Small kana: yōon

Add a small ゃ ゅ ょ after any i-column kana → combined syllable.

しゃしん (photo), きょう (today), りょこう (trip)

Small kana: sokuon

A small っ (like つ, shrunk) doubles the next consonant.

In speech: a brief glottal stop. In writing: a tiny っ that changes the word entirely.

Voicing marks: dakuten ゛

Add ゛ to flip a consonant from unvoiced to voiced. Applies across whole rows.

Rule extends across all 5 vowels: か き く け こ → が ぎ ぐ げ ご

H-row's special twin: handakuten ゜

A small ゜ circle (only valid on the h-row) gives the unvoiced plosive p.

Only row with three forms. The h-row is phonetically unstable, so it picked up an extra diacritic over the centuries.

Or: skip kanji entirely

Pokémon FireRed (Game Boy Advance, 2004). All UI in hiragana & katakana, no kanji.

"あなたの なまえは？". Note the space between あなたの and なまえは, and the entire keyboard is just kana.

What in-game dialogue looked like

→ with kanji: 新人トレーナーの 君は ポケモンと 冒険を 始める！

"You, a rookie trainer, set off on an adventure with Pokémon!"

Why drop kanji?

Each kanji is a 16×16 bitmap → 32 bytes.
A full Jōyō set: 2,136 × 32 ≈ 68 KB of pure glyph data.

For comparison: Pokémon Red & Blue (Game Boy, 1996) shipped the entire game in 1 MB.
156 Pokémon, all sprites, all maps, all music, all code.

A full kanji set would have been ~7% of that cartridge, just for the letters.

Easier: ship the 92 kana glyphs, skip kanji altogether.

But then: readability tanks

こうこうにいきます

Could be: 高校 (high school) / 孝行 (filial piety) / 後攻 (going second) / 高々 (at most)…

Solution Pokémon adopted: add spaces between phrases, a thing Japanese normally doesn't have.

Hardware caught up

The same problem the web solved with subsetting, games solved by waiting for ROM to get cheap.

No spaces. Anywhere.

おすしを食べました

Where does each word start?

お (honorific) + すし (sushi) + を (object marker) + 食べ (eat, verb stem) + ました (polite past)

Need a dictionary and grammar rules to do this.

Compounds make it worse

電車 = 電 (electricity) + 車 (vehicle) = train

図書館 = 図 + 書 + 館 = library

A naïve Ctrl+F for 車 finds 電車, 自動車, 車道, 自転車… you almost never want that.

…but compounds can be poetry

Each character carries meaning. Stack them and you get a tiny picture.

Counters: count by shape

Japanese has no grammatical plural. But it has ~50 counter words that change based on the shape of the thing being counted.

So localization has to know which counter goes with which noun. (We'll come back to this.)

What i18n libraries can't do for you

English needs plural forms. Japanese needs something CLDR doesn't ship.

new Intl.PluralRules('en').select(1)  // "one"
new Intl.PluralRules('en').select(2)  // "other"

new Intl.PluralRules('ja').select(1)  // "other"
new Intl.PluralRules('ja').select(2)  // "other"
						

Japanese has exactly one plural category. The library has nothing to do.

What apps actually do: hard-code per noun, or punt to 個 (ko), the generic "thing" counter.

So how do learners read?

A browser extension that tokenizes + dictionary-looks-up under your cursor.

Yomitan

Yomitan in one sentence

Hover any Japanese text →
it segments words, deinflects them, and shows the dictionary entry.

• 食べました → 食べる (deinflected to dictionary form)
• Shows readings, meanings, frequency, pitch accent
• Bring your own dictionaries

Bring your own dictionaries

Yomitan ships empty. You import whichever dictionary files suit how you read.

Longest match wins

Hover anywhere. Yomitan reads forward from your cursor and picks the longest entry it can find.

So hovering 図 in this sentence pops up library, not "diagram", even though both are valid.

Live demo

→ NHK Easy: ne2026052112586 ↗

Open the link, hover any sentence.

今日は寿司を食べました。

And how do you type all this?

Reading was one side. The other side is input: the IME (Input Method Editor). Built into every desktop, phone, and browser.

Three layers of conversion between the keyboard and what lands in the document.

Picking from candidates

Press space again. The IME shows every kanji string the kana could possibly spell.

Arrow keys or number keys to pick. The IME learns your preferences and reorders next time.

Furigana: kanji with training wheels

Tiny kana printed above (or beside) kanji to show how to read them.

私わたしは学がっ校こうに行いきます。

Where you see it: children's books, manga, learners' material, place & person names with rare readings.

HTML ships with <ruby>

The web has a native element for exactly this, for over 20 years, in every browser.

<ruby>
  漢字
  <rp>(</rp><rt>かんじ</rt><rp>)</rp>
</ruby>
						

Renders as: 漢かん字じ

• <rt>: the small "ruby text" annotation
• <rp>: fallback parentheses for browsers without ruby support (modern browsers ignore them)

Also handy for: chemical formulas (H2twoO), abbreviation expansions, screen-reader-friendly pronunciation hints.

Reading text from images

OCR: Optical Character Recognition. Pixels in, text out.

You use it constantly:

• Google Lens, Apple Live Text
• Searchable PDFs, document scanners
• Receipts, business cards, license plates
• Translating signs and menus on the go

For English: largely solved. For Japanese: still rough.

OCR: harder than it looks

26 letters vs ~6,000 glyphs

Some kanji differ by a single stroke.

Spot the difference

At low resolution? Good luck.

And then there's manga

→ specialised tools like mokuro, manga-ocr.