How big is your Anki deck and how many leeches?

[imron]

Starting from scratch seems very counterproductive to me and I can't see any good reason to do it.

I guess it depends on whether your goal is to do SRS reps or to use Chinese ;) I've found reading to be a kind of natural SRS, but one where you can easily do 2,000-3,000 'reps' in 10 mins. Often, I'll even kick it up a notch and spend half an hour doing 60,000 reps. Not only that, but the reps are automatically sorted by the words that are most useful to me at that exact moment. Looking up 1 in every 10 characters would be a success rate of 90%, but that's a far from optimum rate for real world usage. Currently I find my rate probably closer to 98-99%, which includes a mix of both new and forgotten vocab. This 1-2% of 'failed reps' can then be drilled briefly over a couple of days/weeks with flashcards to cement them more firmly in my memory, and I can be sure that I'm focusing all my effort on the optimum set of words/characters that are precisely relevant to the Chinese I'm likely to encounter. These smaller decks can be purged regularly with any unlearnt words falling back into the 1-2% and I'm not really any worse off. The useful ones will appear again soon enough and before long will be learnt permanently. The non-useful ones I don't care about.

To me, it's the building up of huge SRS decks and spending time doing all those SRS reps that seems counterproductive

[MaLaTang]

imron made a great point. Use reading material as the main learning tool.. Actually seeing those words in context is the best way to help you remember them. But flashcard decks do have their place in learning and currently im using both 'Mastering chinese characters' deck (1800 facts) along with an hsk deck (2000).

[Olle Linge]

I use Anki quite a lot and have used some kind of SRS since I started learning Chinese. That means that I have everything in my decks. Here we go:

Total number of facts: ~17 000

Average time/day: 20-25 minutes

Number of leeches: 0

Backlog: 0

I completely disagree with some people in this thread about leeches. I've done the opposite to turning them off: I've lowered the leech threshold from the original 16 (or something similar) to 7. I have two reasons for doing so. First, if I forget a card seven times, there is something seriously wrong with my mnemonic and it needs to be fixed. Second, SRS is meant to be reviewing, not learning (which is what I'm doing if I learn cards anew all the time). I've written quite a lot about this:

• Killing Leeches
  
• Spaced repetition isn't rote learning
  

I do mostly recall, but what I argue here is even more true for writing.

[WestTexas]

Since someone bumped this, I'll update with my current statistics:

Right now I am at 7889 facts, 132 leeches. My mature success rate is only 77.2%. I don't really see what I'm doing wrong. My mature success rate was at 75% a few months ago, but I started doing two cards, one for pronunciation and one for meaning, for words containing characters with more than one common pronunciation (ie, words like 模样、模仿、无间、着手、 etc), and I've been getting slowly higher. Still, I have a hard time seeing myself ever getting more than 85%. Looking at my leeches, I seem to encounter them in the following types:

1)Words which have a meaning which doesn't really translate into a single English word, so I have to remember a definition instead of a single word

2)Words which have several different meanings which seem completely unrelated

3)Words which have a meaning which seems completely unrelated to the meaning of their two characters

I don't have any leeches for single characters, although I have many single-character cards in the deck, and I don't have any English->Chinese leeches, only multiple character Chinese->English words. There are just some words that no matter how often I see them the meaning just won't come to me.

I was wondering, how do you guys handle words with more than one definition?

One problem I think I have is that some days I will be in a bad mood, and I will be very very picky about giving precise answers, while other days I will accept any answer that seems mostly right. What I mean is, there are English words on the back of the card, and if I think of one of those words or a synonym for them, I will count it right. But some days I will have a different idea of whether or not word X is really an English synonym for word Y, and I think in the long term this results in me being lenient most of the time, the card getting a big interval, and then when I am in a bad mood one day I decide that my previous answer isn't quite precise enough or doesn't have exactly the same meaning and I fail the card, my success rate goes down, and the card interval starts over from 0, because I don't have a consistent idea as to what exactly constitutes a correct answer for a given Chinese word.

Right now the word in my deck with the most misses is 形象, with 31/55 reviews. This card has been suspended as a leech twice. I have no idea why I miss this card so much, as the definition does not seem particularly complicated, and my frequency list says it is a common word, but I can tell you that if someone had shown me this word ten minutes ago and asked me if I had seen it before, I would have sworn in court that I had never seen that word a single time, ever.

