My system for japanese kanji

[Hari-P (Hari)]

This one is for @Jint3x who mentioned spending some time on thinking about applying mnemonics to kanjis.

I’ve been working on japanese again in the last few weeks.

While I’m concentrating mostly on listening and speaking for now, my semi-failed attempts to learn kanjis in the past are a bit of a sore point, so I wanted to « solve » it.

Dividing the kanjis into radicals that you only have to combine (heisig/wanikani) is a great idea but I still had some problems.

1. I wouldn’t remember the order in which the radicals are combined. For example in 明るい I know that it’s sun+moon, but is it « sun-left, moon-right » or « sun-right, moon-left » ?

2. The sentences used to remember the radicals to combine are often way too abstract for me, for example « A king in the countryside uses logic to rule ».
   (But then I tend to do badly with mnemonics I didn’t invent myself anyway).

So I came up with the following system (vaguely inspired by PAO) :

In a kanji the main radical is always on the right and/or bottom.
The main radical is a person.
Examples :

• For water 水, it’s a sailor (capitaine Haddock).
• For the sun 日 it’s Glorfindel.
• For ten十 it’s Kabu (the cabbage-head-scarecrow in Howl’s moving castle).
• For 言 it’s a roman orator.
  Etc.

A radical on the left is an object (tool) characteristic of the person.

• For water/Haddock on the left 氵, it’s a fishing rod.
• For the sun/Glorfindel on the left, it’s a string of bells and jewels (the headstall of Asfaloth, Glorfindel’s horse).
• For ten/Kabu on the left, it’s Sophie’s cane (which Kabu gave to her just before she entered the moving castle).
• For speak/orator on the left, it’s a roll of parchment (where the orator wrote his speech)
  Etc.

The radical on top is another object (most often related to the head) characteristic of the person.

• For water, it’s capitaine Haddock’s hat.
• For the sun it’s burnt hair (think a head emitting smoke), because Glorfindel died the first time when a balrog/fire demon seized him by the hair (you’d think he would bind his hair in his second life but nope he still runs around with floating hair in lotr).
• For ten, it’s Kabu’s top hat (which looks very elegant, but the look is rather spoilt because it’s smelling like cabbage).
• I think 言 never comes at a top radical, but I could use a crown of laurels…

Combined in practice :
計 (to measure, to plot) : it’s Kabu holding a scroll. He’s bringing Howl the scroll with all the PLANS and MEASUREMENTS for the new moving castle. It’s not only a sentence anymore, but in my mind’s eye a scene that plays out, I can elaborate and add details if I need to.

For bright 明 it’s Séléné (from de cape et de crocs) wearing Asfaloth’s bridle like a cross-body bag. I know that sun in on the left and moon on the right, because Glorfindel isn’t the subject, Séléné is.

I began with choosing and drawing people and their objects for the 50 most frequent radicals.

Then I add more on an as-needed basis when I encounter a kanji I didn’t remember and for which I haven’t defined the person-radical yet.

I also add variants. For example, I consider white 白 a variant of sun 日. So it’s Glorfindel again. This time it’s not beaten-up-by-the-balrog-and-lost-his-horse Glorfindel, but Glorfindel the Magnificent : he’s riding Asfaloth (WHITE horse). Left is Asfaloth on his own. Top is long hair floating in the wind staying magically untangled (must be a elvish power or something).

Which gives us for 泉 spring (like in onsen, hot spring), capitaine Haddock with super long hair (which stays magically healthy despite salty ocean water) relaxing in an onsen.

I don’t usually bother with on-yomi and kun-yomi, because I feel it’s easier to simply learn words. I already know the words mizu / water, suiyoubi / wednesday and onsen / hotspring. No need to “clutter” my kanji imagery 水 with random syllables (zu, sui, sen). It’s a question of vocabulary, not writing/kanji and I feel better learnt by immersion.

[Lukelustre (Luke)]

This is really fascinating stuff, amazing job.

I’ve tried to theorycraft how I’d encode everything relevant to a kanji into a system like this, and often find that the on’yomi and kun’yomi is where it breaks.

While I can’t say with certainty as I’m still a beginner, I think you’re right in your approach; nonetheless I’d like to come up with a way to ingrain the pronunciation(s) at some point, but how you do that is beyond me.

When I have more free time I’ll 100% be utilising this system in my learning.

