This project aims to decrease lexical noise for various engines that use the languages algorithm:
Simplifies langer{} so it doesn't have patterns or morphotactics, there is instead a set of universal Initial-Medial-Final morphotactics based on whatever feels the most natural, with words stringed together appropriately. That said a language isn't going to pick every possible combination. Use of diacritics (but within a specified set and used lightly) is also a good idea.
Build this project keeping in mind the potential binary settings. Languages could then just be a boolean flag for each combination and some characteristic of how words form.
Compound words make sense too.
Instead of having weird rules for w and y I could just make them a natural part of the universal morphotactics and then just replace doubled variants with something else. Like Yy just turns into y, ww or w*w turns into w.
Anything in my CI list can potentially have other consonants attached to it.
Reframe the list here.
Have lists also for CM, CF, VI, VM and VF. Consonant diacritics are pretty weird, but vowel diacritics definitely make sense so long as they're isolated and there aren't too many of them -- I'm thinking acute and umlauts only.
The language creator's final step is turning binary data into a language, so I can hook into it directly with a distributive function. However the input should be a number of CI, CM, etc things to pick.
The words function then assembles things according to the universal morphotactics, with the language weighting them. These weights should be dictated in the binary sequence input into that function.
Compound word likelihood should similarly be a function. These should be 2 words only and not too likely.
It's possible that CI, etc will lack morphemes entirely -- make weird backup sets that come into play whenever the list is too small in any case. Maybe pull some of the weirder morphemes out so they only happen in that particular case.
With this project done, my languages will look cleaner and as an added bonus it'll be easier to turn binary data into one of them.
