After 3 years I have about 3000 words and sentences in Anki. It's had gotten to the point where I had about 100 cards per day to review, which pushes the limit of my available study time. So I decided not to add any new cards until the daily count went down. Apparently, this takes longer than I thought. After about three weeks I was doing 70-80 a day, so I loaded a new batch of cards (with Anki adding the default 20 new cards per day). Now I'm back up to 120-130 per day, which is just about too many.

I accept the fact that the review cards Anki is giving me are cards that I need to review. But many of them are unusual readings or easily mixed up readings (is that　大　だい　or たい). And many of them I get right, maybe 85%. I know that I need to review these, but giving up studying new cards seems to be a pretty big price to pay. Also, studying new cards is more fun than reviewing older ones. Has anyone else run into this?

I noticed a similar thing with RevTK. Once I completed all the kanji, I expected my reviews to drop off, but they seem to be decreasing very slowly.

From what I read on this forum, you should start concentrating on word decks only, and parse your old decks against native material (suspend all, and each word you encounter from now on you will reactivate), so you'll have to immerse yourself in native material to do these two things obviously.

I don't know what your deck is like, but I think a likely problem is that your cards are too difficult. I used to do sentences too, about 9k of them, before realizing just how inefficient they are.

In reality, especially when starting out, it's hard to find sentences that have exactly one thing that you don't understand. As a result, I would have sentences that had multiple vocab, and if I missed one of them, I would have to fail the whole sentence. Oh, and sentences took me too long to review as well.

You should just switch to vocab... I average about 6 seconds per card, doing vocab reviews, so your 100 cards a day would take just 10 minutes. Of course as a result, you are able to saturate your process with more vocab (I have daily reviews of 500-600 vocab), so it's not like you really free up time, but the whole process does move a lot faster!

Don't hesitate to hit the very easy button and don't waste too much time on a single card. You'll sometimes come across a card and think, "Oh, I know this! Only a few more seconds and I know I'll remember it!" This kind of thinking is your enemy. It wastes time unnecessarily.

I find it helpful to think of reviewing as a marathon. You can neither take it easy nor push too hard. If you review at a leisurely pace, then your reviews will be painstakingly slow. On the other hand, you're going to burn out and make careless mistakes if you review too quickly. Just like a marathon, you want to reach the highest pace that you can sustain until the very end of the race.

After you stop adding cards, your number of cards due per day drops noticeably only during the first few days (and vice versa).

IceCream Wrote:how often do you use anki? 100 a day after 3 years with only 3000 cards doesn't seem right to me...

I'm curious about this, too.

IceCream Wrote:Other people are going to disagree here, but i've found that reviewing cards to maturity then deleting them works pretty well. Especially when you're still learning very commonly used words. You will need to watch and read things outside anki though...

I suggest this, too, though I think mere maturity is too early to delete a card.

After all, if you keep adding new cards and never stop reviewing old ones, your number of daily reviews will increase indefinitely.

I use Anki every weekday. Sometimes I'll get to it on the weekends and sometimes not (I'm married with two small kids). It's quite possible that skipping two days per week creates a wave of cards that I need to slog through for the rest of the week.

FWIW this is how I get my cards: I made an excel file with all the Core sentences from smart.fm in one sheet, a list of JLPT2 words in another, and the kanji with onyomis and kunyomis in another. I wrote a script where I can select an onyomi and the script will find all the kanji with onyomi, all the words with those kanji, and then all the sentences with those words. Then I pick every word and one sentence per word to load into Anki. These sentences probably aren't as effective as if I had mined them on my own, but it's damn fast.

I just browsed the dates on my cards and found that 90% were added using this method, but only within the last year. The others were manually entered from that blue 漢字と言葉 book. A few of those go back 3 years (OK, two of them), but most don't. So let's say the deck is one year old.

FooSoft Wrote:In reality, especially when starting out, it's hard to find sentences that have exactly one thing that you don't understand. As a result, I would have sentences that had multiple vocab, and if I missed one of them, I would have to fail the whole sentence. Oh, and sentences took me too long to review as well.

This is a problem for me. So I'll lose the non-target words in the sentence, and then once the card ages to certain level I might mix up a しょう　and a じょう　and a じゅう. Also there seems to be this overlap between kanjis with the せい　and しょう, けい　and きょう onyoms.

I like the idea of slamming vocab, especially since I'm signed up for the JLPT2 this December. But suspending that many cards will result in a giant bolus of sentences to review later on. Though in theory, I could finish the vocab relatively quickly, and then the sentences should be easier.

