#linguistic learning model

[heinous-eli]

You know what I find personally annoying about AIs?

Not professionally as a kinda-techbro. Not morally as a human being. Not ethically as someone who trying to be a decent person. Not semantically as a philosophy major. Just personally?

The same behavior y'all hate in people like me, you adore in a chatbot.

[pallases]

okay guys i have calculated it all out and even if i get a big fat zero on this race i will still earn a b in the class assuming i get 100% on the other remaining three assignments two of which are a given for 100% and the last of which is like. even if we get a 75% on it (which i do not really see happening) i can still scrape by w a b-

[raffaellopalandri]

Five things I do for fun

Bloganuary writing promptList five things you do for fun.View all responses That’s quite easy, actually. Photo by Alex Andrews on Pexels.com I do not like spending any amount of time doing things that do not enrich me or make me feel good, so my 5 favourite activities are: Studying and Learning Solving complex problems or issues Repairing, cleaning, or creating complex mechanical objects…

[drawitbooks]

The Importance Of Memory And The Learning Models That Acknowledge It

Is Memory The Most Important Part Of The Learning Process? Take a look at all of these learning models that acknowledge the importance and value of memory. #education #training #performance #psychology #mentality #learningresources #memory #mnemonics

The mind, and more specifically memory, is an incredible thing. But just how important is memory when it comes to learning? There are lots of different approaches to learning; arguably infinite. But by having a more formal approach and analysis of learning processes we can start to recognise the parts they are typically made up of and just how important they are. Then we can start to identify…

[theskydoesgreatthingsnow]

Something I see not talked about NEARLY enough is cultural bias of language ais, not in information but in the way they talk. Kind of a huge deal for a talk-machine.

What does it mean to “sound like human speech”?

These are trained on standard English, reproduce standard English, and will be most useful to people who use standard English.

Messages created with it will best serve audiences that speak Standard English.

Responses will make the most sense to SE speakers, or speakers of the language the ai is trained on.

The language comes across formal, deferential, and polished, in standard English. It takes prompts best that are the same. Not everyone speaks standard English, or is comforted by the peppy, mechanical (white) voice.

Currently, they’re terrible at dialects and accents! Ask chat gpt to talk in AAVE and chat gpt sounds really racist, then after a few passes it will just drop the g on standard English. Ask it to talk like a Jew and it will just say Oy Vey a bunch and use the top ten Yiddish words from every Jewish listicle. Ask it to write a poem in an Irish dialect and it will lose it completely after a few questions.

It picks up the meaning of regional speech of much better than it puts it out. And I would say better than search engines. But it gets tripped up by code switching, non English words, etc.

This is not nearly an accessible technology yet.

Once again the tool of the future reproduces the sins of the past. Tired of these things being afterthoughts, that accessibility is the second step that must fight against the first. Get into it!

[sundrop-writes]

Sundrop's Criminal Minds Masterlist

Lessons For A Genius - Capsule Series (Temporarily Complete)

Note: This is a Capsule Series, so each fic can be read as an individual oneshot. There is no overarching story, and no specific ending.

Lesson One: Slick Silicone - Sub!Spencer Reid x (BAU)Dom!Fem!Reader. Co-Workers to Friends With Benefits. Smut. Despite being a genius, Reid still has a lot left to learn about life. (Mostly sex related.) And he definitely wants to learn from you. His first lesson? Well, a linguistics lesson turns into a hands-on demonstration with a very special toy. (17,200 words.)

Lesson Two: Magic Metacarpals - Sub!Spencer Reid x (BAU)Dom!Fem!Reader. (Pining) Friends With Benefits. Smut. After receiving his first lesson, Spencer is eager to learn more for you. So you teach him the next logical thing - how to pleasure you in return. (26,300 words.)

The Patron Saint of Liars and Fakes - Spencer Reid x BAU!Fem!Reader. Co-Workers to Lovers. Fake Dating. Hurt and Comfort. While undercover inside the Separatarian Sect, you and Spencer realize something important: you can’t live without each other. (8,200 words.)

Push and Pull - Emily Prentiss x Fem!Reader - Co-Workers to Lovers. Smut, Sexual Tension. When investigating Viper, Emily doesn't fall for his tricks, and in fact - spends the night teasing him by showing more interest in you. Little did she know, she was driving you insane in the process. (2,800 words.)

The Perfect Brat - Dom!Elle Greenaway x Dom!Fem!Reader x Sub!Spencer Reid. Co-Workers with Benefits. Smut/PWP. Spencer acts up, so you and Elle put him in his place. (2,900 words.)

