RECENT TOPICS » View all
Almost to 500! 475 to be exact. But what's occurring to me is that the repititions are taking longer and longer.
My question is, is it sensible to carry on with repititions in both Anki AND on this site, as well as learning new kanji? And once you get, say, past 1000 kanji or so how many repititions are you dealing with daily?
Why would you have the kanji in both anki and on the site? That's doubling the workload for nothing.
Plus surely how you answer will be effected. If you have just done this site and been asked 50 items and then goto anki and it asks you some items that are the same you might know them better as you have done them that day on this site. It would be repeating work.
You shouldn't need to review with both Anki and RTK as long as you do your reviews each day and clear out your fail pile each day/regularly. Just let one SRS (whichever one you prefer) tell you what work you need to do each day.
As far as the number of reps you will have each day once you reach 1000+, I suppose that depends on how fast you add new kanji, your % correct in your daily reviews and how well you keep your fail stack cleared. I am at 1650 or so and have about 120 reviews each day. I try to add 15-20 new kanji per day and try to keep my fail stack down. The last couple of days I had less time and didn't add any new kanji and I was down to 75 this morning. Most of the time though it is in the 120-150 range for me. People with better memories might have fewer reviews though... ![]()
With spaced repetition, if you never fail anything and keep adding new material at a constant rate, and the interval increases exponentially (doubling each time), then the number of reviews will increase logarithmically. In other words, if you started a month ago and you are now doing 10 a day, then in another month you will do 11 a day, in another two months you will do 12 a day, in another four months you will do 13 a day, and so on.
However, in practice it will increase faster than that because you will occasionally fail questions.
I've finished RKT1 but my fail stack is about 400 including mostly kanji I seem unable to learn and ones I don't particularly want to waste time on. So I do about 15-20 questions a day on this site to keep up the 1600 that I have learned.
At a point, heck even at the beginning, set yourself a limit: time is the best guage. Let's say two hours.
During those two hours do the following:
1. Review all due cards - if two hours pass and you have cards left over. STOP. If you finish before two hours go to step 2
2. Review added cards from day prior - if two hours pass and you have cards left over. STOP. If you finish before two hours go to step 2
3. Study failed cards - if remaining two hours end and you have cards left over. STOP. If you finish before two hours, go to step 4
4. Begin studying new cards - if remaining two hours end and you have cards left over. STOP. If you finish before two hours, CONGRATULATIONS, you've entered all Heisig cards. From now on, ignore this step.
This process also works when you go to the sentence method. It's self limiting method so that new material is not added unless you're caught up on old material. Plus, it begins early on to spread out new/old material.
The time can vary, but keep the steps about the same above. Don't let the number of reviews dictate how long you study. Let how long you study dictate how many reviews you do.
Using two SRS at once is not useless, it's plain wrong. It defeats the whole purpose of an SRS, that is to show you the cards only at specified intervals, not before, not after.
nac_est wrote:
Using two SRS at once is not useless, it's plain wrong. It defeats the whole purpose of an SRS, that is to show you the cards only at specified intervals, not before, not after.
Seeing something too often isn't detrimental. It's just a waste of time.
I wouldn't say it's useless to use both methods early on in your studies. I did it untill I got to about 350 and found that the work load just got a little too much. I don't really know why I did both this site and Anki at the same time. I guess the first 350 or so kanji is where you are experimenting on different routines and you simply don't want to neglect either/or just in case it turns out to be the best one. If anything it will re-inforce each kanji as if you've seen them out in the wild. I certainly never planned to do it for the entirety of rtk1.
I eventually picked Anki and stuck with that. I'm approaching 600 and doing 20 new kanji a day. I get about 120 cards to review every day and I'm at a ok 83% pass rate. After much experimenting, 20 a day is what I'm gonna stick with for the rest of rtk1.
Wow thanks for all your replies. To be honest, as the last poster said, I've been doing both RtK and Anki because I've just been beginning, and because I wasn't sure which one I preferred. I like the way this site shows you how many times you've passed and failed each particular kanji. On the other hand, Anki's spaced repitition is obviously more complex.
I'll probably go with Anki in the end!
nac_est wrote:
Using two SRS at once is not useless, it's plain wrong. It defeats the whole purpose of an SRS, that is to show you the cards only at specified intervals, not before, not after.
I agree.
