Kongzi vs. Anki

Updated March 19th, 2012

This is a review of the free language learning program, Anki, compared to my own program Kongzi. The original article was written in 2007, but has been updated with the latest information as of March 12th, 2012.

You can find Anki at: ankisrs.net (old link: http://ichi2.net/anki/)

I’ll start by making some general comments about Anki. Over time I’ve realized that development of Kongzi has in some ways been pushed by what I have seen in Anki. That is not to say I’ve copied anything — far from it. I just admire the quality of work which has gone into Anki. Anki has come a long, long way in the last five years and I am of the impression it is currently the most popular free flashcard program in the world. Since Kongzi is about to be released, I would like to compare the two and see if my program meets my own expectations of being “worth releasing”. Why would I release something that is inferior to what’s already on the market now? So this is sort of for myself, as well as everyone else.

Anki vs. Kongzi: Pros and Cons

Pros

  • Anki does flashcards well
  • Anki is Free
  • Anki uses SRS

Cons

  • Anki only does flashcards
  • You get what you pay for
  • SRS is not the best way to learn a language.

Anki does Flachards very well.
Anki is a flashcard program. It does one thing and it does one thing well. This is a good thing. It takes any kind of information you want to remember and presents it to you in a way which helps you learn it and remember it. This is a really great idea and Anki delivers on it’s promise of Flashcards extremely well. I would not be remiss to say that Anki is probably the best Flashcard program in the world right now.

One of the best things about Anki (besides how great it looks) is that it uses Spaced Repetition (SM-2, to be precise) to help you remember. That’s great, it can help you remember the words in the best way science has discovered. In short, “Anki really does flashcards.” Q and A, mathematical formulas, Japanese words, any kind of general knowledge. Of course, the focus here is on Languages. And yes, Anki does in fact excel at helping you memorize a target vocabulary.

But on the other hand….

Anki only does Flashcards
Anki presents flashcards to the user. Okay, and then what?

Nothing. No really. Anki just does flashcards. I could present a very long list of things which Kongzi does that Anki doesn’t do, but instead I will just point out the underlying fundamental way in which the programs were written. Anki was written to take A and B data and present it on the screen in flashcard format. But learning a language is about far more than just memorizing vocabulary. Is Anki capable of pronouncing words from a Native Speaker? No. Can Anki generate sample tests with targeted levels of difficulty? No. Can Anki recognize similar words and phrases and construct meaningful multiple choice questions? No.

Kongzi was written in part to try and solve some of the fundamental problems of Natural Language Processing. For example, when you are giving a Close test and the statement is “Jones bought a [_____] in the bookstore,” you have a, b, c, d answers. You add “book” and then add three random answers. But what if one of those answers happens to be “pen”? The user faces a problem; the user has no way of knowing which answer is correct. This just one of the fundamental problems which Kongzi has addressed in order to help you learn a language — not just memorize a list of facts. Kongzi fundamentally alters it’s behavior depending on which language you are learning. Anki is simply incapable of doing that.

Anki is Free.
Anki is a flashcard program with a very nice UI. It uses spaced repetition to help you remember facts (for example definitions of a Japanese word) as efficiently as we know how. And it’s free. On the surface of it, there’s no comparison at all. You have to pay for Kongzi, and Anki is free. Why on earth would anyone pay for Kongzi?

But on the other hand…

You get what you pay for.
Well, you get what you pay for. The two main issues here are quality control and “what you get”. Anki looks nice but how fast do they fix bugs? How fast do they add features requested by the community? The problem with being free is that what the programmer wants is more important than what the customer wants.

You get what you pay for: example 1.
Take a look at Anki’s Bug List for yourself. There are some issues there which are three years old and older. How does this compare to something you pay for? If a customer calls me up (phone support is nice, isn’t it?) or e-mails me, I will fix the bug the next day, or at least start working on it. Why? Because that’s my job. I work for you. A lot of people need this kind of support. And I am here to provide it. I deserve recompense for that, even if I don’t deserve recompense for the ten years of free time I devoted to writing this all by myself. The entire program fits in my head because I work on it all the time. If I need to change the way something works on a fundamental level, I can do that without breaking the program. And I love paying attention to the community. It’s my job.

