So one of the weaknesses of the standard sentence-recognition approach to Japanese review is that it doesn't provide review of kanji production. Early in the process RtK fills this role, but it's my opinion that RtK reviews should eventually end (at different times for different cards) because the keyword approach begins to be a liability rather than an advantage—an English keyword will reinforce the idea of thinking of the character in terms of English rather than in terms of Japanese, and even a Japanese keyword has the problem of providing a restricted view of the character.

My initial approach to this was to practice production of sentences from readings; the context is usually enough to determine proper kanji. However, this approach sucks pretty hard; it's nonspecific (forget or err on one kanji in the sentence and the entire thing goes back into the queue) and unbelievably tedious in large quantities.

My new approach, which has worked fairly well so far, is to provide a representative sample of words that use the kanji with the kanji itself dekanjified (for example, the prompt for 苦 I've used is "「にがい」「く痛」「く労」「にが手」「くるしい」", sufficient to unambiguously determine the character I'm trying to produce), as well as some descriptive text if necessary (for instance, for 盾 I provide a note that mentions that this is the version without the 木 radical). Answer sections contain the character and stroke count.

Advantages to this approach include decoupling of the kanji from a single specific meaning (especially one in English), low-cost review of kanji production, and passive review of a smattering of readings and words that use the kanji.

Disadvantages include complexity identifying characters suitable for transitioning (have to be able to recognize enough words using the character to disambiguate the character), slow transition process (have to enter all disambiguating words, dekanjify appropriate characters), and poor reusability (disambiguating word choice is heavily dependent on personal vocabulary).

Nevertheless, I've found it useful, so I thought I'd duck back in to mention it in case it might help someone else.

~J

Uhm, I'm getting the grasp of production just with reading.
I guess your system is very nice if you are in a hurry, but production ability do catches up with recognition after some time.

That may not be the case for everyone. For example, back when I was taking Japanese classes, our professor (a native speaker) noted that she wasn't able to write 麒麟, but was able to read it, implying that at least for her production ability hasn't caught up even after several decades.

~J

Interesting stuff. My personal 'method' was/is to just drop reviewing kanji separately and just ask myself, when reviewing sentences, whether I can visualize the kanji easily w/o looking--if I can't, or if it feels iffy or difficult, then I write out the kanji once or twice (depends on how neurotic I am about my handwriting at the time). That applies to compounds/words/etc as well. In other words, post-RTK production for me = mental rehearsal + occasional gestural maintenance.

I guess I would have recommended single word production or dictation in the past, but now I think the above or something like your method would be better.

I'm not sure if continuing to review english key words will be damaging. I think the readings themselves just transend RTK. When I see a japanese word, the Heisig thing basically melts away eventually, and the frequency of the Heisig reviews will be pretty far and few after a while I would guess.

I am just now trying a production card method- Very simple. My deck is all recognition, but I will be taking mature cards with a not due for 6 months or more status (right now I have a pretty small deck, only 375 cards with about 20 due in 6 months or more), and flipping the card (creating a additional card) perhaps having just the audio of the sentence for the question, and the kanji for the answer. I figure it might be a semi-painless thing, as the cards are already well known, and it won't take much effort to make the cards. Maybe every week or two I will go through the deck and see which ones are ready for the "production" phase, and eventually do that daily.

I came up with this idea out of a desire to be equally skilled at production, and the reasoning that it would be more efficient this way than if I created both for each card at it's creation. has anyone else tried this "stagger" method? I am just now implementing it into my studies.

@ Mentat: 2 kinds of production: determining which kanji is in a word and being able to write a kanji. Reading helps with the former (probably not as effective as some active recall), but not with writing. What Woodworj is doing tests both (but with multiple hints as to which kanji).

@Woodworj: welcome back! Sounds like a great way to maintain ability to write kanji and review various uses/readings in one step. btw, are you trying to cover both meanings and readings or just create a more nuanced prompt?

In creating your deck, you might want to take a look at the Kanji in Context vocab list. They've already selected a handful of vocab for each kanji to cover the common readings and vocab. The Tanuki list already has the vocab with the target kanji in kana (plus short defns and sentences). Traditional paper flash cards also have this info on them. Hopefully these resources can speed up the process for you.

This approach assumes one no longer relies on stories to produce the kanji. Some people may opt to use only one Japanese word as prompt in order to a) transition to Japanese prompts, b) preserve their stories as something to fall back on when writing as a last resort, and/or c) speed up reviews. Even with single Japanese prompts, however, multiple words (or sentences) are sometimes needed for disambiguation. Combining your deck with the Japanese prompts deck, someone could still select one word as the [story] prompt and separate the other words as a hint. But learning vocab outside of kanji review is enough to ensure a more nuanced meaning of the kanji will develop/be maintained I think.