Unless you put different material in each SRS. I use this site for RTK and Khatzumemo for Japanese reading practice.
But if you come across a word or something occasionally I find it doesn't hurt.
Tobberoth,
sorry to contradict you, but seeing something too often is just as detrimental as seeing it too rarely, in the SRS philosophy. That's because SRS is build in order to gradually move the information from your short term memory to your long term memory.
Seeing something too often strengthens your short term memory, but does nothing to the long term one.
Right now I don't have the time, but I can find some references to this.
edit:corrections
Last edited by nac_est (2009 January 15, 3:48 am)
nac_est wrote:
Tobberoth,
sorry to contradict you, but seeing something too often is just as detrimental as seeing it too rarely, in the SRS philosophy. That's because SRS is build in order to gradually move the information from your short term memory to your long term memory.
Seeing something too often strengthens your short term memory, but does nothing to the long term one.
Right now I don't have the time, but I can find some references to this.
edit:corrections
If I were to look at all the stories for the 99 expired cards I have today, and then go on to review them, my failed stack wouldn't grow at all today. The cards that I would have failed don't get reviewed as strongly -- they're being produced from short term memory of the mass story reading -- not long term memory, like the cards that I know.
nac_est wrote:
Tobberoth,
sorry to contradict you, but seeing something too often is just as detrimental as seeing it too rarely, in the SRS philosophy. That's because SRS is build in order to gradually move the information from your short term memory to your long term memory.
Seeing something too often strengthens your short term memory, but does nothing to the long term one.
Right now I don't have the time, but I can find some references to this.
edit:corrections
Yeah, a lot of people seem to think so, but it isn't true. If you only do the reviews at this site, you'll do fine. If you do the reviews at this site, in another SRS, reading books with the kanji used constantly etc, that will just make you do it better. After 3 years, you will still remember fine using both techniques.
SRS is built on the idea that if you don't repeat something within a certain amount of time, you will forget it. As you repeat the information, the time it takes before you have to repeat it again increases. The point of an SRS is to minimize your effort by keeping that repetition interval at maximum so you don't review to often. If you WANT to spend more effort, that's your choice, it won't be detrimental at all. If it were, people wouldn't learn anything from normal exposure, yet they certainly do.
If you don't trust me, do this yourself: Write an English word on one side of a flashcard and a turkish word on the other. Everyday at noon, review that flashcard. Do it for a year. Destroy the flashcard. A year later, you will still remember it. Just because you keep it in short term memory doesn't mean the repetitions aren't being put in long term memory as well.
Tobberoth wrote:
If you WANT to spend more effort, that's your choice, it won't be detrimental at all. If it were, people wouldn't learn anything from normal exposure, yet they certainly do.
The problem with this is that the SRS will think you have a great memory, when in fact you're cheating, and will expand the card interval at a very, very fast rate. It won't have the opportunity to gauge the card's difficulty and adjust the easiness factor properly.
Last edited by iSoron (2009 January 15, 7:02 am)
Actually iSoron and playadom hit the mark better than me. The problem is not much in whether seeing something more often will make it stick better, but in the fact that you make the "delay-determining" algorithms totally useless. It's "either you use one (and only one) SRS, or you don't use one at all".
On a side note, however, I'm not too sure I'd remember the turkish word after not seeing it for a year. However both Tobberoth's and mine are impressions, and we wouldn't know which is right...
iSoron wrote:
Tobberoth wrote:
If you WANT to spend more effort, that's your choice, it won't be detrimental at all. If it were, people wouldn't learn anything from normal exposure, yet they certainly do.
The problem with this is that the SRS will think you have a great memory, when in fact you're cheating, and will expand the card interval at a very, very fast rate. It won't have the opportunity to gauge the card's difficulty and adjust the easiness factor properly.
Not true at all. The SRS makes SURE you see it AT LEAST as often as you need to and there is no easy factor regardless, the extra reviews in other mediums are just bonuses. If you click normal every time, it doesn't matter if you see it in other mediums as well or not, the SRS will show it just as many times. It's only if you're stupid enough to pick "Easy" just because you happened to see a kanji the day before that it makes any difference.
A cards difficulty is never gagued. You pass it, or you don't. As long as you pass it, it really doesn't matter WHY you passed it, you knew it when it came up and that's all that matters.
