SoYouLearn – A flexible and advanced spaced repetition platform
soyoulearn.com
soyoulearn.com
A few key differences:
- A central DB (it has over 1.8M flash cards) which anybody can edit. Card owner can accept/reject the change request.
- Courses are just references to flash cards, nothing else.
- You decide what you want to learn "Display me field X,Y,Z and prompt me to enter field 'A'. You can choose any field/audio)
- Efficiently add bulk audio to a course
- A super fast UI (it's an SPA)
I have a few ideas for premium which (as far as I know) would bring features that don't exist yet for any spaced repetition platform:
- Prioritized learning (learn the most pressing ones)
- In advance learning (You'll be gone for the weekend, but it's Friday and you have 1h time: You can prelearn "long term" flash cards that'd come around over the weekend).
Also for some tech background: Frontend is pure Clojurescript + Datascript. Backend is Clojure, Cassandra, Datomic, Elasticsearch. I'd use the tech stack again, hands down.
The best introduction is the video at the bottom of the landing page it show all the features. There are only few courses right now so early adopters should (hopefully) be willing to create their own courses.
Questions welcome.
Where can I download the freely-licensed data?
Windows 10/Chrome browser
Edit: Fixed, Rate limiting was hitting for some reason. :/
Looks really cool, congrats on building it!
Can users make private forks of cards? For example if I liked some set of cards as a base but needed to make some alterations for my own purposes
Also, I was not able to create a new category. I am learning Circuits right now, which is not an available category.
I think Latex support is interesting. I could lazily load it and then render it. Something to think about for the future...
The interesting differences between, say duolingo and what a sign student would want to see:
* Flashcards are okay for some signs but plenty of signs are much better seen animated
* some signs are complicated and you need to see the whole signer from waist up; others are simple enough a simple image would do
* Short youtube embedded clips are okay except loading times are aggravating
* clips should be high res; some finger actions are small and detail can be hard to see
* there are variations between different sign languages, eg BSL and ASL
* source material is hard to get but there are a few sign teachers working on their own, one video at a time. They might be interested in partnering.
I'm a heavy user of Duolingo. Been trying to teach myself Russian for the past 2-3 years. Almost use it on daily bases (I have missed a few days here and there by accident or when I had no connection). Usually my repeat usage is anywhere from 20 days to now 143 consecutive days. My biggest frustration with Duolingo (Russian language) is that none of the grammar rules are explained. So even thought I'm able to repeat something, it's only after I google variations of a verb or a noun and the grammar case, is when I get the answer why something is, for example, ending in a "e", etc.
I have not paid the premium fee for Duolingo. Mainly because the premium adds no value to my learning. I've spent over $100-150 on Russian books from grammar to vocabulary.
If Duolingo added more grammar explanation for premium price, I would pay for that.
It takes Duolingo a few years to roll out a new language. It's hard to do it in-house. And still, with Russian language, they don't have all the grammar hints that exist on other languages like French (I've been told, French has it).
My view is, why not turn Duolingo into a marketplace and let other language expert create additional educational material to augment the lessons and let them sell it to me and follow the Apple's revenue share model. Perhaps a specific lesson might have 3-4 additional learning modules that I can buy from different providers. Each module has a star-rating (you can filter out the fake stars by how often someone is using Duolingo). Also once I find one provider's way of explaining, I'll look for more teaching modules from them.
My point is, if you can look into Udemy and how they turned their video system into a paid learning platform, you might come up with a premium model for your business.
Good luck!
My wife has used all three, and says that grammar explanations are not on iOS and is on the other 2 platforms. She also believes that the explanations make us feel good, but don't actually help. (Then again she's using it to learn Polish and already knows Russian. So the grammar rules are pretty close to what she already knows.)
Plus there are some natural language rules that are just a mass of exceptions. For example try to explain to a non-native speaker why you ride in a car and on a bus.
And furthermore, most native speakers don't know their own grammar rules. For example why do we say "big red truck" and not "red big truck"? Odds are that you've never been taught this order, but you do it correctly:
Quantity or number
Quality or opinion
Size
Age
Shape
Color
Proper adjective (often nationality, other place of origin, or material)
Purpose or qualifierBut I can see it in a case where if you come from a language that has some grammatical overlap with Russian. It may not be as helpful, because she already has a mental language mapping for it. For English speakers or languages that some of these concepts are not native, it's quiet hard.
