Goptjaam (欱攙): the unglossable* language (Cursed Conlang Circus 2)

Kjaangs piës dwen pip lauw kat.
This video is a submission to (and 1st place winner of) @AgmaSchwa 's Cursed Conlang Circus 2 contest: • Who Can Make The Most ...
Watch the judges' reaction here: • Agma Schwa's CURSED CO...
The git repo: gitlab.com/katmistberg/003-go...
*Unglossable using a natural human language
(The part about Sino-Dutch is a joke. Dutch did not systematically borrow words from Middle Chinese.)
================================================
Sections:
00:00 Premise
01:13 Intro
01:21 The basic idea
06:32 Converting to linear algebra
09:16 Abstract linear algebra
15:07 Constructing the base matrix
22:06 Actually constructing the language
26:46 Grammar
30:29 How cursed is this language?
33:33 Opening lines of the Bee Movie
================================================
Assets used:
Irasutoya (in order, I think):
- Tree: www.irasutoya.com/2014/09/blo...
- Fish: www.irasutoya.com/2014/06/blo...
- Cat: www.irasutoya.com/2019/10/blo...
- Eating: www.irasutoya.com/2015/08/blo...
- Cat eating fish: www.irasutoya.com/2016/06/blo...
- Headphones: www.irasutoya.com/2014/05/blo...
- Smirking man: www.irasutoya.com/2013/03/blo...
- Hammer: www.irasutoya.com/2014/04/blo...
- Flying toki: www.irasutoya.com/2020/01/blo...
- Bear: www.irasutoya.com/2016/07/blo...
- Crying: www.irasutoya.com/2014/05/blo...
- Running: www.irasutoya.com/2016/08/blo...
- Chonky: www.irasutoya.com/2018/11/blo...
- Exchange: www.irasutoya.com/2017/12/blo...
Other visual (in order):
- The Rock eyebrow raise: www.tiktok.com/@therock/video...
- Phonology: en.wikipedia.org/wiki/Dutch_p... Morphology: en.wikipedia.org/wiki/Swahili...
- Syntax: By Traced by Stannered - Own work based on: ParseTree.jpg and ParseTree.jpg, Public Domain, commons.wikimedia.org/w/index...
- Fish cat eat meme: 9gag.com/gag/aNgDmOb , / i10sep
- Two buttons meme: jake-clark.tumblr.com/post/10...
- Disintegrating sad emoji: knowyourmeme.com/photos/23699...
- World map: commons.wikimedia.org/wiki/Fi...
- Tang area 742 CE: en.wikipedia.org/wiki/Tang_dy...
- 廣韻: commons.wikimedia.org/wiki/Fi...
- Character for trhjwop: en.glyphwiki.org/wiki/u26463
- Character for kjon: en.glyphwiki.org/wiki/u25d24
- Bee: media.distractify.com/brand-i...
BGM from DOVA-SYNDROME (in order):
- 木陰にて by shimtone: dova-s.jp/bgm/play14849.html
- 激辛ロック by shimtone: dova-s.jp/bgm/play19539.html
- Dianthus by shimtone: dova-s.jp/bgm/play19490.html
- 優しく温かく by shimtone: dova-s.jp/bgm/play19455.html
- アイスクリーム by Addpico: dova-s.jp/bgm/play11777.html
================================================
*Links:*
Mastodon: mathstodon.xyz/@KatMistberg
Gitlab: gitlab.com/katmistberg
Twitter: / katmistberg
================================================
#CCC2 #CursedConlangCircus2 #conlang

[@GLu-tb1pb]

English: Oh I mispelled a single letter, I'll probably be okay Goptjaam: My jam is ugly and sweaters will fly on mars

[@itisALWAYSR.A.]

"Sorry, this is a bad phone line, could you repeat that?" "In a very literal sense: no"

[@chil.6476]

This probably just broke my record for the ratio between how badly i want to share a video and how few friends would understand it. Incredible.

[@mothra3477]

This

[@majejejenta]

the best friends are the ones that are happy that you vibe to this eventhough they don't get shit. i share this and many indecipherable stuff to my friends, and i love every "what tf is this" that they send

[@davidszelbanszkhy9697]

same

[@chickennugget481]

I'm gonna send it to my friends and I don't even understand it

[@babylonastronaut]

I have a friend who is a linguist and I'm currently torturing them with these videos lol

[@blumoogle2901]

The most cursed thing is that you are throwing any Humanities major who wants to study this into having to take pure Math as a prerequisite

[@amadeosendiulo2137]

A good one xD But in the real world various faculties mix too.

[@HaramGuys]

if u are serious about linguistics and want to go grad school, you should know these math

[@EchoLog]

​@@amadeosendiulo2137 this is true. Working interdisciplinarily is like having addition, multiplication, exponents, etc instead of just succession. A jack of all trades may not be a master of any of em, but at least he's not ignorant about that! I'd rather know when someone's smarter than me than think my one magic skill can hammer away every problem-nail I find.

