Conlang Random Thread

[jal]

Fed some conlangs to ChatGPT and asked it to analyse them. It was eerily correct.

Start of its comments about Fake Germanic:

The text you provided exhibits features from a constructed language (conlang) or an older Germanic language, incorporating elements reminiscent of Old English, Old High German, and possibly other ancient Germanic languages.

Start of its comments about Fake Latinic:

The text you've provided seems to be constructed with inspiration from Latin, yet it significantly diverges in spelling, vocabulary, and possibly syntax, suggesting a conlang (constructed language) that draws heavily from Latin or Romance languages.

The fact it knows about constructed languages, and can detect its features (it went on with a detailed analysis of vocab, grammar etc.) is pretty impressive.

I also fed it Sajiwan, but forgot to save the output. It correctly detected it as a pidgin/creole, also assuming it's a conlang, and when pushed noted its similarties with Jamaican. It could also give a fairly accurate transalation of a paragraph of the Hobbit.


JAL

[Ares Land]

Oh, that's fun.

I did a bunch of tests, using an a posteriori conlang, and an alternate history Italian. The para-Italian was designed to be very close to actual Italo-Romance which will prove to have very interesting results.

The first sample was a folk song; ChatGPT thought it was Sicilian (not a bad guess). It correctly recognized it as a folk song and then confidently analyzed it as very typical of Sicilian poetry. Bad guess -- I had translated it from a Sephardic love song, in Ladino.

The second sample was a version of Ambarabai ciccì coccò, an Italian counting-out rhyme. ChatGPT identified the language as Neapolitan (again, that works). Its thoughts on it:

The text does not have a clear literal meaning and appears to be a playful or nonsensical phrase commonly found in children's songs or nursery rhymes.

Not bad!
It did not recognize the rhyme, though.

Next I tried an a posteriori conlang. It thought it was Kurdish, very confidently provided a translation and stated that "the text expresses themes of longing, nature, and traditional Kurdish culture, reflecting the rich oral tradition of Kurdish poetry and music."

In all three cases it provided a translation when I asked one. For the a posteriori conlang, the translation was entirely wrong, of course. For alternate Italian, the results were wrong, of course, but not so bad.

I think it uses the Google Translate API at some point in the process - Google Translate guesses exactly the same languages.

I fed it a bit of grammar. Of para-Italian it said:

The text provides a detailed explanation of the third declension in a Romance language, likely Italian or a closely related dialect. Here's a breakdown of the key points:

Completely unfazed by Italian having such a thing as a third declension

It did realize the a posteriori language was fictional and interestingly, from that point on it did realize the para-Italian was a conlang as well!

I fed it the first text again (para-Italian folk song) -- it still thought it was Sicilian. Curiously, the translation it provided was different. (Though again, not a bad one!)

Then I tried a different approach: I fed it the very same song, telling it upfront it was in a conlang:

The vocabulary used in the text doesn't correspond to any known language, indicating that it's either a completely invented lexicon or heavily modified from an existing language.

One last try: using the same song, I tell ChatGPT it's in a conlang, and ask it to translate it.

Since this text is in a fictional language, I can't provide a direct translation since I don't have access to a dictionary or grammar rules for this language. However, I can try to offer a creative interpretation of the text based on its structure and phonetics:

The translation, or interpretation is completely wrong or rather, creative, just as it warned. But rather curiously, one verse is translated correctly.

[Ahzoh]

I came up with a "shopping list" of features I want to conjugate my verbs in:

• Six Tense-Moods (Indicative/Nonfuture, Potential/Future1, Commissive/Future2, Optative, Counterfactual, Jussive)
• Five Voices (Active, Instrumental/Causative, Benefactive, Malefactive, Comparative)
• Three Persons (First, Second, Third)
• Number (Singular, Plural)
• Polarity (Affirmative, Negative)
• Possibly a relativization marker to create attributive verbs

Now the major issue is how I'm going to make these work together while also having all the funny ablaut and stem shape changes one expects from a "triconsonantal root language".
Mostly, I have try to consider the primacy of grammaticalization of certain features, that is, the order that features have become grammaticalized over time. But it's difficult becuase I suck at diachronics/sound change.