Loverboy - Bratty!Virgin!Spencer Reid x (Dom)Fem!Reader. Co-Workers with Benefits. Smut/PWP. You try your hardest to make Spencer's first time a good one. (3,100 words.)

Black Suit - Dom!Emily Prentiss x Sub!Fem!Reader. Established Relationship. Smut/PWP. After a particularly hard case, Emily takes you home and helps you unwind by showing you exactly where you belong. (2,900 words.)

From Your Lips - Jennifer Jareau x GN!Reader. Established Relationship. Smut, Hurt and Comfort. After JJ is attacked on the Hankle farm, you take the time to check on her and distract her flustered mind. (3,000 words.)

Figure It Out - A Criminal Minds Casefic. Fem!Reader x Gen!BAU Team (Platonic). General Casefic, modelled after a Criminal Minds episode. Angst, Mystery, Hurt and Comfort. When the team takes a case in your hometown - a secret that you have been trying to hide for years comes to be known with a vengeance. (18,000 words.)

Meddle About - Sub!Spencer Reid x Dom!Fem!Reader. Co-Workers to Lovers. Smut. Morgan calls you Reid's 'Mommy', and you don't think much of it - but Reid can't get it out of his head. It accidentally slips out of his lips, and you like how it sounds coming from him. (6,300 words.)

Pathetic - Dom!Elle Greenaway x Switch!GN!Reader x Sub!Spencer Reid. Established Dom/Sub Relationship. Smut/PWP. Spencer gets punished again - in a very creative way. (2,600 words.)

She Keeps Me Up - Dom!Jennifer Jareau x Sub!Fem!Reader. Established Dom/Sub Relationship. Smut/PWP. JJ is very protective of you - and very possessive of you. When an UnSub leaves a tiny scratch on you, she feels the need to remind you exactly who you belong to. (3,100 words.)

Careful (Series - Complete) - Dad!Spencer Reid x Mom!Fem!Reader. Exes to Lovers. Angst, Fluff, and Smut. When you and Spencer broke up, he tried to forget about you. He pushed all of those feelings for you down - until he sees your name on a list of potential victims being hunted by a man who kills single mothers. (57,400 words.) | Series Masterlist | Part One | Part Two | Part Three | Part Four | Part Five | Part Six (Finale)

Jennifer Jareau x Fem!Autistic!Reader Headcanons

Jennifer Jareau and Emily Prentiss Kink Headcanons

Random Spencer Reid Headcanons

JJ Being Protective Of You (Jennifer Jareau x Fem!Autistic Reader)

The Scale of Dominance and Submissiveness in The Criminal Minds Characters (Headcanons)

How would Spencer react to you teasing him with a lollipop?

How would Spencer react to you fainting around him?

How would Spencer, Emily, and Elle react to getting proposed to?

How would Derek, Emily, and Spencer react to your daughter being clingy with them?

Note: This last fic links off to AO3. I'm probably not going to edit it and post it on Tumblr - it's going to live on AO3. So if you want to read it, you can do so at this link.

Burn The Witch - Spencer Reid x (BAU)Fem!Reader. Mutual Pining Co-Workers. Heavy Angst, Smut, Casefic. (Series - Complete.) You weren’t sure which you regretted more: acting on your feelings for Spencer, or writing them down first. But there wasn't much room for regrets when a psychopath was waving a gas can in front of your face and telling you he intended to turn you to ashes. (69,900 words.)

[too-fond-to-be-fearful]

Been thinking recently about the goings-on with Duolingo & AI, and I do want to throw my two cents in, actually.

There are ways in which computers can help us with languages, certainly. They absolutely should not be the be-all and end-all, and particularly for any sort of professional work I am wholly in favour of actually employing qualified translators & interpreters, because there's a lot of important nuances to language and translation (e.g. context, ambiguity, implied meaning, authorial intent, target audience, etc.) that a computer generally does not handle well. But translation software has made casual communication across language barriers accessible to the average person, and that's something that is incredibly valuable to have, I think.

Duolingo, however, is not translation software. Duolingo's purpose is to teach languages. And I do not think you can be effectively taught a language by something that does not understand it itself; or rather, that does not go about comprehending and producing language in the way that a person would.

Whilst a language model might be able to use probability & statistics to put together an output that is grammatically correct and contextually appropriate, it lacks an understanding of why, beyond "statistically speaking, this element is likely to come next". There is no communicative intent behind the output it produces; its only goal is mimicking the input it has been trained on. And whilst that can produce some very natural-seeming output, it does not capture the reality of language use in the real world.