You get what you pay for: example 2.
When you install Anki you are presented with the option of downloading some shared word lists. There is zero quality control here. The user is presented with a word list for (and I quote) “Beginner chines vocabulary”. And get this, it was last updated (as of writing) 1,013 days (2.8 years) ago.

Oh really? I think I’ll pass…    o_o

That level of negligence is completely unacceptable in commercial software. Spelling mistakes in major components of the software 3 years after release, is really just not acceptable to me as a programmer as well. Why doesn’t someone fix this in Anki? It isn’t that they don’t care. They probably do care and this probably will be fixed, one day. But the writers have lives. No one pays them to fix the bugs. So it just isn’t a priority. And some things just slip under the rug.

For years.

So while being free is a benefit, it is also one of Anki’s greatest curses. Additionally, there are hundreds, if not thousands of wordlists. I spent a long time looking for some Chinese wordlists for example. They were unmarked (not categorized) so I had to dig through a very large list of unsorted word lists. I started using one, and I didn’t realize it was simplified Chinese until later. I felt pissed off because I had invested my time into a waste of time because Anki didn’t do something which I felt should be fundamental. My point here is there’s no categories and no user ratings. The shared word list is a huge pile of spaghetti and finding what you want is a bit of trouble. And even if you do find what you want, your results from one word list to another aren’t persistent. I wish this could change about Anki but I don’t think it will. Free software is generally incapable of being both backwards comparable and doing major rewrites. I’d love the Anki devteam to prove me wrong, but I’m not holding my breath — 1,013 days and counting.

Anki uses SRS
SRS stands for “Spaced Repetition System“, which is a system that uses the latest scientific research on the spacing effect to order the flashcards. It’s scientific, and it works. When I say it works I mean, it really works. Using SRS memorization becomes effortless because it adapts to how fast you learn things. Before you know it you will have memorized a hundred, two hundred, maybe even three hundred words (if you can find a quality word list that large). That’s amazing stuff.

On the other hand…

SRS isn’t the best way to learn a language.
Spaced Repetition is designed to help you remember facts. And while it’s true that memorizing vocabulary is great, it isn’t the be all and end all of language acquisition. I’m not talking about the “…battle between lab-tested techniques and conventional pedagogy…“, I’m talking about how I am a professional teacher and how I would be fired if all I did was teach vocabulary. You have to teach a wide range of language skills, not just memorize words. Memorizing vocabulary is only a small part of what I do. In a 2 hour class we spend about 20 minutes going over vocabulary. In the long term, it simply is not the primary focus. And this isn’t even just me, this is industry wide, in highscools, universities and cram schools. Long term, you spend 20% or less time on vocabulary, that’s just how it is. Now sure, you could use Anki to memorize vocabulary and then just hit the books. That’s possible sure. But if you had access to a complete, language learning package, wouldn’t that be better?

Anyways, I admit I have left a little bit of a void above. I stated that SRS wasn’t the best way to learn a language but I didn’t tell you what the best way was. Sorry about that. The best way isn’t SRS, it’s MCI (Massive Comprehensible Input), for example, the TPR Storytelling method. This is not a debate; the research behind MCI is of an order of magnitude larger than SRS. MCI fundamentally changes the structure of the brain to make it easier for you to learn a language. MCI is how we learned our own primary native languages. Further, MCI does not preclude the use of SRS. Spaced repetition can find a very welcome home inside of an MCI system. And Kongzi was written from the ground up to express the ideas of comprehensible input theory. So Kongzi isn’t opposed to SRS — I am looking at making Kongzi more SRS-like (the new F-score system goes a long way to a geometric model of forgetting) but SRS really is not the magic solution everyone makes it out to be for language learning. There’s other, more important research out there, and if we want to be scientific about these things we need to consider all the research.

