Substantial postings about constructed languages and constructed worlds in general. Good place to mention your own or evaluate someone else's. Put quick questions in C&C Quickies instead.
pdusen wrote:* A uniform sound change applier (for language families)
Something along these lines I've been thinking about recently was a sound change applier that changed the phonemic structure of each word at some rate relative to their relative frequency within the language. From what I think I know, sound change doesn't occur nearly as neatly over time as sound change laws would suggest (although they are good models), and there are always a few stragglers or so which later have more conservative structures compared to other words.
Relative frequencies of words could be acquired by feeding in passages written in the language, and then counting each occurrence of some word.
Native: English || Pretty decent: Ancient Greek || Alright: Ancient Hebrew || Eh: Welsh || Basic: Mandarin Chinese || Very basic: French, Latin, Nisuese, Apsish
Conlangs: Nisuese, Apsish, Kaptaran, Pseudo-Ligurian
Znex wrote:Relative frequencies of words could be acquired by feeding in passages written in the language, and then counting each occurrence of some word.
Znex wrote:Relative frequencies of words could be acquired by feeding in passages written in the language, and then counting each occurrence of some word.
How would this account for conjugated words?
I'm not too sure, frankly. The conjugations (well for a single tense) and frequencies would probably best be grouped together as one word, since any more accurate attempt that distinguishes frequency between conjugated words I think would require a lot more data. For agglutinating languages, the groupings could occur according to affix hierarchy.
Native: English || Pretty decent: Ancient Greek || Alright: Ancient Hebrew || Eh: Welsh || Basic: Mandarin Chinese || Very basic: French, Latin, Nisuese, Apsish
Conlangs: Nisuese, Apsish, Kaptaran, Pseudo-Ligurian
So, I've actually implemented a lot of these things at least partially, but I don't have time to release/maintain anything right now. Maybe if I get some help I'll make a webapp (I don't know web development at all).
Jipí wrote:Mh, I think s/he rather means automatic transcription. However, that's not always so easy, and English is a good example here: graphemes may have different phonemic values depending on context, sometimes unpredictably so – e.g. 〈-ough〉 in through, rough, hiccough, lough; it's also why 〈ghoti〉 can't read /fɪʃ/. I suppose the problem either needs a large ruleset or true artificial intelligence to always deliver correct results, depending on the complexity of the orthography.
Probably not so hard if your conlang's graphs are phonemic. Then you could probably just use something like text crawler.
Oh, that makes sense. It's true that conlangs with more complex pronunciation rules are much harder to map to specific IPA symbols.