[Gleaves]

If words have several meanings, I only mark myself wrong if I actually need to know the other definitions. I keep getting 领 wrong because it has quite a few common meanings. Most of the time, though, I'm content with knowing the main definition and figure I'll get the others in context.

I was tiring of my main 2000+ card deck so I was considering nuking it. Instead, I've done heavy doses of deleting while I review and have it down to 700 cards. It feels good. Like when you move and get rid of stuff you don't like and/or need.

[Olle Linge]

Actually, I think one could safely remove words that have reached a certain maturity. It's not as if I'm going to forget words that have an interval of five years or so. However, I keep all words mostly because it's very nice to have a searchable database with all the words I've ever studied (or at least written down). Also, I share several decks and want to keep these updated. Perhaps regularly suspending cards over a certain interval would be useful? What would it be? Two years? Three? One?

[Silent]

Sure, their is no reason to keep words with very long intervals. But then, if you learned them up till a 5 year interval they don't do much harm any more either. Though numbers may add up if you learned a big vocabulary. I think the theory is quite simple. If you learned a word till above the natural repetition rate natural repetition will make you keep remembering it and the word may be taken out of the deck.

In reality it's a bit more complicated. Natural repetition strongly depends on how common the word is used in the context of your daily life which on it's turn depends on your (changing) habits. Words like 我，他，你，etc will be seen on a daily basis so might be removed after learning it beyond 1 day. Some words may occur not at all or virtually daily when involved with a specific subject. It's hard to decide how well you should learn such words before getting them out of your study collection or whether you should learn them at all. I guess the answer also depends on where you are in language learning and what goals you have. In general I see little benefit in learning words well beyond a one year interval, how rare a words do you want to learn? For beginners that interval might even be much shorter as a beginner is learning mainly high frequency words that will be used 'everywhere' on a daily basis.

[Olle Linge]

Good reply!

Let's look at the maths. I have around 10 000 words with an interval above one year. Let's assume the average for these is around a year and a half, or roughly 550 days. That gives roughly 20 words per day. Considering that I know these words quite well, it takes perhaps five seconds per word. I think it's probably worth 100 seconds daily to make sure I still know these words. Some of them are used very frequently (such as the words you mention) and these could be removed, but I would probably have saved more time not writing this post than I would have saved if I removed some common words from my deck. In short, I'm prepared to spend 100 seconds/day to maintain a complete deck and not removing anything.

In general I see little benefit in learning words well beyond a one year interval, how rare a words do you want to learn?

Well, this depends on what you're aiming for. My long-term goal is to have a vocabulary similar to a native speaker beginning studies at university. This entails knowing lots of words that aren't used on a daily basis, so automatically suspending word with intervals above one year would be extremely bad. Which is what you say, I think, because this would not be true for a beginner or someone aiming for natural conversation.

[Silent]

You don't need to actually remove the words. You can mark them as known and make known words not show up again. Open the browser, sort on interval, select the ones with a high interval, add tag 'known'. Time invested will be won back in one day and your deck is still complete, only not everything is shown.

True, what your cut off interval should be depends on your goals. That's also what I stated. But do you really need to go beyond a 1 year interval to reach the level of a well educated native? You learn words by regularly encountering them. A word that occurs less than once a year will be encountered at most about 80 times in a lifetime. That maybe enough to learn it, but when evenly distributed there is a fair chance even a native won't know it. In real life many rare words are learned when a word is temporarily high frequency. That may be due to a hype, some newsevent that makes it temporarily relevant or just because someone takes a temporary interest in a subject. After the word has fallen back into the rare category the word will be remembered, but the relevance has gone.

Your disadvantage as a student is that you don't have a lifetime to learn words like a native does. If you want to get your vocabulary to the level of a native you will need to do extraordinary things. Basicly condensing all interests and phases that a native goes through in his live into a shorter timeframe. Unless you're in an immersion environment that would probably mean reading huge amounts on many different subjects.

Off course the goal of every language learner is to reach the native level. Considering the effort that it would take, vocabulary wise, I see no rationale to do so through study. I think proficiency is well enough. After that improving your language skills should go naturally by using the language for 'useful' things. I don't see a rationale in reading books about quantum mechanics, bio chemistry and boat building, or just (rote) learn the vocabulary, only to increase your vocabulary. It makes sense to do so if you're interested in the subjects.

