Pinyin, stroke, and shape input make different tradeoffs
Aliases: pinyin input · stroke input · Wubi Cangjie
What it is
Logographic scripts such as Chinese cannot type glyphs from twenty-six keys directly; they pass through a layer of encoding. Pinyin maps characters to Latin letters or nine-key digits by pronunciation; stroke codes the order of horizontal, vertical, left-falling, dot, and hook; shape codes (Wubi, Cangjie) split characters into components. The three encodings do not pair with nine-key and full keyboards in the same way. “Larger keys versus more keys” does not cover them.
Why it happens
Pinyin’s syllable inventory is relatively closed, so initials and finals can be packed onto a nine-key pad: reuse adds homophones, but syllable boundaries can still be cut, and a lexicon can work on syllable strings. Full-keyboard pinyin is an almost letter-determined pronunciation string: homophones remain, key ambiguity drops. Stroke codes have about five symbols, so few keys are natural—a nine-key or even a five-key pad fits better than twenty-six; the cost is remembering stroke order, and encoding fails when writing habit disagrees with the standard order. Shape codes need many distinct keys for components, so a full keyboard fits; a nine-key pad would multiplex components again, stacking set-disambiguation on top of decomposition, multiplying learning cost and candidate cost. Pronunciation encoding eats accent and dialect; shape encoding eats unlearned splits; stroke encoding eats stroke order. Keyboard form should align with the encoding’s alphabet size, not with whichever looks more modern.
Studying it
Have the same users enter the same word set (including polyphones, rare characters, and names) with nine-key pinyin, full-keyboard pinyin, stroke, and one shape code. Compare time to learn, keystrokes per character, candidate paging, and failures to encode. Independent variables include encoding, keyboard form, and whether the user was trained in that encoding; dependent measures include success rate, keystroke count, and abandonment. Comparing nine-key and full keyboards only on pinyin users writes properties of pronunciation encoding down as properties of keyboard form.
Where it stops holding
People who do not know Hanyu Pinyin (some dialect speakers, professional typists trained on shape codes) have no nine-key pinyin path. Japanese kana twelve-key pads and Hangul composition on a full keyboard are different alphabet sizes; Chinese conclusions do not copy. Handwriting recognition bypasses the key plane and is not this nine-key / full-keyboard pair.
Applying it
- Default the keyboard to the encoding the user already has: pinyin users get full or nine-key pinyin, stroke users get a small stroke plane, shape-code users get a full keyboard rather than a nine-key pad.
- Allow encoding switches inside one product; do not treat a single “nine-key or full” choice as the IME choice.
- Accept all three encodings on a list that includes polyphones and rare characters; if one fails widely on rare characters, keep another encoding entry.