I do agree many native speakers do not know the rules. But they'e not the ones who would be creating the additional teaching modules. I'm sure there are plenty of Russian teachers who would love to make side money teaching Russian to non-native speakers and create explanation material.
>Plus there are some natural language rules that are just a mass of exceptions. For example try to explain to a non-native speaker why you ride in a car and on a bus.
Some grammar and syntax is more important than other. If I told you "I rode in a bus" I'm sure you'll get immediately what I said. If on the other hand I told you "I rided on a bus" it might make you pause to process what I meant. Learning all the rules of a language is generally a daunting task but getting the basics right can save you a lot of time and trouble. When are you supposed to use the preterite and when the present perfect? What's the difference between "I will do it" and "I'm going to do it"? Etc...
>And furthermore, most native speakers don't know their own grammar rules.
Sure, in the same way that I never wonder when I'm supposed to use the subjunctive or about the gender of nouns when I speak my native french while a foreigner could struggle with that. I don't think it means that grammar courses are not hugely important when you learn a foreign language.
Learning one's mother tongue by being completely immersed in it 24/7 for years as a child and learning a foreign language as an adult in a few hours per week is hardly comparable. If you want to maximize your result you should learn the basics of the grammar instead of going entirely by trial and error like a child would do.
Furthermore as an adult you're more likely to require the use of complex and diverse concepts because you're not actually a 3 year old child. That means that you need to be able to actually "invent" new phrases that are not like any you've encountered before. "I've been asked to get the blueprints and bring them to the 3rd floor, could you tell me where they are?". Eventually as you become fluent the rules fade and you get a more intuitive grasp of the language.
There is nothing inherently wrong with learning grammar (and if you want to, I recommend reading grammar books in your target language). However, it is unnecessary. Memorising grammar rules will especially lead you down the road of approaching language like a recipe. The problem with grammar rules is that there are a lot of sentences that are grammatically correct from a prescriptive sense, but that are idiomatically impossible to understand.
Programs like Duolino are nice because the give you vocabulary in the context of sentences. However, you'll never get very much beyond basic sentence construction because you don't have the larger context. You don't know when to use one phrase and when to use another. You don't know the situations where idioms naturally pop up. You don't know how to organise your thoughts and to express them in the target language as opposed to translating your English ideas.
It's really funny, but when I read a Natsume Souseki book, I can't help but think how similar a writer he is to Dickens, and yet if I try to translate his writing to English it ends up incredibly dry and uninteresting. The charm of it is the expression in Japanese. Especially because Japanese and English are nothing alike, I find that I can't speak Japanese fluently unless I abandon English.
The thing that makes people worry is that they think that their second language is not good enough to become fluent. It's important to understand that fluency and proficiency are different things. Children are not at all proficient in their native languages -- especially 3 or 4 year olds are awful. But they are very fluent. With 2-3 years of Russian, you will also be awful, but there is no reason why you can't be very fluent with what you have -- and that includes reading.
It's too bad we can't have adopted parents in our second language to read us bed time stories because that would be awesome. But you can read by yourself. When I was very small, I had to be made to read because it was difficult for me. Eventually I got to the point where people had to tear the books away from me. The same thing was true in Japanese.
The advantage you have by doing it this way is that you have an endless supply of natural language. Your only task is to try to understand what it is saying. What I used to do when I was learning was to scan a piece of text to find vocabulary I didn't know. I would take about 20 pieces of vocabulary and look them up in a dictionary (I actually wrote software to make that efficient, because I'm a geek). Then I used a spaced repetition drill program (my own) to remember the vocabulary. Then I would read the text. Most of the time I could understand it based on vocabulary alone. If I ran into new grammar I would enter the sentence into the SRS drill with my understanding of the meaning. Rinse and repeat. Every week or so I would go back and reread bits of the book, updating my SRS data with my new understanding -- and it's incredible how quickly you form new understanding. After I read a section of text about 5-10 times, I could read it just like I was reading English.