Because language is not just a set of probabilities - there are an infinite array of other factors at play. And we do not set out only to mimic what we have seen or heard; we intend to communicate with the wider world, using the tools we have available, and that might require deviating from the realm of the expected.

Often, the most probable output is not actually what you're likely to encounter in practice. Ungrammatical or contextually inappropriate utterances can be used for dramatic or humorous effect, for example; or nonstandard linguistic styles may be used to indicate one's relationship to the community those styles are associated with. Social and cultural context might be needed to understand a reference, or a linguistic feature might seem extraneous or confusing when removed from its original environment.

To put it briefly, even without knowing exactly how the human brain processes and produces language (which we certainly don't), it's readily apparent that boiling it down to a statistical model is entirely misrepresentative of the reality of language.

And thus a statistical model is unlikely to be able to comprehend and assist with many of the difficulties of learning a language.

A statistical model might identify that a learner misuses some vocabulary more often than others; what it may not notice is that the vocabulary in question are similar in form, or in their meaning in translation. It might register that you consistently struggle with a particular grammar form; but not identify that the root cause of the struggle is that a comparable grammatical structure in your native language is either radically different or nonexistent. It might note that you have trouble recalling a common saying, but not that you lack the cultural background needed to understand why it has that meaning. And so it can identify points of weakness; but it is incapable of addressing them effectively, because it does not understand how people think.

This is all without considering the consequences of only having a singular source of very formal, very rigid input to learn from, unable to account for linguistic variation due to social factors. Without considering the errors still apparent in the output of most language models, and the biases they are prone to reproducing. Without considering the source of their data, and the ethical considerations regarding where and how such a substantial sample was collected.

I understand that Duolingo wants to introduce more interactivity and adaptability to their courses (and, I suspect, to improve their bottom line). But I genuinely think that going about it in this way is more likely to hinder than to help, and wrongfully prioritises the convenience of AI over the quality and expertise that their existing translators and course designers bring.

[leolingo]

waking up and seeing dream’s rip off project just breaks my heart man what the hell qsmp barely had two weeks to shine and now he’s introducing a VERY similar project in larger scale and uglier graphics and its just “the two are allowed to co-exist?” be fucking serious for a second dude why are you doing this NOW at the height of a project spearheaded by someone that used to call you a friend? like just . logistically speaking comercially speaking when you see how obviously similar these concepts are Why would you announce it now when you know someone else is getting the spotlight for once.

its hard not to call it spite or jealousy or anything of the sort when we cant confirm the timelines of this new project’s development but it REALLY, really feels like something unkind. not only that but it feels really gross to see most aspects of quackity’s passion project warped into something worse.. like LIVE TRANSLATION? really? bc dream of course wouldnt expect people to try and learn the different languages to communicate. he probably doesnt understand how redundant and ultimately hindering it will be to rely 100% on automated translation because 1) he’s not bilingual nor does he make any effort to understand the bilingual experience 2) he has no actual interest in the learning process of foreign languages or the different linguistic communities on twitch and in content creation in general . which makes me wonder WHY he is leading this and very likely profitting off of it when there’s no real reason for him to associate himself with this kind of cultural project other than . wanting to be relevant i guess.

during squidcraft, i didnt see him attempt a single word in spanish. i saw dream use google translate or straight up speak english (fast, idiomatic english at that) to spanish speakers and otherwise not try to meet a communicative middle-ground in any way. if this is how he intends to take on “united SMP” i cant wait to see it fail.

quackity’s project is successful because he cares. its modeled after his own experience and thrives because he as a bilingual host is able to cater to both communities within it and work as a linguistic bridge when need be. which, as we have watched day after day on qsmp streams, becomes less and less necessary because the environment quackity is fostering is actually very concrete INCENTIVE FOR LANGUAGE LEARNING. people are actually interacting and having meaningful linguistic/cultural exchanges that actually LEAD TO LANGUAGE KNOWLEDGE AND UNDERSTANDING. how the fuck is that supposed to happen if theres live translation? ill tell you now, it won’t.