[Olle Linge]

Yeah, I think we agree with each other more or less, even though I think it remains difficult to say where the limit for the interval goes. Marking the cards is of course a viable solution (or simply suspending them?). Nevertheless, a word climbs to one year fairly fast and my spontaneous thinking is that it's too short an interval. Also, seeing the word in context while reading a text is not necessarily the same as reviewing the word itself. I mean, I don't read all words very carefully when I read as long as I understand them. Still, I think what you say is mostly correct; I didn't mean to argue against you, really, more expand and add some things. Let's do that again.

Off course the goal of every language learner is to reach the native level.

This is simply not true. I know lots of people who have other goals, mostly goals that require less work. Also, I think this is a question of how realistic the goal is. I really think that I will achieve this goal in the future. Some people might have this as a vague, distant final goal, without actually believing or planning to get there.

Considering the effort that it would take, vocabulary wise, I see no rationale to do so through study.

You seem to think there is a difference between studying and using the language. I don't think there is, or at least there needn't be. However, even there is, I do think studying (as in formally focusing on certain aspects more or less out of context) can be very useful. Taking English as an example, I know that I have benefited quite a lot from courses I've taken well after becoming fluent (I've only learnt English in school) and I still use Anki to learn words in English. Of course, it's difficult to say that this is the reason my English seems to be gradually improving, but I definitely feel that that is the case. Naturally, my English is at a much higher level than my Chinese, but I still think it's useful to compare studying methods. In short, I think studying is highly useful even at an advanced level.

I don't see a rationale in reading books about quantum mechanics, bio chemistry and boat building, or just (rote) learn the vocabulary, only to increase your vocabulary. It makes sense to do so if you're interested in the subjects.

Well, I think no-one has suggested such an approach? Most native speakers don't know this kind of vocabulary either, so it doesn't really conform to my goal. My goal would be more like knowing about as much Chinese as the "average" first-year student at university, so that would remove most professional language. I only read what I find interesting and most of the words I add to my deck comes either from listening to news broadcasts or reading books, mostly fiction but occasionally other books.

[realmayo]

Interesting discussion. First thing I'd say about the idea of removing or pausing the very easy entries is that it's difficult to know where to draw the line, and the reason I keep the easy ones in my deck is as insurance in case I, for some very unlikely reason, stop studying for two or three years and then want to restart.

I've used Anki for several years now: it gave me a real boost when learning characters, and from the start I always tested myself "both ways": but, a couple of years ago, when I got to the stage where it made sense for me to try reading simple magazines and other texts, and when I was learning lots and lots of new vocab, testing myself on "production" would have been far too time-consuming and would have come at the expense of chunks of that new vocab that I was coming across while reading, and wanted to remember.

So, I stopped testing myself on the writing. Now, though, I'm in a position where I can't spend as much time each day on Chinese, and I've started to forget or mix up lots of words especially those with rather less-commonly used characters. So I've decided to use Anki to re-learn how to write characters (not vocab, just characters, ie I'm given the pinyin and meaning and everything).

I've also deleted my 10,000+ strong Anki deck.

But the contents of those 10,000 cards are stored in a spreadsheet, and I'm slowly adding them to a new "recognising words" deck, alongside my "writing characters" deck. Every week I tell my spreadsheet which new characters I've added to the characters deck, and in return my spreadsheet spits out a list of cards originally from my 10,000 deck which are made up solely of characters that I now know how to write. Those cards then get copied into the "recognising words" deck.

This may seem unnecessarily complicated, but I'm just relearning how to write 3000+ characters, and resetting my main "recognition" Anki deck in step with my progress on the characters. It's time-consuming (I guess 2-3 hours every weekend for the next couple of months, plus the daily reps) but already it's helping me weed out leeches and half-remembered or mis-remembered words. As I re-add words to my recognition deck, I'm giving them intervals of approx 2 years, 8 months, or 3 weeks, and for those which I fail two days in a row I am suspending for now.

I'm already finding it helpful to spend time again on how to write these characters: I'll never need to do so in real life, but for me the time spent on this helps on the word-recognition side. And it's a great feeling to clean out a great big deck and re-add the cards, killing, reworking, or just plain relearning some of the more problematic ones.

But it was very scary stopping with that deck which has been my almost-daily companion for the last four or five years!

[Olle Linge]

But it was very scary stopping with that deck which has been my almost-daily companion for the last four or five years!

Before I say anything else, I want to say that you are very brave indeed. I mean, seriously, I would need very, very strong reasons indeed for giving up my Anki deck. Probably a gun pointed at my head or my family taken hostage or something.

