I'm wondering what other people find most efficient from a retention/utility vs time perspective. At the moment, my cards are one direction only:

question: english word
answer : kana

Lately I've been considering if it makes more sense to reverse direction to favor recognition(passive vocab) at the expense of production(active vocab). I've done forward and reverse in the past, but felt that the more robust retention wasn't worth the extra time invested. Or perhaps it would be even more efficient to do:

question: kanji word(with kana modifiers)
hint: kana word
answer: english word

I should mention that my short term goal is to get to the sentence reading stage as quickly as possible. And yes, I am studying my RTK concurrently.

If your goal is sentence reading ASAP then I would think that recognition is the more efficient choice. That's what you're doing when reading, recognizing words you've seen before. So practice that.

I was doing production before and switched to recognition at around the 1k mark in Core, but I'm still not sure about it. While manga and article reading have definitely improved, I feel I just don't know the words as well. It feels sorta hollow to be able to read it and pronounce it, but not be able to write it if someone asks you. Not actively producing the kanji when SRSing basically means you won't know how to write the word.

For characters in isolation, production gives better results. Having the character on the back means that what you're practicing is producing the character in your head, and if you can do that you can most certainly recognize it when you see it as well.

But it's different for vocabulary because there are usually several ways to translate a certain word. If you've only got the English word on the front, you won't necessarily know which Japanese word it translates to. When you get further along and pile up tons of cards you won't remember all of them individually like you do in the beginning. So it's not just practicing the particular Japanese word, it's also in a way practicing a particular way to translate this English word into Japanese, which is more than necessary (and maybe even hurtful?).

So if you want to do Japanese vocab production it's probably better to have a Japanese sentence cloze deletion, MCD, or something else that makes it apparent on the front side exactly what word it is you're studying.

Now for the actual question... I just do Japanese word in kanji → reading and meaning. I like the simplicity for adding new vocabulary fast. There's an ocean of words to know--going for quantity instead of careful time spent on quality, and then leaving the rest to other studies (listening and reading for me) seems to be effective.

I wouldn't do both recognition and production on the same content. Seems unnecessary, and you're also sort of going against the SRS algorithm by reviewing the same knowledge on different cards with their own separate intervals.

devilsbabe Wrote:If your goal is sentence reading ASAP then I would think that recognition is the more efficient choice. That's what you're doing when reading, recognizing words you've seen before. So practice that.

I was doing production before and switched to recognition at around the 1k mark in Core, but I'm still not sure about it. While manga and article reading have definitely improved, I feel I just don't know the words as well. It feels sorta hollow to be able to read it and pronounce it, but not be able to write it if someone asks you. Not actively producing the kanji when SRSing basically means you won't know how to write the word.

Thanks for your thoughts. At the moment, I have no practical need for writing kanji, so I don't feel like I'm missing much by setting writing to a low priority. It's not that I wouldn't like to learn the writing eventually, but I'm much more interested in reading listening and conversing. It's also something that I can do much more easily(and therefore less excuses. I realize that my own needs are not the same as others so you may weigh writing much higher priority than I do.

Thanks for writing your experiences. I think I may follow your example and reverse my cards for a while and see how that works. I'm guessing I'll be able to increase my new card rate a bit.

TwoMoreCharacters Wrote:For characters in isolation, production gives better results. Having the character on the back means that what you're practicing is producing the character in your head, and if you can do that you can most certainly recognize it when you see it as well.

But it's different for vocabulary because there are usually several ways to translate a certain word. If you've only got the English word on the front, you won't necessarily know which Japanese word it translates to. When you get further along and pile up tons of cards you won't remember all of them individually like you do in the beginning. So it's not just practicing the particular Japanese word, it's also in a way practicing a particular way to translate this English word into Japanese, which is more than necessary (and maybe even hurtful?).

So if you want to do Japanese vocab production it's probably better to have a Japanese sentence cloze deletion, MCD, or something else that makes it apparent on the front side exactly what word it is you're studying.

Now for the actual question... I just do Japanese word in kanji → reading and meaning. I like the simplicity for adding new vocabulary fast. There's an ocean of words to know--going for quantity instead of careful time spent on quality, and then leaving the rest to other studies (listening and reading for me) seems to be effective.