What I am certain of is that modality will be the core of the verb complex,with a basic realis-irrealis mood distinction in the verb, indicated with ablaut:
CuC = a dynamic intransitive verb
CiC = a stative intrasitive verb
CaC = a transitive verb

Code: Select all

Realis > Irrealis
CuC    > CaC
CiC    > CaC
CaC    > CaC

Grammatical voice should probably come next in the order of grammaticalization, as a series of "derived" stems like the Semitic languages do.

---

You know I think it's a bit easier to do the kind of things Semitic languages do with their verbs because they have subject and voice markers as prefixes while having right-leaning stress. Like in Akkadian you get lots of stem shapes like -CCVC, -CCaC , -CaCCaC, -CaCCiC, -CtaCaC, -CCaCaC, CaCC-, CaCiC, etc.

It's just ablaut and prosody effects, but the results are very noticeable, marked even. I want something like this but I don't want a Semitoclone and I don't want a language that gives the impression of being a Semitoclone. And yes, there are lot of languages with person prefixes, such as Ojibwe, but when I, say, have ni-pruḫ (1s-speak), ti-pruḫ (1p-speak), ma-pruḫ (2s-speak), ta-pruḫ (3s-speak), it doesn't feel much different than Akkadian a-prus, ni-prus, ta-prus, i-prus. So I that's why I do paruḫ-ni, paruḫ-ti, paruḫ-ma, paruḫ-ta. But then I feel limited somehow in what I can do, even though I don't see why I would be.

[jal]

Had some more fun with ChatGPT yesterday. Wrote a concise conlang grammar and some vocab, then asked ChatGPT to translate English sentences. It's amazing what it can do, but also where it fails. For example, if you do not explicitly state what a vowel is, it assumes no letters are vowels, apparently. :)


JAL

[foxcatdog]

I fed Chatgpt Amarin and it told me its consonants could be found in many languages.

[foxcatdog]

After further teasing i've seen it compared to Tongan, Samoan, Hawaiian, Fijian and Finnish.

[foxcatdog]

Well after putting in most aspects of Amarin it supposedly compares mostly to various Austronesian languages as well as to like every major constructed language.

[foxcatdog]

Lol
Let's make this language head-initial. In a head-initial language, the head of a phrase typically comes before its complements. For example, in a noun phrase (NP), the head noun would precede any adjectives modifying it. Similarly, in a verb phrase (VP), the main verb would come before any auxiliary verbs or complements.

So, let's consider our example language to be head-initial:

Noun Phrases (NP): Determiner (optional) + Adjective (optional) + Noun

Example: "big house" - adjective "big" precedes noun "house"
Verb Phrases (VP): Verb + Object (optional) + Adverb (optional)

Example: "eat quickly" - adverb "quickly" follows verb "eat"
This structure is typical of many languages such as English.

[Ahzoh]

Those posts of yours could probably be coalesced into one with sections separated by [ hr]. Too late now though.

[foxcatdog]

We should probably make a Chatgpt thread

[Ahzoh]

I seem to have troubles visualizing how an applicative voice would work with a language with secundative alignment (like my conlang Vrkhazhian).

Going by the WALS article, an applicative voice has the following properties/constraints:
1) An applicative does not (often/inherently) modify event structure, it only adds participants; this is why it is more common for applicatives to form on trnasitives only
2) It promotes an oblique argument to direct object
3) The applied object is neither an agent nor a theme
4) It follows a semantic role hierarchy of agent > theme (patient) > goal (benefactor, recipient) > location > etc.
5) The base object is "structurally superior" to the applied object

When I consider all that and consider the structure of a secundative alignment (where R=P and T is distinctly marked), I end up feeling like something is wrong and get confused even though I can't really figure out why it should be wrong.

Going from "I-NOM have a ball-ACC" ("I have a ball") to "I-NOM have-APPL them-ACC a ball-INS" ("I give-to them a ball" / "I make-have them a ball") should make sense given those properties above, and yet part of me also feels like it violates something.

As an aside, to make things more complex, I have come to learn that when it comes to ditransitive verbs, there could be two major categories of lexicalized meanings: send-type verbs that have lexicalized meanings of caused motion, and give-type verbs that have lexicalized meanings of caused possession. I think this also has implications for how applicatives might work in my conlang.

[Moose-tache]

In your example the applicative elevates the recipient to the accusative, but in Vrkhazhian, wouldn't that happen anyway, without the applicative?

Sometimes people assume "applicative" is just a thing that you add to a verb, because that's how it works in some famous examples. In Swahili, for example, there is one applicative: -i, and you figure out what it means according to loose rules like the ones you showed above. But some languages have multiple applicatives with specific purposes. For example, most of the Muskogean languages have separate applicatives for benefactive and locative/instrumental. In these languages "applicative" is more a category of things you can do to the verb.

Since your secundative language presumably already promotes indirect objects to direct objects all the time, perhaps applicatives exist to provide more specific information. Maybe a bare verb can promote a recipient to direct object, while an applicative suffix is necessary to promote an instrument.

[Ahzoh]

In your example the applicative elevates the recipient to the accusative, but in Vrkhazhian, wouldn't that happen anyway, without the applicative?

No, there are no or very few bare stem ditransitives. My "give" verb is a "have" verb in the applicative voice.

Sometimes people assume "applicative" is just a thing that you add to a verb

I think most people understand applicatives as a valency-raising operation, like a causative but for introducing participants in roles other than agent. Like dative applicatives, benefactives, locatives, and instrumentals (which often behave identically to causatives; "I make the man write" is identical to "I write with a pen").

My applicative by default behaves like an instrumental applicative or causative, but it can also promote oblique goals/recipients and beneficiaries to primary object.

[Ahzoh]

I think this is correct to how it should be done in Vrkhazhian (Note: I don't have a separate causative, I just put that in there to show the functional equivalence of the instrumental applicative)

[bradrn]

I seem to have troubles visualizing how an applicative voice would work with a language with secundative alignment (like my conlang Vrkhazhian).

In my view, one first thing to get straight is the meaning of ‘voice’. Generally, when I see that term, it refers to some kind of coherent grammatical subsystem, combining notions of valency, semantic roles, focus and aspect. Thus, for instance, many IE languages have a three-way system of active/middle/passive voice, where each one highlights a different argument with different amounts of agency.

Applicative constructions, by and large, don’t behave like this. Instead, like causatives, they tend to be simple derivational affixes, which derive a new verb from an old verb. You can have systems of multiple applicatives, but they don’t really behave as a distinguished subsystem the way ‘voice’ systems do, so I tend to avoid using the term ‘voice’ for them.

(Yes, I know Wikipedia calls it the ‘applicative voice’. Wikipedia is once again wrong.)

My applicative by default behaves like an instrumental applicative or causative, but it can also promote oblique goals/recipients and beneficiaries to primary object.

When I first wrote this post, I was highly sceptical that such a dual applicative/causative morpheme could exist… but then I looked it up, and it seems I was wrong. (The key connection seems to be that ‘X makes Y do V’ is similar to ‘X does V with Y’.)

Looking further, a commonality between all these systems is that the single morpheme has a causative meaning with some verbs, but an applicative meaning with others. Effectively, the result of the applicative construction is lexicalised. I can’t find any case where it is truly indeterminate between both meanings for a single verb.

Thus, going back to your original post:

Going from "I-NOM have a ball-ACC" ("I have a ball") to "I-NOM have-APPL them-ACC a ball-INS" ("I give-to them a ball" / "I make-have them a ball") should make sense given those properties above, and yet part of me also feels like it violates something.

I think the only really confusing thing here is your gloss. You’ve called it an ‘applicative’ here, but clearly it’s not acting as an applicative, but a causative. So the sentence should instead be: “I-NOM have-CAUS them-ACC a ball-INST”, which seems fine to me.

I think this is correct to how it should be done in Vrkhazhian (Note: I don't have a separate causative, I just put that in there to show the functional equivalence of the instrumental applicative)

I’m struggling a bit to make sense of this table. I feel it would be clearer if you thought in terms of valency manipulations: that is to say, which arguments go where in each derived form. (I know I said that ‘voice’ is about more than valency, but valency is still at its core.)

[Ahzoh]

I seem to have troubles visualizing how an applicative voice would work with a language with secundative alignment (like my conlang Vrkhazhian).