After you get about half way through a novel, you start to learn the idioms that an author uses. You can then often read several books in a row without having to study -- just read them. But then when you switch to a different author, it's back to study mode. Of course I didn't start with novels. I started with picture books (Momotarou!!!), moved on to manga (great for conversational Japanese), then novels for elementary students and I'm currently in a phase where I'm enjoying books that are in the Japanese junior high school curriculum -- but at that point you're pretty functional in the language. I don't actually study at all any more (though I really should practice writing my kanji).
Not sure if any of this is helpful to the OP as well. I spent a lot of time writing my Japanese SRS tool (now sadly a victim of bitrot -- I don't think it even runs any more). I never could find a good way to really efficiently bootstrap native sources of information, but things like a built in dictionary and a deinflection/deconjugation tool are life savers. Honestly, if I were to concentrate on anything it would be that because Anki is already pretty hard to beat on any other front. Focusing on making it easy to create cards based on material found in the wild would be a great differentiator.
P.S. Last word on the grammar thing -- I actually learned Japanese grammar eventually. I had to teach English grammar in Japanese, so that forced me to read Japanese grammar books so that I knew what the vocabulary was. It was actually a lot easier than I thought it would be.
I'm a native English/Russian speaker who took a few Spanish courses in high school and college, and recently decided it would be a waste if I didn't finally learn to speak it fluently, at least like a child. In some sense, I suppose I already do. I can read many things and spoken slowly enough can understand partially -- but there is some mental block, mostly in the smaller, connecting words and other "minor" details that are actually major and completely central to the language. So I drill duolingo from time to time but it seems like mostly a waste of time since I know a majority of the words, and the ones I don't know such as "curtain" aren't critical for me. I always had the idea to visit Spain for several months to force myself to adapt to the language but that's been impractical thus far. I knew reading would be good too and your story of learning Japanese is inspiring, so I think I will find some children's books or Mexican menus. Thank you
And LibriVox is free.
Unfortunately the process of doing that on the Kindle is so arduous that it is unusable.
I'm not affiliated with them in any way but with the number of people I've convinced to use it, I should be getting commission.
Take the entire text and make each paragraph or page a card and make many, many clozes (fill in the blank) for phrases, sentences or words. This will work as a complete solution all by itself but not as fast as doing it and attempting to recite your chosen text.
1. Let's assume: The flash card isn't new. You last review was 100 days ago.
2. It is due in 3 days but you have some extra time today! You have only 10 cards to review but but have some extra time right now and you're are gone for the weekend.
3. It shouldn't matter if you review the flash card today already even though it's only due in 3 days. If it ends up being 100 or 103 days... It doesn't matter. It's all just a probabilistic model anyways with much larger errors.
4. So in premium: After you've reviewed your most pressing/recently-learnt ones you can keep going and pre-learn your old ones.
There is a few more ideas I have but this is basically it.
The app seems incredibly responsive, but adding a card could be easier. And are you planning on adding Chinese language flash cards in the near future?
I'll check it out again later. Thanks for the hard work.
How's Datomic? Are you using the community edition?
Also about databases, are you using or have you considered using a graph for Cassandra like Titan?
I guess also, if you do a mobile app will you use ClojureScript with React Native, like Re-natal?
Thanks.
I have not considered Titan, I looked into JanusGraph the other day but I have intention to switch.
I dabbled in RN + CLJS a few weeks ago and got some toy screens running. So: Yes I'd use that stack for a native app!
How was your CLJS experience with JS interop? Would you still use ClojureScript if you had a lot of JS dependencies?
Interop is not bad, I don't do too much interop. I don't use npm-deps but just bundle the React builds separately and include them in the HTML. They rarely change and I set caching header to immutable so the browser will just read them from memory (my CLJS app changes multiple times a week). If you want to use a lot of NPM deps I'd recommend shadow-cljs, there has been a lot of development by the maintainer and it seems to work incredibly well for NPM deps. Feel free to ping me on the Clojure slack, there is also a beginners channel that is very active. And there is a shadow-cljs channel where the maintainer will quickly respond.
All that said the best way to increase native vocabulary is to read more.
One of my pet peeves about new projects is when there is a complete lack of information on eventual pricing.