I wouldn't do both recognition and production on the same content. Seems unnecessary, and you're also sort of going against the SRS algorithm by reviewing the same knowledge on different cards with their own separate intervals.

Some great points here and I fully agree. Especially the part about quantity over quality. For me personally, quality has to be above a certain level, or retention really falls off fast, but just above that threshold seems to be the sweet spot.

I think you're right about practicing recognition if my short term goal is reading asap. I'm going to reverse my vocab cards for a few weeks and see how that feels. I may try a few with kanji on the front and see how that goes. I'm just a little worried it may take more time than it saves because I'm only at ~250 RTK , but if I can learn some kanji (at least in part) directly along with the vocab, perhaps it's a bit of a shortcut. What do you think? Would you recommend learning more RTK first before putting kanji on the front, or kanji on the front in lieu of RTK?

Thanks for your thoughts!

That's a great question, and I'm sure there are a variety of opinions on it. I'll give you a snapshot of my experience.

I used Anki daily for about three years doing both directions, until about 8 months ago, when I switched to doing Japanese--English (sort of) only. I think this is the way to go, at least for me, and I'll tell you why.

Before that, let me say that when I say "English (sort of)," I mean that I've got a sentence in Japanese with a vocab word in bold that I'm trying to learn. That's the front of the card. On the back is the same sentence with Furigana written above it, and far below, in rather small letters, an English translation. I probably look at the English less than 1% of the time, but doing an initial translation helps ensure I really I understand the sentence, I feel. And of course, yeah, there's a lot of things that don't translate exactly, but I don't get hung up on that.

As for why this is a good method, well, it's much faster, so you can cover more material. Doing English-->Japanese, it's pretty hard to remember the Japanese equivalent to every English expression you want to say. So your failure rate is likely to be higher, and your reviews will probably take longer. On the plus side, most everything you remember, you'll be able to use. So, if you review 100 cards and fully remember 90, that's 90 you've got solid that you can use for output.

But . . . to me, the new best way of studying is to do Japanese-->English (sort of), and review more cards. In the same time, I might be able to do 200 cards, usually with a lower failure rate, including new and young cards. I just can't output them all. So if I study 200, I can read 180, and probably output 90. So the output percentage is the same, but I'm gaining exposure to an additional 90 cards per day. The weird thing is that I don't choose which I'll be able to use for output. My brain does. In other words, some cards I'm able to remember both directions, while others I can only remember one direction.

I'm writing this rather quickly, since I'm on vacation during Golden Week, but hopefully that makes some sense. Maybe?

JapaneseRuleOf7, that was very interesting, thank you for your perspective

I had always wondering about the long term benefits of doing E>J but had stuck to the hard rule of not doing it, so it was interesting to hear about someone doing it for that long. I do my Japanese cards the same way you mentioned I have the word in a sentence highlighted I'm testing and on the back I have the English translation but in a light gray so it's hard to read but there if I need it.

JapaneseRuleOf7 Wrote:On the back is the same sentence with Furigana written above it, and far below, in rather small letters, an English translation. I probably look at the English less than 1% of the time, but doing an initial translation helps ensure I really I understand the sentence, I feel. And of course, yeah, there's a lot of things that don't translate exactly, but I don't get hung up on that.

That's an excellent idea! I think I'll do that myself once I get a few more vocabulary and kanjis under my belt. I totally agree with your reasoning that you're probably learning extra material in the same amount of time. Also, I can see myself being less hung up on getting the exact translation J→E rather than E→J like you said.

You could also try doing the first X number of cards in production mode (E-J) and the rest in recognition mode. That way, you'll have a solid base. I did core 6000 on iKnow! which is by default production based. I usually use it on my android tablet and quickly write out the kanji on the tablet before typing in the corresponding reading. Of course, as you get deep into Core 6000, there will be a lot more synonyms so it'll get a bit confusing. I still like doing it that way since I have those words solidified in my memory. You could do it just for the first 2000 words (ie. Core 2000) for example in Anki to make it a bit simplier. I've been doing recognition via Anki for the remainder of Core 10000 and am about 450 words away from finishing it (finally! :>).

PotbellyPig Wrote:...am about 450 words away from finishing it (finally! :>)

That's fantastic! Congratulations.