I guess 2-3 hours every weekend for the next couple of months, plus the daily reps)

This sounds very optimistic. Won't it take much longer than that? Or am I missing something? One of the reasons I'm scared of changing too much is that changing anything containing more than 10 000 cards takes lots of time. For instance, I would have preferred to have examples in a separate field, but I didn't do that from the start. Changing now is almost impossible.

[realmayo]

Actually it will take longer: I'm only re-adding words to the "word recognition" deck once I have learned (or relearned) how to write all the characters that make up that word. I guess that will take 5 more months, so there will be some words that I can't add until that time.

As I go through the newly added cards in the "word recognition" deck, I'm pressing 4 for about 10-20% of them, which is giving them an interval of approximately 2 years. I'm pressing 3 for over half, for an interval of approximately 6-12 months. The remainder are split between those which go on a 2-3 week interval, and those which are failed. I got through almost 2,000 this way at the weekend.

The next day I reviewed again the ones which I'd failed: if I then got one wrong a second time, or if I decided there wasn't much point keeping with that card, or if it felt very leechy, I suspended it ... I can go back to those next year.

This way, I feel I'm really cleaning out my deck. It might all work out badly: those intervals for the newly added (re-added) cards might not be realistic and that could come back and bite me 6 months down the line. But I think so far it's worth the risk: I'm giving the deck a real clean-through, taking back ownership of it a bit more, getting rid of some dross. And one reason why I was a bit dissatisfied with the existing deck was that there were some cards where I'd been overgenerous to myself in assessing how well I knew it. Indeed, that was one reason for going back and relearning how to write the characters. I'm now being strict again.

[WestTexas]

Before I say anything else, I want to say that you are very brave indeed. I mean, seriously, I would need very, very strong reasons indeed for giving up my Anki deck. Probably a gun pointed at my head or my family taken hostage or something.

Isn't that the truth. Sometimes I feel like I've cut off part of my body and given it to Anki to hold hostage. Except it's not a physical part of my body, it's a part of my daily time. Or Anki is like a tattoo and it's difficult to get rid of. Don't get me wrong, I think Ank is a great program and it has made a huge difference in my Chinese ability in only about 1.5 years, even though I only study a couple of hours a day. But I feel weird in that I may end up using it every day for the rest of my life. It just seems eerie that I have that kind of commitment to a piece of software.

[renzhe]

It is not as bad as you fear.

I used Mnemosyne for a few years and it really helped me get a good vocabulary, but the reviews were turning into a chore, which is why I started skipping days as a rule rather than an exception, and when I got really busy and had some issues with some of the software dependencies, I dropped it. I had finished about 90% of the HSK vocabulary deck at that time, and the last few hundred items were tricky.

I spent a few years without reviewing at all, with only some reading for vocabulary practice. Basically like imron, only reading far less than he does.

I found that most of the words were still there. I've started doing the HSK vocabulary deck in Anki from scratch, and I find that I still know most of them. The ones that slipped are easily refreshed when I see them in context and look them up. I'm adding 100 new cards per day and at this speed I should go through the deck in a few months. This is essentially what imron is recommending -- start over, the important words will still be there.

The million dollar question, of course, is when you are ready to do this. If you give up on SRS too soon, the danger is that you'll forget most of it. I was lucky.

[Silent]

This is simply not true. I know lots of people who have other goals, mostly goals that require less work.

Sure many won't be willing to put in the work, but I've never heard a language student say that he did not want to get fluent in a language.

You seem to think there is a difference between studying and using the language. I don't think there is, or at least there needn't be.

To me study means actively improving your skills. I learned English in school, after that I was forced to use it for my studies and travel (immersion environment). The use of the language made my level near native. After that I used my English for work in a non native, often low level, environment and my English declined again. After secondary school I never actively studied English again. I only used it, at times very intensively.

To me practical use is not study, sure, you gain experience and you may learn from it, but if that's study everything is study and the word becomes meaningless.

Well, I think no-one has suggested such an approach?

No-one suggested the approach. It was just a way for me to elaborate on the fact that you have to do extraordinary things to obtain the vocabulary of a well educated native. A native is stuffed up with useless vocabulary from the past. If you as a student want to get the same seize vocabulary you have to learn useless words or exceed the native in useful vocabulary.

[Olle Linge]

Sure many won't be willing to put in the work, but I've never heard a language student say that he did not want to get fluent in a language.

This is getting a bit off topic, but I just want to clarify that I think there is a huge difference between being fluent and having a vocabulary matching that of a first-year university student. I agree that most people aim for fluency, but I doubt a majority also aims seriously for a university-level vocabulary.