when we study linguistics in college one of the first things we learn in regards to foreign language teaching is that translation methods rarely fuckjng work. by doing that youre limiting human interaction and actually DISTURBING the learning possibilities because youre taking away Real, varied input. dream doesnt know what he’s doing and its so upsetting to watch. dont even get me started on “language rankings” or whatever the fuck the competitive aspect is supposed to be

the project is just so flawed and the timing couldnt be worse. quackity is doing such a great job and? you just try to hijack his idea like this even though you clearly lack both the heart and the knowledge to make something like this work? to me it just appears so sour. so mean-spirited and uninspired. i dont even know man i just dont like it

[markscherz]

if you were to pursue another field of science, (besides herpetology that is), what would it be? (Sorry it’s kind of a basic question)

Depends how far I would have to go. I have dabbled in ichthyology and found it quite okay. If I had to leave a zoology field, I really like evo-devo, and I think there would have been another career for me there. Also parasitology; I once got very into the different malaria species that infect lizards. But if I had to leave biology altogether, linguistics is my shit. I am a regular listener of @lingthusiasm, and I absolutely love that you can apply so many of the same models we use in evolutionary biology to languages, in order to learn about them and where they came from. When I was a student and I would go to flat parties, whenever there was a linguist there, I would sit the whole evening and try to get them to tell me everything. Such an interesting field of study.

[paper-mario-wiki]

id like to know more about your current work and career as an anthropologist. where did you graduate from? what field of study did you focus on? what do you think about the current compartmentalization of the field of anthropology in the US today, removing it from a formerly holistic approach to a more 'separated' and corporate model? are you currently following anthropological news? if so, what region do you have an interest in?

I can't help but feel as though I'm being interrogated right now, but sure I'll answer:

I studied at Doshisha University in Kyoto for 3 years, before withdrawing due to isolation brought on my the pandemic. (bad choice to live across the ocean from your family when you also cant go outside)

I studied cultural and linguistic anthropology, as well as communicative philosophy (which I don't talk about as often because talking about philosophy is exhausting most of the time)

I really don't know what you're talking about here because anthropology is a massive field with a lot of different subsects, and I've noticed no shifts or changes to or from whatever you're calling a "holistic approach" or "corporate model". A holistic approach to what? Cultural anthropology? Biological anthropology? Linguistic? Forensic? Digital? And to what end? In theory? In construction of ethnography? Field work? Bankrolling study? You've said several pretty vague things in this part.

I'm not currently enrolled in any formal academic institution and so I'm not up-to-the-day on a lot of anthropological news, especially since most of my work doesn't demand that I do. That said, I do have a couple of news feeds from various countries I've studied like Japan, and still read ethnographies that catch my interest from time to time. Aside from that, most of my learning is habitually passive instead of active.

[nostalgebraist]

Hello nostalgebraist!

With the recent success of LLMs, I suspect that ML language models are going to have at least some implications for linguistics (or at least, be talked about by linguists) in the immediate future. I would sort of like to get ahead of the curve on this. I was wondering if you might be able to point me to any good resources on transformer models and/or LLMs specifically, for someone without a background (or, failing that, for someone with minimal background) in ML. In particular, I'm wondering if there's anything that is a little more math-oriented/abstract rather than programming oriented/concrete. I don't imagine I'll need to implement any transformer models, I just want to be prepared to talk intelligently about what implications they do or don't have for issues in linguistics.

Anyway, if you happen to know of any such resources, I would be interested in checking them out. Thanks!

If you just want a description of what a transformer is, this should fit the bill?

If you're looking for research that sheds light on what these models can/do learn in practice, there is an embarrassment of riches out there -- including various papers about linguistics, though they may or may not address your questions of interest.

That said, there are likely to be of interest in any event:

Anthropic's circuits research, starting with A Mathematical Framework for Transformer Circuits. But you should get a solid grasp of the architecture (what an "attention head" computes, etc.) before reading this.

Thinking Like Transformers, which introduced RASP, a programming language which can be compiled to transformer weights. Same caveat as the last one, and also, you shouldn't expect real LLMs to look like compiled RASP programs (they do things much more efficiently, using superposition etc). But it may be a useful lens nonetheless.

