Zompist Bboard Again

I’m designing an RPG where certain elements are randomized per playthrough based on a seed, rather as various elements are randomized per dungeon run in a roguelike. One of the things that is randomly generated for each playthrough is the language from which the incantations for spells are taken. (Other uses of this language include the names of other randomized entities deriving from the ancient civilization that created magic, such as dungeons and artifacts.) Each spell’s name is a two to four syllable word generated by combining onsets and rimes from a formula. Thus, for example, the invisibility spell might be KAIGAL, from K + AI + G + AL, or the “physical barrier” spell might be PERTUNUN, from P + ER + T + U + N + UN.

For convenience in setting up the algorithm that does this, I’ve decided I’d like there to be sixteen possible onsets and sixteen possible rimes so that any number of syllables has a round binary number of possible values: 256 unique monosyllables, 65,536 unique disyllables, and so on and so forth. As I also wanted it to be possible to generate a seeded playthrough (for a new game plus, for example) using the language as a password system, I wanted there to be no possible ambiguity in translating between a valid Roman alphabet sequence and a sequence of letters in the conlang’s orthography, and no ambiguity about which syllable sounds belonged to. Therefore, onsets are obligatory.

I’m fairly sure of the set of rimes I want:
/a e i o u ai au an in un al il ul ar er or/
A E I O U AI AU AN IN UN AL IL UL AR ER OR

(The nasal coda assimilates in place to a following consonant; I may decide to have it be written M before M P B.)

I’m also pretty sure I’d like to use at least these onset consonants:
/m n p b t d k ɡ ʔ s h l r/
M N P B T D K G ‘ S H L R

That leaves room for three more consonants. The main ones I’ve been considering are these:
/ts z ʃ x w ʋ v j/
TS Z SH KH W V V Y
(I don’t intend to use more than one of /w ʋ v/.)

So this is my first question: Which of these consonants should I use? I’ve been going for a general “ancient Near East/Fertile Crescent” sound within the constraints of sixteen onsets and sixteen rimes (obviously a better facsimile would have a wider range of possible codas), but I’m not sticking absolutely slavishly to it.

Example vocabulary is coming soon.

I’d add ⟨y sh kh⟩. If you want to avoid digraphs for maximum disambiguation, I’d go with ⟨y w z⟩.

Another suggestion: have you considered using ⟨q⟩ for /ʔ/ instead? That would avoid problems with punctuation, and resemble your desired ‘ancient Near East/Fertile Crescent’ look a bit more. (Plus it’s my favourite grapheme–phoneme assignment, insofar as one can be said to have such a thing.)

Some sample words with my suggestions (as generated by gen):


KARSAR
MORPAL
TINAUBUNRA
BARPORGAU
GERSER
SHOTERGUNDE
TINKHAU
KEPAN
KORDANTU
TORSARPALME
TAUKHIN
BALQIL
SHAIPILBI
PULRAN
KALSUNI
PALQIPUL
QORKHIL
KAIROR
LESOR
HERQOR
RULTOKOKI
DUDAN
KANRO
GOTUN
GILPOLAIGABU
TUL
TARU
DULTOR
QALKERHAU

Here’s some sample vocabulary using the consonant set /m n p b t d k ɡ ʔ s z ʃ h ʋ l r/:

dalduho
tundohin
ror‘or
kehesun
banpor
rorherhu
sunvorpo
bunsu
valmigu
sulzulna
garver
tauzaisil
doserhande
herbodor
lasaino
huldalmor
salre
kerkun
shalpu
kazauvan
maltor
pamarkan
rilrul
vorbor
vunginler
tulzo
valshegi
bairun
hukar
tugar
nelehau
mar‘unlin
telun
banmimer
milru
kanko
beli
tallau
shausaivalshor
rinnau
rilnaitun
lodil
zanril
rodan
gidil
vurar
harmau
lenunnil
hun‘ilrilbor
dinse
hegul
halnor
kuntor
gakerger
mipil
ladan
‘akul
ballila
vor‘e
shezun
kana
tabor
naupal
vebausu
rehun
haga
lullil
zulzer
sarvo
bairin
‘omu
vuri
ranshar
parmar
shanshar
gavarger
‘unkilvul
zai‘ersun
la‘ai
narkibau
tautitai
sulvil
bai‘il
varboshul
maushinzai
nulbalzo
puntil
bullai
lalner
bishi
reshupal
pulrul
peshebu
kulzan
lanuser
laibu
mausan
pabudo
maululnil
hadashin

Here’s another set using /m n p b t d k ɡ ʔ ts s x h l r j/:

silter
kaukor
gobor
talru
ha‘in
yalkhikho
saitsaido
soryeyi
pordaga
pilril
tsunmoyil
karmun
sarhi
peper
gautsatiral
baldor‘a
hoge
royan
gekhortil
ralrulsal
raula
letsun
naltau
pertan
sankun
gauyal
rerkar
dikhal
garmer
mil‘u
biyomin
nilsilhe
yutsuntsal
rarner
yeyinlarko
lakhe
‘ilgaltso
runtsaukalkhul
ganhu
meka
lilma
gulgalla
bornartsa
khe‘a
herko
tuyau
pe‘e
pilli
nortsa
tertsulril
khinbil
‘ulkalin
yaiyernai
yunkil
gulyul
dunkhallin
paidargu
morpunhin
sunbunner
gebun
medebun
dogan
dater
gilger
gagober
lillau
nullul
dunmin
‘itsor
sohernan
gairukin
betsul
tsairodudu
khilbanmi
huntul
satsan
gorminil
nergankarto
gorkhar
hindaurar
sutse
yatsu
datan
khunsil
gipi
sorhul
kanral
hulpaiyin
tsusal
roha
kausun
khaikindo
tar‘i
bortsi
rubul
nirunka
rargankhar
yarmun‘a
kiltin
dutsilsor

Also, feel free to give other feedback as well. The set of onset consonants is just the first thing I specifically wanted to ask about. And yes, I realized I forgot to stick to my convention of all caps for romanized text when I set up Awkwords. Oh, well.

If your going for an ancient Near-East feeling, I think you definitely need ʃ. After that I'd suggest maybe x and w. I'd advise against ts and the labiodentals.

Are all phonemes/rimes going to be equally frequent? Because it feels a little heavy with so many closed syllables ending with one of, was it three? consonants.

I'm assuming you don't want players to "remember" spells from the previous game, hence the randomization. But you also want the spells to fit into a language, otherwise why bother with actual words? If the spell is just a randomly generated label, no different than a barcode, you could just say "and then you found the spell! Well done." If you're going to make them pay attention to the words, give them meaning. So, rather than randomly generating strings of syllables, generate basic elements that can be recombined. You could have an elemental system where "fo" is water and "da" is earth, and "dafo" is mud. Or you could have all animal names end in "gur," so that a "fogur" is a fish, and a "dagur" is a worm. You could give verbs predictable endings as well, but different each game. Once the players find a word ending in "an," they know they've found another verb to use in a spell, but otherwise they've found treasure, animals, etc. The pieces could be the same, just with a new pronunciation.

Hi everyone! Sorry for the delayed response!

Moose-tache wrote: ↑Tue Sep 15, 2020 2:00 pmIf you're going to make them pay attention to the words, give them meaning. So, rather than randomly generating strings of syllables, generate basic elements that can be recombined. You could have an elemental system where "fo" is water and "da" is earth, and "dafo" is mud....Once the players find a word ending in "an," they know they've found another verb to use in a spell, but otherwise they've found treasure, animals, etc. The pieces could be the same, just with a new pronunciation.

As it so happens, this is approximately the system I intend to use. There are a stock of common morphemes between spells, and a given morpheme has a consistent appearance within a playthrough. Thus, say, if the “fire” morpheme is shar, all fire-based spells will have names containing shar. I wanted to talk about that system in a bit more depth after we’d discussed the phonology.

Qwynegold wrote: ↑Tue Sep 15, 2020 1:53 pm If your going for an ancient Near-East feeling, I think you definitely need ʃ. After that I'd suggest maybe x and w. I'd advise against ts and the labiodentals.

Noted.

Are all phonemes/rimes going to be equally frequent? Because it feels a little heavy with so many closed syllables ending with one of, was it three? consonants.

Every possible syllable is equally frequent, because I want every possible sequence of syllables to be a unique input so that the language can be used as a randomization seed. I recognize this isn’t quite naturalistic, although I’m open to suggestions to make the language look more naturalistic within this constraint.

This includes revising the rime set. If you can think of a set of rimes that sounds more naturalistic, by all means share it.

Hmm, even if you allow more different consonants in the coda, you're still going to have the same problem that the language is heavy on closed syllables. Since I don't know exactly how the syllable generation works, and I'm terrible at math anyway, it's hard for me to give any advice on that. But you're saying that if you for example generated a random number between 1 and 6, and 1-3 meant no coda, 4 meant /n/, 5 meant /r/ and 6 meant /l/; that wouldn't work for you?