9

I'm an advanced learner so I already know all the Joyo kanji and maybe around half of the Jinmeiyo kanji. I also think I know several hundred others beyond that. I'm wondering if anyone knows of a method which would produce an accurate estimate of how many kanji someone knows in total.

By 'know', for simplicity I just mean 'be able to read'. I understand that the concept of 'knowing' a kanji is complicated and involves knowing the meaning, readings, how to productively write it, etc. But since it is too complex to include all those aspects, I think that just going with being able to read a kanji is good enough to produce an acceptable estimate.

2
  • By "be able to read", did you mean to "know all its kunyomi, onyomi, and sometimes irregular reading", or just the common one?
    – Andrew T.
    Apr 29, 2019 at 13:37
  • I suppose I mean 'be able to accurately read the character when it is contained in a word'. I'm not aiming for extreme detail, just a fairly good approximate estimate.
    – kandyman
    Apr 29, 2019 at 13:40

4 Answers 4

Reset to default
1

If you want an accurate method to estimate the number of kanji you know, the best way I can think of is to take the list of the kanji that included in the lists of the kanji kentei (https://kanji.jitenon.jp/cat/kyu.html). That's already a pretty inclusive list but https://kanji.jitenon.jp has around 12k kanji indexed as of now so you can use that augmented list.

The test is actually very simple.

  1. Level 1: For each kanji check whether you at least know (and can recall on the fly) one of word containing the kanji.

  2. Level 2: For each kanji check whether you can remember most of its readings and at least one word for each reading.

  3. Level 3: Recall all jouyou reading for each kanji as well as word using that reading. That's already pretty hard which is partially due to the fact that the jouyou list includes some weird kanji that you would not have suspected to be common.

  4. Level 4: Recall all jouyou reading and most¹ non jouyou reading for each kanji as well as word using that reading.

The 1st level is the least requirement in order to know a kanji, the 3nd level is already a pretty strong knowledge how a kanji and the 4th is very strong knowledge of a kanji. This test does not take into account whether you can write a kanji but you excluded this criterion.

Assuming that you know around 3k kanji, and that you need in average 5s to associate a word to a kanji and that you need 2s to tell that you don't know a kanji, you would need around 6 hours to complete this test. That's quite long but at least it is accurate.

¹: Some non jouyou reading are really useless and don't even have an actual word still attached to it.

3
  • 1
    When you say "for each kanji", you mean check 12k kanji? How did you calculate that it would take 6 hours? By my calculations that would take around 24 hours.
    – kandyman
    May 1, 2019 at 0:49
  • I meant for 6k kanji, the calculation is as follows: 5sec * 3000 + 2sec * 3000 = 21000 sec = 5.8 hours
    – 永劫回帰
    May 1, 2019 at 0:59
  • 1
    Do you think taking a smaller sample of these at random would ruin the accuracy of the test? Instead of 6 hours maybe 2 hours instead, ie a random sample of one third of the characters
    – kandyman
    May 6, 2019 at 16:54
3

This site helps you estimate how many kanji you know. It's probably not advanced enough for you though.

https://www.mlcjapanese.co.jp/level_check_kanji.html

2
  • That's a pretty good quiz and I often recommend it to intermediate learners. But the fact that is caps at 2000 is bit dissatisfying.
    – 永劫回帰
    May 7, 2019 at 19:49
  • Yes it's a good quiz, but I need something that goes beyond Joyo Kanji too. Thanks for the suggestion, though.
    – kandyman
    Aug 17, 2020 at 16:32
2

Ideally to get a perfectly accurate answer, you would review a comprehensive list of kanji, but of course that would be quite time-consuming. I suggest using statistics to speed up the process: Test yourself on a representative sample of the full set of kanji you want to estimate from, and extrapolate from that to get an estimate.

For example, what I have done in the past to get a similar estimate of the number of English, German or Japanese words I know is to start with a dictionary and randomly select a set of pages (more pages take longer, fewer pages make your estimate less accurate). Quiz yourself on every entry on those pages to get an average number of correct responses per page. Multiply that by the number of pages in the dictionary, and your result is a reasonable estimation of how many correct responses you should expect you'd get if you tested yourself on the whole dictionary.

You will need to be careful that your selection is actually reasonably representative. Depending on how the dictionary is organized this could be hard - for example if it's organized by grade level then entire pages are going to have a similar difficulty level.

There is a random kanji generator here: http://kanji.fm4dd.com/kanji-random.php that would be useful for doing exactly this sort of analysis. Unfortunately it seems to top out at the JLPT N1 list in terms of difficulty, so while it would be useful for less advanced users, it's not going to be useful for a reader at the skill level indicated in the question.

6
  • 1
    Actually, I don't how the random quiz generator picks kanji but if it uses the JLPT lists from JMDict, then the list of kanji included in JLPT N1 is a bit crazy. Indeed, the way the lists were collected means that some non jouyou kanji have been collected: that includes jinmeiyou kanji but also very rare kanji that would appear only in the two highest level of the kanji kentei. So that may be a good source even for advanced levels.
    – 永劫回帰
    Apr 30, 2019 at 21:41
  • 1
    I played five minutes with the kanji generator and it has some errors (kanji.fm4dd.com/…) 柿 cannot be read こけら, even though the kanji for こけら looks damn the same. The same goes for劃 (kanji.fm4dd.com/…) which is not a jouyou nor included for real in the N1. However, their database does not look like to be exactly the same as the one used by JMDict.
    – 永劫回帰
    Apr 30, 2019 at 21:48
  • 1
    You're right! There are actually 4590 entries in the set labeled N1, so it's got to be at least half non-jouyou kanji.
    – AmericanUmlaut
    Apr 30, 2019 at 21:48
  • I know at least one word where 柿 is read as こけら, namely 杮落とし (こけらおとし).
    – kandyman
    Apr 30, 2019 at 22:02
  • 1
    @kandyman Look closely at the very subtle difference between 柿 (かき) and 杮（こけら）... mamechishiki.aquaorbis.net/wp-content/uploads/2015/10/…
    – 永劫回帰
    Apr 30, 2019 at 22:14
  • I see. Midori lists them as both viable choices for こけら落とし but I can see there is a difference. Is it possible that they are variants?
    – kandyman
    May 1, 2019 at 0:43
0

Do not estimate.

Keep track of the vocabulary words you learn and then do a database search for the different Kanji used in the words you know.

You can do that for example if you learn Kanji alongside with vocabulary and store those words in Anki or such.

4
  • The point of the question was to find a practical method that doesn't involve the painstaking process of documenting each individual instance of kanji. My vocabulary database has over 50,000 entries and I'm not going through it one by one.
    – kandyman
    Aug 14, 2020 at 13:17
  • @kandyman Still, I praise that you keep a database with entries for your vocabulary. 50000 entries is a very impressive figure. My point is that the more systematic you are when studying, the better
    – jarmanso7
    Aug 14, 2020 at 14:18
  • @kandyman Which database are you using? Maybe I can help you with a query. So it doesn't necessarily mean that you go through each entry one by one, but you get a machine to do the tedious job for you.
    – jarmanso7
    Aug 14, 2020 at 14:20
  • I use Anki - it's possible to export in CSV format.
    – kandyman
    Aug 17, 2020 at 16:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .