Between when I registered and August, I used the site and the method quite precisely. In July I began having progressively less success, for reasons that I can't begin to identify in any rigorous way. After a brief bout with attempting to make a directed acyclic graph out of the kanji and radicals followed by a period of computerlessness (ok, I had my XO, but damned if I could find a way to add CJK support without an additional working computer), I abandoned the method entirely for a period of time. I was at about 1600 then.
A few weeks ago, I got fed up with my poor production ability (compounded by the fact that Anki broke its ability to suspend a question independent of the entire fact a few versions ago) and imported the RTK decks. I then went on a process of unsuspending precisely those kanji that fit into one of the following categories:
1: I already know it
2: I almost know it (composed of elements I have familiarity with and strong recognition capability)
3: It's composed of elements I know and is sufficiently common (this is a very loose definition; stuff outside the 2000 most commonly used is usually out unless it's very easy, 1000-2000 most commonly used has a lower threshold, and 1-1000 is in; if I see it a lot, it's also in).
4: I really want to know it for some reason (possibly including "is a component of one or more kanji I want to know").
This has resulted in a pile very distinct from Heisig order, and with a much larger number of immediately useful kanji. Since then, whenever my stack of unseen, nonsuspended cards grows short, I reapply the above criteria with my new knowledge and any additional reading I've done. Additionally, I make liberal use of this technique to cover up for bad keywords.
I also keep an eye out for sentences that incorporate kanji that I've learned that I haven't already obtained sentences for; the end goal is either to turn off the English keyword (which would require a lot of work, as right now I only add disambiguation text if it's necessary) or to simply delete the Heisig decks.
So that's my approach. It isn't a drop-in replacement for RtK (among other things, it relies heavily on the fact that I learned a fair bit of Japanese before doing this), but for me at least it corrects for the two critical problems with RtK: bad keywords and bad order. The first is obvious; arguing the latter is for another topic and has probably already been done.
But yeah. I credit Heisig for giving me the idea, but I think it would be very easy for someone with a DAG to make a better system. (Actually, I think the DAG itself is a better system, as that way you can just progress along whatever branch is most relevant to you at the moment).
~J
Last edited by woodwojr (2009 January 27, 6:10 pm)