[@ninny65]

Extra credit

[@shimrrashai-rc8fq]

As someone who has studied both kinds of field extensively, this is exactly my jam! I'm lovin' it. (Though of course completely impossible to speak, unless you're an AI, perhaps, and interesting seems to suggest that the boundary between "cipher" and "language" is perhaps not as clear as one might think!)

[@SurfTheSkyline]

What i love about the circus is that there are so many ways to make a language cursed that I had never considered. Obliterating intended meaning altogether with a single altered syllable in a long sentence is definitely a contender for as cursed as it gets

[@wilh3lmmusic]

> Obliterating intended meaning altogether with a single altered syllable in a long sentence As opposed to Ithkuil, where changing one syllable only ranges from partially messing up a word to destroying a word without affecting the rest of the sentence

[@livedandletdie]

​@@wilh3lmmusic I mean you can't say that Ithkuil is messed up, it's just that it's roots, and grammar make things a bit too effed up. Some things are easy to convey the meanings of and other things are there to make you want to punch yourself in the face. Attàlûk.

[@drdca8263]

@@whannabiI also doubt it, but, if the language was chosen instead to use a lower triangular matrix, like the transpose of what was dismissed as being not dense enough... well, in that case, the first k syllables would only depend on the concepts used in the first k cases, then that might be *somewhat* learnable?? Like, one could at least learn to recognize something about the meaning of the first word in a sentence, and then like, one could maybe learn how the meaning of the second word depends on the first, and then how the third depends on the first 2, etc. ?

[@sponge1234ify]

@@whannabi ah yes, I love teaching *ring matrix multiplication* to a baby before they can speak

[@angelmendez-rivera351]

@@sponge1234ifyHell yeah.

[@hassibkarim3133]

the most cured thing about this language is that you have no reasonable way to figure out what someone is saying unless you have formed the exact sentence in the past. To decode someone's sentence, you need to memorize the base matrix, the number-syllable mappings, and the concept-syllable mappings. Then you need to transcribe the sentence, convert it to vector, invert, convert it to the concept, and convert it back to a human language sentence while keeping in mind the case tree to figure out how the concepts go together. A conversation using this would only really be feasible through a computer, and in that case, the language is basically just an encryption protocol. Great work coming up with this abomination of a language lol

[@defenestrated23]

holy crap, you have to rely on hash collisions in order to intelligibly understand what was said.

[@icodestuff6241]

this is actually how transformers like chatgpt work, this conlang is essentially just embeddings, which is really practical

[@ianlucas6635]

Technically, you don't need to memorize the base matrix because there was that formula for it where you compute A from B, D, and I, all of which are very simple matrices which are very easy to be memorized. If you do take this route to comprehension, the only other prerequisite is being *very good* at mental matrix computations :)

[@sage5296]

@@defenestrated23 Nah, the matrix is invertible so you can multiply by the inverse to get the original concept vector back, but uhhh good luck doing that on the fly LOL

[@ScienceMeetsFiction]

And people talked about needing to know matrix math for *my* language! This one dialed it up to 11. This is a truly impressive video, which would have been at home in SoME3 just as much as CCC2.

[@KatMistberg]

Thanks! Your time traveler conlang was great too, I quite liked it. Actually, the original plan was to submit this one video to both contests, but it ended up taking too much time and I missed the SoME3 deadline :P

[@asdfghyter]

@@KatMistberg you should add the SoME3 hashtag anyways (unless that breaks some rules), as it was intended as a submission and it helps people find it. 😁 (and because it's fun to see those hashtags together)

[@aedeatia]

Funnily enough, I had this video recommended to me after watching a few videos from the SoME3 playlist.

[@Syuvinya]

there's a surprisingly large overlap between math enthusiasts and conlang enthusiasts

[@Pteromandias]

What's your language, Finnish?

[@ericolson8544]

This feels like the kind of language that 2 AIs would develop to talk to each other without the pathetic humans eavesdropping.

[@m3tr0idgrl]

Just using english is easier

[@angelmendez-rivera351]

@@m3tr0idgrlNot for an AI

[@sponge1234ify]

@@m3tr0idgrl "Without the pathetic humans eavesdropping"

[@hi-i-am-atan]

@@sponge1234ify i mean strictly speaking she ain't _wrong,_ since the ai would be using a language transposed into binary data, like all these comments are, but without the need of transposing it _back_ into human-readable language. humans would still be able to do the latter by running the conversations through an encoder ... assuming the ai _used_ a human-developed encoding, but since the messages wouldn't be intended for humans, the ai could just come up with a pretty basic one and good luck cracking _that_ when all you have to work with is an unceasing stream of 0s and 1s

[@EnginAtik]

If 2 AI use single word sentences often enough while talking to each other, they will reveal their base matrix to pesky humans.

[@emv...]

As somebody studying both Mathematics and Linguistics in university I must say that this is literally the most beautiful thing I have ever seen

[@bojandam963]

I was about to say the same thing lol

[@emv...]

@@bojandam963 I am not alone?? omg how cool:D! How long have you been studying them?