I wouldn't feel that you need to do all of the reviews everyday. You can try changing the card order to focus on the smallest intervals, if you don't get through everything, your at least focusing on what is probably most important and easy to forget. Or maybe you can mix it up, you can try several small sessions focusing on different card orders, like: 5 minutes on smallest interval, 5 minutes on largest interval, 5 minutes on random and so on. Also you can try adding even fewer then 20 cards. Try 5, 7, or 10 maybe. Also, don't be afraid to grade cards using "Very Easy", unless it's a new card you should probably default '3' for most of your gradings and go up or down from that.

Also if you don't have a portable device to review on, be it a netbook, an iOS device like an iPhone/iPad, or an Android phone, consider grabbing one of these so you can review on the go. I'm sure you can find time for a couple reviews throughout the day.

Wakela Wrote:

FooSoft Wrote:I would have sentences that had multiple vocab, and if I missed one of them, I would have to fail the whole sentence.

This is a problem for me. So I'll lose the non-target words in the sentence

with all due respect, if this is how you are making SRS cards, it is WRONG!!!! This way doesn't really work. Each card should only test you on a single (atomic) piece of information. You should pass or fail the card only on that single thing. If there are other words in the sentence you're having trouble with, make a completely different card to test that other word.

Highlight what specifically you're testing in each card. (i.e. do how it is done in core2000)

e.g. 'The DOG walked up the stairs"

pass or fail the above card only on the word DOG. If you have trouble with the word 'stair' for example, create another card testing that word specifically.

Wakela Wrote:I like the idea of slamming vocab, especially since I'm signed up for the JLPT2 this December. But suspending that many cards will result in a giant bolus of sentences to review later on. Though in theory, I could finish the vocab relatively quickly, and then the sentences should be easier.

I just dropped my sentence decks altogether and have never been happier (this was hard, because I had so many sentences, but in retrospect I'm positive it was the right choice). I started blazing through the JLPT vocab deck, and within a month or so, I suspect I had most of the words I learned from sentences in there anyway. IMO it's a lot more useful (and fun) to get grammar/reading practice from books and so fourth, and not the same sentences over and over.

As far as cards that stump me, and I think I'm just on the verge of remembering - I bury them in the deck! So if I still don't remember them the next day I confidently fail them without fear that I was just having a brain fart or something.

I had similar issues with my sentence deck, so I chucked it and mostly just do vocab now. If I want to I read a sentence in the answer column for the word - many of these are the same sentences from my old sentence deck (the smartfm Core stuff). Many people rush through their vocab reviews - I do sometimes, but sometimes I spend a little time with a word, read the example sentences(s) and might practice using the word in my own sentences -let it "sink in" - thus I can get the grammar practice of sentences if I want, or the speedy review of drilling the vocab - it all depends on my mood and the amount of time I can give it.

I also advocate learning words from things you have seen "in the wild". I just un-suspend the words I come across in jdoramas or wherever and learn those (from that core list), or make new cards when needed. Then I can replay the source material later on having studied (notice I don't say "learned") the individual words. It really helps with recall in the SRS too. For me learning from a list for "efficiency" is anything but - having a richer context of origin for the words (then just example sentences from a list) is way better for me. I think learning is a dynamic process of which SRS is only an important piece. Having too many things coming from a relatively weak context just muddled things up for me.

I don't see the value of rushing through the language (for me)- maybe you can pass the reviews, and quickly have a massive deck, but that's just a start - does it hurt to slow down and really digest things? Some might say "yes, that's inefficient." And it may be for them - I can only asses my own process, and offer those insights.

Ooh I like the bury the card "brain fart" idea LOL.

I think one of the reasons for just plowing through some sort of raw word list is so that you get some base amount of vocab you can apply to reading native material. Without knowing a couple of thousands of words, I don't think it's possible to read anything and actually enjoy it. Looking up every other word kind of takes the fun out of reading a novel. I also refuse to read things that I wouldn't be reading in English (things written for little kids).

I've hit the anki wall many times as well.
What I've found that helps is...
Since it's more fun to add more cards/info, for missed cards, I try to add more example sentences to the card.

If that fails to do the trick then just add multiple cards for the same info picking different examples then delete those extra cards after you learned it.

FooSoft Wrote:I think one of the reasons for just plowing through some sort of raw word list is so that you get some base amount of vocab you can apply to reading native material.

