Speaking as a total ignorant from a coding perspective. But I guess that wouldn’t be the hard part, considering that most of Duolinguo is just boxes and text inputs. How difficult it is to create a database of competent linguists with an efficient training who can progressively enhance your understanding of languages?
Duolingo is pretty bad at teaching you a language so I don’t think we really need to make an open source alternative. If you want to actually learn a language, just use anki (anki is open source) for flashcards and get a textbook. I say anki because it uses a spaced repetition system which is the only way to effectively study more than 100 flashcards and there are browser plugins that allow you to create new flashcards from a couple clicks on a new word. Once you get far enough you won’t have to use the textbook and will be able to just sentence mine for words and have to google the occasional grammar point.
Well I used Anki for quite some time during my studies and I really didn’t like it.
- First of all I’m a visual guy and the UI is pretty ugly imo. For example it’s default card formatting feels very unintuitive for me.
- Secondly the spaced repetition algorithm Anki uses isn’t as effecitve for me as Quizlet’s. Which puts me into a bad situation as I refuse to use Quizlet because it’s closed source.
- And finally the learning quality is heavily depending on the deck. Some are really making use of the various features which make Anki so versatile but most decks don’t.
I think it’s highly personal what suits you best but I really like the idea of a rewarding system or some fun repetition options just like Quizlet has it.
Repetition of words in isolation (ie flashcards or what anki offers) does nothing whatsoever to teach you a language.
Duolingo is far from perfect but it certainly does more than basic flashcards (which are fine if you’re ok with just vocab). What people constantly miss about Duolingo is that it also offers lessons (to teach you how grammar works for example) but people have to read them and take the time to understand them. Which isn’t what they normally do because it takes time and it doesn’t give you xp (it’s not gamified so everybody ignores it).
It’s how school teaching works (no it doesn’t work great either but that’s because this part is only meant to teach you about the basic layer of language, not the rest).
So Duolingo and anki aren’t designed to do the same thing at all. But if you’re serious about learning a language, Duolingo is certainly a better start IF you do it right. A combination of the two is a better bet.
Duolingo open source? Doable but you need teachers to open source their lessons and vet them. Huge amount of time and probably costly which is where the cookie crumbles.
Well you don’t just use the flashcards, You have to use textbooks and practice on supplementary material. Those aren’t really software that can be open source.
Also Duolingo removed its grammar lessons a while ago, so its really not that different from flashcards now. EDIT: nevermind they only removed grammar on desktop.
Doable but you need teachers to open source their lessons and vet them.
If an OS alternative was trying to completely replace duolingo, it would need far more than that. Duolingo has had extensive work put into listening and speaking lessons. Almost all lessons have a listening componentwhich is a ton of content to make up for. They have significantly better voice recognition than my phone. The amount of effort to get something like that working for a language, let alone dozens of languages is a high bar.
Take a look at any of the job postings that duolingo has, they’re only looking for Google employee level of skill for a reason (aside from how fucked the job market is).
It’s not impossible for duolingo to be replaced with an open source version, but it’s a giant undertaking.
Yomichan (primarily for Japanese but may also work for Chinese and Korean): https://github.com/FooSoft/yomichan
Yomichan itself is no longer maintained, but an actively developed but still beta fork Yomitan exists: https://github.com/themoeway/yomitan
Setup for Japanese: https://aquafina-water-bottle.github.io/jp-mining-note/
I think this question is a XY problem. You want to learn a language and not build an app for learning languages. To learn a language you just need motivation and resolve to stick with it for long periods of time. Just grab a text book, some video courses and flashcards.
The technology is probably the smallest part of the problem. Most of it will be getting it critical mass of users, and expanding that user base, cross multiple languages. So that’s advertising, politics, social networking, promotion.
Yeah and their staff have been very dogged about promotion in language communities, including crowdsourcing the content (i.e. getting their users to produce it for free)
I think a combination of fast advancements in multilingual open source TTS models and maybe LLMs? could help. The problem is LLMs habitually lie. It also may be unneeded. Duolingo seems relitively simple iirc (2019 was the last time I used it) a combination of simple phrases and good gamification of topics. Add some anxiety and there you go, Duolingo lite or sumthin
I’m not convinced that model of learning is really effective anyway. At best, it’s a fun-ish time killer where you learn a bit about language.
I’ve been learning French on there for awhile now and it’s been extremely effective por moi. They teach you as much French as you would get in a 4 year university program. Plus having an AI powered practice set is like having a teacher who knows exactly what you know and what you need to learn more.
Try taking a break for a month and see how much you actually remember. In my experience it was depressingly little, and I’m not generally bad with languages at all.
I think that would go for most learning methods. When you don’t practice a skill you’re always going to get worse with it over time, especially if it is a language.
As someone who is at around a high B2/low C1 level in Dutch now and moved to Belgium and used dualingo in the beginning I have a bit of insight into it.
It doesn’t do shit for grammar and sentence structure, but it builds vocabulary. If you learn only with dualingo, you will probably make a lot of mistakes, flipped adverbs, verbs in the wrong place, sentence structure errors, etc…
It definitely made me feel like I could speak Dutch because I could read it MUCH better after 1/2 of the dualingo course, but then when I moved, speaking was pretty bad and I had only moved up to an A2 level with the vocab of maybe B1. You definitely cannot become semi-fluent or fluent with dualingo. It doesn’t teach, it only helps practice what you have already been taught.
Conversely, I’m at a similar level in Dutch and I’d attribute a lot of my grammatical knowledge to Duolingo, especially modal verbs. However I have scoured dutchgrammar.com many a time because I learn well from reading textbooks. It hasn’t helped me at all with output though; I was quite bad at that until I moved to NL.
