(13-10-2020, 02:20 PM)geoffreycaveney wrote: You are not allowed to view links. Register or Login to view.In the course of my research on West Slavic languages in the medieval period, I have come across interesting descriptions of the early orthography of Old Polish in the medieval period. A short summary can be found You are not allowed to view links. Register or Login to view..
Unfortunately such changes in spelling are insufficient to bring down the conditional entropy close to what is seen in the VM. For instance the entropy of the Pan Tadeusz text I used is 4.47 for standard spelling and 4.25 for simplified spelling, conditional bigraph entropy for standard spelling is 3.55 and for simplified spelling is 3.49. This reflects mostly reduction in symbol number not reduction of the information, as the entropy normalized for maximum for given character set actually increases for the simplified spelling.
The most likely culprits for reduced entropy in VM in my mind are: 1. pig-latin like tricks, 2. intentional duplication or multiplication of letters (so combinations of glyphs stand for single letter of underlying language), 3. attempts to code phonemic-rich language (something like Ubykh with 88 consonants, language where syllables are perceived as phonemes by native speakers (like in Japanese) or language with rich tonal phonemics) using 20 or so glyphs, 4. we do not kow what the glyphs actually are and we are over-analyzing them in the current interpretation (eg. c and cc are in reality something like a and b). Adyghe (phoneme rich language) written in cyrilic (36 glyphs or so) has indeed somewhat lower conditional entropy due to abundance of bigraphs and trigraphs. Another example of lowered conditional entropy: replacing each letter with quadruplet of the letters in Pride and Prejudice drops conditional bigraph entropy from 3.59 to 1.65