Home›Typing guide›Wubi

The four rules for splitting characters in Wubi

September 20, 2026 · about 814 words · Wubi

Knowing the radicals is half of Wubi. The other half is deciding how to cut a character apart, and this is where most errors actually come from — not from forgetting a radical, but from cutting in the wrong order or stopping at the wrong component.

Four rules cover the overwhelming majority of cases.

Rule one: take components in writing order

Split the character the way you write it: left to right, top to bottom, outside to inside.

• 明 = 日 + 月 → J + E
• 想 = 相 + 心 → S + H + N (相 itself splits further)

This sounds obvious and it resolves most characters immediately. The failures are cases where visual grouping tempts you to grab a shape that is not written in that order.

Rule two: prefer the larger component

When a character can be cut two ways and one way yields a bigger, recognised component, take the bigger one.

The classic example is 章. You could split it as 立 + 早, or as 音 + 十. Both are valid cuts. The rule says take the larger piece, so 音 + 十 wins: U + J + F.

A useful way to apply this: after splitting, look at what you got. If a piece is itself a common character or a well-known radical, you probably cut correctly. If both pieces are fragments, you probably did not.

Rule three: when components overlap, take the one that keeps strokes together

Some characters have components that visually interleave. The rule is to prefer the cut that keeps a component's strokes contiguous rather than one that breaks them apart.

This is the rule people find hardest, because it is genuinely ambiguous in places, and different versions of Wubi resolve some cases differently. When you hit one: look it up once, note it, and move on. There are perhaps a few dozen genuinely contentious characters, and they are learned individually.

Rule four: the recognition code for short characters

A character with fewer than four components does not fill the four-key code. To avoid an avalanche of candidates, you add a recognition code (识别码) as the last key: a single letter encoding the character's final stroke and its overall structure.

The structure part has three values:

• 1 — left-right (左右)
• 2 — top-bottom (上下)
• 3 — enclosing / mixed (杂合)

The final stroke picks the zone (1–5 as in the radical chart), and the structure picks the position within that zone. So a left-right character ending in a dot gets Y; a top-bottom character ending in a dot gets U.

Worked example: 只 is 口 + 八 → K + W, which is only two keys. Its last stroke is a dot (zone 4) and its structure is top-bottom (position 2), so the recognition code is U, giving KWU.

This is the single most useful rule for reducing candidate lists, and it is the one beginners skip — which is exactly why beginners see so many candidate windows.

Why four keys, and what happens with more

The code is at most four keys long. For a character with more than four components, the rule is:

• First component → first key
• Second component → second key
• Last component → final key

Everything in the middle is discarded. So a complex character still types in four keystrokes.

For characters with exactly four components, all four are used and nothing is dropped.

Where the rules genuinely conflict

Rule two (larger component) and rule one (writing order) occasionally disagree, and there is no derivation that settles it — the answer is specified per character. This is the honest limitation of Wubi: it is a designed system, not a logical one, and roughly fifty to a hundred common characters have an arbitrary splitting decision baked in.

The practical response is not to agonise. Look it up, type it a few times, and it becomes automatic within a day. Every long-term Wubi user has a personal list of characters they had to memorise individually; the list is short and it stops growing.

Debugging a code that will not come out

When you type a code and the character does not appear:

1. Check the split before checking the radicals. Wrong split order is the more common cause.
2. Check whether you needed a recognition code. Characters of fewer than four components usually do.
3. Check the version. If your reference says one code and your input method produces another, you may be reading 86 rules while typing 98, or the reverse.
4. Use the wildcard. Z substitutes for any unknown key. Typing ZZZZ or a partial code with Z in the unknown position will usually surface the character and show you the correct code.

That last one is the most underused feature in Wubi. The wildcard exists precisely for the learning phase, and using it is not a crutch — it is the intended way to look a code up.