Hi Guillaume, welcome.
I've read a few files on your GitHub repository.
Your mapping is well optimized to generate Latinish words, but not yet at the level of a "Lorem Ipsum" generator (fake Latin, but not entirely fake). Too many 2,3,4,5-letter nonsensical words, far too many "eius".
I haven't used the Perseus Latin Dictionary, but 265.000 forms is quite small and (very probably) only covers classical Latin, not medieval Latin (words and variant spellings). For example late medieval Latin words rarely had any "ae" in them, with few exceptions, like "aer". Instead of "ae", the e cedilla "ȩ" was getting reintroduced during the Quattrocento in Italy, by the Humanists, but everyone else wrote only "e". So words like "cerae", "cassiae", "aeque" and "aequi" are not plausible if the text was wriiten in the first half of the 15th century.
BTW your data/dictionaries/latin.txt is actually English.
I've read a few files on your GitHub repository.
Your mapping is well optimized to generate Latinish words, but not yet at the level of a "Lorem Ipsum" generator (fake Latin, but not entirely fake). Too many 2,3,4,5-letter nonsensical words, far too many "eius".
I haven't used the Perseus Latin Dictionary, but 265.000 forms is quite small and (very probably) only covers classical Latin, not medieval Latin (words and variant spellings). For example late medieval Latin words rarely had any "ae" in them, with few exceptions, like "aer". Instead of "ae", the e cedilla "ȩ" was getting reintroduced during the Quattrocento in Italy, by the Humanists, but everyone else wrote only "e". So words like "cerae", "cassiae", "aeque" and "aequi" are not plausible if the text was wriiten in the first half of the 15th century.
BTW your data/dictionaries/latin.txt is actually English.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)