RECENT TOPICS » View all
Language can be reviewed either naturally through exposure or artificially through an SRS. Both certainly have their pros and cons. IMO you need both but not 24/7. Which brings me to my question...
When does a deck become obsolete?
That is to say "At what point have you reviewed so much that you'd rather delete it than even bother opening it to review the 30 or however many reviews it gives in a whole week now?"
I ask this because, although it's not a pain to review a deck which costs almost no time to review because you've really done it to death and it's old and grey now. I'm guessing by such a time natural review through constant exposure should be enough to sustain the knowledge you've gained anyway.
I get sick of decks I'm "finished" with kinda easily. It's not a good thing but it happens and I don't want to be bogged down by SRS all the time. It's just that if you have a deck which is several thousand cards big that you've completed in 2 - 3 months then it leaves you with 100 cards a day for the next few months once you've finished it anyway and oftentimes that's enough to make me not want to add anything more to the SRS.
So is it a question of too much too fast? Or is it keep reviewing til every card has been reviewed to 100% of it's maturity then delete the deck?
I'm guessing this isn't so much of a problem for those with sentence mining decks which are an ongoing thing in which the adder has added around 30 cards a day?
Anyways... if you have anything to share, please.
It sounds to me like you've been leaving everything in your deck regardless of the information contained. (Correct me if I'm wrong)
Anyway, I think as long as you carry on "maintaining" your decks they will just "die out" naturally.
I have never completely deleted a deck, but the decks have gone through cycles.
As I come across cards that I no longer need or feel that they are just preventing me from wanting to do any reps, I just delete that card.
Of course as time goes on it is possible that all of the original cards in the deck will have been deleted, but since you keep on adding to the deck you now have this new "Cirlce of Life" to go through ![]()
So to answer the question of "When Does A Deck Become Obsolete?", my personal opinion is that the deck itself will never be obsolete, and that as long as you maintain existing deck (deleting and adding) the cycles it goes through will naturally become obsolete (and thus discarded) as time goes on.
Basically, they become obsolete whenever you feel they are obsolete.
No sooner, no later. ![]()
Once cards have an interval of several years, reviewing them is probably redundant. However don't make the mistake of thinking that just because you know it well now, you won't forget it after a while.
@Nyanda: Excellent reply! I think perhaps I should weed out the things that I feel are almost ridiculously easily or well ingrained by now with the delete button (or perhaps suspend). The circle of life concept is a very good point, applying to sentence mining decks (as opposed to crusade decks which is what I'm talking about).
@Codexus: Here in lies the quandry... something I know well know but if I didn't seen it again for another 6 months may forget, may well be forgotten in taking such an approach. However, given extensive reading and listening in an immersion environment, if I don't come across that word naturally in a whole 6 months it's going to do me little harm to forget it anyway. Right?
The way I feel about it is... after I've completed a crusade deck, once the interval on every card is about 6 months I would probably delete it.
I guess if I could get 95% retention reviewing a deck till intervals were at 1 year + on every card vs. reviewing it til all intervals were 6months + then doing natural review through extensive reading via immersion and got 80 - 85% retention (which is very likely) then I'd find the latter far more appealing.
I'm kinda starting to realise the balance needed between Anki and extensive exposure to natural language.
Last edited by mezbup (2009 December 01, 4:19 am)
IceCream wrote:
It kind of feels like, once you've made a certain pathway, it's only a matter of reinforcing it through context. But, the pathway takes a little while to make so it doesn't disappear completely.
I definitely agree here. I'd also like to add that whilst you might pass something in Anki reviews you might not have actually learned it well enough to understand it in a different context in the wild anyway.
That article/essay about how extensive reading is needed in order to learn collocations of words really got me thinking about how we do learn to produce natural sentence patterns and the more I think about it, the less importance I seem to put on SRS. In saying that I'm definitely not against SRS though... just... how much is too much?
