/sci/ - Science & Math

Too autistic for me but I'm intrigued by it.

>current phd student here
Tell me about linguistics. What insights does the field offer?

>>11535443
>Linguistics
not science or math

>>11535443
Is there a way to determine whether a given string of phonemes, phones or allophones are pronounceable or not?

>>11535482
>is not aware of computational linguistics, and more generally of the connections between linguistics and mathematics
nigger

>>11535482
It's basicly math

>>11535443
I am a biz phd student and we want to steal your stuff. Wet are just using tools like LIWC

Questions:
1. Are there any other linguits for dummies tools out there to impress my peers
2. Are there any good texts on computer linguistic for dummies that help me to bullshit my way through the methodology section?

>>11535477
desu a lot of it is bullshit "anyone" could do--the heavy theory/borderline philosophy people like chomsky do at UMass Amherst, MIT, and Maryland is work that is heavily rewarded in linguistics but probably forms a minority of all the work that is done in places that are called linguistics departments

>>11535482
if biology counts, linguistics writ large also counts--we're trying to theorize about a domain of natural phenomena, viz. human language. i might agree with you about certain areas of linguistics though, like discourse and conversational analysis (which IMO is really just linguistic anthropology with a different culture)

>>11535489
i'm not a sound person but i think the response would be that this question presupposes a level of definition that phone{t,m}ic transcription doesn't have devoid of context. for a given language and transcription system, of course you could decide whether it's licit or not, but that's not what you're asking--you're asking about whether it would be pronouncable in general, right? aside from individual sounds that are illegal combinations of our articulatory parameters (e.g. pharyngeal nasal), the answer to the question 'is this string (e.g. [zt͡ʃbl̩tszt]) pronouncable?' is 'idk, probably'

>>11535500
this too

>>11535521
1. do you know the basics of python? go use this and have em shit their pants https://spacy.io/
2. this is the NLP/CL bible, i'd start here: https://web.stanford.edu/~jurafsky/slp3/
might be a little advanced for what you're talking about--also see the nlp for python book (nltk), free here: https://www.nltk.org/book/

>>11535540
>if biology counts
Biologists do not use the scientific method.

I'm obsessed over how much ambiguity natural languages entail. Why we don't want to fix it when it is so essential to knowledge?

>>11535540
No no no!
We need clickity click tools. Think SPSS.

Look at LIWC. You don't even have to create a dictionary. Just feed text in, get segmentations etc out.

Is there, for example, a tool that makes it easy to train an algo? Eg that I can classify 100 phrases and it runs thought it and applies it to the rest. Or creates predictions from it.

>>11535480
oops forgot to respond to this one. there's a lot to say about this, but i'll try to keep it short: for most of history, grammarians have mostly been interested in the question "in a given language, what are the grammatically correct expressions and their meanings?" but these days, linguists want to ask questions like "what are the reasons (genetically inherent, environmental, or otherwise) why all human languages have the grammatical and ungrammatical expressions that they do, with the meanings that they have?"

let me demonstrate with an example: any English grammarian before 1950 could have told you that (1) is fine, but they definitely would not have come up with a sentence like (2), which turns out to be wrong, but for interesting reasons. And it turns out that to explain cases like (2) in anything resembling a principled way you need to develop a somewhat detailed theory of what kind of structures are used in human language. (If you're curious about this example, google "adjunct island constraint")

(1) Which movie did John say Bill saw?
(2) * Which movie did John have dinner before Bill saw?

now back to your question: the most high-minded in the field like Chomsky say that linguists are ultimately cognitive scientists and want to develop an empirical understanding of that part of the brain/mind that is responsible for human language--the scope of its responsibilities, its architecture, its interfaces with the rest of the brain/mind, etc.

Linguistics is an incredibly motley field though, containing people who would describe themselves as anything from an anthropologist to a computer scientist to a logician. A lot of linguists would grate at Chomsky's research statement and just say that they're interested in describing human languages in less rarefied, less remote terms.

>>11535554
ok

>>11535570
what kind of ambiguity are you thinking about in particular? "I shot an elephant in my pajamas" ambiguity, "you" in English not indicating number ambiguity, all of the above?

>>11535584
oh hmm--that's pretty hard. well first of all, for memes, you should check this out: https://talktotransformer.com/

there's a lot of software that corpus linguists has developed that might kind of be like what you're looking for if you're interested in concordancers or collocation studies. CQPweb and AntConc are two things to look at in that area. A machine learning GUI, though, I don't think I've seen

>>11535606
>"I shot an elephant in my pajamas
Yes, syntactic ambiguity, mostly. A regulated use of punctuation could solve many, for one thing?

