SRS is best used to reinforce things you learned outside the flashcard environment. So put things into Anki that you have some real-world context to anchor that memory to, and use Anki to reinforce that memory, not to create new memories that lack any vivid context in your mind outside of the drab flashcard app window where you first discovered the term in isolation. Ideally you're consuming native media, talking with natives, etc. and drawing from that to create your flashcards, since you'll come across a lot of terms that you learn once and then have otherwise forgotten by the time you see them again in the wild; this is where SRS helps.

I make software to help language learners generate flashcards for SRS while reading or watching videos: http://readlang.com

It works as a browser extension for reading web-pages or you can upload texts and whole novels to read via the web-app. You can then click-to-translate words and phrases that you don't understand, and it generates cloze flashcards for use in it's own flashcards, or for export to apps like Anki.

Would love to hear feedback from anyone who tries it!

I glanced over the site, and was really taken by the simplicity of the approach! I've used Aki in the past, but resorted to scraping the sites I was learning from to create flash cards. This looks like it really takes the pain out of putting material in.

As a memory researcher, I'd love to see a good example of how you (or someone else) implemented readlang to learn a language. The site seems good at conveying how readlang can get flashcards cooking, but it would be interesting to hear how it was used as part of someone's language learning process (the big picture). I saw on the about page that you used it to learn Spanish, and there are a bunch of posts on the site, so it may be there and I missed it.

I'll definitely look more into it later this evening, but a few questions/thoughts I had were..

1. is there a good way to programmatically pull out flash card information (as say a JSON object)? Is the export to Anki as a csv/tsv?

2. How fleshed out is support for Chinese? What should I expect from it being in beta?

3. I was intrigued by the video player functionality, clicked the "Find something to watch now" on the features page. Clicked blindly. Arrived on a page of text in Spanish. Backed up. Realized that it was a mix of text articles and video articles. Scrolled down to a video article. Was very impressed with the player, but thought with a little less patience I may have missed it. This seems like an incredible feature (similar to fluentU's approach), and one that the link should take people to with as little friction as possible!

Sorry if any of this should have been clear from a more thorough read. I didn't have a lot of time to look, and was really impressed, so thought I would fire off my impressions before giving it a more thorough look :).

Thanks for the feedback!

I agree it would be nice to include examples of how Readlang fits into a different people's language learning process. Here are a couple of articles I found online:

- http://www.languagesurfer.com/2015/01/14/readlang-review-six...

- http://www.alexstrick.com/blog/2015/9/surviving-middlebury-h...

To answer your questions...

| 1. is there a good way to programmatically pull out flash card information (as say a JSON object)? Is the export to Anki as a csv/tsv?

Export is by CSV (or you can specify your chosen delimiter) and you can choose from a number of different fields. You can't export data from the spaced repetition algorithm since I felt it would make the UI confusing. But you can access this data via the API: https://github.com/SteveRidout/readlang-api

| 2. How fleshed out is support for Chinese? What should I expect from it being in beta?

Chinese, Japanese, and Thai aren't that well supported at the moment. The main omissions are:

- Lack of "word" boundary detection (these languages don't use spaces to separate words) - Lack of Pinyin translations - Lack of word frequency lists to prioritize flashcards by usefulness.

| 3. I was intrigued by the video player functionality, clicked the "Find something to watch now" on the features page. Clicked blindly. Arrived on a page of text in Spanish. Backed up. Realized that it was a mix of text articles and video articles. Scrolled down to a video article. Was very impressed with the player, but thought with a little less patience I may have missed it. This seems like an incredible feature (similar to fluentU's approach), and one that the link should take people to with as little friction as possible!

Thanks, glad you like it! I agree these should be more discoverable. BTW: These videos are all added, sync'd, and shared by Readlang users using the web-app, here's a short guide: https://www.youtube.com/watch?v=szcvArpfxWI

Word boundary detection in Thai is indeed a thorny problem.

