What is the most optimal phonological spread possible?

[bradrn]

To me, the optimally-small phoneme set is /p t k m n s~ʃ~h r~ɾ~ɺ~l w~ʋ~v j~ʝ/ for consonants and /a i~e u~o/ for vowels...

I feel obliged to link Darren’s five-phoneme language. (And more importantly, his exhaustive survey of very small inventories.)

[HolyKnowing]

To me, the optimally-small phoneme set is /p t k m n s~ʃ~h r~ɾ~ɺ~l w~ʋ~v j~ʝ/ for consonants and /a i~e u~o/ for vowels...

I feel obliged to link Darren’s five-phoneme language. (And more importantly, his exhaustive survey of very small inventories.)

Even though Darren's 7-phoneme (where are you getting the number 5 from?) language is probably the logically correct answer (that is, the literal smallest solution set that satisfies all of Joseph Greenberg's conditional requirements), intuitively speaking, it is probably Travis B.'s 10-phoneme answer that we visualize as the smallest of them all.

[Torco]

zomps right don't be mean to the newb, guise, come on.

Holy, I think what you want is lemmatization? cause like, yeah, english and natural languages are totally ambiguous, but also people are not going to learn a new language to talk to Bard GPT. but what you can do, in order for some LLM to parse it correctly is to... well... have a library that parses it correctly, reducing this ambiguity and accessing meaning more directly. current lemmatization, at least in spanish, is kind of crummy (anyone know a good spanish lemmatizer?). i haven't tried it english. once you become able to correctly tokenize speech, you can start parsing NLP inputs semantically and, for example, explain to your LLM that people do not, in fact, eat rocks even though it has seen that sequence of tokens... which won't be helped by either including or excluding clicks lmao why even have a concrete phonology for this idea?

i'm sure someone's thought of the idea, but wouldn't it be the logical thing for an interlang or such to have, say, three vowel vague phonemes/graphemes and ten vague consonant phonemes/graphemes and just let different people pronounce it the way it was easier for them? within reason, of course, I feel as if "incorrect" (highly divergent) pronounciation, for example in second language speakers, doesn't hinder understanding as long as grammar and vocab are mostly "correct". if an american who knows zero spanish reads spanish text, I very often can understand it. try phonological anarchy, coming soon to a supermarket near you!

[bradrn]

Even though Darren's 7-phoneme (where are you getting the number 5 from?) language

Ah, sorry… five consonants, three vowels. But three-vowel inventories are well-attested and pretty unremarkable; the most interesting part is the consonant inventory.

[HolyKnowing]

zomps right don't be mean to the newb, guise, come on.

Holy, I think what you want is lemmatization? cause like, yeah, english and natural languages are totally ambiguous, but also people are not going to learn a new language to talk to Bard GPT. but what you can do, in order for some LLM to parse it correctly is to... well... have a library that parses it correctly, reducing this ambiguity and accessing meaning more directly. current lemmatization, at least in spanish, is kind of crummy (anyone know a good spanish lemmatizer?). i haven't tried it english. once you become able to correctly tokenize speech, you can start parsing NLP inputs semantically and, for example, explain to your LLM that people do not, in fact, eat rocks even though it has seen that sequence of tokens... which won't be helped by either including or excluding clicks lmao why even have a concrete phonology for this idea?

i'm sure someone's thought of the idea, but wouldn't it be the logical thing for an interlang or such to have, say, three vowel vague phonemes/graphemes and ten vague consonant phonemes/graphemes and just let different people pronounce it the way it was easier for them? within reason, of course, I feel as if "incorrect" (highly divergent) pronounciation, for example in second language speakers, doesn't hinder understanding as long as grammar and vocab are mostly "correct". if an american who knows zero spanish reads spanish text, I very often can understand it. try phonological anarchy, coming soon to a supermarket near you!

What isn't represented in the consonant system, gets transferred in word length instead. Languages with simple consonant systems look and sound like Japanese (even though WALS sordidly sorts her under "moderate" consonant complexity). With long verbs and ambiguity.

Also the bigger the sound inventory gets, the more potential grammar you gain: vowel harmony, which is a stronger type of ablaut, becomes possible with vowel systems at least the size of Finnish, but remains ablaut with vowel harmony the size of Spanish, e.g.

[bradrn]

Languages with simple consonant systems look and sound like Japanese (even though WALS sordidly sorts her under "moderate" consonant complexity). With long verbs and ambiguity.

