On WaniKani

Ideas

This is a post in its idea generation phase. I'm collecting notes for what to write about. Proceed at will.

Good

  • Visual design: colors, URLs
  • API design: that extensibility is possible
  • Quirkiness
  • Teaching you just a couple readings
    • Contrast w/ vocab on Anki deck
  • The split b/w radical, kanji, vocab (and that it’s not just kanji).
    • It’s actually funny ppl focus so much on learning kanji; now that i’m doing it later, i’m way more interested in learning vocab
  • Coming up with mnemonics is actually a ton of work. Try doing it on your own for vocab when starting. Will make you appreciate WK doing it for you, when some would otherwise seem half-baked.

What I’ve learned

  • You need to be learning Japanese separately, or else it won’t just be unmotivating, but it won’t really make sense
    • Seeing things actually used turns this from super artificial to helpful
    • I forget exactly where I saw, but I saw implied you can “frontload” WK to speak and then learn Japanese. I don’t think this would work for most people sustainably. (Lang takes a long time.)
      • Oh, they have it on one of Tofugu’s pages about starting to learn Japanese! They reference it in a pull quote, and I think they even explicitly say to do it in their guide.
  • Mental model for radicals and kanji being abstract
  • Mental model for why so many readings: think about assigning characters to words after the fact

Bad

  • the frustrating thing about constantly referring to onyomi and kunyomi isn’t that we have to do it, but it’s that he’s constantly using information that we didn’t learn, so you’re constantly a bit confused. it’s not like “oh we have to use K for X and O for Y,” it’s like “i didn’t know (and won’t remember, because it’s not in a mnemonic and not something we’re trying to learn) which are K and O, so your explanations aren’t referencing something i really know.” but these explanations are like, for every single vocab.

    • Like “The reading for this word is all kun’yomi despite being jukugo. Just think of sketchy things going on in the downtown, just like how there’s sketchy kun’yomi readings for this word. You should know both the kun’yomi readings from the previous level, though, so if you do you know the reading for this word as well.” (for 下町)
  • You can tell global inconsistencies from, e.g., teaching 何月 and 何日 where he acknowledges the multiple readings, vs 何年 where he’s like “you should know this”

  • Don’t differentiate between similar vocab (probably have this below)

    • Production different — mnemonics don’t help (can tell using kaniwani). mnemonics are only for kanji part
    • Some explicit comparisons would be helpful
  • Extra lessons re, e.g., patterns about intransitive vs transitive

    • Can have some bonus distinguishing types
  • Example sentences are almost always not useful

    • It’s cute that they’re silly, but I’ve found only maybe two out of hundreds that I can read.
    • This means that you don’t get any context for the vocabulary. Seeing / hearing vocabulary in context is huge. I think this may be the weakest thing about WK—you learn isolated sound bytes and words, but you don’t hear them in the Japanese language. This means you have an entirely separate phase of learning where you need to see / hear words in actual context. That phase could be introduced in WK if the example sentences were more useful.
    • This would be completely acceptable if they actually read the sentences out loud, but they don’t
  • High level takeaway is that it seems that the people who made it didn’t actually need to use it to learn Japanese, because some of the confusing things happen very early on

  • First many things you learn are OY, then don’t use. Frustrating. Esp. frustrating because…

  • Not teaching you whether learning KY vs OY

  • Mnemonics are not helpful for KY vs OY

  • Don’t take advantage of sequential learning or patterns

    • e.g., numbers, months
  • Can’t adjust SRS timing

    • I want to have to repeat more. My memory isn’t as good as it used to be.
    • Most motivated people seem treat it like a game and cram through as much as fast as possible. While I felt that way on previous (failed) attempts, I don’t at all this time. (It’s also funny to treat it like a game that’s “gating your progress” because it’s just a nice flash card app with really good cards you whose intervals you can’t customize)
  • seems like random number of synonyms (and sometimes alternate spellings?)

    • Wish they’d separate what to memorize from what they accept
  • Don’t distinguish things that seem to have the same meaning

    • 女の子、女子、
    • 文字、字
    • There are good cases of this, e.g., description of 学ぶ There is another word that means “to learn” that is going to be more common, which is 習う. You’ll learn this one later on. 学ぶ sounds a little more serious compared to 習う, so keep that in mind when you use it.
  • Doesn’t consistently distinguish mnemonics for the same meaning. E.g., first 死 (death), then 亡 (death). The second mnemonic is a generic “death”-related wacky thing.

  • The whole avoiding kunyomi thing almost works, but doesn’t — you still benefit greatly from learning patterns, but have to do it without their structure

  • Same mnemonics with different pronunciations

  • Be more data driven

    • See when people actually forget
    • e.g., lvl 60 guy that says his secret was to always review after learning. i also find i need a refresher before my review is available and i can do it. this means the forgetting is not calibrated.
  • Don’t pronounce kanji

    • this might not be possible. but it might.
  • Don’t pronounce word use examples or context sentences

    • As a beginner, almost always words or Kanji I’m not sure about, and having in-context readings would really help how to say them.
  • Radicals are clearly composed of others, but they don’t show it that way

  • No reverse (could join with next point)

    • Fortunately, kaniwani exists
  • Should build (or generate) quizzes vs similar / commonly confused radicals. It’s easy to guess which chapter something is from so you don’t get much overlap of confusing things in order to practice distinguishing them. This is esp. important because they get morphed later on.

  • Mnemonics can be half-baked

    • It’s hard to come up with good ones, I found making them myself
    • but still you’re paying for them
    • especially, which part of the word is used? e.g., “ke” or “keta” for “ketal”
  • A handful of useful, obvious features are missing

    • A free frontend app someone made even has them!
    • You have to install userscripts which clash and break
    • My main ones are:
      • pitch info
      • katakana (they say it’s obscure and for dictionaries; I’ve seen it commonly elsewhere, and really need katakana practice)
      • meaning and reading together (note: they quote research to support not doing this, but the research does not support this actual point. plus, it’s a setting as a learning one should get to choose)
  • adding this but might be part of another: not helpful for remembering which reading goes with which. example: 小皿 vocab, alt kunyomi for 小 (ko), and the mnemonic is totally unrelated to this word

  • even when there’s two ones we learned, not helping distinguish. e.g., vocab 不正, we learn しょう first for 正, and no explanation when the other is used.

  • nothing helping distinguishing radicals that are similar conceptually

    • 金 gold vs 宝 treasure
    • sometimes they do w/ e.g. run vs foot radicals
  • some radicals seem clearly wrong

Semi-unrelated

  • EtoEto just radio silence dropped, not even online any more, which was part of my incentive to buy WK. Loved the real lang + breakdowns of various depths w/ hotkeys. Wish they just kept what they had online.