Even if you don’t yet know the pricing you will end up with, are you always going to have a free core set of features?
Are we going to lock in to your system only to face big charges when you leave beta, etc.?
It was a fantastic idea that ultimately requires a lot of volunteers/manpower in constantly keeping things updated and pruning/merging duplicates. Eventually they (Memrise) moved to curated dictionaries that course creators could then pull from to make their own courses without affecting other course creators' copies. Creators can add new words but their words will not be automatically added to the curated set.
I wish you the best of luck! You may want to find curators, equivalent to higher-standing Wikipedia editors to make sure the word databases stay (1) accurate and (2) everyone can actually benefit from it without fear of selecting the wrong "Monday" in a list of 19 "Mondays".
A very large portion of the "original" Memrise community were people who had been abandoned by smart.fm, a free-to-use site which later became iknow.jp (subscription model). Our motivation for helping was a mixed bag of mostly the following three reasons, though I don't pretend to have known every original contributor or their motivations.
1) For our own educational use
2) Because we wanted to be able to help others learn our language or a language we had interest in
3) Because we wanted to see the smart.fm community stick around in this new home, Memrise
Personally, I fell mostly under the first reason. I never found Anki good to use and preferred having a website to navigate to and easily sync across devices (even in the early days Memrise had plans for the app) by well... letting someone else handle things. So smart.fm/Memrise were more convenient for me and my learning.
I've played some part in many of the Japanese courses with 50,000+ users. So the work I contributed has helped at least some portion of them begin their studies and I take satisfaction in being able to have done that for them.
E:
I didn't answer all of your questions, so in short. Yes, their data was closed, but they were willing to share large portions of it (except user data, even anonymized user data!). My contributions have my "by {Username}" for my courses - although many of the courses I built/contributed to became standard "Memrise" courses under their name I don't mind getting or not getting the credit - so long as they continue to help people learn!
I had a bit of confusion with the Spanish course. I read through the basic categories, and decided to say "Claim Known" for one of them. It then asked me to "Choose a way to learn." This was odd, because I was just saying I knew the words, not in a particular direction (confusion #1).
I chose that I knew them English => Spanish. I then went up to the top to set "English => Spanish" for the whole course. This jumped me into the course, which I did not expect (confusion #2).
I hit "back," went back to the landing page, searched for the course page, and went to the course in order to try and learn a particular set. I scrolled down to one I didn't know, and clicked on "English => Spanish" to start the set. Unfortunately, the "Claim Known" button was already active, which I hadn't realized (surprising, since I felt like I re-entered the course from scratch). So clicking on "English => Spanish" sent those words off as "known" (confusion #3).
Now I have a list of words that I don't know marked as "known," and I don't know exactly what this means (confusion #4) and how to find these words to tell the system I don't actually know them (confusion #5), and I'm not allowed to learn those words even when I click "Learn Now" because it tells me I know them already (confusion #6).
Thanks for the feedback! This is very helpful. There are definitely many rough edges in the app that will be improved. Regarding the UI: The React components state are actually always "remembered" so if you're in the middle of doing anything (like creating a course) you can go anywhere else and come back and find your UI in the exact same state. So it's feature but seems in your case was probably not optimal. I'll try to improve the UI around course learning!
A central database is also a disadvantage as far as I'm concerned as it means that copyrighted material is made significantly harder to study.
When you add to these the fact that Anki can be rewritten with Python plugins s that can change just about every hit of behaviour, interface, and add completely new functionality, I'm sorry but there's really no reason for this to exist, besides having a slightly nicer web interface.
That said, while we're on the topic of Anki did you know the cards can execute javascript? (Since Anki uses webview to display the cards).
Python for plugins with css/js in the cards themselves gives incredible power and flexibility to your flashcards.
How would you advise a motivated, but time-limited person who wanted to learn a language to B2?
So although I like Anki overall I would welcome an alternative to it.
You can view ankis webapp here https://ankiweb.net/. There's almost nothing on the homepage. Just a login. Not exactly enticing for users looking to review their cards online. Anki's web interface is extremely barebones. Its more or less the equivalent of a dropbox folder or an amazons3 storage box. Not a place to share, debate about the writing / content on cards, a forum where people suggest new features on ankiweb, etc. Its also not the place you preview your cards either. Ankis webapp doesn't let you upload images through there. Everyone puts images on their cards via desktop app
Ankidroid, a 3rd party app, makes anki extremely useful. I currently make all my cards on desktop, and then preview all my cards in the morning going to work everyday.
Anki is also extremely client-side customizable, which is one of its biggest pluses. I currently use about 5 or so plugins to make data entry more streamlined. I have some other plugins that read/writes to the database that affects both ankidroid and ankidesktop app.
Comparing this to "soyoulearn" or "memrise" or "duolingo" the site is much more inviting for community driven features on its webapp.
However, the problem with a webapp with a centralized database is two fold.
First, it degrades the quality of the cards and card deck. Imagine you had to study for test in college. You could write your own cards, or borrow your friends card who took this test last year. The content of the test might be exactly the same, so the cards are still relevant. However, since you didn't create the cards yourself, you might not necessarily understand the context of why it was made. What specific situations that card is useful, and the hard earned experience to make it. You don't have the same appreciation of a card created by someone else than if you made it yourself. You don't get the "aha!" moment when reviewing someone else's created card, e.g. "Oh I made this card in the university Library and my friend so and so said this great idea that I jotted down on the card".
Second, it encourages laziness during card creation. Why create a deck when someone else has made one? Especially if its community driven and approved.
---------------------------------- ----------------------------------
TL-DR
In summary - a community driven flashcard webapp (soyoulearn.com) is only better than a customizable client-side app (anki) for the following reasons:
- If you don't want to make cards, and don't mind shallow content behind each card to save yourself time
- If the cards you are making are predictable in nature and well established( e.g. cards for a specific course, languages, etc)
- If you don't mind purely memorizing a set of cards without having a physical memorable moment attached to that card, and understand that these cards have less value than ones created by yourself.
A client-side app like anki will always be better for the following reason
- If your card was hard-earned, e.g. your a programmer and made a error, you want to remember how you solved it
- You read a niche science paper that only a handful of people look at, and draw meaningful insights from it
- If your card's content is extremely niche, meaning there's going to be few people making cards related to that content
- If your card might possible be outdated overtime (e.g. the changing state of javascript / webdevelopment - your card regarding a specific library might be irrelevant next year), and you need the ability to fine tune your deck easily (suspend card).
- Your cards are created at random, they aren't created based off of a syllabus or university course
- You want to have the highest quality deck possible and scale it upwards
I think both have their pros and cons. For me, I would definitely prefer to dump all my language learning on a webapp / use premade decks, but all my programming / hard earned experiences on anki. As of now I seperate my cards based on the context of how they were created - were they created in a bulk, batch process (e.g. memorize a set of APIs), or did I have an actual question, and I created that card as a result?
Otherwise I mix all my cards in one deck
I have far too many things ranging from professional to personal to menial that I need to jot down. A lot of it I want to remember, some of it I just want to be able to look up, but ultimate there's just far too much information in my life and I need a personal Google.
I'm not sure this product is my exact needs, but I'm really happy to see more people thinking about personal knowledge/data retention. It seems so important to me!
Also, how well does it work on a phone? I'm assuming it's a PWA...?
Also, is there a way to "favorite" courses? From a UX perspective, "My Courses" feels like it should be "courses that I am currently learning". I don't expect that to mean "courses that I have created", especially since a small minority of users will actually do that.
It does not yet use PWA, though I'm intending to add it, so that I have offline support. It does however work pretty well on mobile right now. The entire SPA is rather small ~180KB (+ React). On my old phone it refreshes the entire app (initial load is slower) in 1 to 1 1/2 seconds. Reviewing on the phone is pretty good and I use it myself.
The course you're learning will be under "notebook". My course is really only the courses you created.
how to leverage internet/web to enhance education and self-life-long-learning is indeed still waiting for more creative adventures.
for me I'm still looking for a markdown-native CMS with authentication, not the static site generator as they do not have login thus not a CMS, not the "old" CMS with html editing etc either as they do not have native markdown support. so far I found none though.
Markdown makes sense for technically inclined folks.
Educators who are not tech saavy are likely still better off with WYSIWYG.
Edit for the downvoters: Speak up - I have to work with both tech skilled and low digitally skilled educators.
Care to share your stack/toolchain? front-end/back-end/API's?
Backend: Cassandra + Datomic + Elasticsearch + Redis + Clojure.
SAAS: Mailgun, Datadog.
Editor: Cursive IDE (hands down best Clojure editor)
Saw this in the website:
> Note: All current functionality will remain free.
> Premium functions are available with a monthly or yearly subscription.
> With SoYouLearn premium you will be able to learn even faster and better.
edit: add the free bit from the website and formatting
But there is a feature I want badly: a KDE5 plasmoid widget to show the cards on my desktop, switching them every now and then so they will catch my attention and get memorized as I use my PC (I use it about 10 hours a day and I believe many other hackers do the same) without me having to dedicate time for this. I also believe other people might also appreciate similar desktop widgets for Windows, Android and other platforms.
Here's a bug: Your search uses regular expressions without validating them.
1. Go to "https://www.soyoulearn.com/app/#/cards/search/"
2. Type: c++
3. The autocomplete errors and page goes blank
4. Inspector Console: "SyntaxError: Invalid regular expression: /c++/: Nothing to repeat"
Here is what I learned in the last 10 years in this field. I hope you will find this valuable when developing SoYouLearn.
- Language Content vs Features: We as developers like to focus our time and energy on implementing great flashcard features but often the language content itself is a second thought. Odten we do not curate the content ourself and let users do it or we prefer to use copyright free content. From our experience, the quality of the content is extremely important. Your students will quickly become frustrated if they see errors in the language content and then think that your whole product is of low quality -even if you have a great technical implementation! So make sure you have high quality content!
We have invested heavily in the content of our app and even wrote Grammar notes for nearly 20 languages -not because WE wanted to do it (we probably would have been much happier coding ;) but because OUR USERS where asking for this -repeatedly.
Also from our experience, students are more willing to pay for high quality content (e.g. Top 10000 Spanish Flashcards) then for advanced features. This is true especially for Flashcards because ANKI is well known, working very well and free. So why pay except for great content?
- Focus on Languages OR other Areas: Yes in principle, spaced repetition works for all kinds of contents (languages obviously but also if you learn geography, anatomy, heck even for learning coding you can use flashcards). However once you want to built a powerful and more importantly user friendly system, there are important differences. E.g. when you learn languages like Chinese you need to provide a field for Pinyin, you should have audio associated and maybe also an example sentence. So what I want to say is if you focus your system on ONLY language learning you can make your UI much more user friendly.
I think you might come into troubles if some users use your product for language learning and others for learning other things as then your users will have different priorities on what features they need and what you should implement.
- Include Motivational Aspects within your App: Learning something completely new is tough, Language Learning is even much tougher! A very high percentage of language learning app users give up very quickly. It took us many years here at L-Lingo to realize this and we are now spending approx. 50% of our time and effort to keep our users motivated to follow through on their language learning.
We do this in many different aspects, gamification of the language learning, learning reminders, giving our students tips on effective language learning etc.
Keep in mind: Even the best flashcards system does not work if your students are not motivated to use it regularly!
- Beta testers and early adopters I think one of the key successes when we launched L-Lingo was to actively looking for a small but very engaged group of early users. And we focused on a small niche (Thai Language Learners!). I think this helped us to create value for this group early on, to understand what was really important (e.g. the quality of the content). So you made the first right step with this hackernews post but now try and find real permanent users that you can engage with regularly.
Thats all for now. Feel free to ask questions if you have!
Also, it would be nice if I could choose the correct answer (e.g. a county name for a flag) by clicking it among a set of options instead of typing it in manually all the time.
I don't currently have an option for multiple choice learning. That's something I might do in the far future.
Yes, enter should submit the form. I think I had `onSumit` defined at some point but it gave me issues so I reverted it. I'll look into it again and make sure it works. And yes, please feel free to contact me!
What will the pricing model be?
I'm reluctant to contribute to something that is going to be used as a profit center.
more_vert This is not a valid session ID. Note that this URL ist not valid across browser reloads.