Skip to content

Wang Yongmin and the Chinese Input Wars

Abstract

The QWERTY keyboard has 26 letter keys; written Chinese has tens of thousands of characters. In the early 1980s this mismatch was widely described (inside and outside China) as an existential threat: either the Chinese language would keep China out of the computer age, or computing would force China to abandon its script. The engineer Wang Yongmin answered in 1983 with Wubi, an input method that decomposes every character into a sequence of at most four component shapes, typed on an unmodified Western keyboard. Wubi let trained typists enter Chinese at speeds comparable to English typing and helped carry a 3,000-year-old script into the digital age, before being overtaken, in one of computing’s quieter dead ends, by the statistically-smarter pinyin methods that power a billion phones today.

The Problem: A Script vs. an Industry Standard

Chinese had collided with Western information technology before. Telegraphy forced Chinese into four-digit number codes memorized by clerks; Lin Yutang’s MingKwai typewriter (1947), which selected characters by combining shape components, left its inventor deep in debt. When the PC arrived, the stakes rose from inconvenience to exclusion: every layer of the stack, from character encodings to keyboards, assumed a small alphabet (see The Globalization of Software). In the same years, China’s state “748 Project” (launched 1974) attacked the output side (Wang Xuan’s laser phototypesetting rescued Chinese printing) but input remained the bottleneck: how does a human get 60,000 characters out of 26 keys?

The typewriter had been the first test. The American missionary Devello Sheffield showed a machine with 4,662 characters on a disc in Paris in 1897; Zhou Houkun built one at MIT with 1,500 characters on each of four cylinders (reported in Popular Science in 1917); Shu Zhendong’s design of 1922, made by a Shanghai publisher, held 2,500 types in a sliding tray from which the typist picked one with a pointer. That tray principle, patented in Japan by Kyota Sugimoto in 1915, won: Japanese machines sold as Wanneng (“all-purpose”) pushed out local designs during the occupation, East Germany’s VEB Optima in Erfurt reworked the design in 1952 for a Shanghai factory, and from 1953 the national standard was the Double Pigeon, built as a uniform model (the DHY) from 1964 and made until 1992. Lin Yutang’s MingKwai, patented in the US in 1952, never went into production; its only prototype turned up in a New York basement in January 2025 and went to Stanford. West Germany’s Olympia showed an electronic Chinese typewriter with an inkjet printer in Beijing in 1981 and sold only about 600 of the production model 1011 from 1984.

Serious voices proposed the obvious surrender: romanize Chinese, as Vietnam had romanized Vietnamese. The input-method pioneers were, in a real sense, fighting for the survival of the script itself.

Wang Yongmin’s Five Strokes

Wang Yongmin (born 1943), an engineer from Nanyang, Henan province, spent five years from 1978 on a brute-force cognitive project: analyzing thousands of characters into recurring components, the roots and radicals a hand writes in a defined stroke order. His Wubizixing (“five-stroke character shape”) method, completed in 1983 and released for the PC in 1984, mapped these components onto 25 letter keys organized into five zones, one per fundamental stroke type: horizontal, vertical, left-falling, right-falling/dot, and hook.

To type a character, the typist keys the character’s first three components and its last, at most four keystrokes for any character, and frequent characters in one or two. Because the code follows shape rather than sound, Wubi has two properties phonetic methods lack: it works identically for speakers of any Chinese dialect (or for typists who don’t know a character’s pronunciation at all), and it is nearly unambiguous, the typist almost never pauses to choose among homophones from a candidate menu. Experienced Wubi typists are reported to reach about 160 characters per minute, which answered the claim that Chinese was unfit for the machine age. Through the late 1980s and 1990s, Wubi (marketed as “Wang code”) became the standard of China’s professional typists, government offices, and typing schools, and Wang Yongmin was named a National Model Worker in 1988. In 1989 he founded Wangma, a Beijing software company, to sell it.

Dead End: The Pinyin Counterrevolution

Wubi’s strength (encoding shape with mechanical precision) was also its weakness: it must be studied, like shorthand. Memorizing the component-to-key mapping takes weeks; fluency takes months. That cost was acceptable for professional typists. It was unacceptable for the hundreds of millions of ordinary users who came online in the 2000s, and by then, software had removed the reason to pay it.

Pinyin input, typing the romanized pronunciation every mainland pupil learns in school, then picking the intended character from a menu, had always been trivially easy and painfully slow. Statistical language modeling changed the trade: IMEs began predicting whole words and sentences from context, engines like Sogou Pinyin (2006) built their dictionaries and rankings from web-scale usage data, and the choice menu almost disappeared. Pinyin became fast enough with zero learning cost, and the smartphone finished the argument. Today the overwhelming majority of Chinese text is entered phonetically (or by handwriting recognition and voice); Wubi survives as a professional niche and a cultural cause, its partisans argue, with some scholarly support, that purely phonetic input is eroding Chinese speakers’ ability to write characters by hand (“character amnesia”).

The deeper history, as the historian Thomas Mullaney has argued, is that China never simply “adopted” the Western keyboard; it subverted it. Every Chinese IME turns typing into human-computer collaboration: the keystrokes are not the text but a query, which software resolves into writing. That architecture (predictive, candidate-based, model-driven) anticipated the autocomplete-everything world that Western text entry only reached with the smartphone.

Legacy

Wang Yongmin’s five years of manual character analysis belong beside the great feats of pre-digital data engineering. Wubi proved, at the moment of maximum doubt, that the Chinese script and the computer were compatible, buying time during which encodings (GB 2312, later Unicode; see Unicode and UTF-8), fonts, and phonetic methods matured. The war Wubi won was not against pinyin; it was against the proposition that a billion people would have to change how they write to use computers. That war stayed won. China’s subsequent software industry (see China’s Tech Industry) was built by people typing Chinese.

📚 Sources