I'm a python, java, and javascript programmer, and also very interested in writing a fantasy novel. I'm a begginner, but i didn't think you could literally use the coding to BUILD the language. This is just genius.
You could go crazy with this, build a tokenizer, generate random text of some conlang, give it some chosen meaning and run it through an LLM fine tuning process to make the LLM output any conlang of your choice.
That's absolutely awesome. I do research in (academic) computational linguistics, so it's cool to see this kind of application. Are you planning on writing rules for sonority scales, etc?
My plan right now is to use the English sonority scale, just because I want it to be easily pronouncable / sounded out by English readers. At some point I want to do some investigating with other sonority scales that are different but still easily pronounced! I'm very much a beginner to all things conlanging so I'm trying to take it slow.
You should join my discord! There are a few conlangers in there and a few people with their own programs. Having another person to collab with would be great! If you send me a message on my website contact me page, I can send you an invite link!
It seems to me that your generator is giving you lots of words you can't pronounce and you are choosing sounds you like. From what you're saying you'd like a clear naming distinction between the invaders and the invaded. I would pick up a pen/pencil and paper and use a sound chart (pg 33 Art of Language Invention) and generate some names from that. Then write your code. For example, we could have the name Galed (Gal Ed or Ga Led depending on stress, Gal Ed for me) and move the stops forward: Daleb, move the vowels back Dolab, changing the voicing of one consonant, Tolab or both Tolap, which I don't like so let's go with Tolaph or Tolaf or Tolof because I keep writing that sound. If we had CV structure, then maybe Galeda female, Galedo male. I hope my advice is welcome. And, whatever approach you use I'm enjoying watching you create. Thank you.
I may give that method a try and see if it works better for me! I struggle with the IPA chart and going at it from the symbols but maybe I can figure it out!
@madelinejameswrites The IPA isn't immediately accessible, which is why I recommend that sound chart, as a lot of the symbols are the same as we use in English. 😊 For quite a while, I looked at the IPA as something too academic and full of odd symbols. But it's nice when you get the logic behind it. I think it's a good idea of yours to focus on sounds that English speakers are comfortable with. And an easy way to get a language with an non-English feel is to reduce the vowels (I think English has fourteen, certainly my Northern English has a range) and remove 'th'.
@madelinejameswrites Mark Rosenfelder in his Language Construction Kit book, page 39, has a nice figure of the vowel sounds in American English. Peat, Pate, Pit, Put, Boot, Boat, uh, Putt, Pet, Bought Pat and Pot. If you are just using e, i, a, u and o you're limiting yourself...butttt. that's how we write. The vowel sound in bird is different to bid although we use the same symbol, i and burred sounds closer to the vowel sound in bird. If you're keeping the five vowel symbols then you could do what you've been doing making some sounds more likely. So, you could say e and i are more likely than a, which is more likely than o and u. Has that answered your question? 🤔
doesNotContainSyllableBuildIssues should clearly be done in a more sane way. I suggest Stream.of("iy", "qr", "tm", ...).noneMatch(stringToValidate::contains)