I think what I will do is review KO for one more month which is when everything hits 6+months and then it's out the window. Blazing that 2kyuu grammar deck worked like a charm too, I'm noticing plenty of it come up in the wild! Think i'll keep that deck around.
A deck becomes obsolete when I can't bear to review anything in it ever again. I've had quite a few obsolete decks. I can say with confidence that I've never regretted deleting a deck that's become stale, monotonous and boring. When in doubt, chuck it out!
Codexus wrote:
Once cards have an interval of several years, reviewing them is probably redundant. However don't make the mistake of thinking that just because you know it well now, you won't forget it after a while.
I absolutely agree and would take it even further: If you haven't actually used that knowledge within a year, there's no point in maintaining it in an SRS anyhow. And if you have used it within a year, the SRS isn't needed. It's obviously just slowing you down, and you can always look it up the once-every-few-years it appears anyhow.
So anything with an interval of more than a year should be an auto-delete.
What about Kanji decks?
My RTK deck is killing me slowly every single day. I have around 80 reviews a day with around 90%+ correct answers and it just won't stop... Maybe its because I don't use the easy button at all or maybe because I've 'finished' it too fast or maybe because I got myself in 900+ due ditch about a month ago and I'm still paying a price for that.
Thing is I'm supposed to learn all this Japanese but instead of that I'm still doing RTK most of the day... It's not that 80 reviews takes a lot of time (30mins usually) but its so hard to make myself do it. I usually read/watch something while doing that to help me distract myself and it works so well that I'm doing 80 reviews for 3-4h and then there's so little time for anything else.
To make matters worse usually I can barely notice the connection between a kanji -> keyword (I train RTK the other way around keyword->kanji) so when I'm doing sentences I see all those familiar kanji but if I want to recall their individual keywords I struggle or sometimes can't do it at all.
thurd wrote:
What about Kanji decks?
My RTK deck is killing me slowly every single day. I have around 80 reviews a day with around 90%+ correct answers and it just won't stop... Maybe its because I don't use the easy button at all or maybe because I've 'finished' it too fast or maybe because I got myself in 900+ due ditch about a month ago and I'm still paying a price for that.
Thing is I'm supposed to learn all this Japanese but instead of that I'm still doing RTK most of the day... It's not that 80 reviews takes a lot of time (30mins usually) but its so hard to make myself do it. I usually read/watch something while doing that to help me distract myself and it works so well that I'm doing 80 reviews for 3-4h and then there's so little time for anything else.
To make matters worse usually I can barely notice the connection between a kanji -> keyword (I train RTK the other way around keyword->kanji) so when I'm doing sentences I see all those familiar kanji but if I want to recall their individual keywords I struggle or sometimes can't do it at all.
I deleted my RTK deck like a month or so after finishing it. Same as you, couldn't force myself to do it any longer, reviews piled up. So I freed myself. Has it been detrimental? To some extent yes but it's not that big of a deal for me to re-do RTK. You also mentioned that you can't even recall keywords when you see a kanji... try a quick run through RTK going kanji - keyword. Bet you all that will change.
One suggestion I read in another thread was to review some decks weekly... I thought that was a good idea to help ease review loads daily but problem is you need to get the decks to the point where weekly is an option... hmm.
Really interesting replies guys...
IceCream wrote:
For instance, that core6k i finished? deck's deleted now. i got sick of it and wanted to do more dramas.
O_O That doesn't seem like a wise thing to do. That deck isn't old by any standard. You barely finished going through it recently and you did it in a record time. That's closer to the "learn once and forget" approach than SRSing. Have you really assimilated all that vocabulary in such a short time?
All righty, I guess I'll be the contrary one at this party. I follow this rule:
If I dislike a deck, I indefinitely suspend every card I fail. You'd be amazed how fast the reviews drop off.
(TL;DR: Conclusion at bottom.)