[AutumnBreeze]

Wow, I really love the system you came up with! I don’t know Japanese at all – but from reading your description the approach sounds super solid for the purpose you had in mind

Especially I like how you don’t try to turn EVERYTHING into an image, but use “regular” language learning and mnemonics together to sort of complement each other.

I wish you all the best with your learning And I’d for sure be interested to read any updates you have on this

[Jint3x]

Interesting topic!

I haven’t done much mnemonic research on languages, I do have a couple of ideas, but they are not really fast enough to work in real-time, when having to speak with people.

Something I’m trying now is to visualize the kanji in their actual forms rather than apply mapping like PAO. I am hoping to be able to “connect” words [meaning] with their kanji, so the meaning can very quickly pop up in my head instead of having to decode it with a technique.

Mnemonic techniques for languages definitely require focus on speed or very good ordering rather than accuracy due to the real-time requirements. At least these are some of my initial thoughts.

[Hari-P (Hari)]

I continue to apply the system to kanjis (going through my long-abandoned anki deck).
It works well: I no longer struggle to find the meaning of kanjis like 防 妨 坊 彷 肪 or mix them up.

I changed a few of the objects I had defined at the beginning, because I had way too many hats at the top and too many tools/weapons on the left.

I tried to “personalize” the people as much as possible. For example, 土 is not a generic farmer anymore, but Mr. Stock (from enchanted glass). Because if I already have a “backstory” in my head, I know how the people are going to react in such and such situation and the mnemonics stick so much better.

I added the rule that when there are three radicals “in a row” (or “stacked”) the “main” radical, so the “person” is in the middle.
For example, 舎 (hut) : Mr. Stock 土 (person) comes back home to find a box on the floor 口 (bottom object), with Gandalf’s hat 𠆢 (top object) on it (knowing Gandalf it’s probably fireworks and the farmer isn’t happy because he fears his hut is going to explode…)

I am however “cheating” sometimes when an image comes immediately to mind and makes my life easier.

• For example, 朝 (morning) would be Selene + Asfaloth bridle + Sophie’s cane twice according to my system, but it’s way too complicated and “Selene (the moon goddess) going away on her chariot in the morning” is way too good an image to pass up. So I’m keeping it even if it’s not exactly 車 on the left side.
• Same for 亭 (restaurant), what I see immediately is a table with a head on it. So I don’t bother with my system because “a restaurant serving a human head as a dish” is a perfectly good image. It also works well for 停 (stop / person+restaurant), because people are, of course, abruptly stopping in front of the table when they see that.

I also decided to not bother using my system for kanjis that I already know well. So I have “capitaine Haddock” for 水 (left: fishing rod, top: captain’s hat , bottom: small boat) since it’s used as a radical in a lot of other kanjis, but I did not bother with anything for 氷 , 永, and 泳, because they are already obvious to me (and the “variations” seem logical).

Because, after all, the idea is to make my life easier and learn quicker. Not to spend extra time using a system which is only a tool, not an end.

[Hari-P (Hari)]

I continue to learn with my system a few kanjis everyday (I didn’t post daily updates because it wouldn’t be very interesting).

It’s going well. More and more often I don’t use the mnemonics anymore and recognize the kanji instantly. But if not the mnemonics always bring back the meaning quickly.

I made several observations which may be of interest for other learners.

Using my “kanji-people” to feed (back) my PAO system

With time and using them the people and objects naturally « evolved ». For exemple :

• The fishing rod acquired a fish, and is now sometimes replaced by a fish.
• The mover (person for 口), who was in the beginning only a very muscular person transporting boxes, has become a sort of super-handyman able to solve any problem.
• My doctor (person for 心) is very depressed, even suicidal sometimes, has a gambling problem (poker) but can’t stop working to become a flutist like he’s dreaming of because he needs money to put his son through college. Needless to say he’s rather neglectful and doesn’t do his job very well, quite often it’s the mover-super-handyman (口) that saves the day. But he stays a fundamentally honest man, who refused to steal 100 millions yens when he had the chance.

In fact, some of them acquired so much personality (or biography) that I feel I know them better than some of the random celebrities I had to take from the internet to fill up my PAO (because I honestly don’t understand why people care about celebrities and follow their life… so I just don’t know so many).

Now, that’s an idea: since I already have a couple of “imaginary friends” from childhood in my PAO (people created for the comics we used to write), why some of these “kanji-radical-people”? (Plus I’d be able to choose their names specifically to fill in any gaps hmmm…)

Isn’t it basically the Heisig/wanikani system?