What field of linguistics enables the shortest path to teach a unique person (someone you just met) a given concept?

As in what is the linguistic function of exchanging awareness and understanding of consent, utility, and duration.

>>11535628
Btw, in some language like mine, even lexical ambiguity is too strong now that I think about it, even with a given context

>>11535627
>>11535627
Uhhh, CQPweb is interesting! Thanks so much!

I played around with AntConc in the past. Have you seen recently any interesting papers using it? Is there any more general intro do what AntConc does than the manual? Like on the different types of significance that it uses for the keyness measures.

>>11535443
Did you do NACLO when you were younger, anon?

>>11535628
>>11535640
right, like in the famous "To my parents, Ayn Rand and God." This isn't an area I know too much about but I think the consensus right now is something like this: there's a tension between production and comprehension. We want production to be as easy as possible, e.g. to just say "aaaaaaaa" and have it mean what we want it to mean because that way we don't need to deal with retrieving and manipulating different symbols. On the comprehension side, we want a message that is easy to hear/see and understand (i.e., not ambiguous). In the real world it seems that these two competing forces put us somewhere in the middle: we *do* have rich vocabularies and grammars, but with an appreciable amount of lexical and syntactic ambiguity. It's obvious that most of the time this ambiguity is effortlessly resolved, which means that ambiguity tends to arise only where it doesn't pose a serious and recurring problem for understanding--or else, we might presume, it gets snuffed out. For instance, it's ok for "bank" to mean 'place with your money' and 'place next to the river' and 'place where you deposit a certain kind of thing' ("food bank") because humans are usually able to apprehend which sense is meant.

Sorry that was kind of a ramble, I think if you're interested in this you should take a look at work by Ted Gibson and Piantadosi.

>>11535631
uhh, not sure if this is really a properly linguistic topic, but i know there are some people whose work spans linguistics and education. sorry, i've never seen this topic before

>>11535646
np! yeah it's too bad there isn't more in the GUI area--language data is becoming very important in all kinds of areas but for the most part it seems you still need to have some data science chops to work in depth with it.

It's been a while since I've seen a paper with antconc but check out the papers that cite it, maybe some of them will be insightful: https://scholar.google.com/scholar?as_ylo=2016&hl=en&as_sdt=20005&sciodt=0,9&cites=17509006618635360556&scipsc=

>>11535652
actually no, i was too busy jacking off and playing video games. But it looks like a lot of fun and I wish I had. I wasn't even going to do linguistics until a friend of mine encouraged me to major in it instead of computer science. (i ended up doing both)

Are there linguistic theories that deal with context? Some sentences that are ambiguous in isolation become unambiguous when interpreted in a certain context.

In face-to-face communication, the content of one's speech is annotated with things like tone-of-voice, cadence, body gestures, etc. Do linguists take those into account? Is there an abstract theory of communication that considers all communication factors?

Do linguists deal with the meaning of sentences? What does the sentence, "She is beautiful." mean if every person has a unique, subjective interpretation of the word "beautiful"?

>>11535679
Well how much of linguistics becomes impactful or powerful beyond studying to explore more complex aspects of how other people use words?

Or is language adoption truly just a resource constraint to end-user utility adoption thing?

>>11535698
To me a language is when the silent observer of your existence attempts to reach out to another silent observer in order to describe or explain or teach or seduce, but ultimately it is a silent observer trying its best to describe whatever it can reduce silent observations down into expressions and signals/signalling.

I'm really enjoying watching these deep learning language models continue to BTFO linguists.

Chomsky is probably sweating bullets as these things slowing encroach on every theory he's had.

>>11535443
How would a linguist recommend someone learn another language? What's the best way to grasp basic grammar and vocab easy?

>>11535443
based thread, anon. thnx for the answers! i'm on the genetics side of the evolution of the language. i was a researcher for a limited time on that topic and i wrote a paper about the FOXP2. linguistics have many mysteries that are waiting to be solved and i'd like to dive into the more specific subjects one day.

>>11535679
Is there some semi GUI tool that is a good intro into doing things more manually? For example the machine learning thing.

>>11535679
Or some really idiot guide for getting into that whole coding world with a strong focus on linguistic stuff.

Scott Adams says that the best way to persuade someone is to use simple language in short sentences. Is this true?

>>11535861
Are you trying to persuade me to something right now? If so it is unambiguous.

>>11535482

Certain aspects of linguistics are not math or science, others are. Such as computational linguistics, whether the statistical and empirically derived models of language, or the theoretical language constructs built from CFGs and other sets of production rules.