[@mastershooter64]

​​@@emv...been studying mathematics for about 2 years, been studying linuguistics for about the past 2 minutes lol i have been interested in linguistics since a couple months ago but never found the time to actually sit down and go through a proper textbook

[@emv...]

@@mastershooter64 amazing! i've been studying both for a year. linguistics is really interesting once you get into it, i really recommend watching more conlang-related vids for easy content, since they do touch a lot of linguistic concepts and really broaden one's horizon to everything that's out there:)!

[@zoc2]

Me too! I'm taking linear algebra and phonology this sem :3 next sem I'll take syntax and probably mostly computer science classes, bc that's my real major :) diff eqs over the summer and one more upper div elective and I've got my math minor!

[@Tudsamfa]

Finally, a language that is incomprehensible if you don't listen to the whole conversation, or mishear something slightly. Don't you hate it when you have to stop gossiping since the person you talk about walked in? Not any more! Worried somebody is eavesdropping? Not when a slight muffling effect will make them hear complete gibberish. Truly ideal for any office space.

[@cameronhunt5967]

When I realized at 9:00 that this language is an encryption from world space to language space I laughed for like a minute straight

[@MenkoDany]

exactly!

[@endmark_3447]

this is such a perfect but deranged sounding description of what the language is that it genuinely sent me into a fit of giggles

[@stevend285]

Quite literally an unlearnable language, unless you can memorize a 4kx4k matrix and do fast modular matrix vector multiplication in your head. In which case, this would be fun to learn.

[@paulkanja]

Plot twist: this is zuckerburg's mother tongue

[@stevend285]

@@paulkanja nah, this is the language of Paul Erdos or someone like that

[@milkwater1204]

Not only that; to understand other speakers of the language, you would need to know how to invert the 4kx4k matrix and operate on the sentence fluently.

[@paulkanja]

@@milkwater1204 and you'd need to figure out how often they operate n sentences. DO they operate on 1 sentence at a time, or 1 phrase at a time, or full paragraphs at a time.

[@bacicinvatteneaca]

@@paulkanja i think an end marker is needed.

[@eddierosenblum1218]

I love how you still recite it in 7 syllable chunks like it's an actual chinese poem despite any group of 7 syllables having a completely unrelated meaning on their own

[@hughcaldwell1034]

So I wasn't just imagining that it had the texture of a jueju...

[@AndreaColombo-fx1wh]

And he can speak Chinese... And Old Chinese And Middle Chinese

[@ShankarSivarajan]

The "Sino-Dutch" thing is great! It reminds me of the Finno-Korean Hyperwar.

[@robinder_]

is this sino-dutch thing some sort of meme/inside joke

[@jackkennedy98]

I completely believed it for a second

[@dominicpancella3012]

​@@robinder_Alternate history where the Dutch replace either the Portuguese (Macau) or English (Hong Kong) on the Chinese mainland? I don't know

[@estherbrown4084]

One alternative universe regarding Sino-Dutch (and other Sino-European art-languages for that matter): Some Europeans, who disapproved of the Christianization of Europe by force in even the slightest bit, got transported somehow into a Taoist abode and a Buddhist abode with a portal between them. Such Europeans come up with syllables and writing systems for reading the Classical Chinese literature in at least the manner of ruby characters or the Japanese language's furigana. They have also access to at least the Tibetan and Kawi writing systems and all of the books Plethon had as well.

[@pianojay5146]

I feel that each cursed conlang is turning into elaborate incryption methods

[@erkinalp]

Congrats, you invented a language that is both comprehensible and easily speakable by computers, but difficult to learn for humans.

[@Pramerios]

"Difficult"

[@Frau_Brotchen]

@@Pramerios ?

[@VanByyrg]

@@Pramerios Found a bot

[@JakobLastname]

This language feels like the unholy spawn of Ithkuil and Lojban. Truly cursed and also ingenious. Absolutely diabolical.

[@StrategicGamesEtc]

Ithkuil and lojban are both way less cursed than this

[@JakobLastname]

@@StrategicGamesEtc I know. This is like if Lojban and Ithkuil had a horrifically deformed child

[@drdca8263]

Ithkuil, Lojban, and the esolang “Malbolge”.

[@StrategicGamesEtc]

@@drdca8263 yeah, this is nice Malbolge, though. It tries to make "cracking" the "encryption" "relatively" easy.

[@Apostate_ofmind]

Not my poor Lojban 😭😭😭😭

[@MooImABunny]

when you're not sure if you should submit to the cursed conlang circus or SoME3

[@tomkerruish2982]

There's another comment along these lines. The poster responded that they wanted to also submit it for #SOME3 but missed the deadline.

[@warpspeedscp]

We're going to need checksums to make sure info is correctly conveyed!

[@coarse_snad]

You know the language is good when you need a checksum to know if you heard someone correctly.

[@sponge1234ify]

...Unless you're speaking to a person of lower status, in that case checksum is implied.

[@nio804]

This brings to mind large language models, but the model is the language.

[@sponge1234ify]

Hmmmm today I will learn ChatGPT's language "You mean learn language with ChatGPT, right?" No, *ChatGPT's Language.*

[@unlomtrash]

@@sponge1234ify hahahahah xDDDD

[@rkvkydqf]

Seriously, can some mad (data) scientist actually fit this conlang for a corpus of English so that you can actually translate any English sentence into it?