As this topic is still not too old, I did not feel like opening a new one just to ask a related question.

So lately Anki has become more and more of a burden to my daily life than being supportive in terms of my Japanese studies. Soon I will need to abandon it anyway, because I will be occupied with studying for university, but until then I want to make sure to make as much use of Anki as possible.

I want to change the model of my anki decks. For the time being, I want to change the model only for the core6k model. I want to go E->J, but I have no idea how to create these cards, without starting from words like 食べる again I rated to be that easy that they would not come up at least in the next five years...

Also I am not sure how to utilize the audio of the deck then. I have monolingual definitions on the backside now -- where are they supposed to go logically? I feel they have to remain on the back to be read..

I also heard about clozed deletion, but I cannot handle the plugin at all (not sure what to press T-T).

Grateful for any advice.

Are your cards currently J->E? If so, and all you want to do is reverse direction from J->E to E->J open your deck in the browser and click on "Cards" then copy the code from front to back and vice versa. Keep the bit on the bottom, where it says something like {{answer}} and <hr id=*****>. That should be close but I'm sorry if it's not exactly like this as I'm on my work computer now.

I have to say caution you that switching direction has been a bit of a speed-bump in my learning. A significant amount of words I know E-J do not know know J-E. So if you are hoping to make the best use of a short period of time, I suggest sticking with what you are doing.

Yes my cards are currently J-E, they have been always.
Thanks for the method. I will definitely give it a try, although I know it will be a speed bump. Still, I reached 13k unique words and I have to force myself adding new cards. It is really hard because it seems as if I am addicted to learning new vocab T-T

Super EDIT for space:

Where applicable, both is definitely best in terms of speed and retention. Doing E->J successfully is tricky though. You need to make sure there is enough context on both sides of the card to lead to the word you're looking for. Honestly this is much easier in monolingual study, as dictionary definitions are fairly descriptive. In bilingual study you have to rely on example sentences from trusted sources.

Here's my recommendation using the word list method.

1. Pick out 5-7 words at a time, no more no less.

2. On the J side, kana/kanji + example sentence

3. On E side, translation of word + english translation of example sentence

4. If it helps, underline the specific word in both sentences.

5. Run through the 5-7 words in anki J-> E

6. Run through the 5-7 words in anki E -> J

7. Repeat with another set of words as desired.

This is how I do it anyways. Its great for retention, as the production reinforces the reading you just did. If you don't have access to reliable sentences avoid E > J altogether. Context is so necessary for E > J to work. Definitely do both for kanji though.

amtrack Wrote:1. Pick out 5-7 words at a time, no more no less.

For exercising your long term memory 5-7 items is not a good number. They are handled quite well by your short term memory. To make sure that you are committing newly learned items to the LTM you need at least twice as many. (BTW, I seem to remember - sorry, I can't be bothered to look it up now - that Anki's creator suggests 20 items at a time.)

Inny Jan Wrote:

amtrack Wrote:1. Pick out 5-7 words at a time, no more no less.

)

Just chiming in here. I also find 5-7 a good number of cards to start with exactly because they are handled quite well by short term memory.

Once I get 5-7 into short term memory, I learn the next 5-7 cards, etc... and then review everything. I have a filtered review deck for this which searches every card rated hard in the past 2 days(rated:2:1). This review deck is usually around 50-60 cards which try to review a few times throughout the day. During the review is when words go into long term memory.

I've tried 20 new cards at once and I just kept cycling through them and most wouldn't stick in my memory, short term or long. This is just my experience, so maybe people just learn differently.

^I found that 20 new cards in one block is too many for me too and after some experimentation I settled on 14-16.

My approach to learning new cards is slightly different from yours though. Instead of reviewing them in the usual way, I open the browser and try to memorise each card on a list that has 14-16 new cards I want to learn (because the cards are displayed in the browser, I can see them all at once). Once I’m confident that I can remember the whole the list, I finally review them in the normal way. I should add that I do the same for the cards that I failed on that day too (I review early in the morning and learn late in the evening) so the list of the due cards can be longer than 16 but since the failed cards are much easier to re-learn than the new ones to learn that is not of particular problem.