[st-just]

Even though we are most conscious of the big, widely spoken languages like English and Chinese, most languages are spoken by a much smaller number of people. Because of the sheer number of languages in the world, something like Estonian is much more typical. To a first approximation, nobody in the world ever learns Estonian. If an Estonian person wants to talk to someone who’s not Estonian, they speak English. In the past, they might have spoken Russian. But nobody is learning Estonian. And Estonian is very complicated, with 14 noun cases and all kinds of other trouble. But as Trudgill points out, even Estonian — with its ~1 million speakers and solid K-12 education system and some broadcast media — is much more widely spoken than most historical languages. The vast bulk of human history consists of language communities that had no television or radio or even writing, speaking face-to-face mostly to other people they actually know. Communities like that are Petri dishes of linguistic innovation. When I used to host “The Weeds,” Dara Lind liked to jokingly say things like “Teds Cruz” to mean “people like Ted Cruz” modeled on the pluralization of attorneys-general. That became an inside joke that Sarah Kliff and Jane Coaston and I picked up. And Trudgill says this is how language change works in small oral communities — someone makes something up because he or she thinks it’s clever, and if other people like it, they copy it. There aren’t a lot of entertainment options, language play is fun, and there’s no influx of foreigners learning your language and needing to simplify it. So you end up developing a kind of anti-creole, with lots of weird flourishes.

-Matthew Yglesias, Political lessons from Jamaican Patois

[max1461]

@triviallytrue asked "what's Chomsky's deal?"

Put briefly, the initial thing that made Chomsky famous is that he argued, and managed to convince everyone, that B. F. Skinner was wrong about language acquisition. B. F. Skinner believed that kids acquired language because when they made grammatically well-formed sentences they got positive feedback and when they made grammatically ill-formed sentences they got negative feedback, and by operant conditioning they learned how to speak. Chomsky was basically like "wait. That's not fucking true", and he was right, it isn't. His argument is usually called the Poverty of Stimulus argument, which has been used to justify many dubious conclusions since then, but which basically says that kids are not getting nearly enough actual input of the appropriate types for them to learn language this way.

Chomsky was also interested in modeling the computational properties of language, and that's where the Chomsky hierarchy comes from. The idea is to find a formalism (for syntax, for phonology, whatever) capable of generating all and only those sentences in fact found in natural language. So this formalism would have various parameters that you could set to specify the syntax of an individual language, and then it would have an input lexicon (the vocabulary of that language), and then it would spit out exactly the set of grammatically valid sentences in that language. This approach is generally called (lowercase-g) generative linguistics. It has been a very productive good idea which runs into certain empirical hurdles that Chomsky was not prepared to solve.

Chomsky is interested in this generative approach not because he really wants to model natural language as an observed phenomenon, but because he believes, more or less, that when an appropriate formalism is found it will be reflective of the innate cognitive processes which give rise to natural language in the first place. These processes are generally called Universal Grammar or UG, or sometimes "the language organ". Chomsky's beliefs here are technically more nuanced than this but not in a useful way. His claim about this is basically a deepity, which he and his acolytes try to pass off as actual science with a bunch of philosophical mumbo-jumbo. Every Chomsky paper starts with like fifteen pages on the history of science and Newton and Kepler and blah blah before he makes his various unfalsifiable claims.

The overall upshot, in my view, is that Chomsky had a lot of really important ideas about human language that he simply was ultimately not prepared to follow up on. His work, especially his early work, has reshaped the entire field of linguistics and lead to a series of extremely productive research programs. On the other hand you should not take most of what he's said in the last 40 years too seriously.

[echo-bleu]

shine still brighter (2/?)

Chapter 1 | On AO3. Deaf!Artanis bullet-point fic.

And I'm back with some linguistics! I barely have an idea where I'm going, but writing this AU is a lot of fun.

Three weeks later, Arafinwë brings little Artanis to Fëanáro’s office. She’s immediately entranced by all the shiny gems and strange little contraptions that are everywhere and she tries to touch them, and Arafinwë is terrified that she’ll break something and Fëanáro will explode.

“Let her,” Fëanáro shrugs. “There’s nothing in there that I can’t afford to replace. It’s mostly old prototypes, anyway.”

Right. His twins are a year younger than Artanis. He’s used to little children running around and being curious.

He brandishes a sheet of paper. “This is just a very rough sketch, and sign language is terribly frustrating because you can’t really write it down, but I’ve thought of ways to go beyond the basic mimic gestures and into the symbolic, which is really what you need for a language to express complex thoughts. We can use spatial variation to express basic grammar, such as tenses. A flexible word order can also take us a long way. Using the entire body opens up an incredible number of fascinating options, think of facial expressions alone! A smile or a frown could be used to modulate any statement into a question or an affirmation, or even something else entirely! We could have a specific mood for reporting speech whose origin is doubtful, for example. And the potential for spatial morphology! I really need to talk to some dancers about this, they might have new ideas. Or theatre comedians, maybe. Oh, and I’ve also devised a signed alphabet based on my Tengwar, for direct translation. It won’t be immediately useful, of course, but you’ll be able to teach her to read and write more easily, and it can be used for names and maybe homonym disambiguation.”