With Chinese, though, it should be simple. Just allow users to make flashcards of any number of contiguous characters. Unbound morphemes aren't something you need to worry about and even a single character part of a larger word could reasonably be a vocabulary item.

That is exactly how it works now - users drag to select a sequence of contiguous characters. I haven't tried learning Chinese myself so I don't have a good handle on how pleasant this is to use.

Chinese traditionally had no bound morphemes and even now tends not to. I think dragging a sequence of contiguous characters should be good. Most words are two characters, but some are one, three or four. Many four word characters contain words within them.

This system should work.

I love it! I was looking for something exactly like this a while ago but I didn't find your site.

I'm especially impressed by the accuracy of contextual translations. I've just done a few simple texts, but it always provided the correct translation in context even for words with several meanings. How does this work?

It's simpler than you think. It uses Google Translate to translate the word or phrase you select and the context isn't used. This works surprisingly well most of the time. Of course it does fall down, and can be confusing beginners who don't have a good grasp of the basic grammar. For intermediate and advanced users who just need the occasional gap filled it it's great, and they can usually tell when the translation isn't correct and either expand the phrase to add more context, or make use of the additional sidebar dictionary: https://readlang.uservoice.com/knowledgebase/articles/279539...

Asks for premium before I really tried.

True, this can appear very quickly for users who translate a lot of phrases.

You get unlimited single word translations and flashcards for free, so the free plan is actually very usable. But if you want to translate more than 10 multi-word phrases / day, you need to upgrade to premium.

I think you're absolutely right there: constant immersion is needed to provide the context for learning. My mistake was that I wasn't reading enough Japanese to make use of the memorised characters from Anki, so they were quickly forgotten.

There still remains the problem of review card creep, though. After adding about 5 new cards a day for a year and reviewing them every day, I ended up having to review hundreds of cards a day. Your suggestion might fix this too, though, now I think about it. If I'm using the memorised information outside of Anki (by reading Japanese articles, etc), I'll remember them better, leading to less reviews of those cards in the future.

For me SRS worked splendid inside the flashcard environment so far. I'm pretty solid on 500 japanese characters now, but I also learn related words of the characters simultaneously.

> soon becomes a pain when you have hundreds of cards to review every day

If you have hundreds to review each day then you've gone nuts adding new items too fast. I know, because I've done exactly that and have seen tons of other people make the same mistake. Adding something like 5-10 new items per day (max) keeps things manageable in the long run. I would suggest that if you're spending more than 10-15 minutes in review that you stop adding new items until your review is under control.

Another problem I've had that seems in common with others is that using SRS is somewhat addictive, feels productive, and gives you nice concrete numbers to gauge your "progress". Many people, like me, fall into the trap of spending more and more time doing SRS, displacing other parts of the curriculum.

> However, the characters I learned while living in Japan, in context, are still fresh in my mind.

So much this. I haven't lived abroad, but I listen to podcasts and other media in my target languages. Things that I've heard and looked up and then heard again multiple times are pretty much permanent now. However memory works, this seems to trigger "this is important and useful, don't lose it!"

For learning foreign languages, I found TPR (Total Physical Response)[1][2] to be very effective.

A great intro to it is an article called "Leave me alone! Can't you see I'm learning your language?"[3]

[1] - https://en.wikipedia.org/wiki/Total_physical_response

[2] - http://www.tpr-world.com/mm5/merchant.mvc?Screen=CTGY&Catego...

[3] - http://wolofresources.org/language/letmealonech1ch2.htm

Thanks for the heads up, these look really interesting.

I'm learning Russian in my spare time, and have used the goldlist method to build vocabulary. The goldlist method is great, because 1) There's no cramming. Nothing is worse than cramming, and 2) you can have breaks as long as you want. I reached 2200 words the first year, while having many week long breaks. I kinda stopped the regular vocabulary building around 2014, but even after months of inactivity, I can easily pick up the goldlist and continue where I stopped.