>>11535477
>too autistic

that's the appeal, friend.

>>11535443
I'm currently investigating the syntax of the definitions of words to see if there is a universal syntax for word definition to make classifying and ranking words and their meaning easier.

>>11535910
>if there is a universal syntax
soooo what do you think?

If we made a virulent and inclusive enough language that basically pushed out all those who didn't want to learn it because it was 'too autistic to even penetrate' then I think that'd be a win for science.

>>11535918

Big surprise but I haven't found it yet. Not sure whether if you can prove if it's possible or not, either. I'm a little out of my league with most of this stuff so it's more of a hobby for me than anything. Key word is investigation. ;-)

>>11535961
well, keep investigating then, anon! we would like to hear more from you when your research give you some ideas.

>>11535890
Persuade you for what? To answer my question? Just curious since Scott Adams is a trained hypnotist.

>>11535974

Thanks man, I certainly will

>>11536004
So you were asking for corroborating evidence to support the claim, not attempting to make the sentence true by suggestion?

>>11536009
Correct.

>>11535540
>i'm not a sound person but i think the response would be that this question presupposes a level of definition that phone{t,m}ic transcription doesn't have devoid of context. for a given language and transcription system, of course you could decide whether it's licit or not, but that's not what you're asking--you're asking about whether it would be pronouncable in general, right? aside from individual sounds that are illegal combinations of our articulatory parameters (e.g. pharyngeal nasal), the answer to the question 'is this string (e.g. [zt͡ʃbl̩tszt]) pronouncable?' is 'idk, probably'
Thank you so much for your answer! I am quite a noob, so I am not sure if I know what you mean by
>phone{t,m}ic transcription
but might I be correct that this is has the form of a sequence of phonemes which is associated with a particular language? How can I determine whether a sequence of phonemes is licit? Where can I find a list of illegal combinations? What is the relationship between what is licit and what is pronouncable? Does pronounceable imply that it is licit? Or licit imply pronounceable?

>>11535443
Far from an expert, if anything quite the opposite, just someone with a slight interest in it, but seeing how languages move the world and are moved themselves over the course of history and events always amazes me. At first I never thought much of it but now seeing how systematic, pattern-based, and scientific the study can be, it makes me honestly depressed how proportionally few people in the world are interested in things like it.

>>11536021
>Where can I find a list of illegal combinations?
In English

>>11536024
So the logarithmic distribution of how language adoption and use rose and fall during renaissance periods, war, and the like? If you included etymology you might even be able to find a pattern in how certain languages collected and dispersed or discover what conditions were required for one regional language to dominate another or how certain expression sets within a language affected birth patterns.

Humans can't interact with other humans without using language so when you think about it language would be an entirely underutilized data point for historical events and reasoning. Could even identify moments where new words gained dominance and etc.

How many languages can you speak or read/write?

What do you think about Chomsky? Universal grammar?

>>11536511
It is pretty obviously wrong at this point. In my opinion.

Chomsky key observation is that language learners learn the restrictions of their language, despite usually never being told these restrictions explicitly. He therefore proposes that an innate universal grammar which provides these restrictions.

However, Deep learning language models are very forcefully challenging this idea. We know these models have no kind of innate universal grammar, yet they are still able to infer language rules and generalize within the rules of the grammar. These models are able to do precisely what Chomsky said is impossible.

Chomsky basically got things backwards. He was puzzled by how efficiently humans can learn something which seems so complex like language. So he came up with this idea of innate universal grammar, which act as an inductive bias to accelerate learning.

I think the answer to this problem is in the languages themselves. Human languages evolved to be simple to learn. If a language was hard to learn it simply wouldn't survive. The properties of language have gone through a selection process that prioritizes learnability.

>>11535443
My ex bf was a grad student in linguistics. Great guy

>>11536591
Well that would mean languages were born from a need to satisfy some unique resource constraint over some trial and error attempts at communication in order to reach stable reproduction of identifying what an origin point is, what a cycle is and when it phase shifts, and being able to identify when cycles overlap or are distinct.

I guess some cave people would base their language around some logarithm of stone and the memory of pressure of that material against their body and muscles if we add the nerves of our body to the mix as influence over a language, but that might just be syntactic sugar ultimately. The language of the tool is interesting but for spoken language to arise the language of the tool would have come afterwards.

A language would default prioritize by virtue of natural selection things like food growth cycles and the ability to distribute this knowledge/awareness to other members. So we have members who want to repeat the bare minimum for tasks that enabled survival. The more important the task the shorter that ritual of information transfer would become over time as it became a core expression for saving time.