Well, that’s the way I go about learning new cards – maybe it can work for others as well.

Inny Jan Wrote:

amtrack Wrote:1. Pick out 5-7 words at a time, no more no less.

For exercising your long term memory 5-7 items is not a good number. They are handled quite well by your short term memory. To make sure that you are committing newly learned items to the LTM you need at least twice as many. (BTW, I seem to remember - sorry, I can't be bothered to look it up now - that Anki's creator suggests 20 items at a time.)

I simply do 5-7 blocks at a time until I reach 20+ for the day. In anki those cards will be reviewed in total the very next day (so 20+ at a time), thus meeting the long-term memory requirement.

Long story short, day 1 gets several groups of words into my short-term memory. Day 2 will give me all the words I studied on day 1 at once, thus working words into my long-term memory. The only reason I do it this way is because I cannot process 20 words at once the first day I study them.

EDIT: And in fact I can study much more than 20 words for the day and successfully blast through them all in review the very next day. I normally don't because, quite frankly, I get too bored.

A quick side question in relation to this: has anybody gone down a route primarily involving image based recognition as opposed to primarily word/kana/kanji prompts? For example, having an *image of a dog, or a type of tree, etc. on the front, with the Japanese term for it on the back?


*purely for concrete objects, etc. which have a general 1:1 correlation between the image and the word.

uisukii Wrote:A quick side question in relation to this: has anybody gone down a route primarily involving image based recognition as opposed to primarily word/kana/kanji prompts? For example, having an *image of a dog, or a type of tree, etc. on the front, with the Japanese term for it on the back?


*purely for concrete objects, etc. which have a general 1:1 correlation between the image and the word.

Try: http://www.towerofbabelfish.com/

Oniichan Wrote:Try: http://www.towerofbabelfish.com/

Read through the page, method, etc. I've read a few different scientific studies on related contextual/non-contexual vocabulary experiments, etc. most of it rather interesting. The "base vocabulary" section of that website is similar to something I've been experimenting slowly with the core10k deck. Creating a second deck as I go through it in a similar style, focusing on imagery. Personal anecdotes was really what I was interested in, but didn't want to create a new thread. Anyway, thanks for the link.

Oniichan Wrote:

uisukii Wrote:A quick side question in relation to this: has anybody gone down a route primarily involving image based recognition as opposed to primarily word/kana/kanji prompts? For example, having an *image of a dog, or a type of tree, etc. on the front, with the Japanese term for it on the back?


*purely for concrete objects, etc. which have a general 1:1 correlation between the image and the word.

Try: http://www.towerofbabelfish.com/

I have done this a few times in the past because the study materials I was using included them. This may be different for different people, but I found them distracting. I know other people have different methods that suit them, but I try to keep my cards simple without any extraneous information on them. My cards have only a Japanese word on the front, English on the back or vice versa.

I have however started adding Kanji to my kana vocabulary on the front. This way, if I happen to glance at the kanji, even if I don't know it, I'm becoming more familiar with it and more likely to recognize it next time I see it. Also, if I happen to know the meaning of the kanji, it gives me a hint as to what the kana means.

uisukii Wrote:A quick side question in relation to this: has anybody gone down a route primarily involving image based recognition as opposed to primarily word/kana/kanji prompts? For example, having an *image of a dog, or a type of tree, etc. on the front, with the Japanese term for it on the back?


*purely for concrete objects, etc. which have a general 1:1 correlation between the image and the word.

I have not, but do be diligent about the images you choose if you go that route. The more you know, the more ambiguous images become. Heck if I showed you any random image, you could probably come up with a good 5 words in english to associate with it.

^Oh, it isn't something I'm doing myself. The question was put out there purely out of interest to see if anyone had any anecdotes about personal experiences with a similar methodology. Interested after reading various academic/scientific papers on changing trends in vocabulary "teaching" in class settings. Personally, my internet connection isn't reliable enough to source enough images at the rate I'm interested in studying. Core10k is good enough; I've divided it up into thematic groups and semantic groupings for the verbals/adjectival terms/phrases.

I didn't think the initial question would garner such response. For the sake of the original post question, I hope you aren't offended if further discussion on this topic is left without response. It would probably suitably discussed further in a thread of its own.