Arafinwë has not understood any of that, except that Fëanáro is very excited.

Fëanáro has never been excited at him before.

It’s a very intense experience.

“…can you teach us?” he asks, a little winded.

Fëanáro once spent a decade learning the languages of various wild animals, one after the other, so he could in turn teach them to Tyelkormo. Of course he can teach them.

“It’s not a complete language yet,” he warns. “I can’t make a language for her. She’ll have to make it her own.”

“…okay.”

“I’m calling it Mátengwië.”

‘Language of the hands’. Fair enough.

He goes to sit cross-legged in front of Artanis.

She puts down the shiny brass model of a windmill she was playing with and looks at him.

She doesn’t instantly scream in his ear, which is a good thing, because Arafinwë has clear memories of Fëanáro excusing himself from meals because of the noise he and his siblings were making.

“Hello,” Fëanáro says, deliberately moving his hands into signs. “I’m your uncle and I’m going to teach you some signs.”

Arafinwë’s heart jumps at “uncle” (Fëanáro has never forgotten the “half” before, when he even bothers to acknowledge them as family).

Most likely he hasn’t invented a sign for “half” yet, but that seems like a strange oversight on his part, given his insistence.

Artanis is fascinated.

“We’ll start with simple words.”

Fëanáro is speaking slowly, because he’s not fluent with the signs yet, but he doesn’t baby-talk. Arafinwë isn’t sure what Artanis actually understands of this – she can recognize some words from their lip-shape, but not consistently, and definitely not whole sentences.

The signs don’t seem to look like anything, not like the ones Findaráto made up. Those were all easily understandable in context.

But within a few hours, Artanis and Arafinwë both have a handful of new signs for everyday items and tasks.

Fëanáro uses clever ways of mimicking and pointing to explain them to Artanis, and she seems to catch on immediately.

Then she spends the rest of the lesson pointing at various things around the office for Fëanáro to name.

Artanis’s signs are a bit sloppy and simplified, because she doesn’t have much dexterity yet, and Arafinwë’s are self-conscious (because doing literally anything in front of Fëanáro makes him self-conscious), but they’ve communicated more in one afternoon than they have in the last two years.

And it’s thanks to Fëanáro.

Ñolofinwë is never going to believe it.

And Fëanáro was bearable the whole time.

Scratch that, he was nice. He teased a little, but it was never mean, and never directed at Artanis. And he laughed at his own mistakes just as much.

Arafinwë actually had a good time.

They go back the next afternoon.

And the next.

And the next.

They get to basic grammar and full sentences.

Artanis is opening up again.

She still gets frustrated a lot, and she’ll slam the door and lock herself in her bedroom whenever that happens, but she retains and uses each sign that Fëanáro shows her.

Arafinwë does his best to keep up.

Findaráto is still not doing too well, but he notices the changes, and after a couple of weeks, he begs for permission to come with them.

Fëanáro seems a little doubtful at adding a teenager to the mix, but Findaráto, if he has sufficient motivation, is an excellent student.

He takes to signing like a fish to water, faster than Arafinwë, and faster even than Artanis, who doesn’t have the benefit of translation.

Within a few more weeks, Fëanáro and Findaráto, and Arafinwë to a lesser degree, are capable of basic conversation in the sign language, allowing Artanis, by imitation, to start moving beyond naming objects and easily demonstrable actions, and into the abstract.

It’s beautiful to witness.

It’s still not a complete language by any means. Fëanáro repeats that warning several times per session, though Arafinwë doesn’t completely understand why it’s important.

It’s important because as they make up more and more sentences, they’re starting to hit at the limits of what Fëanáro has built.

It is not long before Artanis and Findaráto are inventing their own words, at first by combining signs or miming things, but soon enough they’re using their instincts and coming up with brand-new signs. And sentence structures. And grammatical elements.

It’s fascinating to Fëanáro.

(Contrary to popular opinion, he’s not a prescriptivist. The thorn issue is specifically sensible to him because it relates to his mother and he’s entirely irrational about it, but he’s otherwise endlessly happy to watch language evolve and he’s tracked all of his sons’ linguistic progression from when they were born, with charts and all, well into their adulthood, recording all the teenage innovation that other elves tend to scorn.)

Findaráto’s innovations in sign language are mostly based on Quenya, making up signs to translate words from his mother tongue.