Let's see why. Assume I just reviewed a card increasing its interval to one year. This card, being well-formed, takes less 10 seconds or less to review. It has an Anki ease factor of 2.4, the average from my RTK deck. For now, assume I remember it every time I review it.
Review # / After interval / cumulative time retained
1 1.0y 1.0y
2 2.4y 3.4y
3 5.8y 9.2y
4 13.8y 23.0y
5 33.2y 56.2y
A sixth review will be required 135 years from the present.
Suspending failed cards puts a maximum limit on the amount of life-time review time. Let's say I start with the complete RtK 1, RtK 3, Core 2000, and Core 6000. Two cards per sentence, one card per kanji gives 15,005 cards. This totals no more than 12,500 minutes of review, or 208 hours.
However, if we let each review thin the deck by permanently suspending failed cards, here's what happens (using my 77.7% mature card retention rate from my RtK deck):
Years in the future/ Pct reviewed / Cumulative reviews / Pct retained.
1.0y 100% 1.00 78%
3.4y 78% 1.78 60%
9.2y 60% 2.38 47%
23y 47% 2.85 36%
56y 36% 3.21 28%
Pruning the deck along the way, I can still retain at least 28% with 48,200 reviews (or 134 hours). I expect that fraction to be higher, since retention rate should increase with time as difficult items are expelled.
So, that gives a rough estimate of the review time required.
Next century: 134 to 208 hours (28% retention or greater)
Next decade: 99 to 125 hours (47% retention or greater)
Remember, these are time estimates starting with 6000 sentences (read and listen) and 3005 characters. Learning that much material to the 1-year mark that wccrawford mentions is going to be something like 90k to 150k reviews, or 250 to 416 hours worth of study time.
Conclusion
If you can find the time to finish huge SRS crusades like Core 6000, you sure as hell have time to review just the easy stuff! As I've mentioned before, SRS focuses your attention on your absolute weakest points, giving you a false view of how difficult a deck is. If your goal is no longer 100% coverage, that's fine. You just have to use the SRS differently. Keep only the easy cards, indefinitely suspending everything that you fail. That way, the SRS will trace your metamorphosis from "Super Kanji-Sentence Ranger" to "Just Maintaining," keeping track of the difficult subset of cards to which you've said "not now." If you ever change your mind and thirst for a targeted micro-crusade, it's just an unsuspend and reset away.
No need to ever "re-do RtK." Seriously, that would suck: mountains of reviews of characters you already know, wasting your precious time.
P.S.: Please use the Easy/簡単 button before complaining about Anki having lots of reviews. Good/普通 really means "thank you sir, may I have another?"
P.P.S: Review daily. Weekly habits are much harder to keep than daily ones, and the cost of missing a review session is higher. If you have multiple old crusade decks, merge them, so you only have to open one deck for review, but do review daily. It'll make cute little puppies happy, I promise.
Key point: SRS isn't there to judge you. SRS is there to sort facts depending on how well you remember them. By default, most people pour extra attention into the hard facts--but you don't have to do this.
Consider these three categories.
1) Facts you'd remember anyway or forget and re-learn outside of Anki.
2) Facts you remember thanks to Anki refreshing them.
3) Facts you forget, suspend, and don't re-learn.
Category 1 is a wash, you remember them either way.
Category 2 is the most obvious advantage of continuing review. I agree with the consensus, this category will eventually shrink as your overall language skill improves. However, while "it'll eventually happen" is a good excuse to ignore hard items, it's a lame excuse to ignore easy ones. Easy is time-cheap (that's what all the math's about). Just do it. Don't waste time thinking about it.
Category 3 is in my opinion where the greatest advantage lies. Frequency is the key. Language--especially vocabulary--eventually runs into the law of the long tail: there's a lot of good to be had from a broad knowledge of rare items. Once you run out of high-frequency vocabulary, then what?