In my view, one first thing to get straight is the meaning of ‘voice’. Generally, when I see that term, it refers to some kind of coherent grammatical subsystem, combining notions of valency, semantic roles, focus and aspect. Thus, for instance, many IE languages have a three-way system of active/middle/passive voice, where each one highlights a different argument with different amounts of agency.

Applicative constructions, by and large, don’t behave like this. Instead, like causatives, they tend to be simple derivational affixes, which derive a new verb from an old verb. You can have systems of multiple applicatives, but they don’t really behave as a distinguished subsystem the way ‘voice’ systems do, so I tend to avoid using the term ‘voice’ for them.

(Yes, I know Wikipedia calls it the ‘applicative voice’. Wikipedia is once again wrong.)

Well yes, my causative-applicative is actually a lexical derivation, and not an inflection. It could just be called a "general transitivizer", though I wanted to have multiple derived stems or one derived stem whose function is specified by markers on the accusative noun.

I think the only really confusing thing here is your gloss. You’ve called it an ‘applicative’ here, but clearly it’s not acting as an applicative, but a causative. So the sentence should instead be: “I-NOM have-CAUS them-ACC a ball-INST”, which seems fine to me.

The same morpheme is used for "I-NOM write-APPL/CAUSE pen-ACC letter-INS", you can say "I make the pen write a letter" or "I write the letter using a pen"

I’m struggling a bit to make sense of this table. I feel it would be clearer if you thought in terms of valency manipulations: that is to say, which arguments go where in each derived form. (I know I said that ‘voice’ is about more than valency, but valency is still at its core.)

applied object = direct object = primary object = recipient = causee = patient
base object = indirect object = secondary object = theme

In a normal monotransitive/bivalent verb the theme is the primary object (AKA direct object), in the causative/applicative/whathaveyou ditransitive/trivalent verb the recipient is now the primary object (AKA direct object) while the theme is demoted to secondary object (AKA indirect object).

This is how it works for normal ditransitives in Secundative languages like Greenlandic where "give" is identical to "provide with" or supply with; He give Nisi with money, he supplied Nisi with money, he provide Nisi with money, etc.

But it's hard to extrapolate that understanding to verbs that aren't like give, maybe send-type verbs are different. "Send him (with) a gift" sounds ok, "Send New York (with) him" sounds awkward.

Relatedly, This pdf I read suggest two (likely not only) categories of ditransitive verb, give-type verbs or "verbs of caused possession" and send-type verbs or "verbs of caused motion".

From my understanding, verbs of possession lend themselves well to secundative alignment, while verbs of caused motion lend themselves well to dative alignment

But this is why I call it an applicative, because it's not merely a causative in some circumstances. But it increases valency all the same. Also for this reason why I ponder a split-ditransitive system where some verbs have secundative alignment where the theme takes the instrumental and some verbs have dative alignment where the recipient takes the dative.

[bradrn]

I think the only really confusing thing here is your gloss. You’ve called it an ‘applicative’ here, but clearly it’s not acting as an applicative, but a causative. So the sentence should instead be: “I-NOM have-CAUS them-ACC a ball-INST”, which seems fine to me.

The same morpheme is used for "I-NOM write-APPL/CAUSE pen-ACC letter-INS", you can say "I make the pen write a letter" or "I write the letter using a pen"

Yes, this is one of those in-between cases where you could take either interpretation. (My reading suggests this is called a ‘sociative’.) But, in most other cases, it seems like such an affix must ‘choose’ a single interpretation.

Thus, with ‘I-NOM have-CAUS/APPL them-ACC a ball-INST’, the causative and applicative meanings are quite different:

Causative: ‘I make them have a ball’
Applicative: ‘I have a ball with them’

It seems that you want to choose the causative interpretation in this particular case, but the applicative in others.

(It may be more helpful to think of this kind of affix as in fact being two separate ones, a causative and an applicative, which just so happen to have the same realisation. This is in fact how Dixon describes this situation in his Basic Linguistic Theory.)

This is how it works for normal ditransitives in Secundative languages like Greenlandic where "give" is identical to "provide with" or supply with; He give Nisi with money, he supplied Nisi with money, he provide Nisi with money, etc.

