@CB:
I just examined the file from your not-yet-released program, and I have the following suggestion:
At character Heisig X, it is far too restrictive to try to generate words only composed of characters 1 to X; the Heisig order is too different from the natural character frequency order for that. Hadamitzky and Spahn do restrict their words to 1 - X in their book "Kanji & Kana", but they can do so only because they present the kanji in a frequency-based order. Naturally, then you allow for that condition to be broken, as for instance in line 18 of your output, at which none of the kanji with 冒 are to be found before it. That is discouraging to a learner and defeats the purpose of the list.
Have instead the user specify a range 1 - R among which words including X < R could use any kanji from 1 to R, not only from 1 to X.
This would allow people to do Heisig in steps, namely alternating between learning kanji and learning real Japanese words. For instance, a hypothetical learner could push up to Heisig 1000, then take a break from kanji learning and use your tool to generate words to reinforce all characters previously studied. Obviously R would need to be large enough to allow for generating a nice word repertoire.
Then that hypothetical learner would push for 500 more, re-use your tool this time with R = 1500, etc. This learning in steps would be better, and perhaps would prevent the phenomenon of people doing Heisig once, forgetting it all, and having to re-do it again.
In the event where no viable combination occurs with elements within 1 - R, then you just output nothing.
御苦労様。
Last edited by louischa (2013 June 24, 11:37 pm)