[@adakarga]

a large language.

[@kepler_22b83]

​@@rkvkydqf Nah... I'm too lazy for this. Maybe, someday, when I'm bored enough? Perhaps someone else is up to it?

[@orangereplyer]

I love that you can't really tell whether someone has finished a sentence, ever, since to say many things, one after the other, you can't jut say them sequentially over time. This would make for a great Borges story concept.

[@voxelsofsorrow]

I literally said "oh no" out loud when I realized where you were heading with this. Like, I could handle the idea of using embeddings, but the sheer horror of sentences being spoken vectors just got me. arithmetic coding is also making me shiver.

[@nice3294]

This has got to be my favorite CCC entry, I didn't even register this was a conlang video, I just saw what looked like a math vid thumbnail and clicked it. Loved the math, loved the cursedness of this.

[@thewhitefalcon8539]

this talk could plausibly be at Chaos Communication Congress - if they have a parody section

[@peregrineperry]

in my opinion, this is more cursed than babelingua's seraphim. i mean,,, infinite vectors??? infinite base matrix?? infinite cases?? i have a migraine

[@alexeyvlasenko6622]

This is great. Seems more like an encryption algorithm than a language!

[@U20E0]

this could probably be used to hide the presence of information pretty well.

[@abdillahahmad7025]

It's a cipher that takes in pure information as the plaintext instead of a string

[@ayamii37]

now i kind of want to make this into a working program-

[@OMGYavani]

This is the best cursed language ever. The fact that cryptographic concepts are involved really proves that it is the utmost detriment to communication. I love both maths and conlaging so this video really struck a chord - literal euphoria when diffusion was showcased in the final language 💕

[@AJMansfield1]

Dammit, now I'm gonna have to actually go finish the cursed conlang I've been working on based on Reed-Solomon error correcting codes...

[@the11382]

Cyberpunk fan?

[@binathiessen4920]

Finally a cryptographically secure language. It would be more practical if you worked in some form of error correcting code, but who are we kidding, the point was never to be practicality. Because of the way sentences combine by expanding the tree, and the modulus of the language being finite, only finite texts are possible, because eventually the case multiplier will alias.

[@drdca8263]

Huh? I don’t follow. For any finite sequence of concept-IDs (some of which may be 0), we can apply the matrix to it, and then truncate the result, and to get back the original list we just apply the inverse matrix? (And truncate) The entries in the vector to be encoded are always less than p, nothing wrapping around, right? Edit: ah, I think you maybe thought “we multiply a vector by the case value”, but the case value just tells us which column to use. The multiplication is multiplication by concept-id .

[@ShamelessDuck]

This feels like trying to build a house out of gauge blocks. Edit: No, actually it feels more like building a house out of 4d printers which generate a model of what a person needs based on the hash function which take in the request and the exact state of the universe at the moment of request, so that every time you want to sleep an object gets printed out of thin air which isn't really a bed, but you can sleep on it alright.

[@MuradBeybalaev]

I'll opt to live in the Tower of Babel instead, thank you very much. I laughed my ass off when you explained math and just when it felt like you're about to return to unmath - you instead warn us that the following two sections go deeper into math.

[@bloodspatteredguitar]

I've never come across Hokkien, but it sounds like its word for "cat" is "meow", so I now want to learn it.

[@zoeyaurem9285]

utterly incomprehensible, I love it 10/10

[@jaxsyntax]

1:03 "What if we tried to break this concept" "What if we tried to break it using linear algebra"

[@catgirlQueer]

bra-fucking-vo, holy shit the sino-dutch reading part is, horribly cursed, and that's not even the most cursed part of this entire thing

[@biglegs]

Even without the language, I have to respect this video just on the quality

[@LudoCrypt]

This is one of the most fascinating ways to not only structure a language but the process of creating the language at all. This is inspired. I love this

[@calvincrady]

- About half of the initials' pronunciations are duplicates - About half of the finals' pronunciations are duplicates - Each spoken syllable has about 4 possible numbers it could correspond to - Each spoken sentence of N syllables has on the order of 4^N possible interpretations, not counting homophones in the original Middle Chinese - Because of the confusion property, all ~4^N interpretations mean something completely different

[@the11382]

Imagine giving a speach and everyone in the room has a significantly different interpretation. Nobody can agree.

[@StuffandThings_]

@@the11382 English majors would be having a field day

[@aloysiuskurnia7643]

OK just read the entirety of this video. This is practically a linguist's equivalent of Malbolge - Perfectly logical, can be constructed--forwards and backwards, small change spreads everywhere, unreadable by this puny human mind.

[@selladore4911]

what's a Malbolge?

[@splittydev]

@@selladore4911Malbolge is an esolang (esoteric programming language)

[@Bobbias]

​@@selladore4911Malbolge is a programming language so cursed the author never wrote a single program in it. The first program want even written by a human, but found using an automated search algorithm. It was found 2 years after the language was publicly released. We're still not sure if it's actually Turing complete, meaning we're not sure if it's actually capable of computing all computable answers, something that is a requirement for any general purpose programming language.

[@selladore4911]

@@Bobbias i like your use of the verb "found" when talking about how that first program came to be. cursed indeed. i don't know much about programming but the fact that we even know what answers are computable or not is interesting

[@Bobbias]

@@selladore4911 they literally used a search algorithm, one called a beam search (it has a wiki article). We know what is computable and what isn't, but we don't know if Malbolge is actually capable of computing all computable answers or if it's more limited. An example of a problem that's not computable is what is called the "halting problem". If you're interested there are quite a few good videos on KZbin about it.

[@spok_real]

This is the coolest hash algorithm I've ever seen, it works on ideas instead of files and it makes language instead of data

[@norude]

No. The author was specifically creating something that can be decoded to the original text. Hashes are specifically created as a one way function. This is more of a cipher

[@RainShadow-yi3xr]

OK, what the hell is this, a few of the entries I've seen have been really clever and interesting almost to the point of not being cursed. This is not that. This is extremely cool. And extremely cursed. good job!

[@aiocafea]

everyone can say 'i want many sounds' or 'i want an impossible phonology' but tHIS? this is a masterpiece

[@paulamarina04]

this might honestly be my favorite for this years ccc so far. everything about this is so well put together, i love it!!! honestly i might take inspiration from the tree-case part for my serious engineered conlang projects!!! ive actually been searching for something like that for quite a while good luck on the contest, i hope you win!!! edit: YES YOU DID WIN LETS GO

[@carloselfrancos7205]

I think it's so good that it's not really a cursed language anymore. As a fellow math and conlanging enthusiast, this is exactly my type of content

[@gaeel330]

Would a truncated utterance "make sense" in Goptjaam? If so, it could be fun to see what the intermediate utterances mean as each word is spoken. In most natural languages, it's possible to finish somepne else's sentence because the beginning of the sentence gives enough context to be able to guess how it might end, but in Goptjaam, you need to wait until the utterance is complete.

[@blahblah3347]

This could be poetry in Goptjaam, when the meaning morphs beautifully with each syllable until end of utterance.

[@gaeel330]

@@blahblah3347 Oh, I really like this idea! Riding the ambiguity out. Every new syllable telling a new story. Would be a real challenge to compose though!

[@niklasbrandt7415]

​@@gaeel330This should be possible, if computer assisted. You just have to get a list with all the options that the next sillable offers, and pick the one that is nicest, though I doubt that you'd really be able to plan out what your poetry will be about. If I find the time I might go and program a poetry writing assistant

[@nerdycatgamer]

this is like end-to-end encryption on a spoken language

[@Chiken1]

and people say math isn’t it’s own language

[@MQTate]

No idea why the algorithm decided to bless me this video, but I'm so glad it did. Amazing work man

[@minamozna]

i study abstract math and i love linguistics, you just made my night! also, #SoME3

[@jacool2565]

23:09 biggest plot twist in History lol I wasn't expecting that

[@panzerschreck4906]

eerm... whats the plot twist???

[@thelxinoe5293]

Amazing entry! One of my favourite cursed conlangs that ive seen this year!

[@Sqone1]

great video! btw as a dutch speaker you basically nailed that dutch accent lol (actually like how did you do that do you speak dutch or)

[@KatMistberg]

ja ik spreek nederlands (maar niet als moedertaal) :)

[@Sqone1]

@@KatMistberg knap hoor!

[@victorwindahl4903]

​@@KatMistbergI found the fact of middle chinese spreading and influencing dutch really cool! But I couldn't find any good texts on the subject. Could you share some sources for further reading?

[@indoorcoyote]

@@victorwindahl4903 it was a joke, dutch vocabulary is not actually 50% sinitic and 30% tamil : p

[@victorwindahl4903]

@@indoorcoyote haha 🤣 that explains things! That it was believable says something about dutch!

[@talitek]

Absolutely vile, I love it

[@glitchy9613]

Just finished watching it, as a nerd of both math and linguistics, this conlang is a dream come true.

[@notpencil1338]

no way is this really glitchy from all dimensions wiki

[@crep50]

God damnit, even when i try to be normal for once i get hit with “LINEAR ALGEBRA”. Yes i will be watching all of this. No i wont regret it.

[@Commentwrecker]

Congrats!

[@aloysiuskurnia7643]

I was watching this video seriously with full enthusiasm and maximum nerdery, and then when 23:48 arrives my sides just ROT13'd itself XDDDDDDD

[@pikapuffin368]

This is cursed in a way similar to last year’s winner: this is not a language a human uses to communicate. This is a language a machine community would use. 11/10.

[@user-xi6by2we2i]

I cannot even describe how much I love this. Amazing work!!

[@chickengogo1683]

linguistics, cryptography, abstract linear algebra, history...genuinely insane video

[@milkwater1204]

You've taken the main difficulty of learning a logographic language, exacerbated it, and applied it to a spoken language instead. Bravo. Edit: After finishing the video, it also occurs to me that in order to speak this language, you have to have precognition of what you are precisely going to say before you say it.

[@warpspeedscp]

Youd need to form each sentence in its entirity before expressing it, and if you change your mind well, youll have to redo it all

[@milkwater1204]

@@warpspeedscp it would be a good cipher

[@warpspeedscp]

@@milkwater1204 indeed, since both you and the other would have to agree on a particular set of matrices first. Good otp.

[@sealpiercing8476]

This is incredible. It's also my first exposure to Cursed Conlang Circus so I'm really excited to see what else people came up with.

[@Wielorybkek]

That's one of the most awesome video about languages I've seen! Great content!

[@leo9463065]

Using binary tree for both the syntax and generating case number is genius. One slight nitpicking though: the concept referenced by relativizer doesn't seem to be unique, since you can just keep going up and taking the parent node. Maybe adding a rule or for that, like adding a concept that marks the concept in the tree being modified?

[@KatMistberg]

How I envisioned it working is that the relativizer references its grandparent node if the relativizer has no children (i.e. if it is a noun), and it references its parent node if it has children (i.e. if it is a verb). Maybe I didn't explain it clearly enough in the video. But I believe that way there is no ambiguity there. Let me know if that's not what you meant :)

[@leo9463065]

@@KatMistberg ah that would solve the issue. Further, I think that would make the tree of the sentences with center embedding pretty interesting, like "The rat the cat the dog bit chased escaped. " and the tree just keep growing on one side. That would have some interesting consequences in this conlang due to large amount of the skipped case number. Like, since every noun retlativizer add 2 layers, I think for that sentence above, despite only using 8 nodes, have 6 layers and max case number of 17.

[@KatMistberg]

Yeah that's pretty interesting! If there are n concepts, each of which is the left child of the previous concept, then the number of Goptjaam syllables would be 2^(n-1), which is kinda crazy.

[@kaedotmoe]

Poetry and wordplay in this language would go insanely hard though. Like you can make rhyming phrases that have radically different meanings. It would be practically impossible to come up with these intricate rhymes, but the language in general is impossible for any speaker to produce on the spot anyway

[@rkvkydqf]

I can't even begin to describe how perfect this video is... This is the most beautiful and cursed thing ever to grace mine eyes... Thank you!

[@coarse_snad]

Assume you could trivially understand this language. If you are reading a book written in this, you would not understand anything until you finished the book.

[@oddbirdMusic]

Finally, a lossless language.

[@cmyk8964]

Aw darn, I was expecting each word position to have its vector multiplied by a power of 2, so that the sentence “the cat eats the fish” might be V(cat) + 2V(fish) + 4V(eat), “the fish eats the cat” might be 2V(cat) + V(fish) + 4V(eat), and perhaps even, “the cat fishes for something to eat” might be V(cat) + 4V(eat) + 2V(fish)?

[@cmyk8964]

I was, to put it simply, not galaxy brain linear algebra-pilled enough.

[@KatMistberg]

That's a decent idea, and I did consider it, but it doesn't work as well as how I have it in the video. If we want to be able to have arbitrarily long sentences, since there are only a finite number of syllables (p = 3877), you end up eventually running out of powers of 2 to use. It's a similar thing to how (p+2)V(cat) = 2V(cat) in arithmetic with modulus p, so the maximum number of concepts in a sentence is p. (Of course, in practice that will not be a problem, but I didn't want any restrictions on the lengths of sentences, and having infinitely many cases is really cool, so I ended up not doing it like this :))

[@fortidogi8620]

For how messed up of a concept it introduces, you made this video really, impressively accessible. Thanks!

[@05degrees]

This is actually incredible; also neat a posteriori elements!

[@JohnSmith-of2gu]

This is beautiful madness, so much creativity and work gone into it! I would love to see a separate video just on this alternate universe Sino-Dutch. You created that monstrous creole and it's a mere footnote!

[@johannbauer2863]

dude, this is an insanely amazing concept :D Not very practical, but how it disregards / is incompatible with some concepts of linguistics is amazing and innovative XD!

[@dymaxion3988]

Watching this feels like SCP-3125 using the 4th dimension to reach around my skull and irreversibly imprint its reality-altering alien worldview on my brain. Great work!

[@txikitofandango]

This is the kind of idea I can't believe I didn't think of myself! It's so simple and twisted. And well explained and explored in this video, nice job

[@TheArtOfBeingANerd]

I applaud you. Couldn't really find any math videos that looked interesting and havent watch a handful of times, so i settled on what i thought would just be another cursed conlang. I am impressed

[@Comonad]

Incredible! Thank you so much for this!

[@hwangsaessi2335]

Great stuff! Something that came to mind is a connection to error correction coding as used in digital communications. Having the information be diffuse can actually be useful against "burst errors". In the case of natural languages, if a whole morpheme/word is erroneously transmitted it can in principle make the message totally incomprehensible or wrong. Natural languages can handle this by having a large degree of phonetic redundancy + things like context, but ignoring that, your diffuse language technically allows for robustness to this type of error. Not without building in redundancy, but that should be doable by adding some sort of parity checks, most of which is also "just" some linear algebra. Of course at this point, your language contains no redundancy so a wrong syllable is even more catastrophic than in natural languages lol. Anyway Im not knowledgeable enough to give more concrete recommendations but it could be something for a possible future iteration of this language.

[@KatMistberg]

Yeah, definitely! I did think of the idea of adding error correction, which can be implemented easily as truncating the base matrix not as a square, but as a rectangle that is longer than it is wide; or, using a square truncated matrix but then afterwards multiplying it by a rectangular matrix. I didn't end up doing it because that wasn't the main point of the conlang (and I also ran out of time), but yeah, maybe in a future conlang? :P

[@zidanez21]

I can’t believe someone was able to mix my two neediest hobbies Conlang and Maths, probably one of the best videos I’ve seen on KZbin ever

[@jaysonbunnell8097]

This is hilarious and happens to be a perfect intersection of my hobbies lol. Will be sharing this with my math-nerd-linguist-friends and starring the repo

[@cee_yarr3177]

Wait, is this a CCC2 entry or a SOME3 entry?

[@adiaphoros6842]

Why not both?

[@Huzairhzrm]

Yes

[@damnerd]

I haven't seen _all_ #CCC2 videos, but this one is the winner for sure. Congratulations on the great (?) work!

[@maigowang]

Oh my god what ecstasy it is to encounter a video that's such a feast of math and linguistics and realize I understood all of it!

[@dyllanmccreary688]

Truly incredible perspective on languages!

[@jerryplayz101]

How on earth are you so good at pronunciating all these languages? And if you can speak them, that's amazing!

[@damian_madmansnest]

When one can pronounce Hokkien and Dutch at native level, the ability to pronounce all other world languages is unlocked.

[@abarette_]

@@damian_madmansnest meanwhile xhosian language families lurking in the corner

[@joseftrojan7664]

Good job! I love this!

[@jordensjunger]

this is one of the coolest things i've seen while still being totally cursed 😍

[@extrachipper]

The plot of Arrival 2: Once the Heptapods left, the Goptjaapods showed up and our linguists all collectively quit their jobs. In the end, it was the mathematicians that saved us.

[@iantaakalla8180]

The lesson of Arrival 2 is that every single moment you have is important, so cherish every aspect of your life (a lesson neatly encapsulated by this cursed language).

[@este_2013]

This was certainly a trip. I enjoyed every second of it.

[@gameofpj3286]

I was smiling the whole way through this video :D This is so amazingly cursed!

[@aceofhearts37]

Oh my god. In my heart, you win. I'm genuinely sad that I can't share this video with anyone who would appreciate all aspects of it.

[@MarcelYT16]

This is hilarious and very creative. You've basically created something adjacent to a linguistic hash function. Well done 👏

[@pity2672]

That note about an agreed upon ordering of concepts with (optional) particles to mark the places of nouns is exactly what Lojban does - could be interesting to further implement elements of loglangs into it!

[@largepimping]

That's Numberwang.

[@Cydonius1701]

You, sir or madam, are a genius! 😂

[@taimunozhan]

Here's an admittedly less cursed idea for a conlang based on some of these ideas. First of all, concepts will be mapped to real-valued vectors of finite but fairly large dimensionality. This could be based on word-embeddings such as Word2Vec or GloVe, which are commonly used in Machine Learning tasks. I've worked with 300-dimensional Word2Vec and GloVe vectors before and they did a good enough job for some Natural Language Processing (NLP) applications so let's say that concepts are represented as 300-dimensional real-valued vectors, although the larger the number of dimensions (and the lower the number of concepts), the better this language will work. We'll normalize those vectors, that is to say, each concept vector will have the same length, but pointing to different directions in an 300-dimensional space. We'll also have vectors corresponding to semantic roles, basically corresponding to case particles. We'll likely need less than 300 of those particles to make the language work so we could have these particle vectors be all orthogonal to each other. This makes it so a combination of roles would correspond to an unique direction, if a noun is both the agent and the patient of a verb, its reflexive role could be indicated by combining the agent and patient vectors. All role vectors (including combinations) will be scaled to have a length of 1/2. In order to form a phrase, we'll start with a vector corresponding to the verb (this would be a verb-heavy language, with adjectives and noun copula being handled as verbs as well). The verb vector (verbtor for short) and semantic role vectors (SRVs) will be used to define a number of auxiliary vectors for potential verb arguments. The SRV for a certain role r and a verbtor v will be constructed as (1/2) v + r, so all SRVs will live within a hyper-ball of center (1/2) v and radius 1/2. Now we'll need to map concepts to those arguments. Luckily for us, matrices are an excellent way of mapping vectors to other vectors. We'll try to build a matrix that transforms each concept into the intended SRV while leaving the verbtor in place (the verbtor will be an eigenvector). Unless we have as many dimensions as concepts (which would allows us to represent all concepts as linearly independent vectors), there's no guarantee it will be possible to construct such a matrix, so you'll have to rephrase things in some other way that is hopefully possible. This matrix will be the representation of our sentence. The main verb could be identified as an eigenvector with eigenvalue 1 (there might be multiple, of course, but every language is entitled to some ambiguity). That is, as long as the sentence is affirmative, as negations could be constructed by multiplying the verbtor by -1. But, of course, not everything is an affirmation or a negation, we could also have questions with no expected truth value or hypothethicals. These more complex sentences could be handled by multiplying the verbtor with a unit complex number, with +i for pure questions, -i for hypotheticals without a bias towards either truth value (such as when pondering existance as 'to be or not to be') and intermediate values for affirmative or negative-coded questions or hypotheticals. The main verb could still be identified as an eigenvector with an eigenvalue of norm 1. Thus, simple phrases will be represented as complex-valued n×n matrices. Special concept vectors could be reserved for referencing 'the previous phrase' or 'the following phrase', allowing for recursive structures. Thus "I created a language so horrible that it made me cry" could be expressed as a sequence of matrices corresponding to "creation: 1s-AG language-PAT", "horrible: {previous_phrase}-THEME" and "weeping: 1s-PAT {previous_phrase}-CAUS".

