Typing Japanese: The IME

Japanese uses three scripts and thousands of kanji, yet it is typed on an ordinary QWERTY keyboard. The trick is the IME — the Input Method Editor (日本語入力, nihongo nyūryoku). You type the sounds of a word in romaji; the IME turns them into kana as you go; then you press the space bar to convert (変換, henkan) the kana into kanji, choosing from a list of candidates. Once you understand this pipeline, two things fall into place: the practical keystrokes for tricky characters, and the reason the keyboard makes you spell certain words by their kana, not their sound.

The pipeline: romaji → kana → kanji

Typing Japanese is a three-stage flow, and each stage has its own key.

1. Type romaji. As you type, the IME shows hiragana with an underline, meaning it is still "composing." Type nihongo and にほんご appears.
2. Convert with the space bar. Press space and the IME offers kanji candidates: 日本語, 二本後, and so on. Press space again to cycle through the list.
3. Confirm. Press Enter (or just keep typing) to lock in your choice — this is 確定 (kakutei, "settling"). The underline disappears.

To produce 日本語 above, you type nihongo, watch にほんご form, hit space, and select 日本語 from the candidate list. The whole word is entered by its reading and converted — you never look up the kanji by shape.

Katakana and the long-vowel bar

Loanwords take katakana, and there are two easy ways to get there. Type the reading and press space until the katakana candidate appears, or press F7, which instantly converts the current composing string to full-width katakana. The long-vowel bar ー is typed with the ordinary hyphen/minus key (-).

To produce コーヒー, type ko-hi- — the two hyphens become the two ー bars — giving こーひー, then press F7 (or space) to get コーヒー. The mapping is literal: every hyphen you type is one ー. This is the first hint of the deeper rule below — the keyboard cares about the written form, so a bar is a - and nothing else.

Small っ and doubled consonants

You never type the small っ directly for ordinary words. Instead you double the consonant, exactly as Hepburn romaji writes it: kitte → きって, gakkou → がっこう, matte → まって.

切手 is entered kitte, 学校 is gakkou. To type a standalone small っ (rare — e.g. a trailing っ showing a cut-off "！"), type xtu or ltu. The same x/l prefix forces any small kana: xya → ゃ, xa → ぁ.

The moraic ん: type it "nn"

The kana ん causes more typos than any other character. Before a consonant it usually resolves on its own — shinbun → しんぶん, because the b tells the IME the n must be ん. But before a vowel or や/ゆ/よ, a single n merges with the next sound: type honya and you get ほにゃ (ho-nya), not 本屋. The fix is to type nn for a guaranteed ん, or add an apostrophe: honnya (or hon'ya) → ほんや → 本屋.

禁煙 must be typed kinnen (or kin'en); typing kinen gives きねん (記念, "commemoration") — a different word. The moraic ん and its apostrophe are explained in full on the moraic ん page; the typing rule ("type nn") is the practical takeaway.

The particles you spell by their kana, not their sound

Here is the rule that surprises every learner. Three grammatical particles are pronounced one way but written with a different kana, and the IME follows the writing. So you must type their historical spelling:

The topic は in 私は is typed ha — type wa and you get わ, the wrong kana. The direction へ is typed he, and the object を is typed wo. This is not the IME being awkward: it preserves the very kana-spelling rules the reader must know. Because を is written を, it is typed wo; because ー is written ー, it is typed -. The keyboard is a mirror of the orthography, not the pronunciation. (The particle-spelling rules themselves are covered on は, へ, を as particles.)

づ and ぢ: type "du" and "di"

Because じ/ぢ and ず/づ sound alike, the IME can't guess which kana you want from the sound — so it borrows the Nihon-shiki spellings (see romanization systems). To reach づ you type du; ず is zu. To reach ぢ you type di; じ is ji or zi.

三日月 (mikazuki) contains づ, so it is typed mikaduki; typing mikazuki gives みかずき, which will not convert to the crescent moon. This is the one place everyday typing quietly uses a non-Hepburn spelling.

The step you can't automate: choosing the right kanji

The IME converts by reading, and many readings map to several kanji. It will offer a ranked list, but the machine cannot know which meaning you intend — you must confirm the right one. This is the orthographic responsibility the IME hands back to you.

Both 着る ("wear") and 切る ("cut") are read kiru, and the IME lists both for the same keystrokes kiru. Pick 着る in the first sentence and 切る in the second — swap them and the sentence becomes nonsense. The same trap waits in はし (橋 "bridge" / 箸 "chopsticks" / 端 "edge") and かえる (帰る "return" / 変える "change" / 蛙 "frog"). Modern IMEs use context to rank candidates well, but the final choice, and the responsibility for it, is always yours.

Common mistakes

Nearly every one of these comes from typing a word by its sound when the IME wants its kana — wa for は, o for を, zu for づ — or from accepting the first kanji candidate without checking the meaning. Type the spelling, and read the candidate list.

Key takeaways

• The IME flow is romaji → kana → kanji: type the reading, press space to convert, Enter to confirm.
• Small っ is a doubled consonant (kitte, gakkou); ー is the hyphen key; F7 forces katakana.
• Type ん as nn before vowels/や-ゆ-よ and at word end (honnya, kinnen).
• Type particles by their kana: は = ha, へ = he, を = wo; and づ = du, ぢ = di.
• The IME can't read your mind — you choose the correct kanji among homophone candidates (着る vs 切る).