I guess pain would have been a much later thing for any language to develop when one thinks about it. Even animals basically just howl and screech and haven't learned the ability to calm down so they can provide more detailed behavior. Might be the difference in language ultimately between animals and humans, the ability to heal and care for the pain in another that an observer is not experiencing is a pretty complex system.

>>11536595
Hope it was an amicable separation and an amazing relationship that helped you get to new places with nothing but hope and love behind every step.

>>11535443
i'm currently studying for a germanic linguistics degree with my main interest being philology
of course i'll probably never be able to make any money off of it so idk what i'll do professionally
what field are you going into anon?

bump

>>11535443
Is Sapir-Whorf still valid in current year?

hi OP here, i'm back

>>11535698
check out https://nlp.stanford.edu/~manning/papers/probsyntax.pdf for a sketch of a context-aware theorization of syntax, and https://www.glossa-journal.org/articles/10.5334/gjgl.635/ for an example of how formal semanticists are trying to account for gesture. more broadly to your question yes, historically there was less emphasis placed on not-strictly-linguistic matters like those, but recently intonation in particular has been a very hot research area. this isn't something i know too much about though so maybe someone can fill in

> Do linguists deal with the meaning of sentences? What does the sentence, "She is beautiful." mean if every person has a unique, subjective interpretation of the word "beautiful"?
i'm actually reading a book about this right now--the answer's yes, and you can look in here for an idea: https://global.oup.com/academic/product/subjectivity-and-perspective-in-truth-theoretic-semantics-9780199573684?cc=us&lang=en& (you can find the pdf on libgen)

>>11535731
> Well how much of linguistics becomes impactful or powerful beyond studying to explore more complex aspects of how other people use words?
> Or is language adoption truly just a resource constraint to end-user utility adoption thing?

these terms are all a little more high-level/vague than ones linguists usually deal with, you might want to ask a psychologist, anthropologist, humanist, or philosopher of language instead

>>11535758
not saying anything about my own feelings on the matter, i think chomsky's totally unfazed by recent advances in NLP. There are a lot of good reasons to claim that the way an NLP model learns language is very different from how a human learns language (and also that *what* they learn internally is different as well)

>>11535777
this is applied linguistics, which i'm not too aware of--the keyword you should search for is "second language acquisition". sorry i don't know personally

>>11535780
aww thanks! glad you like it. i wish i had more time/ability to work on the interdisciplinary stuff like that, it's very cool

>>11535807
not that i know of :(

>>11535861
a psycholinguist might have an answer but idk personally--there's probably some psych studies that could be getting at this

>>11535910
cool! what do you mean by word definition? are you talking about a logical denotation? what's some of the foundational literature in this area?

>>11535482
>Trolling outside of /b/

>>11536021
> but might I be correct that this is has the form of a sequence of phonemes which is associated with a particular language?
for a phonemic string, yes--you're transcribing linguistic form not as it was actually pronounced but in terms of the abstract phonological units that you're positing people that speak that language work with. For instance, the plural morpheme /s/ in English sounds like a [z] in "meals". phonetically we would transcribe that as [milz] but we would transcribe it phonemically as /mils/
> How can I determine whether a sequence of phonemes is licit?
in general you need to construct a phonological analysis of the language. there are heuristics you can use though. it's interesting to look at what happens when loanwords enter a language: if the loanword has an illicit sequence, speakers will map it to the "closest thing". this is really obvious in Japanese, which really really prefers that every syllable have the structure CV, where C is a consonant and V is a vowel. The German word "arbeit" came into Japanese, but Japanese speakers didn't like how the [rb] were right next to each other without a vowel, and how the "t" came at the end without a vowel. They repaired this by adding vowels to yield "arubaito" (アルバイト).
> Where can I find a list of illegal combinations?
answered above, but most major languages will have an authoritative reference grammar that you could check for a phonological analysis. For English, that'd be The Cambridge Grammar of the English Language
> What is the relationship between what is licit and what is pronouncable? Does pronounceable imply that it is licit? Or licit imply pronounceable?
no citation to back this up, and it's probably an open question, but i think most linguists would agree with the conjecture that "anything pronouncable could in principle be licensed by some language's phonology". this depends on what you count as "pronouncable", though, which I won't get into lol

>>11536024
yeah, the first time I was impressed by this is when I read What Language Is by John McWhorter as a wee undergrad and he talked about how radical immigration in ancient Persia led to the grammatical simplification of the language. Another surprising result to me in undergrad was the pretty robustly shown fact that ethnicity, race, and language are all quite free to vary independently.

re: interest in this kind of thing, is it still depressing if you consider that most people need to prioritize earning potential, and that there's very little economic benefit in studying this kind of thing? (it's also not necessarily something you can easily study just from materials you'd find in a library, which raises the bar for hobbyists)

>>11536511
I won't say exactly which since that'd give away my identity :) but the answer is that I only speak one language natively, I have decaying conversational proficiency (still have good reading though) in another major world language, and have reading knowledge of two other languages. As any linguist will tell you, though, you can be a very good linguist even if you know just one language.