Some people may be satisfied with re-learning low-frequency items once every couple of years. That's not what I aspire to. Once I've exhausted the frequent-enough, I can still keep collecting the less frequent, and one place to look is the "heavy fraction" of suspended cards. In them, they'll be cards that I've picked up elsewhere (grade easy, easy is cheap), the less-frequent cards I'm looking for, and cards that are either still hard or were poorly formulated. The last part gets separated into an even lower fraction, re-factored, or deleted.
It's the "General MacArthur" approach to difficult SRS items: I may retreat from a difficult item, but I shall return. Retreat is tactical: to put effort into easier victories, not an abdication of my right to use rare words, grammar, or characters.
You won't always be in the mood to mine the heavy fraction (and there's no obligation to), but thats no reason to throw it away blindly, especially if that means tossing the entire deck. Just separate it out and save it for when you're really good (you believe you'll be good one day, right?) and looking for things to learn.
I'm of the "Let the SRS obsolete itself". I'm also like Nyanda in that my decks tend to evolve either in what it contains or how I test it. My vocabulary deck started with UBJG, then moved onto Tae Kim, then Cloze Deletion Tae Kim, and now I'm adding Kanzen Master sentences. My RTK deck started on RevTK, then moved onto Anki with a more verbose card set (Katsuo's spreadsheet with yomi and various meanings) in addition to recognition cards. My vocab deck started with 2k1KO but became Core then Core 2k and 6k, then both recognition and writing, then I added Tanuki list, now I removed the Core 2k photos.
The benefit for me in all this is I'm introducing more and more Japanese into my life, just slowly. For someone like Icecream that's getting a much more major dose of real Japanese can handle deleting decks cause she's likely to run into the material more frequently.
So, if you're going about it slow, do not delete. If you're the 24 hour always in Japan mode, then deleting is not a big hazard. Know what's right for you and your lifestyle and learning style.
I just keep 1 deck, that way I only have to delete items.
Some people say you can't learn through the SRS and I used to disagree but now I agree in part. One thing that's really hit home lately is that you can use SRS to actually learn new vocab items crusade style where Anki is the first place you see the word. You can't however learn the words collocations through Anki which is 50% of the knowledge you need to truly (and I mean truly) know a word. Which brings me to my point and i'll use Icecream as an example.
"Learn 6000, retain 4000 solidly, re-learn 1000 with bit of exposure, completely re-learn 1000".
The second, very important step of learning those words is to learn their collocations which only happens through exposure so the amount of reading (thus exposure) one must do to pick those 2000 words back up will be far more beneficial than reviewing the entire deck in Anki til you have a "95% retention rate" (which is isn't a true representation of anything).
The notion of suspending something that was too hard kinda surprised me at first because I thought well you won't learn it if you just suspend it. Then I thought back to when I did KO and there were times where a word was giving me too much shit and I just clicked easy and said "it's not important" turns out when more than half those cards hit maturity I knew the word by then haha then there's the rest of those cards where I'm like ahhh it's a word I didn't give a shit about. So I guess that was my way of doing that and yeah it wasn't so bad. Definitely a good point that you could work through your suspended list to pick back up on some rarer items at a later stage in your learning.
wildweathel wrote:
No need to ever "re-do RtK." Seriously, that would suck: mountains of reviews of characters you already know, wasting your precious time.
I have 800 RtK cards due, the oldest of which was due 5.2 months ago. Is there really a way for me to not have to just start over? I still have around 150 cards that aren't due yet, but I feel like the recall rate will be off (since I haven't been reinforcing them with newer cards). I figured I would just resume with the lowest-numbered card that's due, and skip the ones that aren't due yet.
thurd wrote:
It's not that 80 reviews takes a lot of time (30mins usually)
Actually, since (if I understand correctly) you're only maintaining your existing kanji repertoire, I think you're spending too much time -- 80 reviews in 30 minutes puts you at 23 seconds per card!
I'd say for any card except ones you've recently learned, you should spend 5-10 seconds maximum. If you're still completely blank after 5-10 seconds, just turn it over, mentally recall the story when you see the kanji, and fail the card. For easy cards, try to do it as quickly as possible; if the stories fade and you start doing recall without stories, great!
