5

I am creating a language curriculum for Standard Chinese (none exist in my area). If someone gets a question/phrase/word incorrect, for example a phrase from a module on greetings, they should obviously see that same phrase or a different phrase with those same words, again until they don't forget it (using spaced repetition):

A student should meet a word every x, x+5, x+10 minutes to ensure retention

enter image description here

But my question is... what if they get the phrase correct on the first try? Can I assume they've got it and move on? Should I assume it was a fluke and verify? If so, how many more times and does it still follow the same repetition curve/timing?

For example, some implementations of SuperMemo algorithm (such as Anki) take into account getting things incorrect and correct. How is this done?

  • 1
    (1) Could you expand a bit on what a "module" is? Especially how big? (2) What else does the course consist of except the content for spaced repetition? – IkWeetHetOokNiet May 28 '17 at 18:14
  • 2
    @ChristopheStrobbe thanks for asking; I've provided clarification above. The course is actually a game on teaching foreign languages. A module is an exercise that teaches a set of words (e.g., Introductions), and my question is asking if someone gets a certain phrase incorrect (e.g., Hello, my name is ___) – Growler May 29 '17 at 16:56
3

According to research that I cited in this answer, "a typical learner should meet a word about 8 to 10 times to obtain full word knowledge" (F. Rasouli and K. Jafari: "A Deeper Understanding of L2 Vocabulary Learning and Teaching: A Review Study", International Journal of Language and Linguistics, 4.1 (Jan. 2016), emphasis added).

So it seems that you will need to present correct phrases several times before you can assume that they are stored in long-term memory. However, it seems a bit strange to present an entire module again if such a module contains several phrases (all of which presumably need to be correct), one of which has not been learnt correctly. From the point of view of space repetition systems, the modules may not be "atomic" enough.

Since phrases that were memorised correctly still need to be tested again, the next question is then what algorithm to use for determining the time between the previous "test" and the next. There are a number of algorithms for this, for example:

I cannot give you a recommendation for a specific algorithm, since this requires research that compares their effectiveness, which I do not have at hand.

  • Thank you for this. Re However, it seems a bit strange to present an entire module again if such a module contains several phrases - are you suggesting only the word should be tested again? I've tested how DuoLingo works - they present the same exact phrase again if you get it incorrect. They present / reuse words in different sentences if you got a previous set of words in a phrase correct – Growler Jun 1 '17 at 5:24
  • @Growler If a module contains several cloze tests, it seems strange to present the exact same module again if only cloze test was answered incorrectly. (Re-representing the same or a very similar cloze test in the context of a different set of phrases would seem preferable.) If it's just a single (short) cloze test, I think you can re-present it as is. – IkWeetHetOokNiet Jun 1 '17 at 9:37
0

You just need to come up with algorithm, why not? Take a look at some research about the problem you are facing, and implement the results of research in your code. Why not?

Example. I read somewhere, sorry I couldn't find the source, that if one wants to memorize the word for a long period of time, phrase in foreign language, s/he should face the word, phrase more than 80 times.

On the other hand, insisting learners to say 80 times the phrase they have already got, can cause other problems as losing your users because of boredom of your software.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.