>>11536591
> However, Deep learning language models are very forcefully challenging this idea. We know these models have no kind of innate universal grammar, yet they are still able to infer language rules and generalize within the rules of the grammar. These models are able to do precisely what Chomsky said is impossible.
I'll just say two things here, not having any particular investment on either side: (1) state of the art NLP models are trained on *billions* of sentences and have an *enormous* number of parameters. GPT-2 has 1.5B parameters and trained on 40GB (~40B words) of training data; (2) state of the art models still haven't solved many linguistic tasks humans can effortlessly handle, like long-distance movement and coreference resolution

>>11536691
you must be cool--the people in the german dept are always some of the most intellectually curious and vigorous i meet! it's good you're working on philology because the area is all but dead in linguistic depts, sadly, even though it was all linguists did in the 19th century. what areas you looking at in particular?

i'm still undecided between buying my tenure track lotto ticket and going into industry. i do enough computational stuff that i could probably get a job at a tech company, so i think that'll just be my fallback. what about you?

>>11538200
the consensus for all but the weakest versions of it is a firm "no" (great article: https://plato.stanford.edu/entries/linguistics/#Who ), but lera boroditsky is today's whorfian standard bearer, so you should take a look at her work

>>11539233
>plato.stanford.edu
bro that website is my addiction along with the IEP. do you have some other sites like that to rec? it can be about philosophy or linguistics or anything you enjoy to read, it doesn't matter if the articles written by professionals.

>>11539289
yeah SEP is based. don't know of any other sites with such high and consistent quality but i'd recommend western linguistics: an historical introduction by pieter seuren for similar reading

>>11539690
forgot to add: just remember that by "an historical introduction" he means "a very typically european-ly combatively opinionated overview"

>>11535570
fix what?

Im a computational linguist at a major company, hi friends

>>11539709
what kinda problems do you work on? how related is it to what you did your phd in? how intellectually stimulating is it compared to what you did as a grad student?

>>11539690
>>11539696
thnx for the rec, i'll read that.

>>11539173
Then you are just a professional professor and not a linguist.

>I am a language, not a degree or a doctorate.

>>11539173
Also I fucking hate how they have theoretical and applied as descriptors to degrees.

It goes:
Field > Principle > Discipline > Practice > Person

Constantly redescribing the eternal interaction between exploration space and participation space is beyond dumb. The more language differentiation we as a species perform the less agglutinative our distributive psychological boundaries and potential experience field becomes.

>>11535443
I am a cognitive linguist

>>11540951
You are now a thinking potato head translator.

>>>/lit/

>>11535663
Good post

>>11535777
Not a linguist(whatever that means) but it depends on what you really want. Do you actually want to speak and understand the language or do you only want to memorize tourist phrases? If you want to actually become fluent then the answer is exposure. Increase exposure and you'll learn more of the language. Language learning cannot be intellectualized. We don't use the intellectual part of our brain when learning a language. Learning language from a textbook or a phone app is notoriously difficult for this reason. Expose yourself to the language and things make sense over time. Eventually you will be able to construct sentences yourself and know when something sounds right or wrong.

>>11535443
I larp on /sci/ but I love me some language. Kinda larping there too but yeah!! I know I'm not the only linguisticfag. I welcome this, not that I'm anyone.

>>11540572
Fun fact : I only have a bachelor's in linguistics
Took a few comp ling classes and learned myself.

Its nothing special, I train ai models, make grammars. We use shell and mix of python and java. I work on nlg and nlu. I dont program much except for the grammars.

>>11541029
How would you define a translation process as an arbitrary cubic equation?

>>11541035
The same way word embeddings or vectors are much more tasty than mcchickens

>>11541110
What about grammatical imprinting via vectorized wordplay?

>>11541119
I'd say its the same as using activated almonds instead of using the activation function. Same results.

>>11541136
How would you describe an auto-learn language function that could be described as an arbitrary +1?