An 8 second average means 11 minutes for 80 cards -- much better, right?
To make matters worse usually I can barely notice the connection between a kanji -> keyword
I'm planning to quickly (5-10h total study time?) go through the complete deck in the opposite direction (kanji -> keyword) when I'm done with RtK 1, so why don't you give that a try? While most kanji will be fairly easy, it might be a useful thing so spend your time on. (I'm waiting until I'm done learning RtK 1 because I don't want recently learned kanji popping up as recognition questions -- I'm worried my visual memory might take over and make the whole process less efficient in the long run.)
I don't think it's possible to practice recognition (kanji -> keyword) using this website, but here's how you do it with Anki:
1. Open your Heisig deck (or click the Download button in Anki and select RtK from the list if you haven't used Anki before).
2. Go to Settings -> Deck Properties -> Edit Model -> Card Templates, and create a "Recognition" template (be sure to keep the Heisig number out of the question box when you copy-and-paste, or it'll serve as an undue memo aid ^^). Close the dialog boxes.
3. Go to Edit -> Browse Items, Actions -> Generate Cards, and generate cards for the new "Recognition" template.
4. In the card browser, select "Recognition" from the filter, mark cards 1-2042 (or 1-3007, or wherever you are in RtK), and hit Actions -> Reschedule to randomly schedule the cards for a 0-28 day window or so, so they don't hit you all at once.
5. Hint: You may want to mark and suspend (Actions -> Toggle Suspend) the remaining recognition cards.
Happy reviewing!
Last edited by epsilondelta (2009 December 01, 6:22 pm)
mezbup wrote:
I'm kinda starting to realise the balance needed between Anki and extensive exposure to natural language.
mezbup, I think this is the most important thing you've said.
From my perspective, I don't understand these super-anki-blitz-the-whole-deck-method. Anki for me is something that takes about 30-45 minutes a day, max. I can't remember the last time I reviewed 200 cards in a single day. I'm going through the KO deck right now and to be honest, I'm only seeing about 500 new cards a month because I'm only doing about 100-150 reviews when I review, and occasionally I miss a day. But anki is only one facet of my study- I'm also reading native material, watching TV and playing video games.
I love the positive results people like you and icecream are having with these deck blitz marathons, but I would burn out very fast doing something like that. I've been doing the KO deck for 3 months and I've only seen 2000 of the 6000+ cards, but that doesn't bother me- I've still done 10k reviews which have significantly helped my reading ability. (I always let the audio play, which means I average 20 seconds/card)
Gambit,
Here's how I recover a neglected deck with Anki.
1) Pre-test. Set show failed cards last. Get through the reviews, but don't start re-learning failed. If you have to do 30 minutes a day for a week failing 70% of them, that's how it goes. (But, in my experience it will probably go better than that.)
2) Once you have just failed cards due, reset them by opening the browser, selecting all due cards, and using the "reschedule" option.
3) Pick up where you left off. By default, Anki will display new cards in the order they were added, which should be RtK order.
With some luck, this will require only a few days but save you re-learning something like 300-500 kanji.
Capital, Mezbup,
Totally agree. Fast progress happens when you mix SRS-organized study with extensive input. I've found a good pace at 30 minutes of Core 2000 per day and another 10 to 15 minutes of RtK. In practical terms, that does translate to about 30 sentences and 10 kanji a day. Fast? I dunno. 'tis certainly comfy. Flying through easy listening cards is really fun.
I am considering an RtK 3 + 人名用 + '10 常用 blitz at 50 characters a day, but I now know I don't really have to do something like that to make real progress. I think a higher priority, though, is taking a couple weeks to get more comfortable with a monolingual dictionary.
@captal:
The idea behind the crusade mentality is to bootstrap you to a point where bombarding yourself with native material all day long doesn't stress you out because you can read, listen and understand enough to get by and really begin learning things in their proper context.
