dashstofsk]

'a[g:j]': 1, 'dchdaldyykedy': 1, 'lkedylkeeody': 1,

Your approach is very intriguing. But here note that a[g:j] is not a word. Probably the writing is not clear in the manuscript. It might be ag or aj. Also  dchdaldyykedy, lkedylkeeody  are very rare. In fact they are so rare that they don't seem to appear in ZL3a-n. ( I suspect your program is not handling '<->' correctly. Have you considered working with GC2a-n. This is a simpler transliteration that doesn't have '<->'. ) But is your program really looking at all words? Might it be better if you were to exclude unknowns and rare words. Perhaps then it would not take 'one hour on my poor old pc'.