Artanis’s innovations are astonishing. Entirely new ways of expressing concepts, of stacking signs on top of each other, of using space and her body to explain abstract ideas.

She takes Fëanáro’s basic concept and elevates it in a way he would never have thought about.

He hasn’t felt the rush of shared creation since he was Mahtan’s apprentice.

He can feel it with Nerdanel when they try something entirely new that isn’t either of their fields (like, say, making children) but in his chosen fields, everyone else is too far below his level to follow him.

And now this tiny child, who is far from being able to keep up with his linguistics knowledge, is making leaps and bounds that he would have never imagined.

He is obsessed.

Arafinwë is getting a little concerned.

He’s also getting frustrated, because he was never good at the word invention games that many of the Noldor are so fond of, and now he’s getting left behind in his children’s learning.

Angaráto and Aikanáro are learning signs bit by bit, enthusiastically, uncaring about having atrocious grammar and form in the way only children can. Artanis frowns and corrects them with a serious face that’s absolutely adorable.

Eärwen is struggling because of her fatigue, but she’s better than Arafinwë at getting to the essentials, at mastering the phrases and signs that she needs first without getting into complex, abstract things. It means that she misses some of Artanis’s rapid development, but at least she can tell her daughter that she loves her

And to stop screaming in their ears to get their attention.

Generally, things are getting better. Findaráto is coming out of his shell, Artanis gets frustrated far less often, and astonishingly, Fëanáro is being nice to Arafinwë even outside of the lessons.

The lessons are really more of an excuse for Fëanáro to document Artanis’s progress, she doesn’t actually need his help any more, though she’s surprisingly open to his suggestions to make a turn of phrase more elegant, or a sign more economical.

Surprisingly, because she’s not taking anyone else’s advice.

On anything.

Being able to communicate hasn’t made her any less stubborn.

She insists on doing everything herself, and now that she has a language of her own, she’s started to resent people who don’t sign.

Findaráto’s translations, even though he tries hard, aren’t good enough for her.

She refuses to play with anyone who can’t sign to her satisfaction.

Understandable reaction—but unfortunately impractical, because she has little patience for anyone who don’t sign as well as she does, which means the only people she’ll voluntarily spend time with are Fëanáro, Findaráto and maybe Arafinwë, on a good day.

Fëanáro has shown an incredible amount of good will so far, but he’s very busy. Mátengwië may have become one of his special projects, it’s still only one of them.

Specifically, aside from his princely and fatherly duties, he’s working on ways to capture light inside gems.

He can’t spend all of his days with a child that isn’t even his.

Findaráto is about to start university and needs to focus on his studies, however much he loves his sister. And socializing exclusively with a child isn’t very good for him, coming out of several years of depression.

“Eärwen and I have been talking about tutors,” Arafinwë tells Fëanáro one day. “Artanis is more than old enough to need one now, but none of them can sign with her. And she doesn’t read or write yet.”

“Reading will be a challenge,” Fëanáro confirms. “She doesn’t know Quenya, she will need to learn an entirely new language and medium at the same time. But she’s very bright, she’ll pick it up.”

“But who can teach her? I tried to start, but didn’t make any progress, she lacks any patience for what she doesn’t understand.”

“That’s not strictly true,” Fëanáro chuckles, remembering hours-long conversations with little Artanis about subjects as varied as which of her brothers is the most intelligent and what should be the right hand-shape for the word “turtle”. “But this particular challenge is understandably frustrating. I will teach her.”

“Truly?”

“Yes. As for tutors, I suggest Tulcasar, once she’s proficient with writing.”

“The loremaster? They’ve always refused to tutor any of us, I know Father asked them.”

Fëanáro laughs. “They tutored me before you were born. They will only accept the brightest students, they dislike children who cannot keep up with them. They lasted two weeks with Findis.”

Arafinwë tries very hard not to feel offended. Fëanáro isn’t even saying it as an insult, he’s so confident in his own superiority that it doesn’t register to him that it might be belittling.

“They’re tutoring Morifinwë and Curufinwë part-time right now,” Fëanáro continues. “My eldest two were never as interested in academic pursuits. Tulcasar will enjoy the challenge of learning Mátengwië, and Artanis is bright enough to keep them on their toes.”

“Alright,” Arafinwë says carefully.

“In the meantime, for the other subjects, you might ask Nelyafinwë or Morifinwë. You know Nelyafinwë adores her. And Morifinwë could use the challenge. I think he’s been feeling a little inadequate since Turkafinwë was accepted into the Hunt and Curufinwë got me to promise him an apprenticeship. He hasn’t found his craft yet.”

