Language Learning for the Dispassionate | Hacker News Reader

Language Learning for the Dispassionate

langliter.com

112 pointswkrause45 comments

Do you want content for Chinese? I have a lot of data I've gathered for use with http://pingtype.github.io that I'd be willing to share (and I admit that your web design is better than mine).

Learning Chinese has a lot more initial barriers, which I explain more in this blog post:

https://pingtype.github.io/docs/failed.html

Eventually yes, my vision for the app is to effectively use NLP to solve some common challenges readers have when learning a language. I'd like it to be more than a news reader with a flashcard system.

So to do that properly, I would need to learn a bit more about Chinese to understand how it would fit. Your docs do a good job explaining the problems you're solving, but since I lack even a cursory understanding of the Chinese writing system, I have a hard time following it all.

Impressive app btw, I clicked the "Advanced" button and can see there's a lot of functionality there.

Wow. What an odd and interesting collection of useful stuff on your site. I'm about to get lost in it for a while.

Looks like OP is the author behind the app, so I would suggest adding a "Show HN" to the title, for more kudos and feedback.

That said, congrats for the good work! I really like the idea, and I feel this problem very much. Currently picking up Swedish, and unable to find easy enough news worth reading every day, so I'm a bit stuck at that level "able to understand a bit but not enough to read everything".

Do you happen to have a roadmap about the addition of new languages? If Swedish was there I would totally go Pro.

Glad you like the concept and thanks for the suggestion. I actually had a blog post lined up more befitting a Show HN: https://langliter.com/blog/launch/

But I guess the combination of a graph and curmudgeonly article title proved too irresistible and this one got some play.

The only languages I have planned at the moment are Italian, English, German and Russian, roughly in that order. I think Swedish will eventually be possible, but I don't want to spread myself too thin.

Jag är också tränar hur man pratar lite svenska.

What resources have you been using? Klartext is a service that delivers news in accessible Swedish (sort of like simple English Wikipedia) along with pronounciations that I’ve found to be very useful - sverigesradio.se/klartext

I’ve also changed my phone’s native language to Swedish and I’ve been watching English Netflix shows with subtitles which has really helped with picking up new words. I’m also using the SVT Play app to watch Swedish language programs with Swedish subtitles to develop oral comprehension and improve pronunciation.

There’s probably no substitute for total immersion but since I don’t plan on moving to Sweden anytime soon I’ve found the above resources to be helpful.

Looks cool, thanks for sharing.

Duolingo Spanish has a new "Stories" section that quizzes you in the context of a story.

https://stories.duolingo.com/

I'm also digging their new Podcast, which tells stories, NPR-style, alternating between Spanish and English.

Nit pick: shouldn't marginal utility be the first derivative of total utility? That bend in the marginal utility near "ventriloquist" looks wrong to me given the total utility shown.

Ha ha, yeah you caught me. I was fighting with the Vector tool in Sketch. My old econ professors would be disappointed.

I have been working on something very similar for the past year.

I first build this prototype with the goal of generating sentences and then analysing the user translation because I wanted a feature to see if the user knows the word even though he misspelled it.

http://ling-academy.com/

I finished that 3 months ago, received feedback but its just a proof of concept and since then have been building v2 which is very similar to what you have build - the idea is to have users learn vocab and then match them up with fitting text. This v2 is almost done but not deployed yet.

I do however plan on going much much further then this. I have been looking for a co-founder who is also invested into e-learning and language learning. Maybe we should team up? Email is:

Michael . baumgarn at gmail . com

Thank you so much for this. I'm an unabashed linguistic philistine. I see language at best as a useful tool and find no more beauty in it than I do in a hammer. But I'm also a long term expat and I feel kind of rude barging into these countries, taking the jobs and the women and not adapting the culture at all. This is exactly what I need.

Before now I've half-heartedly tried to write a script that would give flash cards based on a combination of word frequency in common texts and the amount of times I've gotten the word correct in the past. It didn't work. I should try your app.

Funny you should mention that, in about a week I'm planning to make public a frequency list collected from a corpus of over 40k Spanish news articles (I've been working on Langliter for a long time). Since I'm finding the lemma of every word in the corpus and I'm tagging each word with its PoS, it's actually a really valuable list. I plan to update the list quarterly as my corpus grows.