What tobberoth is saying is pretty much common sense just because you see something more often than what the SRS decides doesn't meen it will stay in your short term memory if this was the case nobody would ever learn a thing, all an SRS does it give you a much more efficient way to remember something for the long term instead of reviewing it each day you review it when your about to forget as if you review the information every day your workload would be HUGE an SRS just makes life easier only seeing what your about to forget.
If you happen to see it more than the SRS dictates then this is a good thing unless your "actively" reviewing it each day which is a waste of time.
Last edited by tibul (2009 January 15, 10:38 am)
No matter what the case may be, a single SRS is clearly the most efficient way of doing things. Doubling the work for little or no benefit is not a good idea.
Tobberoth wrote:
It's only if you're stupid enough to pick "Easy" just because you happened to see a kanji the day before that it makes any difference.
This is the problem. It means that you have to start doing the job instead of the algorithm. The user is supposed to just score how easily he's remembered the answer, very quickly.
If you have to stop and try to judge whether you were able to remember the card because it was in your short term memory or in the long one, then you can just throw the SRS away and do the whole job by yourself.
Anyway, since we all agree that it's not a great idea to use 2 SRS's for the same thing, I'm cool with leaving the discussion as it is, because it doesn't make a lot of a difference ![]()
nac_est wrote:
Tobberoth wrote:
It's only if you're stupid enough to pick "Easy" just because you happened to see a kanji the day before that it makes any difference.
This is the problem. It means that you have to start doing the job instead of the algorithm. The user is supposed to just score how easily he's remembered the answer, very quickly.
If you have to stop and try to judge whether you were able to remember the card because it was in your short term memory or in the long one
Which is why one shouldn't ever use any Easy button. Either you remember an item or you don't, the effort to remember it shouldn't matter since, like you said, it should be done quickly. If it takes you too long, you obviously don't remember it well enough and should fail it.
My view of an SRS is that when you have entered something into it, you SHOULD remember it. You should in theory never fail anything if the SRS is optimal and you have spent enough time learning things from the start. And thus, WHY you remember something when you see it in the SRS is completely meaningless. If you remember it, you remember it and you've remembered it until now since you entered it. If you saw the fact a few days before, you should have known it just as well then, so it shouldn't have made a difference towards the algoritm.
Last edited by Tobberoth (2009 January 15, 12:46 pm)
tibul wrote:
What tobberoth is saying is pretty much common sense
And like most common sense, it's probably wrong. This article (with included citations) supports the idea that increasing review rate decreases long-term retention.
~J
woodwojr wrote:
tibul wrote:
What tobberoth is saying is pretty much common sense
And like most common sense, it's probably wrong. This article (with included citations) supports the idea that increasing review rate decreases long-term retention.
~J
Um, that article agrees with me. It says that the benefits of overlearning decrease with time, but not even once does it state overlearning is detrimental.
The idea is not about 1 item, but 2000 or 10,000 items or more. Yeah, it's not efficient to use 2 SRS's. However, if both SRS programs are good then they'll compensate for the fact you're using two.
See, in both, the items will keep spacing out that you remember. That means you're seeing them less often even with two SRS. Sooner or later, an item not in long term memory (and with 2000 items, one of them is bound to be that) will reach the point where between review on 1 SRS will be too far away from review on another SRS. One spaces farther out, one comes back down.
I agree about not using the "easy" button. I agree if you take too long to think of the answer, mark it wrong, I agree one should not use two or more SRS for the same items, I agree you should study the same thing everyday. These are things I'll apply to myself and suggest to others. It does not make them right, as what works for me may not work for another.
In all this, don't forget that the SRS will not be your only exposure to Kanji, or Japanese or whatever other information. As you see real Japanese (movies, books, etc) you're going to see words and kanji that you're also reviewing. Do not be so pig headed about SRS purity that you say "I can't look at Japanese, I'll corrupt my SRS spacing of Japanese cards". Don't put the process ahead of the reason for the process.
Tobberoth wrote:
Um, that article agrees with me. It says that the benefits of overlearning decrease with time, but not even once does it state overlearning is detrimental.
From the footnote to Figure 3: "For any value of [Retention Interval], an increase in [InterSession Interval] causes test scores to first increase and then decrease". Other similar statements are scattered around. This article has similar data and conclusions laid out slightly better (I didn't see on first glance whether the data was from different studies).
~J

