How hard is it to build an open source alternative to Duolinguo?

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)
Plus you'd have to hire some psychopathic assassin who'd work pro bono.

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.

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.
Permanently Deleted
- 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/
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.

Check out LibreLingo and these

I did know about LibreLingo, but it still feels like it's taking really small steps! And the other alternatives don't seem to offer the semi-holistic experience that Duolingo provides.
- Join LibreLingo instead of creating your own scratch. Unless you're planning to use Rust lol
- Elaborate on “small steps”?

Let's not overstate Duolingo's effectiveness for language learning.

The technological challenge to adopting a self-taught language learning method into an app is rather small. You just need the content. Either you develop the course under a Free Culture license, or you purchase the rights for an existing method and you port it. Plus maybe some volunteers to handle user-interaction.

A good example is the VHS Lernportal which implements three levels of German class in a way that actually has some pedagogical merit. It's killer-feature is nothing technological, but that they have some teachers in the backoffice that will read your occasional text-production exercises and offer corrections (no, language tool wouldn't be able to replace humans in that case, because language tool doesn't know what you are trying to say and therefore gives you multiple guesses but no way to know which one you actually need).

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.

As jet said the difficult part is the content, not really the "technology" (which is mostly just flashcards tech + quick "fill the void" exercises, and we already have software like Anki). There's probably the gamification aspect too.

There are lots of free language learning materials on the internet. For Japanese for example, there's Tae Kim's course: https://guidetojapanese.org/learn/ (his grammar guide has a CC BY-SA 3.0 license, his "complete" guide has no license yet as he hasn't finished it yet). It would be great if course authors could allow open-source devs to build a unifying app on top of their courses (it does seem that Tae Kim is okay with it, since he allowed an Android dev to make an (now discontinued) app version of the course: https://play.google.com/store/apps/details?id=com.alexisblaze.japanese_grammar).

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.

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.
- Thanks for the insight. I can see that, maybe a vocabulary builder.
  
  Though I'd bet your time would be better spent doing other things like Anki or consuming media in the language instead, though some people just need/want the gamification.
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.
- I've been learning French on there for awhile now and it's been extremely effective por moi.
  
  I wouldn't normally comment on a spelling issue, however, in this case...
- 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've used Language Transfer with good success. Don't even need to sign up, you can just go, click on a course and start streaming lessons. You can also download them locally.

https://www.languagetransfer.org/free-courses-1

The idea is to create rules that help you "transfer" words from the source language into the target language, hence, "Language Transfer".

For example, going from English to Spanish: Words in English that end in ation, will end in acion in Spanish.

Confirmation -> Confirmación Conversation -> Conversación

Further, words following this rule are "ar" verbs. Confirmation -> Confirmación -> confirmar

Another one is words ending in al, which came to Spanish and English via Arabic... are the same. Just said with an accent.

Normal -> Normal Formal -> Formal

A few rules might get you a few hundred words. And while some words might be more formal than how something is typically said, you should still be understood.

They're completely donation based, ad-free, and no sign-up required.

Looks really good! I'm sad they don't have Dutch though, that's what I'm struggling to learn
+1 for LT. The guy that runs it certainly has an open source ethos. The German one despite being a "Complete" series is frustratingly very incomplete, but that aside it was a useful way into the language. The word order explanations were particularly good. Everything is always free and the project as a whole is expanding with the help of volunteers and donations. It's a good thing to be a part of.
No mandarin offered though

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 don't think you need linguists for an early version (well ever, because linguists don't do language learning), I'm guessing early duolingo plagiarized existing language learning materials and crowdsourced the correctness of it to whoever they could get for cheapest. For larger designs beyond that, I'm guessing they used a mixture of instructors and data they'd collected. There would be labor costs associated with that, as well as infrastructure costs of storing data.

On the app side, it's more complicated than just text boxes because the app accepts voice inputs. Voice to text is pretty good, but dealing almost exclusively with people new to the language might complicate things significantly.