#linguist problems

For years, language experts and learners have been requesting that the app include languages such as Icelandic and other languages with relatively low populations of native speakers. additionally, while Duolingo has been credited with "playing a key role in preserving indigenous languages," they have yet to fulfill their promises of adding additional at-risk languages. Specifically,  Yucatec and K’iche, which the app faced "setbacks for." Even worse, in my opinion, is the fact that they are utilizing AI to create language courses in Navajo and Hawaiian.

The ethics of using AI to model and create indigenous languages cannot be ignored. What are their systems siphoning from? Language revitalization without a community being involved and credited is language theft and colonization. (I can't even get into the environmental impact of AI).

Instead of working with more language experts, hiring linguists, and spending more on their language programs, more and more money is being poured into their marketing. While they have a heavy team of computational and theoretical linguists, there seem to be fewer and fewer language experts and social linguists involved.

Their research section has not had a publication listed since 2021. Another research site Duolingo hosts on the efficacy of Duolingo has publications as recently as 2024, but only a total of 5 publications (2021-2024) listed were peer-reviewed and only 2 additional publications were independent research reports (2022 & 2023). The remaining 9 publications were Duolingo internal research reports. So, while a major marketing feature of the app is the "science backed, researched based, approach" there is much to be desired from their research setting. Additionally, the manner on how they personally determine efficacy in their own reports, as written in this blog post, has an insufficient dataset.

And while they openly share their datasets derived from Duolingo users, there are no clear bibliographies for individual language courses. What datasets are their curriculum creators using? And what curriculum creators do they even have left considering their massive layoffs of their translations team (10%) and the remaining translators being tasked with editing AI content?

Imagine I played you a podcast in the language you are learning and I slapped you every time you heard a word you do not know. On average, how often would I slap you? What a weird question to begin with, right? It may not be obvious yet, but the answer to that question is a very good way of measuring your language abilities. If I were cocky, I would even say it’s the best method.

“Okay, Mr Cascade Languages, prove it,” I hear you say. “My pleasure”, I respond. The measuring of language abilities has at least one thing in common with the measuring of rainfall. With rainfall, you first need to choose the part of the world whose rainfall you want to measure. With language abilities, you first need to specify the environment against which language abilities need to be assessed. Let me use an example to demonstrate what I mean.

Hiroshi is a Japanese high school student who is studying English so he can study Philosophy in Australia. ASR is a good method for measuring Hiroshi’s English abilities (I’ll get to what ASR stands for in a minute). The environment in the case of Hiroshi is that of a Philosophy student in Australia. The first step of ASR is to sample the environment. Australia already has Philosophy students and sampling the environment would simple consist of recording the speech and language interactions of Philosophy students in Australia. We find some Philosophy students (say, 5 or so), then stick recording devices on them and record all the English they are exposed to for a semester. This will include everything they say, hear, read and write. The second step of the ASR method is the most fun (for us, the assessors).

Having compiled a recording that is representative of the language environment of Australian Philosophy students, we assess Hiroshi against it. We play the recording and whenever Hiroshi is exposed to a word he is not familiar with, we slap him. Not only that, we keep track of the frequency we slap Hiroshi. When Hiroshi is still a beginner, we would be slapping him at a high rate (e.g. 200 slaps a minute). As Hiroshi’s English abilities improve, the rate also reduces. After Hiroshi’s abilities have improved to a level where we would consider him proficient, the slap rate will be low (e.g. 5 slaps an hour). Oh, ASR stands for Average Slap Rate.

Turians, as they stand visually, have a gap between their lower and upper jaws that shows their teeth, as this photo shows and has largely been the case with most turians for the last nearly two decades.

Here's where my undying love of linguistics comes in. How then, pray tell, are they able to call their home planet Palaven? Plosives (some pronunciations of 't', 'd', 'k', 'g', and most importantly, all pronunciations of 'p' and 'b') require something to stop on. For most letters, it's usually a tightening of the larynx, but for 'p' specifically, it's a hell of a challenge to not contain some air in your mouth and still have the sound come out correctly.

like lets say you want to say "she is beautiful" right? the gender is that sentence is only present once (she), in spanish it would be ellA es hermosA, gender is present twice in the same sentence

this makes me feel like english is a much more gender neutral language in general which i find more comfortable even if as a cis girl, like yeah im a girl but i dont want it to be the only present characteristic about me and in romance language it often ends up feeling like it is, specially when speaking in plural because plural spanish is either the masculine gender pronouns or the new gender neutral pronouns that were invented somewhat recently (elle/le) and it still feels like the feminine pronouns are the odds ones out, like if you want to say they are beautiful you're going to say ellOs son hermosOs or ellEs son hermosEs, and i dont know i kinda wish we could say ellOs son hermosEs and for gender to only be present once in the sentence like in other languages :(

I really love learning about words and their origins.

practicing ipa sounds that aren't in your native language at 3am

looking for minimal pairs by candlelight

trying to find anything remotely related to your field of study in an old library

drawing syntax trees with fountain pens

analyzing formant structures on praat and forgetting your tea until it's cold

"yeah using Wikipedia on this lab is fine because let's face it, no one is going on Wikipedia to mess with the reports of which vowels are used in Japanese"

"well now I know what I'm doing this weekend"

spilling coffee on your notes for how R recognizes ipa symbols

(some of these may or may not be from personal experiences)

randomly picking a language and doing a super deep analysis of it for funzies

correcting people who correct other people's grammar because grammar by definition cannot be incorrect

that weird love-hate relationship with grammar/literature teachers for simultaneously perpetuating bad stereotypes of elitism in language and also giving you the will to live and introducing you to Hamlet

awkwardly watching two professors argue over how many vowels and consonants a language has

wearing a blazer and plaids to go play scrabble during midterm season

doing a historical linguistics study and finding out that's what shakespeare said