[@taimunozhan]

So, as a very simplified example, let's say we wanted to translate "A cat eats fish" in an implementation of that language using only dimension 3 vectors (so phrases are 3-by-3 matrices). Let's have some random vectors for the concepts of 'cat', 'eat' and 'fish': >>> eats (0.6198586474198046, 0.6619178078938269, 0.421473691719805) >>> cat (0.8054980726669024, 0.09697303954384269, 0.5846101987919248) >>> fish (0.6462296240342479, 0.6312587920556468, 0.42883517867940835) Let's say that agents have the role vector (0.5, 0, 0) while patients have the role vector (0, 0.5, 0). This means that the arguments of eating are: >>> eater (0.8099293237099023, 0.33095890394691346, 0.2107368458599025) >>> eaten (0.3099293237099023, 0.8309589039469134, 0.2107368458599025) So "a cat eats fish" would be translated as a matrix A such that verifies the following matrix-vector products: A × eats = eats A × cat = eater A × fish = eaten The resulting matrix (and thus, the translated sentence) would be something like this: ((-18.5772, 1.32247, 26.7515), (10.9018, -0.0327644, -14.431), (-11.5527, 1.23298, 16.0746)) Now, in order for the listener to understand it, first they'll need to find out what the main verb vector (the verbtor) was. That will be an eigenvector whose length and eigenvalue had a norm equal to 1. Fortunately, the matrix only has two such vectors, one that corresponds to "eats" and one that corresponds to "- eats". The eigenvalue is +1, so the verb is "eats". Knowing the sentence is about someone eating something , the listener can look for its arguments. The listener will know that the subject and object will be two vectors such that: A × subject = eater A × object = eaten So, in order to understand the sentence, the listener would need to invert the matrix, obtaining a matrix B = A^(-1) such that: subject = B × eater object = B × eaten And that would allow them to find out that the subject is the same as the concept vector for "cat" and so on. Problems might arise if the listener doesn't guess which arguments the speaker encoded for the verb, though. If the speaker simply meant to say "the cat is eating", the vector the listener would get as "B × eaten" would be completely spurious, although the chances of it being a valid concept vector (and having length 1!) might be acceptably low.

[@samkimber6747]

Another fun thing is that, by definition, every utterance is ambiguous and complete unintelligible until you reach the final syllable.

[@sebas31415]

This is merging my two lifetime hyper fixations; love this!

[@vladthemagnificent9052]

wow now this is an entry to both Conlang Circus and #SoME3

[@MaxwellNieberger]

I am in awe. I need to go lie down and process this for a bit

[@VanByyrg]

A few quirks of this language IMO: * In order to understand what the other person is saying you have to begin listening when the first sound of the sentence was spoken, otherwise it will be interpreted as a pretty much random sentence * You have basically zero idea what the fuck the other person is talking about until they finish the sentence completely, because the meaning of the whole sentence changes completely after each spoken syllable * Due to the previous point speakers of Goptjaam need to have an extremely vast and precise memory, as well as being extremely patient and polite. Any kind of interruption in a society consisting of Goptjaam speakers would probably be on a level of yelling racial slurs in our society in terms of social acceptance * Due to the points above any verbal use of the language will probably disappear and people would resort to using blocks of text on paper/in letters/internet and using only very short phrases in spoken language. Or the communication will evolve into weird start-stop dialogue like: " Person 1: I Person 2: Yes. P1: Will P2: Yes. P1: Go P2: Yes. P1: To P2: Yes. P1: The P2: Yes. P1: Mall P2: Yes. P1: To P2: Yes. P1: Buy P2: Yes. P1: Milk Person 2: Yes. Person 1: Yes. (Indicating the end of his sentence) Person 2: Ok. "

[@volodymyrkilchenko]

the fan thing that the words are infinite and you can continue and continue never stopping

[@roastyou666]

太強啦⭐️

[@muninnodinnsraven4193]

finally, my two favorite things combining, linguistics and linear algebra!