There seems to be more and more touch screens as the years pass and eventually someone will have the wonderful idea of making a kanji SRS that will tell be able to tell you if you wrote the kanji correctly. It could even be able to tell if you mistook one kanji for another and increase the review rate for both accordingly.

It's possible to make it now but drawing with the mouse is a pain, how available are touch screens these days? Maybe someone's already made it and I've just not heard of it.

Web based Kanji OCR (Optical Character Recognition): http://chasen.org/~taku/software/ajax/hwr/
Web based Kanji SRS: http://kanji.koohii.com
Cheap Touch screens: [never been good at finding hardware]

there's also a possibility of using a tablet.

Nintendo DS + any of the kanji learning software.

Within five years, technology is going to make it so fluency in Japanese can be achieved in 6 months while having fun. I think I can achieve it in two more years with what tools I have available to me now online. With what was available when I finished high school in 1987, I could have studied for ten years and still not gotten there.

I have a $1200 Dell Tablet PC that I'm -very- happy with. Woot.com recently sold the newer model of it for $600, and Dell sells the base mode for $900+.

When I was trying RTK, I used Anki on the Tablet and had the Windows handwriting recognition be the bottom part of the screen. I'd try to remember the kanji from the keyword, then draw it. Then I'd match it to the answer and see if I got it right.

It was working -very- well.

In fact, I was sitting here researching the API for that handwriting system to see if I could write a kana/kanji drilling software with it. If anyone knows how to hook into it with C# or Ruby (prefer Ruby) and would be willing to give me some pointers, I'd be happy to have them. The project will be released as Open Source if it ever gets anywhere, so there's no money involved.

bodhisamaya Wrote:Within five years, technology is going to make it so fluency in Japanese can be achieved in 6 months while having fun. I think I can achieve it in two more years with what tools I have available to me now online. With what was available when I finished high school in 1987, I could have studied for ten years and still not gotten there.

Not likely. There's no way around the time needed to be invested. Regardless of how good technology we have, 2 hours of studied Japanese is still 2 hours. Getting about 10 000 hours invested in Japanese in 6 months will not be possible until we get some form of The Matrix solution where data is simply downloaded into your brain.

I see no benefit in touchscreens. Gimme a pen and a notebook to write my kanjis, thanks.

Tochscreens suck.

Tobberoth Wrote:Not likely. There's no way around the time needed to be invested. Regardless of how good technology we have, 2 hours of studied Japanese is still 2 hours. Getting about 10 000 hours invested in Japanese in 6 months will not be possible until we get some form of The Matrix solution where data is simply downloaded into your brain.

Well, the idea behind advances in language learning is always that you will progress more with less time invested. So, the idea is that what used to take 10,000 hours could be done in a lot less.

A lot of people on this site seem pretty critical of classes, for example. But if you have a good class it can dramatically cut down on how much learning time you need. I took a class for three years, and only did 10 or so hours a week, for 9 months out of the year. Not a whole lot of time invested, but I learned a ridiculous amount of Japanese at the time (I attribute it to being a good program. The professor was one that used to run the Falcon program at Cornell, along with native Japanese teachers who were trained as Tanaka lecturers by Jorden. Naturally, the textbook was Japanese: The Spoken Language. Oh, and the class was fun!). Now I live in Japan, spend all my free time studying (with SRS, RTK, reading simple books, and a private tutor) and am not making as much overall progress (with the exception of reading, which is improving faster than it was in the class).

However, I myself would guess that the amount of time needed to invest will decrease faster for reasons other than technology. That is, serious studies done about effectie ways to teach have only been in abundance for the last 30 years, so we don't know a whole lot about good teaching. And then, studies done about memory should help develop techniques to remember things easier. These studies should play a big part in language learning the in the future.

Nevertheless, I'm interested in how technology can help language learning too.

Tobberoth Wrote:There's no way around the time needed to be invested.

Um wasn't this forum created around the Heisig method, which we all agree is an incredibly efficient method of learning how to write Kanji, and the implementation of an SRS system which is an incredibly efficient way to handle loads of information? Those two things alone prove your statement to be false.

Now I'm not saying the 6 month to fluency claim will be true within 5 years but the belief that things can not be done more effectively in a quicker amount of time has been proven false about 1000000000000000000000000000 times throughout history.

Oh and BTW most of the popular netbooks (ASUS EEE, MSI Wind's) have touchscreens that you can install for about $90.

vaerer Wrote:Web based Kanji OCR (Optical Character Recognition): http://chasen.org/~taku/software/ajax/hwr/

Awesome! Just switched to ubuntu linux and really missed this from the windows IME. (Got 64 bit and can't seem to install Tomoe)

Here is a web-based OCR-like tool, which tells you to write a specific Hanzi, and then makes it pretty: looks pretty cool and has gained some popularity among the Chinese-learners, however my internet is too slow to really check it out
http://www.skritter.com/

Last time I checked out Tomoe its recognition engine was not anywhere near as good as the commercial ones (no surprise -- handwriting recognition, which incidentally is not the same thing as OCR, is a Hard Problem and the commercial engines have a huge head start in development time). And of course for use with reviewing kanji you have the problem that these engines are generally designed to be forgiving, to make best guesses based on sloppy handwriting or common miswritings, which is exactly what you don't want for deciding whether to pass/fail a review.

Anybody seriously interested in doing work on this ought to compare Tomoe and Zinnia to see which would make a better base. This guy did and he reckons Zinnia is both faster and more accurate.

I would also remark that if your recognition engine is only getting a 77% first match accuracy it's not going to be much good for kanji review scoring.

The Kanji LS app uses iphone's touch screen for practicing and testing "fingerpainted" kanji. In test mode, you are given the english meaning or the kana. You produce the kanji. Then you compare the character side by side with the app's version of the character.
No OCR, so you are the judge of the side-by-side kanji. Good 'nuff.

You can also sequence through the stroke order to see how your stroke order matched the official stroke order. It does keep track of the order in which you wrote the kanji. The app has severe limitations (no SRS or customized decks, for starters), but the drawing part of it works pretty well.

I agree there is nothing like pen and paper, but I don't always have those handy, and I do always have my phone handy.

pubbie Wrote:I thought I'd mention that mobile anki can work on cellphones with touchscreens.
http://ichi2.net/anki/wiki/AndroidAnki "White board"-mode. I'm going to buy a Google Android phone any day now and having anki in it makes me want it so much more. Anki doesnt do OCR sadly. You have to use your own, unbelievably good, OCR device, namely your brain.

I have a G1 and I just use the 'online study' from the site. That means no audio, but I'm okay with that... I just set all the audio cards to 'noweb'.

HerrPetersen Wrote:Here is a web-based OCR-like tool, which tells you to write a specific Hanzi, and then makes it pretty: looks pretty cool and has gained some popularity among the Chinese-learners, however my internet is too slow to really check it out
http://www.skritter.com/

I would love something like this for Japanese. -sigh-

pubbie Wrote:You have to use your own, unbelievably good, OCR device, namely your brain.

Ha, ha, ha... yes, that's exactly it. At first I was disappointed that Kanji LS app didn't tell me if I was right or wrong, but after using it for a few days I became really glad that I am the judge and jury. This way, I will never find myself arguing with my phone over whether I drew the kanji correctly or not. I am the decider.

The character recognition for the DS games (like Kanji Sono Mama) is incredible. If only it could be exported!

pm215 Wrote:Last time I checked out Tomoe its recognition engine was not anywhere near as good as the commercial ones [...]

Anybody seriously interested in doing work on this ought to compare Tomoe and Zinnia to see which would make a better base.

last time I checked zinnia there was no front end available, but one has recently been released: http://www.mblondel.org/journal/2009/02/...of-tegaki/

Does anyone know how to install Zinnia/Tegaki? I'm kind of new to the linux game. I'm not dumb with computers, but when you've been using windows since you were 5, it's hard to get used to something else.

And I really miss drawing in kanji

I recently found RTK, and I've been thinking about these things.

Automatic scoring could be very useful. I have played memorisation games in various subjects that check your answers, and this seems to speed up the testing a lot compared to simple flashcards.

These games didn't have SRS features, but it's easy enough to imagine SRS with the same sort of interfaces. For questions with simple answers, such as clicking the right place on a map, it may even work for the system to rate the difficulty of a question just by how long it takes you to answer. I'm not sure what could be done with drawing questions, but there may be potential there.

The method practised at this site means a lot of time spent doing review questions. So if the same questions could somehow be tested more quickly, with the same effectiveness, this would reduce the overall time to learn all the kanji.

I notice, when using flashcards, I do need to write the answers. If I try to answer questions in my head, then when I flip the card, I'm not always sure whether I got it right. The answer on the card can sneakily replace the one in my head before I've finished comparing them. ...Now it comes to me that I could speed up the testing a little by not putting my pen down when flipping and rating the card ;)

If I'm going to make a serious attempt at RTK, I'd also hope to test the stroke order. I could easily enough make flashcards with stroke order diagrams as the answers. But checking my answer is quite time consuming. And once I've written the character, my brain can start playing tricks on me like in the last paragraph, and make me think I drew it in a different order than I actually did. Computer entry that records stroke order has an advantage there.

It seems to me that checking correctness is a different problem from normal handwriting recognition. As pm215 mentioned, the latter attempts to be forgiving; and it gives you several options to choose from. Perhaps a partially automated solution could work. If the program is sure you've got it right, nothing further is needed; but if it's not sure, it can show you the two side by side and let you decide. It may also be possible for the program to check the stroke order while assuming the shape is correct.

Do any of the existing kanji drawing games have SRS and customisable decks?

I've never heard of any game using a SRS algorithm.

mentat_kgs Wrote:I've never heard of any game using a SRS algorithm.

Slime Forest Adventure uses one.

wccrawford Wrote:

mentat_kgs Wrote:I've never heard of any game using a SRS algorithm.

Slime Forest Adventure uses one.

Slime Forest Adventure is really useless though, at least that's what I felt when using it... but then again, I already knew several hundred kanji and it still kept sending hiragana enemies after playing for quite some time so...