But it's hard to extrapolate that understanding to verbs that aren't like give, maybe send-type verbs are different. "Send him (with) a gift" sounds ok, "Send New York (with) him" sounds awkward.

I don’t really get what you’re saying here. To me, ‘give Nisi with money’ and ‘send Nisi with money’ sound no different. ‘Send New York with him’ does sound awkward, but only because the gloss tricks you into thinking in terms of English, where dative shift is impossible in this particular case. (I believe the PDF you linked describes this case.)

Probably the better way to think about this situation is that, in your language, the verb supplies the destination of the action. This is even clearer in one of my own languages, which uses a serial verb construction for the theme:

Be siwe ŋay paatli Nuyok.
I take it give New.York

From this viewpoint ‘send New York’ makes perfect sense: the object of ‘send’ is the recipient.

Thinking more broadly, this is also an opportunity to reconsider the semantic range of verbs in your language. Do you really need the distinction between ‘give’ and ‘send’? In English it makes sense, because the distinction is in how their objects are affected. But if the object is instead the recipient, maybe the distinction is less relevant. Instead, your primary distinction could be, let’s say, the affectedness of the receiver: trying to think up some examples, your primary distinction could be between verbs like ‘transfer’ (recipient is nearly unaffected), ‘absorb’ (recipient is completely affected), ‘receive’ (focus on the recipient), and so on.

(I believe this is related to the idea of verb framing vs satellite framing, although that generally focusses on verbs of motion, rather than ditransitives.)

Relatedly, This pdf I read suggest two (likely not only) categories of ditransitive verb, give-type verbs or "verbs of caused possession" and send-type verbs or "verbs of caused motion".

This looks quite interesting; I’ll have to read it properly.

Although note that a lot of this is English-specific. Dative shift is a fairly uncommon phenomenon, so it’s hard to know how the phenomena they describe generalise. Just because English ‘send’ has certain restrictions on dative shift, it doesn’t mean that other languages would have that restriction, especially if they don’t have dative shift in the first place. (This also goes back to what I was saying about semantics: you might not need a ‘send’ verb at all.)

[Ahzoh]

Yes, this is one of those in-between cases where you could take either interpretation. (My reading suggests this is called a ‘sociative’.) But, in most other cases, it seems like such an affix must ‘choose’ a single interpretation.

You mean this? https://www.degruyter.com/document/doi/ ... 9-0011/pdf
Though this kind of sociative involves social interaction.

Thus, with ‘I-NOM have-CAUS/APPL them-ACC a ball-INST’, the causative and applicative meanings are quite different:

Causative: ‘I make them have a ball’
Applicative: ‘I have a ball with them’

Becuase of the multifaceted meaning of "with", I always use "using" or "by means of" so it doesn't get confused with the comitative "with" or ornative (possessing) "with".

They don't have to be different. In some languages these are functionally the same, and they are the same in mine. Although I suppose verbs regarding the nature of possession (or statives in general) are strange, where the distinction is more separable than verbs like "write".

(It may be more helpful to think of this kind of affix as in fact being two separate ones, a causative and an applicative, which just so happen to have the same realisation. This is in fact how Dixon describes this situation in his Basic Linguistic Theory.)

Maybe I should think of it as a derivation, like eat > feed. Then it doesn't really matter the underlying syntactic behaviour because it could just be generalized.

I do have for example rabad- "guard, protect" > rabbad- "assign (to guard)".
And it is a modified stem, involving gemination, such that yanud- "hunt_fish\ACT" becomes yannad- "hunt_fish\APPL".

Though unlike "guard" > "assign", sometimes there is less obvious lexical meaning, such as "hunt_fish\APPL spear-ACC" = "I go spearfishing" versus "hunt_fish\APPL son-ACC" ="I made/had my son go fishing".
My language does have split ergativity between animate nouns and inanimate nouns, so I suppose whether it's causative or applicative depends on the primary object's animacy.

This looks quite interesting; I’ll have to read it properly.