Thora Wrote:@Woodworj: welcome back! Sounds like a great way to maintain ability to write kanji and review various uses/readings in one step. btw, are you trying to cover both meanings and readings or just create a more nuanced prompt?

The main purpose is creating a nuanced prompt; I tend to go overboard in terms of coverage, both because kana lookups are relatively slow in my dictionary (discouraging checking to see if collisions actually exist) and because I'm bad at at not letting covering meanings/readings creep in. In principle, the latter task is left to sentences and general experience.

Quote:In creating your deck, you might want to take a look at the Kanji in Context vocab list. They've already selected a handful of vocab for each kanji to cover the common readings and vocab. The Tanuki list already has the vocab with the target kanji in kana (plus short defns and sentences). Traditional paper flash cards also have this info on them. Hopefully these resources can speed up the process for you.

I appreciate the advice; the main slowdown comes more from selecting a variety of prompts that I know I'll recognize and are collectively distinct from other possibilities, plus the actual data entry.

Well, that and the sheer volume of cards this is being done for.

Quote:This approach assumes one no longer relies on stories to produce the kanji.

This is an excellent caveat that I forgot to mention, thank you. That's pretty much the criteria I use when selecting cards to transition (other than "do I know enough words using this kanji"): cards where I no longer need the story to produce the character.

(I also moved away from stories for most new characters some time ago, which is probably why I forgot to mention it.)

Quote:Combining your deck with the Japanese prompts deck, someone could still select one word as the [story] prompt and separate the other words as a hint. But learning vocab outside of kanji review is enough to ensure a more nuanced meaning of the kanji will develop/be maintained I think.

This may be a mileage-may-vary point; I've found that I'm better at dissociating the meaning of a kanji in the context of a single word from the greater "possible meanings of this kanji" than I am at dissociating the English keyword from the latter. The "nuance" I talk about actively seeking above is nuance in disambiguation, such as picking compounds that aren't shared between kanji that share a word (戦 and 闘 both make たたかい, but 作戦 doesn't have a 闘 variant)—the only official goal is maintaining and improving ability to write the kanji on demand.

~J

I started doing something similar because, despite begin able to read words, when I tried to write them myself I couldn't remember the kanji to use. Though I just have the word (or a small phrase) in kana, and have to write the whole word in kanji. But for some words I did add an English definition to help tell words with identical meanings apart (though I'm considering adding Japanese definitions/example sentences instead).

But since I've not been doing it for too long, I can't say much about the benefits of it.

woodwojr Wrote:... I've found that I'm better at dissociating the meaning of a kanji in the context of a single word from the greater "possible meanings of this kanji" than I am at dissociating the English keyword from the latter.

Sorry, I wasn't very clear. I meant select one Japanese word as a prompt and keep the other words for disambiguation/hints. The Japanese prompt would be as close a possible to the RTK keyword so existing stories can be used. (Only a few English keywords don't work.) I have the same issue with English keywords getting in the way.

I started looking at Japanese prompts again and completed about 800 (in KiC order) with that kind of information before tiring of it. Yeah, checking for conflicts/homonyms is time consuming! I'll see if I can finish more and send what I have to Wrightak who is currently revising his list. Perhaps we can share info at some point to create a deck that others can customize to their liking. I see it as a great way to maintain hard-earning writing ability post RTK. Thx for posting.

albion Wrote:I started doing something similar because, despite begin able to read words, when I tried to write them myself I couldn't remember the kanji to use. Though I just have the word (or a small phrase) in kana, and have to write the whole word in kanji.

Sounds ideal, but doing production for every word is time consuming. One way to save some time is to go through an entire list of production cards - deleting a card whenever you get it correct. I found that production cards were good for identifying weaknesses, but multiple reviews usually weren't necessary. During recognition review, I'd remember having failed the corresponding production card at least once and that seemed to be enough reinforcement.

Quote:But for some words I did add an English definition to help tell words with identical meanings apart (though I'm considering adding Japanese definitions/example sentences instead).

Would JLPT type kanji exercises be a good source of ready-made sentences with target vocab written in kana? I imagine they're written so that the meaning is clear from context. Tanuki has searchable brief sentence + Jdefinition, but the mixed kana/kanji would need to be cleaned up. [pls excuse me butting in if your post was directed at W]

Personally I feel keeping up the RtK reviews is without a doubt the best solution. It's easy, it's effortless and it's an investment you've already made. You get less reviews with an old deck than a newly created one, so why would you want to stop doing RtK reviews just to do the same thing all over again?

I started migrating to Japanese keywords for the kanji instead. I simply add a japanese keyword whenever I review a kanji which I feel I need Japanese keywords for. This way, the effort I have already put in is maintained.

As for remembering which kanji to use for which words, that's a different skill altogether and I wouldn't group it with kanji production. It can also be trained in tons of ways but it usually takes almost no effort... 3-4 reviews in Anki where you have to write the word and you'll remember it for years, most of the time.

woodwojr Wrote:My new approach, which has worked fairly well so far, is to provide a representative sample of words that use the kanji with the kanji itself dekanjified (for example, the prompt for 苦 I've used is "「にがい」「く痛」「く労」「にが手」「くるしい」", sufficient to unambiguously determine the character I'm trying to produce), as well as some descriptive text if necessary (for instance, for 盾 I provide a note that mentions that this is the version without the 木 radical). Answer sections contain the character and stroke count.

I had a similar idea, just automatically but indiscriminately adding all the jlpt vocab that includes the kanji:

http://forum.koohii.com/showthread.php?p...1#pid30041 http://forum.koohii.com/showthread.php?p...8#pid30748

That has the advantage of being low-effort, definitely. I'm uneasy about the idea of shifting over to it myself, because my vocabulary is still small enough that I'm only likely to know a handful of words for a given kanji and I'm nervous about the possibility of introducing a cue that calls to mind the wrong kanji, but I've done nothing to make sure that this risk actually exists.

~J

Thora Wrote:Sounds ideal, but doing production for every word is time consuming. One way to save some time is to go through an entire list of production cards - deleting a card whenever you get it correct. I found that production cards were good for identifying weaknesses, but multiple reviews usually weren't necessary. During recognition review, I'd remember having failed the corresponding production card at least once and that seemed to be enough reinforcement.

I have only just started, but I have just been doing it normally at the moment. I don't add a lot (it's only set to 5 new cards a day at the moment), so I don't have a lot of reviews to do (at the most it'll be around 50-60) right now. But there are several where I've made a mistake after getting it right once or twice before. Some of it is a problem with forgetting kanji which is probably something finishing RTK would help with, but others are forgetting which kanji to use.

Quote:Would JLPT type kanji exercises be a good source of ready-made sentences with target vocab written in kana? I imagine they're written so that the meaning is clear from context. Tanuki has searchable brief sentence + Jdefinition, but the mixed kana/kanji would need to be cleaned up. [pls excuse me butting in if your post was directed at W]

So far I've just been adding random words I thought were particularly problematic (I added 部屋 after not being sure on the reading after seeing it in a book). But I have thought about trying with something more structured. I might try it with some JLPT material.

Yeah, I think it is a good idea. I'm just saying it is not really the only way to do it.

For the 麒麟 example, well, the Kanjis are really not hard to write because the りん and the き radicals are very common, and off course he would know 鹿, so I think your professor was just being modest.

@cangy - ths for the field overwrite plugin. I read the detailed instructions (11/08) you provided and am going to try it (though I'm still slightly intimidated). I assume it works with the latest version of anki?

@mentat - I understand your point that we can use other bits of knowledge when trying to write kanji, but I think you overestimate its usefulness a bit. For eg, it doesn't really help if a radical is common if you're uncertain which of 3 common radicals to use. Phonetic components offer very little help in writing (btw I don't see how the RHS elements in 麒麟　would be apparent.）There's a reason people can read far more kanji than they can write. And they forget ones more common than 麒麟. (Is that one normally written in kana?)

But whether reading is enough to maintain an ability to write kanji is a slightly different issue than whether Japanese learners should invest any effort into writing. Those who don't practice writing because they'll "never to need to write Japanese" are missing out on one of the most effective learning tools imo. Writing helps reading. And the ability to determine and write the correct kanji is better developed by production.

As far as writing kanji, do you ever have a mental image of its layout, shape and density, but one part is too blurry to decipher? Sometimes I can rely on a story to figure it out (yeah RTK!). More often, if I stop analyzing it and just start writing, my hand seems to complete it. I equate it to playing a musical instrument - you can often play through a forgotten passage if you just stop thinking about it. I think this goes beyond muscle memory - it's more like accumulated patterns, sounds, shapes, movements, knowledge are internalized and the answer just comes to you. Reading alone doesn't create as deep a pool of experience to draw from.

No right or wrongs here - I guess I'm just trying to present a view of writing as part of the process rather than an end result. Before simple IME, the benefits of writing were unavoidable - now it's a personal choice. [edits - I've forgotten English :-)]

This is a lengthy topic, but Khatzumoto once suggested in a post to arrange your sentences deck so that the question side has the sentence 100% in kana with spaces to separate words, then kanji on the answer side, which you would have to produce to "pass".

For example:

Question: よい　てんき　も　ながく　は　つづかなかった.
You write down what you think it is with kanji, make sure you understand it.
Answer: よい天気も長くは続かなかった.
Profit!

I've considered doing this eventually. Any thoughts on it?

sethg Wrote:This is a lengthy topic, but Khatzumoto once suggested in a post to arrange your sentences deck so that the question side has the sentence 100% in kana with spaces to separate words, then kanji on the answer side, which you would have to produce to "pass".

For example:

Question: よい　てんき　も　ながく　は　つづかなかった.
You write down what you think it is with kanji, make sure you understand it.
Answer: よい天気も長くは続かなかった.
Profit!

I've considered doing this eventually. Any thoughts on it?

I (and others I think) tried these kind of cards for a while and found that it takes way to long to get through reviews, making it too tedious to be worth it. I prefer making one word in kana that I have to write out, or just using wrightak's JRTK deck

Thora Wrote:@cangy - ths for the field overwrite plugin. I read the detailed instructions (11/08) you provided and am going to try it (though I'm still slightly intimidated). I assume it works with the latest version of anki?

yep, still works for me at least...

these are still the best instructions, though with the built-in plugin downloader now you can ignore the bit about downloading, and also where it says "The plugin is run when the deck loads." note that when anki starts up it seems to load the current deck, then enable plugins, so this means you need to close the current deck and then open it again for the plugin to activate

also if you are planning on using those jlpt lists, I can build another one with furigana readings and including iknow core vocab too

sethg Wrote:This is a lengthy topic, but Khatzumoto once suggested in a post to arrange your sentences deck so that the question side has the sentence 100% in kana with spaces to separate words, then kanji on the answer side, which you would have to produce to "pass".

For example:

Question: よい　てんき　も　ながく　は　つづかなかった.
You write down what you think it is with kanji, make sure you understand it.
Answer: よい天気も長くは続かなかった.
Profit!

I've considered doing this eventually. Any thoughts on it?

Yep. It's the method I describe in my second paragraph in the original post, the one I used before this. Verdict is "sucks pretty hard", and obliterating the sentence-production cards was required for me to dig myself out of the thousand-due-card (~seven hundred after you remove sentence production) hole I'd gotten myself into.

I don't recommend it, attractive as it was for its simplicity (I could get it free with any sentence I entered, instead of all of this painful data-entry).

~J

sethg Wrote:This is a lengthy topic, but Khatzumoto once suggested in a post to arrange your sentences deck so that the question side has the sentence 100% in kana with spaces to separate words, then kanji on the answer side, which you would have to produce to "pass".

For example:

Question: よい　てんき　も　ながく　は　つづかなかった.
You write down what you think it is with kanji, make sure you understand it.
Answer: よい天気も長くは続かなかった.
Profit!

I've considered doing this eventually. Any thoughts on it?

Yep. It's the method I describe in my second paragraph in the original post, the one I used before this. Verdict is "sucks pretty hard", and obliterating the sentence-production cards was required for me to dig myself out of the thousand-due-card (~seven hundred after you remove sentence production) hole I'd gotten myself into.

I don't recommend it, attractive as it was for its simplicity (I could get it free with any sentence I entered, instead of all of this painful data-entry).

~J

Haven't read the entire thread, so sorry if this has been said already - assuming you're learning use the sentence-picking method (AJATT), how about just picking *one* kanji or kanji word from a sentence you've already been using, and dekanjifying that? Then, provided you do that for 2000-odd sentences, you can have practice producing each kanji. I'd suggest highlighting or colouring the word you want to produce too, to distinguish it from other types of card, as well as resetting the card's interval. That way, there'd be no ambiguities in the sentence, and you could focus just on the writing (since you've already learnt the sentence).

Thoughts?

It could work, but I have (possibly unfounded) concerns about disambiguation, and it doesn't seem like it would save much effort. Still, it seems like it could be roughly comparable in effectiveness if disambiguation isn't as much of an issue as I think.

~J

sethg Wrote:For example:

Question: よい　てんき　も　ながく　は　つづかなかった.
You write down what you think it is with kanji, make sure you understand it.
Answer: よい天気も長くは続かなかった.
Profit!

@uberstuber @woodwojr

What about using this method but with one kanji in kana in the question?

Edit: What about practicing writing real Japanese instead of SRS reviews?

Sorry about just asking, I'm still doing Tae Kim's

Going all kana to all kanji sucks way too hard and takes way too much time. Just picking one word from the sentence to go kana - kanji is much faster and equally as effective but tbh it still kinda sucks.

Maybe its just a necessary evil?

I was doing it that way for a while before I started doing KO. One thing I can definitely recommend is just making production cards out of already mature sentences you've collected. That way you can 90% of the time guarantee that you're at least familiar with the word, how to read it and what it means. Instead of having to assimilate all of that and how to write it straight of the bat. I think that's where I had the most trouble.