The thing to remember is that Kongzi doesn’t just do flashcards. It helps you learn a language. So while SRS may tell the program you know a word, how does this help the program determine whether or not you understand the word in a sentence? Or whether or not you know the complete definition of the word, or are capable of using it in a conversation? It can’t. For skills like that, SRS is meaningless and MCI has been shown to be a better training method.

I’d like to close on some positive points. After all, Anki isn’t all bad.

Anki looks Great
My first impression of Anki was about how great it looked and how user friendly it looked. Anki really nailed it in the looks and user friendliness department.

Anki has some very advanced features
Anki can do some things that other programs simply can’t do. When it comes to flashcards, Anki really has it nailed. It can do anything. Custom styles of flashcards, plugins, a highly searchable and configurable database. Did I mention the look ‘n feel? Anki can do graphs. That’s amazing. It keeps track of everything. Stats freaks, and I am one, love this sort of thing. It really gives you confidence to see everything you do reflected immediately in the graphs and stats. Kongzi times your quizzes and gives you scores too, and keeps track of a bunch of stuff, but Anki really gives you a lot of cool stats.

Anki has a large Japanese learning community behind it
Although Anki does not support any one language “specifically”, as time has gone by a large number of Japanese learners have adopted Anki. Thus the community support (i.e. shared word lists) are pretty good, for Japanese learners. So if you are interested in Japanese, you should check out Anki for sure.

Bells and Whistles / It’s been around for a long time

Anki has been around for a long time and as a result it looks very polished and has many bells and whistles. Let me give two examples. The screenshots for Anki show graphs for card intervals and cumulative views of due cards. Wow! I love stats! Never mind that since SRS works with geometric rates of forgetting, that after a while it will nearly always look roughly the same, I just love stats! Are these graphs really important? Why were they coded in? How are they helping the user to learn Japanese? I don’t care! They’re amazing. Anki also has advanced search and browse features. It has many little options. The user interface isn’t just polished, it’s helpful and complete. Kongzi isn’t that great looking or helpful because it’s just new, and has had no user feedback. In that sense Kongzi’s UI is many years behind Anki.

More on stats. Anki keeps track of the total and average number of seconds spent looking at every card. it has mysterious sounding features like “current factor” numbers. Are these just mysteries for the sake of creating some kind of meme or internet culture? Apple does that, Blizzard does that, Valve does that, and other companies, to build hype over their products. And it works! People GO for this kind of stuff. I wish Kongzi could have that kind of culture but it doesn’t. Kongzi is just a tool you can use and trust — like how you trust your bike when you go riding. Someday, I am sure Kongzi will have all the bells and whistles Anki does. But not today.

Conclusion

If you want to memorize a list of facts, or you want to cram for your JLPT exam, give Anki a try. If you’re interested in learning a language, give Anki a try. You can’t lose anything, it’s free. So there’s no reason not to try Anki and come to your own conclusions.

http://www.wired.com/medtech/health/magazine/16-05/ff_wozniak?currentPage=all

Advertisements

7 Responses

  1. You’ve completely and TOTALLY missed the point of SRS.

    Renli responds: Au contraire, SRS is great for flashcards. However (and especially due to) Anki’s “Review Now” feature it boils down to seeing cards you don’t know very well before you see cards you know well. Further, Kongzi does more than just Flashcards, and SRS does not apply to much of what it does. Anyways as I said to Lior below I’ll have a look at the article later (and I do follow up on topics like SRS over time), but I will counter-suggest that you totally missed the point of Kongzi 😉

  2. Let me tell you that I’m baffled by your arrogance. You clearly haven’t understood the spaced repetition system. I propose that you read a really good article about SRS and all the research behind it (more than 30 years). I hope you’ll post a followup on your article after.