I'm currently only using it to calculate a modified version of the Dale Chall readability score. Figured people could make good use of it and it would be a good way to drive people the the site.

That would be absolutely amazing. I can't wait for the German addition to the site (current learning project).

How much hand tuning does this process require? If you had a corpus of German news articles, for instance, could you just run the software and receive a sorted frequency list?

I benefit from the fact that articles are processed and tagged at the time of collection. So getting this list is just a matter of looping over the structured data that make up the articles I send down to clients.

Just letting you know that I've uploaded the first frequency list. Just Spanish at the moment, but will be adding more over time.

https://langliter.com/blog/frequency_lists/

Can you go into any detail around how you curate the Portuguese content? The icon has both the Brazilian and Portuguese flag. One thing I've had difficulty with is that there's just enough differences between the two that mixing them can be confusing while learning. Lack of European Portuguese is why I basically dropped Duolingo and settled on just reading books/listening to music. The dialects are super similar to those who already know the language, but when even basic nouns differ it can get hard!

Just downloaded the app and it's awesome. Thanks for making it!

Glad you like it so far, and that's valuable feedback regarding Portuguese as it's not a language I'm personally studying.

I'm leveraging RSS feeds from popular papers in the languages I'm supporting with some weights applied so no single source is overly dominate in the feed. That said, I need to add more Portuguese and French sources, which is something I'm planning to take on this weekend. Right now the only paper from Portugal I've included is Correio da Manhã. If you have a favorite paper, let me know and I'll look to add it.

In terms of control, you can filter by country of origin. Please let me know if you think that is sufficient to address your concerns or if there is more I should be doing to separate the European and Brazilian Portuguese.

Thanks for this! It's nice to see more tools out there for advanced language learners because, like the article states, it sure is a long slog isn't it. I especially am happy to see that this app adds the Lemma https://en.wikipedia.org/wiki/Lemma_(morphology) of the new vocabulary to the spaced-repetition flashcards rather than the particular form it appears in on the page. That has taken up a lot of my time with other similar apps.

Thanks, the lemma issue is literally the reason I started building Langliter.

The idea sounds great, but the app description definitely should have mentioned what languages are available. I spent five minutes fighting with the faulty OAuth only to find that German is not offered.

It is stated clearly on the front page (through a list of flags)

Sorry about that, I really should keep an updated list of supported languages with the call to action download at the bottom of the article. Thank you for bringing this to my attention.

I'm looking to support German the first half of this year. It's currently number 2 or 3 on my new language priority list. If you'd like to get an email when it's available, there is a form at the bottom of the homepage that let's you fill out what language pairs you're interested in.

I would make that form a bit more visible. If I wouldn't have read this comment I would have left the website and didn't come back.

Looks like a great app; while I'm learning german I will definitely sign up when it becomes available.

Any plans to support Asian languages?

The short answer is no, because I don't know enough about those languages to deliver a valuable experience.

I think the value from Langliter comes from two areas:

1. The NLP processing and tooling built on top of that processing 2. Its superior mobile experience

If I just used a simple form of tokenization, I think I could very quickly deliver a superior mobile experience than what is currently available from similar apps. However, I'm not sure if PoS tagging and some of the other techniques I'm using translate to those languages or if they do, how much value they'd add to the learning process.

But it's something I'll certainly revisit if the app ends up being popular and I have the resources to do it properly.

Awesome work! I am learning a language and have been fantasising about a tool like this.

What about support of non-latin alphabet languages? I'm learning Dari at the moment. It uses Arabic alphabets, the words are a string of alphabets, there is tenses, verbs, nouns, adjectives. It's different to English only by the alphabet used, the vocabulary and the ordering of the verbs/nouns/subject.

Is it possible for it to be opened up to third parties to contribute new languages?

It's definitely something I hope to get to explore. Langliter is only a side project at the moment, so I'm focusing on languages that I have a little experience with. But as long as the licensing is compatible, I've built Langliter to be pretty flexible in terms of the types of tags it could support.