If you're trying to figure out Goldlist and having a hard time, this video helped me understand the actual method much faster than reading the 22 step list.

https://www.youtube.com/watch?v=Ixxq8moh4pg

I think you aren't supposed to start with hundreds of cards. Start with 10, once their frequency drops to once/week or less, add 10 more, repeat ad nauseum. Since the frequency of cards drops integrably, you'll always have a constant and manageable number of cards to review.

I was adding 5 new cards a day, which seems to be quite common amongst language learners. I figured the SRS algorithm would take care of giving me a manageable frequency.

Interesting. Have you ever tried Heisig's remembering the Kanji method? I'm using that for the characters + skritter, and they're sticking fairly well. Though it's only the writing + keyword so far, no drawing.

I think Heisig says you will forget eventually though, and need to read some Japanese daily for life to retain them, ultimately.

I'm only using Anki to reinforce a Pimsleur course I'm doing, orally. I'm not sure that I'll use it for anything else, depends on the use case. I'm finding reviews pretty tedious.

Anki is still a great system, there are just limits to keep in mind.

Edit: The RTK method for Kanji involves making a story with the component elements of the character. So you use the power of imaginative memory to retain the characters.

I used the KanjiDamage method, which is similar to Heisig but available online: http://www.kanjidamage.com/. The advantage is that the mnemonics also help you remember the on-yomi for each character. It's not for everyone, though: the author has a filthy sense of humour! Personally, I found that made the mnemonics even more effective.

I've found that mnemonics really only help at the very beginning of learning. But once you even get to 100 things you need remember, a completely unrelated extra thing you have to remember on top of what you actually need to learn becomes too much.

For kanji, just remember the radical readings, and memorize vocabulary with the different readings of the kanji you're trying to learn. Knowing that 貢献 sort of kind of sounds like cocaine does nothing to help me remember that it means contribution. But knowing 献血 and 献金 help me memorize that けん is a reading of 献

That looks like a very clever method. With mine I'll have to learn the readings separately. But to be clear....that's what did not stick?

I think my failure was more due to me using Anki, then abandoning it rather than anything else. It's a really good system. I tried Heisig for a bit, but KanjiDamage worked a lot better for me. Perhaps if I'd have combined it with something like the goldlist method it would have worked better for me.

I've been working on an iOS app that tries to make learning vocabulary a little more fun (ie. feel less repetitive). At the moment, I only have simple games like Match, Concentration and Hangman, for example. The goal is create more complicated, and enjoyable games, so it's possible to fire up an app and play games that'll help you practice the most common words. You can see pics on my site: http://h4labs.com

DuoLingo is probably the best thing if you're starting from scratch but I feel like something is really missing to help people maintain what they've learned.

Well, the characters are useless by themselves. I made myself a deck that contains both kanji and the words made from these kanji [1]. This way I can learn how to read the characters within various words, although it would take much longer to learn all the kanji. I'm currently only up to the 4th grade level or so after many months of learning, but the ones I did learn, I feel like I know them pretty well and can easily read them in various contexts.

[1] https://ankiweb.net/shared/info/831167744

Not a brag, just a point: I learnt that many characters in a year too, without Anki. I also devoted an hour a day to learning.

My point is: devote an hour a day to any endeavor and you'll go far.

I find that link very frustrating. The writer just babbles on about this "method" but I can't find anywhere he concisely defines it. Well, spells it out clearly at any length -- I don't think "concise" is in his vocabulary.

It certainly isn't the solution to all problems. But it's good for getting a lot of words in a short time. Then you use and apply that knowledge and that is much easier after having a few hundreds words more in your repertoire.

I think part of the trick that's commonly overlooked is mixing the two approaches. Use SRS so that you remember the cards tomorrow and can use them in Japan.

I think this could also be partially solved by making better cards.

That goldlist method is fantastic! Thank you for the link. Low tech is much closer to our bio than electronic flashcards.