It's only after recently getting to that point where I'm noticing a different kind of learning starting to happen. I've sort of shifted from "vocab learning" to real "language acquisition". I've noticed that at times i'll hear a phrase and understand every nuanced detail of it and it'll really click solidly in my mind that that's what you to say to express that particular thought/idea/emotion/situation and it becomes part of my active output repertoire.
@wildweathel: monolingual dictionary using can take a lot of getting used to! The reason being it uses words to define a word and to understand it you have to know all the words in the definition, if you don't there's more and more look-ups to be had. Not necessarily a bad thing and definitely not a reason to steer clear of it! My point is it's not something you can really get "used" to per se owing to the fact unless you already have a huge vocab you'll always have to look a word in the definition up.
My best advice is to have a fixed strategy for using it. Personally, I try it monolingual first and if I understand the whole definition then great! If there's one word I don't know i'll try and look that up monolingually too but that's as far as I take it. If I can't understand the second words definition monolingually then I look it up bilingually and go back to deciphering the word I originally looked up. Saves too much time being wasted looking things up endlessly but i'm sure you'll develop a system that works for you.
One thing I notice is that often times words that i've looked up monolingually when I come across them they feel more Japanese to me because that's the only way I know them!
Last edited by mezbup (2009 December 01, 9:34 pm)
mezbup wrote:
@captal:
The idea behind the crusade mentality is to bootstrap you to a point where bombarding yourself with native material all day long doesn't stress you out because you can read, listen and understand enough to get by and really begin learning things in their proper context.
I see- I'm already at that point so I guess doing the KO deck at a slow pace is more comfortable for me. I don't think I could have handled what you and others did- trying to swallow 4-5+ new words per sentence would have put me off rather quickly. That took some perseverance on your parts.
I do remember when it was frustrating to watch/listen to native material- it lasted quite a while- so perhaps your method is better.
Last edited by captal (2009 December 01, 10:49 pm)
captal wrote:
mezbup wrote:
@captal:
The idea behind the crusade mentality is to bootstrap you to a point where bombarding yourself with native material all day long doesn't stress you out because you can read, listen and understand enough to get by and really begin learning things in their proper context.I see- I'm already at that point so I guess doing the KO deck at a slow pace is more comfortable for me. I don't think I could have handled what you and others did- trying to swallow 4-5+ new words per sentence would have put me off rather quickly. That took some perseverance on your parts.
I do remember when it was frustrating to watch/listen to native material- it lasted quite a while- so perhaps your method is better.
Actually that was an interesting feature of that method that believe it or not wouldn't mind carrying across to sentence mining. I guess when sentence mining you can pick and choose but if you went with the "no word left behind" method (then pruned things that just won't stick AT ALL) you'd wind up with a lot of stuff learned pretty fast.
I'm not even using Anki for a lot of vocab learning now because words I learn one day I usually see again the next or sometime that week and again within the month ![]()
I'm still going through RTK, but I already have a single deck which starts with kana, then RTK, then Tae Kim waiting for me when RTK is done. With Anki it's easy enough to tag things, do bulk operations etc., so I don't see a need for more than one deck - that keeps everything simple in terms of syncing between Anki and AnkiMini etc.
Maybe this view will change when I start getting past the 5-10,000 mark on cards - might get a bit more complicated, but by then I could just delete individual cards due over a year or whatever, or move them to an archive deck to save space.
I've taken to tagging my decks in a very structured manner now as to keep track of and separate out useful data. It actually took me a very long time to work out all the little things you can do with Anki and just how versatile it is.
That's a good question. I've been thinking about this too; is it best to keep SRSing the same cards forever, or does it make sense to discard them after a point. I mean after a while, you should be able to maintain things almost indefinitely without SRSing like with your own language.
Last edited by vosmiura (2009 December 12, 7:03 pm)