“Does he even need a craft?” Arafinwë asks. “I don’t have one. Findaráto is showing no sign of choosing a single field, and neither has Findekáno. Or Father, for that matter.”

“He thinks he does, at least,” Fëanáro says. “Perhaps Nerdanel and I have encouraged that a little too much. He persists in learning to paint, thinking it will please his mother, but I doubt it will ever be more than a hobby. If tutoring Artanis could help him realize that his strengths are more in academia, I would be grateful.”

“Fine, I will ask him. On one condition.”

Fëanáro raises an eyebrow—they both know that Arafinwë isn’t the one doing him a favour, here. But Arafinwë persists nonetheless, because he’s been meaning to bring up the topic.

“Let Maitimo finish his apprenticeship with Ñolofinwë. You know Father is not a good teacher, and he dislikes statecraft, for all that he is the King. Your hang-ups with our brother are hindering your son.”

He fully expects Fëanáro to get angry, only hoping that he’s accumulated sufficient goodwill that it won’t be the end of what friendship they have managed of late.

But Fëanáro laughs.

“You have been away from court for too long, Ara. Nelyafinwë has been shadowing Ñolofinwë for years.”

Arafinwë frowns. “The change hasn’t been acknowledged.”

“Does it need to be?”

Maybe it doesn’t. Let Fëanáro keep his pride and his misplaced grudge intact. He’s been fairly quiet about Ñolofinwë lately, no need to push him into another bout of paranoia.

And so Artanis starts taking reading and writing lessons from Fëanáro in the morning and spends many afternoons with Maitimo or Carnistir. Arafinwë and Findaráto come along the first few times, but it quickly becomes clear that she’s in good hands, and that their presence is hindering her more than helping. Arafinwë starts spending more time at court, since the family are now in Tirion a lot more.

Fëanáro and Ñolofinwë are actually being polite to each other. It’s quite a sight to see.

Things are going quite well, really.

[stormsthatrage]

Important PSA: Do not ask linguists how many languages they speak. Linguistics is the scientific study of language, investigating things such as (but not limited to): how language can be modeled as a complex logic system; the cognitive processes and neurological structures that support the acquisition, storage, and usage of language; the biological and acoustic mechanisms involved in the production/perception of language; how language influences society and how society influences language.

Other things linguists do: work with communities to document or revitalize dying languages; develop dictionaries; work on language modeling software; consult for copyright litigation; reconstruct dead languages.

Really lucky linguists get hired by Hollywood to create fictional languages for sci-fi and fantasy blockbusters.

What linguists do not do: spend all day learning foreign languages. (Alas, if you are a linguist who enjoys learning foreign languages, you must do it in your free time, not during work hours.)

Also! Another important PSA: Any linguist who has learned, like, anything about language, WILL NOT JUDGE YOUR GRAMMAR! Do NOT apologize to a linguist for how you speak. Remember: no dialect is "more correct" than another. There is literally no objective criteria with which you can compare two languages or dialects and decide one is "better" than the other. If you are communicating successfully, you're doing language right!

[transgenderer]

Editor's Summary: How do young children learn to associate new words with specific objects or visually represented concepts? This hotly debated question in early language acquisition has been traditionally examined in laboratories, limiting generalizability to real-world settings. Vong et al. investigated the question in an unprecedented, longitudinal manner using head-mounted video recordings from a single child’s first-person experiences in naturalistic settings. By applying machine learning, they introduced the Child’s View for Contrastive Learning (CVCL) model, pairing video frames that co-occurred with uttered words, and embedded the images and words in shared representational spaces. CVCL represents sets of visually similar things from one concept (e.g., puzzles) through distinct subclusters (animal versus alphabet puzzles). It combines associative and representation learning that fills gaps in language acquisition research and theories.

Abstract: Starting around 6 to 9 months of age, children begin acquiring their first words, linking spoken words to their visual counterparts. How much of this knowledge is learnable from sensory input with relatively generic learning mechanisms, and how much requires stronger inductive biases? Using longitudinal head-mounted camera recordings from one child aged 6 to 25 months, we trained a relatively generic neural network on 61 hours of correlated visual-linguistic data streams, learning feature-based representations and cross-modal associations. Our model acquires many word-referent mappings present in the child's everyday experience, enables zero-shot generalization to new visual referents, and aligns its visual and linguistic conceptual systems. These results show how critical aspects of grounded word meaning are learnable through joint representation and associative learning from one child's input.