Kanji overlap

The problem of ambiguous flash cards

A general-purpose flash-card application like Anki is a good way to study the Japanese writing system. However, if you work entirely from broad English renderings of each kanji (as clues) to individual kanji (as answers), the clues soon run together.

There are notorious problems with the kanji as a system of writing, such as misleading visual similarities. For the limited purpose of flash-card clues, one of the most pernicious problems is semantic overlap: Not visual similarity but similarity in meaning. This is when two or more characters mean very similar things, sometimes to the point that they are interchangeable, even though they may look completely different.

For example, gjiten 2.6 gives the following readings for one pair of characters:

While it is useful enough, this particular mapping between a kanji and its meaning is most useful when you have the kanji in hand and you want the meaning. It’s worse when you only see the list of meanings and you have 2000 kanji to choose from. In that usage, lists of simple meanings become dangerously ambiguous.

Continuing the example, given a clue with “method” and “system” in it, you guess 法, but the flash card says 式. You’re technically wrong, but your error is not meaningful or conducive to learning.

Such ambiguity is bad for flash-card applications, particularly when your main interest is to learn the shapes and common words. Some kanji are genuinely broad in meaning, while others have multiple, sometimes unrelated uses. In modern Japanese, the “real” meaning of an individual kanji is not important. It can be inferred from compounds, native words and ancient Chinese etymologies, to the extent that it can be determined at all, but it is not central to the language.

It can be tricky to catch that whole mess in a couple of words of another language. In many cases, you can find more distinctive, less-common English words that point to the nuances of a particular kanji, and this can make for good flash cards. However, this is effectively naming the kanji in English, not learning Japanese.

For this reason, in my own use of the All in One Kanji Deck for Anki, I edit some cards to include selected wrong answers. As a step toward improving this dubious practice, here’s a big list of duplicate glosses for all the kanji in that deck as of 2019-08-27.

15 repetitions

13 repetitions

12 repetitions

11 repetitions

10 repetitions

9 repetitions

8 repetitions

7 repetitions

6 repetitions

5 repetitions

4 repetitions

3 repetitions

2 repetitions


The list itself has problems:

I may eventually build some kind of web app or flash-card deck to improve the situation. For now, this is barely a curiosity.