Although note that a lot of this is English-specific. Dative shift is a fairly uncommon phenomenon, so it’s hard to know how the phenomena they describe generalise. Just because English ‘send’ has certain restrictions on dative shift, it doesn’t mean that other languages would have that restriction, especially if they don’t have dative shift in the first place. (This also goes back to what I was saying about semantics: you might not need a ‘send’ verb at all.)

The pdf is titled "A Crosslinguistic Verb-sensitive Approach to Dative Verbs" which made me think that it asserts that these are crosslinguistic/meta-properties of some ditransitive verbs and not merely an English-specific analysis of how those verbs work specifically in English.

[bradrn]

Yes, this is one of those in-between cases where you could take either interpretation. (My reading suggests this is called a ‘sociative’.) But, in most other cases, it seems like such an affix must ‘choose’ a single interpretation.

You mean this? https://www.degruyter.com/document/doi/ ... 9-0011/pdf
Though this kind of sociative involves social interaction.

No, that’s completely different. I’m talking about the ‘sociative’ category introduced in Shibatani & Pardeshi’s chapter The Causative Continuum (in The Grammar of Causation and Interpersonal Manipulation, ed. Shibatani).

EDIT: actually, no, taking a second look it might actually be the same category. Hard to know as I can’t seem to access the article.

Thus, with ‘I-NOM have-CAUS/APPL them-ACC a ball-INST’, the causative and applicative meanings are quite different:

Causative: ‘I make them have a ball’
Applicative: ‘I have a ball with them’

Becuase of the multifaceted meaning of "with", I always use "using" or "by means of" so it doesn't get confused with the comitative "with" or ornative (possessing) "with".

I intended the comitative reading there, since that’s the applicative function which tends to be closest to causatives (again cf. sociatives).

Though really, for my example it doesn’t really matter which applicative you choose (comitative or benefactive or instrumental or other); they’re all quite different from the causative, here.

(It may be more helpful to think of this kind of affix as in fact being two separate ones, a causative and an applicative, which just so happen to have the same realisation. This is in fact how Dixon describes this situation in his Basic Linguistic Theory.)

Maybe I should think of it as a derivation, like eat > feed.

Yes, this is precisely what I’m getting at.

My language does have split ergativity between animate nouns and inanimate nouns, so I suppose whether it's causative or applicative depends on the primary object's animacy.

I don’t really see why those two would have anything to do with each other.

This looks quite interesting; I’ll have to read it properly.

Although note that a lot of this is English-specific. Dative shift is a fairly uncommon phenomenon, so it’s hard to know how the phenomena they describe generalise. Just because English ‘send’ has certain restrictions on dative shift, it doesn’t mean that other languages would have that restriction, especially if they don’t have dative shift in the first place. (This also goes back to what I was saying about semantics: you might not need a ‘send’ verb at all.)

The pdf is titled "A Crosslinguistic Verb-sensitive Approach to Dative Verbs" which made me think that it asserts that these are crosslinguistic/meta-properties of some ditransitive verbs and not merely an English-specific analysis of how those verbs work specifically in English.

Yeah, you’d think so, but I don’t see much cross-linguistic exploration in that article… every single example seems to be from English.

[Ahzoh]

I intended the comitative reading there, since that’s the applicative function which tends to be closest to causatives (again cf. sociatives).

That can't be correct.That doesn't make sense. The instrumental applicative should be what is closest to a causative since an instrument is semantically closer to a causee (the thing being caused to do something). And a causative is defined by a causer affecting a causee that results in the causee doing something.

"Agent does using/by means of instrument" is somewhat identical to "agent makes instrument do" which in turn is functionally identical to "causer makes causee do".

A comitative applicative would imply a co-action and co-actor(s) which a causative does not inherently imply ("fight alongside with" ≠ "make fight/provoke"). Moreover, a comitative applicative would thus be more akin to a reciprocal ("fight alongside with" ≈ "fight with/against")

This is also why the papers people write on this topic talk about how languages with instrumental applicatives have those applicatives behave identically to, and be conflated with, causatives.

My language does have split ergativity between animate nouns and inanimate nouns, so I suppose whether it's causative or applicative depends on the primary object's animacy.

I don’t really see why those two would have anything to do with each other.

The instrument semantic role would be assigned to the less animate thing while the causee semantic role would be assigned to the more animate thing.