Japanese Visual Mnemonics

Last updated today

An experiment using Midjourney to generate visual mnemonics for learning/remembering words in Japanese.

Layout of a bunch of visual mnemonics, even looking back at these a year later I can still remember most of the words each image is associated with

The core idea here is to create a mnemonic peg system with every sound in Japanese, making three lists: people, actions, and clothingIn retrospect I should have chosen “objects” as the third category, after coming across the PAO system (person-action-object) I discovered that this is actually an existing and prevalent technique among memory champions

. For every sound in each list I would associate with an English word that starts with similar phonetics (for example susuperman).

Then for any Japanese word, I take the first, second, and last sounds and map these to people, clothing, and actions. So the word kurumaisu becomes kungfu panda (ku), rubber boot (ru), surfing (su), ignoring any other syllables or sounds within the word. With those three words, I used GPT-3 to generate a description of visual imagery incorporating the meaning of each particular word which then I piped into Midjourney to generate a final image.

An interface I made in Google Sheets which generates (with GPT) mnemonics which I can then paste into Midjourney


Next Steps