    It’s on Wired:
    http://www.wired.com/medtech/health/magazine/16-05/ff_wozniak?currentPage=all

    Thanks,
    Lior

    Renli responds: I’ve used Anki and frankly, it didn’t work as well as I had hoped. At any rate Anki is not SRS, it just uses the algorithm. For example, Anki uses SRS2 because it is incapable of using SRS3, because Anki only contains knowledge as wordlists which the program itself does not understand. It is only capable of presenting pre-digested knowledge. SRS is perfect for a program like that. On the other hand, for a program like Kongzi, SRS is useless because Kongzi represents the underlying language in ways unique to that language. For example Kongzi can transliterate between Hiragana and Katakana, and can construct flashcards for you to help you learn Similar sounds, kanji, etc. automatically. it also tests muliple core skills directly, such as phonics and grammar. My suggestion is that if you really want to learn a language, I suggest you forget about SRS and look into MCI and other more recent L2 theories. Anki is a great program but it’s just not a complete package, and SRS alone won’t save it.

    Anyways thank you for pointing out that article to me. I’ve read it and considered it deeply in revisions to this article I have made over the years. Thanks again and good luck.

  3. Perhaps a really good article to read about SRS will show you the benefits: http://www.wired.com/medtech/health/magazine/16-05/ff_wozniak?currentPage=all

    Renli responds: Thanks, Max!

  4. I think you would benefit from investigating SRS learning more. The basic idea that looking at a card less frequently as you get to know it better (automatically managed by the system, e.g. Anki/Pauker/Mnemosyne/Supermemo/etc) causes the information to move into your long-term memory more quickly and consistently has massive benefits, and it really does work.

    Renli responds: sure, I will look into SRS later. But I am not planning on working on Kongzi again for quite some time.

    Instead of studying 100 flashcards every day, you can learn faster and better by only reviewing 10 per day. This of course means you can increase your throughput and learn far more cards in the same amount of time.
    Personally, I haven’t experienced the problems you quote with Anki. I use it for learning Chinese and it has caused me no problems, on OS X anyway.

    Renli responds: Learning faster and better is more aptly done with MCI theory and not SRS. However, SRS does have it’s benefits. That being said, Kongzi’s method can already approximate SRS as you describe it with the right configuration.

    It does not boil down to user-friendliness or “perspective”, although Anki has a much nicer interface than Pauker and Mnemosyne (plus you don’t have to compile it from source like Mnemosyne, which is why I stopped using it). It’s the simple fact that an SRS scheduler hugely increases learning speed compared to more naieve (e.g. look at as many flashcards as possible every day) strategies. It’s definitely worth adding a spaced repetition scheme to Kongzi.

    Renli responds: For commercial software, it more than boils down to user-friendliness and perspective. Anki is free; it can do what it wants. Kongzi, if I ever release it, will cost money.

  5. Bwahahahah!!!

    It’s the dumbest review of SRS I’ve ever read. And as has already been pointed above, it demonstrates that you Renli have absolutely no idea what SRS is.

    A good SRS program will never allow you to do what you wrote you’d like to do…

    If you want to review whenever you want, why don’t you create paper flashcards? It’s quite easy, you know? Then, you will be able to review the same paper deck all day long, 7 days a week…

    Renli responds: I deleted a lot of the comments on here, especially the more insulting ones (even some less insulting than Lior’s above) however I let this one in particular stand the test of time because of how ironic it is. Anki has since allowed you to “do what I wrote I’d like to do”. And, once again, I’d like to point out that I do in fact know quite a bit about SRS, and that does not mean I agree with it, and, that Anki is not SRS. They’re two separate things.

    But yeah, thanks for being a really good straight man, the opening “Bwahaha” really made your post in particular all the more omoshiroi~

    At any rate… I’m sorry if you feel that I don’t understand SRS. Please feel free to not use my program.

  6. Oh, Renli, you didn’t like my comments about the dumbness of your review of Anki and deleted my post. [snip]

    Renli replies: If all you have to say is that my review is dumb, of course I will delete your post. Responding to such comments is generally a waste of my time. If you have anything real/useful to say, then do so.

  7. I believe there is now a cram mode for Anki where you can study a lot more cards than the normal mode. You might wanna check it out.

    Renli responds: Yes, thanks for pointing that out. I edited the original article to include this information. Thanks!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: