Comment by emtel | Hacker News Reader

Comment by emtel | Hacker News Reader

On the topic of learning to remember characters - Japanese students attempting to learn kanji (very close to traditional chinese characters) often use a book called "Remembering the Kanji" by James Heisig. The method devised by the author for learning the characters is truly ingenious.

He developed this method after moving to japan in the early 80s for a position at a research institute. In a few months, he developed and applied his method to learn english meanings (but not japanese readings) for all ~2000 standard-use kanji. When his colleagues asked him what he'd been spending all his time on, and he told them that he'd finished learning all the kanji, they didn't believe him. Not because they didn't believe he'd done it so quickly, but because they had never seen any foreigner ever learn more than a few hundred. After demonstrating to their satisfaction that he'd really managed to do it, he was told "go write a book about how you did it".

I've been working my way through RTK (with the help of anki) for several months now. The method really does work. I'm at about 1500 out of 2200. It's getting harder the further I get, but all told I don't think I've spent more than a hundred hours or so over the course of 4 months.

For chinese learners, he's produced books on remembering Hanzi, simplified and traditional. I think if Moser had had access to these books, he might have been a bit less despondent about the task of learning the characters.

"Remembering the Kanji" does work. I'm re-visiting it right now. I used it many years ago (when I was younger and my memory was a bit better..), and at that time I managed to, with relatively little effort, learn between ten and twenty new Kanji every day. I spent maybe half an hour every day, or at least never as much as a full hour. The trick isn't really flash cards - because it's not really about imprinting them to memory. The trick is to create a working association, something that calls forward not only the meaning when you see the Kanji, but also tells you how to write it when you think about the meaning. With the right association it'll stick after one try, or, at worst, after two or three.

When I started visiting Japan a couple of years later I found that when I ran into a Kanji I knew, on some sign or whatever, it immediately transferred to learning the sound where it was used, and remembering the meaning of words with Kanji I knew (e.g. parking lot entrances and exits, to name a couple of simple ones).

But for various reasons I stopped working with the Heisig book after I knew less than two hundred of them, so I've started it up again. As I'm older now I've noticed that it's a slower process than before, and I have to work a bit more to get working associations. The net result is that I can't do as many Kanji per day as the previous time, and sometimes I run into bumps where I don't get the associations really working, and then I have to work on that for a while before I can continue. So there may be a week with no progress.

When that's said - I'm not sure this would be that useful for Chinese characters. The set of characters to learn is just so much larger. The reason for using the Heisig method to remember the Kanji is simply because you can learn all of them (as used in Japanese) in a relatively short time, and it turns out that when you already know them it's a much shorter step to learn them for real (as in when you start learning to read and write Japanese). And Heisig himself mentioned that, he noticed that Chinese students of Japanese learned to read and write Kanji much faster, because they already knew the meaning. Even though the languages are totally different. And, as I know a Chinese who is now learning Japanese, I've seen the same first-hand.

I'm not sure what you're seeing is an effect of being older. Learning the first 200 kanji in RTK is easy. So easy it feels like magic.

Things get much, much harder by the time you get to 1000. You start to have very similar overlapping words (e.g. Earnings and Profit and Assets and Savings), primitives that share the same name as an graphically-unrelated kanji, kanji that are made up of a grab bag of completely random primities, e.g. Dissolve, which is made of ceiling, mouth, hood, human legs, spike, insect.

I'm also skeptical of the idea of learning these things permanently on your first try, as Heisig claims. He even used flashcards himself, and describes how to review with them properly. There are certainly some characters that I have such vivid stories for that they stick easily, but coming up with 20 such vivid stories per day seems impossible.

It might be somewhat handy to have the rough meanings memorized, but it's a very long way from there to actually reading and writing Japanese. I never tried it myself, but I've heard many advanced Japanese learners say they did it early on and later felt it was mostly pointless.

I read and write Japanese fluently and never studied kanji alone; I learned them along with the vocabulary they appear in, which is the approach recommended by the best advanced Japanese schools.

Just wanted to +1 this. I also read Japanese at a high level (JLPT N1) and I've always learned characters in context of the vocabulary they are in. Spending months learning 2000 "meanings" for characters isn't a great way to spend your time in my opinion, the hard part of learning a language is remembering tens of thousands of words. Kanji is relatively not a big deal.

The exception I suppose is if you really want to be able to write characters by hand, then the Heisig method has value. If you can live without being able to handwrite well, then it's hard to justify spending months up front remembering unhelpful meanings than actually learning the language itself...

The time spent to learn Kanji via RtK is in my experience only a very little part of the time spent on everything else related to learning Japanese. The RtK way (and I don't do flashcards) is actually a fun diversion, and I find that I enjoy drawing the characters. And that's from a guy who can't stand writing by hand in general.. I started using a typewriter at 14. And I can't draw pictures to save my life. Unlike the previous time some years back, this time I'm using a cheap click-to-erase mechanical 'magic' drawing pad. Much better than the stack of paper I used back then.

Funny enough "learning characters/words in the context of the vocabulary they are in" is exactly what NLP machine learning models use to learn "rich" word/text representations based on the "distribution hypothesis" which states that the choice of words in the same context share a common meaning.

This is certainly true - I thought about emphasizing this point but didn't want to make my comment longer.

The approach is interesting, but I'm not sure it works for everybody. I tried the Heisig method, but I found that while I could get good at flashcards, they didn't really stick.

Probably helps if you are also exposed to Kanji on a daily basis in their true context.

Reply on news.ycombinator.com