Returning to Anki decks, I've also experimented with trying to learn to write some characters I previously only knew how to read. However, I've done this mostly with either separate, more temporary decks, or simply writing a lot more (like writing example sentences or just writing the characters that come up instead of just translating/pronouncing). This is of course a flawed method, it's random and lacks structure.

Still, I think that learning to write is mostly a matter of practise and I'm not sure I need a specific deck for that. I mean, if I can write the 4000 individual characters in my deck, I'm okay, I don't need to know how to write all the words. Knowing individual characters and then knowing which characters create a specific word is enough. Wouldn't it be quite convenient to just have a "write individual characters" deck and keep all the words in the recognition deck? What do you think?

[realmayo]

I've never heard a language student say that he did not want to get fluent in a language.

I'm going to study some Korean for a few months next year before visiting the country. But I don't want to be fluent, unless one can develop fluency in Korean with only 50 hours of studying.

Snigel, that's what I'm doing at the moment -- one deck for recognition (of words and characters), another for production of characters. As I've said elsewhere, the main reason for learning to write characters is because there were too many times when reading magazines or books that I was mixing them up with similar ones, and because these days I don't have time to read a great deal, reading on its own is unlikely to really help.

In another thread I wrote that words like <consume / initially / lowlife / superstition / landlady / ghastly / calculate / uneasy> have a frequency in English of around two per million. So to do "SRS-by-just-reading-lots" an English learner would have to read War and Peace (which has over 500,000 words) just to see one of those words again once.

That's a lot of reading!

[imron]

So to do "SRS-by-just-reading-lots" an English learner would have to read War and Peace (which has over 500,000 words) just to see one of those words again once.

That's a lot of reading!

It seems like a lot when you measure reading in units of War and Peace, however let's break it down into more practical units:

Average reading speed: 250 wpm

Minutes required to read 500,000 words at 250 wpm: 2000

Days required to read 2000 minutes at 30 min/day: 66.67

So, if you read at an average speed for a total of 30 minutes a day, it would take about 2 months to see one of those words once (and actually, if they all have a frequency of two per million, then you'd probably see them all in that same span of time). And, you don't necessarily have to read War and Peace to do that either, you can read whatever books, magazines or other content takes your fancy, just remember to mix it up a little to get broad exposure to different types of language usage. Compare that with how often such words might appear in SRS reps, especially once you've "learnt" the word, and you'll probably find you'll see those words more often by reading than you would by using SRS, plus you'll be seeing them in real-world usage rather than just as part of SRS reps.

But that's for English, and it's just a bunch of estimates, so let's bring this back to what a learner of Chinese can achieve, based on actual hard facts.

Last year, I read 17 books in Chinese, totalling over 4 million characters, which might sound like a lot, but worked out to spending about 30 minutes to an hour a day of reading. So far this year I'm on book 12 and over 3 million characters, also on 30 mins to an hour per day.

The books I read usually have about 200-400 pages and I get through a little over one a month.

Obviously reading speed has big effect on how much you can read, so spending a little bit of time improving your reading speed will pay huge dividends.

Anyway, I'm constantly surprised at how words and characters in one book will appear again in other books that have completely unrelated themes/styles. In coming across such words in the new book, it's great to be able to think back to the other book and compare usage of such words. I'm also pleasantly surprised at how well some of these words stick in memory after such long periods.

Now, granted, such words won't be appearing at optimally spaced intervals to maximise the amount of time required before the next revision in order to prevent the word from falling off my forgetting index, but I can live with such small inefficiencies considering I find reading a much more enjoyable pastime than doing SRS reps.

[anonymoose]

Last year, I read 17 books in Chinese, totalling over 4 million characters, which might sound like a lot, but worked out to spending about 30 minutes to an hour a day of reading. So far this year I'm on book 12 and over 3 million characters, also on 30 mins to an hour per day.

Last year, if you spent an average of 45 mins per day reading, and read 4 million characters, that gives you an average reading speed of 4000000/(45×365)=244 characters per minute.

This year, if you spent an average of 45 mins per day reading, and so far have read 3 million characters, that gives you an average reading speed of 3000000/(45×298)=224 characters per minute. (Today is the 298th day of this year.)

Obviously reading speed has big effect on how much you can read, so spending a little bit of time improving your reading speed will pay huge dividends.

Evidently how much you read has a big effect on your reading speed. Read too much and you will kill your speed.