What I’m doing is nothing new. Yes, it’s basically Heisig/wanikani, but with a difference: I make EVERYTHING concrete (I have to remember myself to be strict with it for the system to work well). For example:

• 音(sound) is Glorfindel with a water bottle. He’s making sound/music by blowing on it (like a pan flute). So when 音 comes up as a radical in other kanjis, instead of using « sound » (which is abstract) I’m using a « pan flute ».
• 九 (nine). It’s easy to talk about 軌 as a “track formed by nine cars going through a field”. But why not ten cars? Or seven? Wouldn’t seven cars already have formed a track? If the exact number makes no difference to the scene, it’s an unclear mnemonic to me. I’d rather have a Nazgul (“one of the nine”), to represent 9. There’s no way I’ll mix him up with somebody else.

I’m now convinced that learning on/kun-yomi with the kanji is an utter waste of time (and possibly one of the reasons I “failed” in the past)

I went back and forth for a while trying to include on/kun-yomi. At first I included them when a mnemonic came spontaneously to mind that 1. fits well with the scene and 2. does not complicate/lengthen the scene.

But I now understand (which I did not initially, when I made my kanji deck a few years ago), that, if you want well-made anki cards, you should include only ONE answer for each card → so for one note you would need, 1 card with the meaning (and mnemonics), + 1 card with on-yomi (and mnemonics), +1 card with kun-yomi (and mnemonics). The situation is even worse when you get several possible kun-yomi!

I think it may be because I didn’t understand this principle that I gave up the first time I tried to learn kanji: I was trying to review cards with way too much information, which asked for TONS of answers, instead of focusing on one essential thing.

So it would mean multiplying the number of cards (and thus of reviews in anki) to learn an information… that may or may not be very useful! Because you’ll never know if the vocabulary words you learn later are an exception to the rule (have a different kun/on-yomi than the one you learnt).

You may as well learn vocab directly and let yourself notice patterns naturally (“another word with this kanji was pronounced SHOU so it’s probably also the case for this word using the same kanji? Oh yes it is!”). And you’ll remember it even better because you had to think about it.

So what I do now, is to look up the kanji in the dictionary, to see which words are formed with it, to get a more precise idea of its meaning. Because 1 kanji = 1 translation is a terrible imprecise thing. Even if I only keep 1 keyword per kanji, I need to “get a feel” for what this kanji is used for “in real life” (by this I mean excluding obsolete and rare words).

I then put these vocabulary words on the card, since I looked them up, but they are not part of the answer: I do not have to remember them, I’m just reading them after having given my answer and checked that it’s correct. Sometimes this exposure is enough to learn the word (when for example it’s just a matter of noticing “oh! so sabishii is written with THIS kanji”)

I fact, I wonder if I should have only learnt kanas as a beginner and waited a bit more to learn kanji? Kanjis are so much easier when learning them only means “put an image on words I already know”.

[z6am (Zunaid)]

Very helpful update because I knew almost none of the references you used (Haddock, Howl, Kabu, etc.), but I like the “kanji people” concept which I can readily generate myself. As well because the concepts of making everything concrete and using one reading per card should help to improve the method dramatically for me from the start. Thanks!
Curious to know your progress a half year later.

[Hari-P (Hari)]

I’m still using the system, with no changes compared to last time. It’s one of my projects I consider a success, because using it feels now natural to me.
I don’t need to plan anything anymore, I just see a new kanji, name its components, tell myself the story that results in the meaning and that’s it. Often I don’t even take the time to write it down, because things seem so clear.
If there are new components/radicals I choose Person+Objects on the fly. Now that I get the ball rolling I rarely need to agonize about my choices it feels easy.

[KofiSazdin (Kofi Asante Sadzin)]

This is really interesting. When I was learning kanji (lived in Osaka for 7 years), I started with the Heisig method, which attached a meaning to the radicals of the kanji. From there, one was to create their own associations based on the radical meanings, usually as a short sentence (story). As I got more used to Japanese, I ended up using vocabulary to study kanji. This basically consisted of me using the meaning for each kanji (which are easily searchable) in a word and then creating a story that gave me the sound and the meaning. On top of that, I was reading A LOT of manga and lite novels. It got to the point where I could decode words without knowing their pronunciation.