I could mention numerous counterexamples, but for now I’ll simply give you a sample of Iau, with six consonants and almost entirely monosyllabic words:

/ai⁴²³ɪ²³ | di⁴²³ a³²fa⁴³⁻³² bɪ⁴³bɪ⁴²³ du⁴³bɛ⁴³ | bʊ⁴² dɔɛ⁴³ du⁴³ | a⁴²³ tɔ⁴² di⁴²³ a⁴²³da³ bɛ⁴³si⁴³ aɛ⁴²³ ɪ³ || bɛ⁴³fɛ⁴³ bi⁴³si⁴²³ ba⁴³bʊ⁴²³ bʊ⁴² dɔɛ⁴³ a⁴²³ ɪ⁴²³ | a⁴²³ fi⁴⁵au⁴³ fa⁴²³ fɛ⁴⁵ bɛ⁴³ | i⁴³ fɔ⁴²³ aɪ⁴⁵/

(The numbers represent tones.)

EDIT: better example in the usual orthography, from https://www.yavba.net/en/iau-stories-an ... au-stories:

A⁹ da⁸du⁷ si⁶ o⁷ da⁸dv⁹ y⁸ bo⁴ Ba⁸ti⁸vs⁸ ty⁷ Ba³kv³si⁶ bui² i⁷ be⁸ y⁸ bo⁴ a⁷se⁹ be⁸sy⁹ Ei⁷⁻⁸fo⁷ Bv⁴si⁹ be⁷ ba⁵sui⁴ da⁸dv⁹ a⁹ a⁷se⁹ A⁸da⁷ bv⁸ bi³. Bv⁸ bi⁸ da⁸dv⁹ y⁸ bo⁴ a⁷sy⁹ da⁷ da⁸ i⁷ da⁸dv⁹ fi⁴au⁷ da⁸ i⁷ da⁸dv⁹ y⁸ bo⁴ a⁷se⁹ be⁶ o⁷ da⁸dv⁹ du⁹ a⁷se⁹ a⁹ foe⁴ ba⁷ da⁸dv⁹ bv⁶ a⁷se⁹ ai³ka³ bu² da⁸dv⁹ bv⁶ a⁷se⁹ ai³ka³ bu² da⁸dv⁹ a⁷se⁹ u⁶ sui⁵ be⁷ by³ a⁷se⁹ dae⁷ avy³.

[Darren]

What isn't represented in the consonant system, gets transferred in word length instead. Languages with simple consonant systems look and sound like Japanese (even though WALS sordidly sorts her under "moderate" consonant complexity). With long verbs and ambiguity.

There is no significant corellation between phoneme inventory size and level of morphological complexity. Wichita has 13 phonemes and is one of the most polysynthetic languages on Earth; Iau has 14 and is isolating and monosyllabic. In varying places between you've got Japanese, Hawaiian, Orokolo, Buin, Puinave and so on.

Here are some more examples of small consonant inventories that sound nothing like Japanese – firstly Wichita:

Hira:wisʔiha:s kiyari:ce:hire:weʔe hikaʔaca:kikaʔakʔicaki hira:rʔ tiʔi na:kiɾih.

And Puinave:

Ná’a wɤyú juiwɤkbatdi bɤ́njɤ́i janakmádináa wɤyú ju. Óyat kamokma ipɤ́iot mot, dukjínot.

And lovely Karajá:

Nawiihikə̃ hɛka hɛmə̃lalawana wimə̃ radɔnə̃mə̃hə̃re. ɗahɛ nawiihikə̃ rʊnə̃mə̃hə̃ ɔwɔrʊɗɨrɛkɪ rəma rɪθamə̃.

[HolyKnowing]

I could mention numerous counterexamples, but for now I’ll simply give you a sample of Iau, with six consonants and almost entirely monosyllabic words:

/ai⁴²³ɪ²³ | di⁴²³ a³²fa⁴³⁻³² bɪ⁴³bɪ⁴²³ du⁴³bɛ⁴³ | bʊ⁴² dɔɛ⁴³ du⁴³ | a⁴²³ tɔ⁴² di⁴²³ a⁴²³da³ bɛ⁴³si⁴³ aɛ⁴²³ ɪ³ || bɛ⁴³fɛ⁴³ bi⁴³si⁴²³ ba⁴³bʊ⁴²³ bʊ⁴² dɔɛ⁴³ a⁴²³ ɪ⁴²³ | a⁴²³ fi⁴⁵au⁴³ fa⁴²³ fɛ⁴⁵ bɛ⁴³ | i⁴³ fɔ⁴²³ aɪ⁴⁵/

(The numbers represent tones.)

EDIT: better example in the usual orthography, from https://www.yavba.net/en/iau-stories-an ... au-stories:

A⁹ da⁸du⁷ si⁶ o⁷ da⁸dv⁹ y⁸ bo⁴ Ba⁸ti⁸vs⁸ ty⁷ Ba³kv³si⁶ bui² i⁷ be⁸ y⁸ bo⁴ a⁷se⁹ be⁸sy⁹ Ei⁷⁻⁸fo⁷ Bv⁴si⁹ be⁷ ba⁵sui⁴ da⁸dv⁹ a⁹ a⁷se⁹ A⁸da⁷ bv⁸ bi³. Bv⁸ bi⁸ da⁸dv⁹ y⁸ bo⁴ a⁷sy⁹ da⁷ da⁸ i⁷ da⁸dv⁹ fi⁴au⁷ da⁸ i⁷ da⁸dv⁹ y⁸ bo⁴ a⁷se⁹ be⁶ o⁷ da⁸dv⁹ du⁹ a⁷se⁹ a⁹ foe⁴ ba⁷ da⁸dv⁹ bv⁶ a⁷se⁹ ai³ka³ bu² da⁸dv⁹ bv⁶ a⁷se⁹ ai³ka³ bu² da⁸dv⁹ a⁷se⁹ u⁶ sui⁵ be⁷ by³ a⁷se⁹ dae⁷ avy³.

That's crazy.

[bradrn]

I could mention numerous counterexamples, but for now I’ll simply give you a sample of Iau, with six consonants and almost entirely monosyllabic words:

/ai⁴²³ɪ²³ | di⁴²³ a³²fa⁴³⁻³² bɪ⁴³bɪ⁴²³ du⁴³bɛ⁴³ | bʊ⁴² dɔɛ⁴³ du⁴³ | a⁴²³ tɔ⁴² di⁴²³ a⁴²³da³ bɛ⁴³si⁴³ aɛ⁴²³ ɪ³ || bɛ⁴³fɛ⁴³ bi⁴³si⁴²³ ba⁴³bʊ⁴²³ bʊ⁴² dɔɛ⁴³ a⁴²³ ɪ⁴²³ | a⁴²³ fi⁴⁵au⁴³ fa⁴²³ fɛ⁴⁵ bɛ⁴³ | i⁴³ fɔ⁴²³ aɪ⁴⁵/

(The numbers represent tones.)

EDIT: better example in the usual orthography, from https://www.yavba.net/en/iau-stories-an ... au-stories:

A⁹ da⁸du⁷ si⁶ o⁷ da⁸dv⁹ y⁸ bo⁴ Ba⁸ti⁸vs⁸ ty⁷ Ba³kv³si⁶ bui² i⁷ be⁸ y⁸ bo⁴ a⁷se⁹ be⁸sy⁹ Ei⁷⁻⁸fo⁷ Bv⁴si⁹ be⁷ ba⁵sui⁴ da⁸dv⁹ a⁹ a⁷se⁹ A⁸da⁷ bv⁸ bi³. Bv⁸ bi⁸ da⁸dv⁹ y⁸ bo⁴ a⁷sy⁹ da⁷ da⁸ i⁷ da⁸dv⁹ fi⁴au⁷ da⁸ i⁷ da⁸dv⁹ y⁸ bo⁴ a⁷se⁹ be⁶ o⁷ da⁸dv⁹ du⁹ a⁷se⁹ a⁹ foe⁴ ba⁷ da⁸dv⁹ bv⁶ a⁷se⁹ ai³ka³ bu² da⁸dv⁹ bv⁶ a⁷se⁹ ai³ka³ bu² da⁸dv⁹ a⁷se⁹ u⁶ sui⁵ be⁷ by³ a⁷se⁹ dae⁷ avy³.

That's crazy.

Yes, Iau is phonologically one of the crazier languages around. I love it!

(If you want less crazy, Darren’s examples are good too.)

[Torco]

yeah, the actual diversity of natlangs is much crazy, and much greater than what one might think before studying it. i recommend it!

[Darren]

(If you want less crazy, Darren’s examples are good too.)

Hey, Karajá is at least as crazy as Iau