This is a great tool! Can I ask what kind of database and backend setup you're using?

Do you choose articles, then run nlp over them? How do you get dictionary results?

I really like being able to quickly and easily tag words for learning. Thanks. I'm a polyglot and completely relate to the premise of your app and system.

I used the iOS version but an curious, are these native? React Native?

I'm running it on Google Container Engine, Go for the main API layer and CockroachDB for the DB. I actually have a blog post in the works for why I chose CockroachDB. The tl;dr Langliter is offline first so most of the data is synced in the background. Therefore some additional latency makes no real difference and the resilience of a CockroachDB cluster is really impressive.

And good eye, it is React Native app. I should probably look into following platform style guides eventually ( I wouldn't count UI design as a real strength on mine :-) )

At least, as far as Korean is concerned, there is a version of the Penn Treebank for Korean: https://repository.upenn.edu/cgi/viewcontent.cgi?article=102...

And of course there are the Universal POS tags, although I'm not sure if they are precise enough for your intended purpose.

Well if you ever get a chance to do this for Indonesian (uses the roman alphabet), I'll deff use the app or at least be willing to try.

As someone living in Indonesia for about two years who has tried other apps, buying grade school books, nothing is interesting to me once the novelty wears off, and my incentives for learning more than navigational stuff and greetings is pretty much nil.

I am currently working on a project that lets you paste any chinese text coming from any source (for me it's mostly from news media and wechat moments), and provides you the translation, automated reading, word-by-word translation, and allows you to select words to memorize by a spaced repetition system.

For me the real deal is in the memorize function that gives you a word plus a sentence coming from your collection of text to get some context. You get the same word-by-word translation on the sentence, you can play it out, you are not only working on your words but discovering new words from the sentences and improving your listening comprehension.

For those interested in languages, I’ve got an iOS Word Search game that allows you to play in English, French, German, Italian, and Spanish.

https://itunes.apple.com/us/app/h4labs-word-search/id1311744...

I noticed that there is some interest in Russian here. I can add that in the future. There wasn’t enough room on the controller so I left it out.

I hope that language education services don't forget the desktop experience. As someone who spends 8h a day online, I have times where I go home and don't want to look at a screen.

I want a good desktop experience. I'd love something that would ping me every few hours to make me exercise my language practice.

The Internet + Yomichan + Anki integration is basically this for Japanese. You read real texts, Yomichan helps with missing words and with one click you can add them (with context) to Anki.

Add support for Russian and I'm in. This looks like exactly the language app I've been dreaming of--real language-native content with quick lookups for stuff you don't know.

Glad you like the concept! Russian is currently language #4 on my list, and I hope to add it in the first half of this year. There's a form at the bottom of Langliter's homepage that lets you select which language pair you'd like to see supported. That'll let me can shoot you an email when it is ready.

The concept seems really interesting and I've signed up to try and improve on my Spanish. I currently live in Amsterdam though, and there's no option in the drop-down for learning Dutch. Is it likely to make it onto your list?

Thanks for checking out the app. Dutch support would probably be a ways out. I think I'm going to need to eventually decide if I want to support languages before I have a lemmatizer in place. For me lemmatization is Langliter's killer feature and it's core to a lot of what's going on in the app.

But others might be okay with just a reading app that works offline, provides PoS highlighting and let's them build a vocab list (even in a less refined way).

Nice. I'm also making a language learning application and getting "real" content (websites, tweets?, news, etc) into the app is definitely one of my main targets.

What's your language priority list? I'm learning Polish

Italian, English, German, Russian

German might happen before English depending on how much ground I cover supporting more navigation languages.

Beyond that I'd need to do some research on which languages to support.

Any chance for Arabic in the future? Looks like a great app!!

Thanks! I'm not planning to support Arabic any time soon, but from a tools and data set perspective it's possible. I'd just need to invest the time to research the language and talk to users about challenges specific to that language. Hopefully the project is successful enough that I get a chance to do that.

Installing on my phone now. My major beef with duolingo was that it felt too much like a game. Goddamnit, learning isn't fun and don't try to make it be!

Reply on news.ycombinator.com