Yeah I get the reason behind this - it just didn't work as well for me as I thought it would (did about 1500 words this way). I felt like I had too shallow a grasp on too many words, and after a while it got more and more inefficient. I also learned a bunch of so far not so useful words (no big harm done) and it really caused some interference type of confusion for me (kind of like RTK key words that are too similar to each other, or obscure -> amounts to lots of fails) It does give you a base though - and gives a strong familiarity with readings. And really all I'm doing now is getting a slightly less shallow grasp to begin the learning process - but I find better context origin equates to better anki reviews - they really stick better for me this way (oh gads I'm repeating myself here).

Have you tried splitting your study time?
Instead of 1 hour straight, have you tried 30 minutes in the morning and 30 minutes in the evening? I figured that with this way it clears out more cards or at least splits them in sessions so instead of one 100-card session I'll have one 70-card session and one 45-card session or something like that.

I'm working my way through KO2001 and I'm finding that the more cards I add, the easier it's getting because I know more words and kanji and I'm giving more cards "easy" on the first pass.

Keep in my mind that my method for Anki is a bit strange since I load all cards in it to study instead of studying them first, then putting them in SRS. My "first-seen" success percentage is 59.2%, but surprisingly my mature success percentage is 97.6%.

Great replies, guys. Thank you so much. I do love this forum.

Kainzero: Yes, I do split it up sometimes. I used to do 50 cards at a stint, but for me it went a little better when I switched to a time limit. Not sure exactly why. Maybe when I'm not in the mood and space out a lot the time runs out no matter how many cards I've done, but if my goal is 50 cards, then it just seems to drag. And I use Anki the same way you do, I just throw everything in there and study as I go. I get almost all of them wrong the first pass.

TaylorSan / FooSoft: There is a balance isn't there. At one end of the spectrum you have 1000 words to memorize for an exam the next day, and at the other you sit down with an Atogawa Ryunosuke novel, look up every word and write every sentence into Anki. Neither of these are very good ideas. So we all pick our own sweet spot somewhere in the middle. I don't have a lot of free time, and when I started the only way to get sentences into Anki was to enter them manually, so the idea that I can press a few buttons and load hundreds of cards is very tempting. But TaylorSan is right. The sentences have no resonance with me. To be honest, I don't even know if reading the same sentence over and over again until you get it is a good idea, since that's not how we read in the wild.*

I'm liking the idea of getting a good base of "normal" words from the JLPT2 list, and then adding words I see in the wild. I have a few issues of National Geographic in Japanese, and those articles seem pretty good. They're interesting and use more challenging grammar than newspaper articles.

*This gave me an idea. A software application into which you load the kanji/words you want to study and it finds example sentences on the internet. So the sentences you study from are always new to you. Of course it would return a lot of poor quality sentences, but maybe if you had users rate them in terms of studyability, you could eventually generate a "corpus" of good, unique sentences. Just a thought. I'll never do anything with it.

Wakela Wrote:A software application into which you load the kanji/words you want to study and it finds example sentences on the internet.

There's an example sentences plugin for Anki which does this. The sentences are sourced from the Tanaka Corpus, which isn't perfect but is probably more reliable than a random internet search.

Wakela Wrote:The sentences have no resonance with me. To be honest, I don't even know if reading the same sentence over and over again until you get it is a good idea, since that's not how we read in the wild.

I found with sentences, that I would memorize the sentence itself quite often, and that the words were learned "chained" inside, so seeing them in other sentences was weaker then it ought to have been (or even a ridiculous fail). By studying the words individually I have noticed a markedly better ability to read them "in the wild". Having example sentences on the card is a benifit too, as you can practice/read them or not, and you need not fail the card based on not getting some part of the sentence. Sometimes I paste several example sentences in there (usually from my mac dic) which will have varying levels of difficulty - some are way over my head, some are easy, ect. But really I don't read the sentences a majority of the time, and review fairly quikly - however my average time per card is much longer then maybe most, as I do play around/meditate/practice-talk on certain words quite often. Or on some days I just blaze through the mah....

I like you're idea of "sweet spot" - this is a great way to describe it. Finding that took me a lot of reading the forum/experimenting, and even tracking my progress - and it of course still continues. But retrospectively by far the best changes I've made were going to vocab-centric and context based sources for the SRS. I do still SRS some phrases and the like, but it is a much smaller focus.

Oh shit - I better go do my reviews (damn forum!).