Four-Corner Code Lookup

Reading a character by its corners: how the four-corner code works

A Chinese dictionary needs an order, and characters have no alphabet. Radicals and stroke counts are one answer; the four-corner system is another. It looks only at the corners of a character: the shape of the stroke at the upper left, the upper right, the lower left and the lower right, each turned into a digit from 0 to 9. 陳 is 7529: a corner at the upper left, a piercing stroke at the upper right, a vertical at the lower left and a "small" shape at the lower right. Many codes add a fifth digit after a point, read from the stroke just above the lower-right corner — 陳 is 7529.6.

Ten shapes, numbered 0 to 9

The digits stand for ten stroke shapes: 0 頭 (a dot over a horizontal, as in 亠), 1 橫 (a horizontal, which also covers a rising stroke and a right hook), 2 垂 (a vertical, which also covers a left-falling stroke and a left hook), 3 點 (a dot, which also covers a right-falling stroke), 4 叉 (two strokes crossing), 5 插 (one stroke piercing two), 6 方 (a square, as in 口), 7 角 (a corner), 8 八 (two strokes in the shape of 八) and 9 小 (the shape of 小). The groupings for 1, 2 and 3 are as Taiwan's National Central Library sets them out in its classification scheme.

0 does two jobs. It is the 頭 shape, and it is also what a corner gets when it has nothing new. In the rules printed in the Japanese dictionary 字通, a stroke that already has a number is 0 at a later corner, and when the top or bottom of a character is a single stroke, it is numbered at the left corner and the right corner is 0. So 一 is 1000.0: one horizontal at the upper left, and nothing new at any other corner. 亠 is 0000.0 for the same reason: the 頭 is coded once, at the upper left, and the other corners have already been used. A 0 in first place is always a 頭, but after that the code alone cannot say which job it is doing (諒, 0069.6, has a real 亠 at its upper right), so this page gives both readings for a later 0.

The corners are read from the handwritten regular-script (楷書) form, not from a printed Song or Ming typeface: both the National Central Library's rules and 字通 say so. 戶 begins with a dot when written by hand, so its first digit is 3, whatever the font on your screen shows.

Why so many characters share a code

Four digits allow ten thousand codes, but characters do not spread evenly across them. Unicode's table gives codes for 16,916 characters, and they use 3,546 distinct four-digit codes, counting both codes of the 149 characters that have two. Only 1,134 characters, 6.7%, have a four-digit code to themselves, and the code 4422 is shared by 97.

The fifth digit helps, but less than you might expect. 14,746 of the 16,916 characters have one, and with it 4,994 characters, 29.5%, have a code to themselves, so most still share. Unicode also almost never gives a simplified-only form a fifth digit: of the 2,380 characters in GB 2312 but not in Big5, 2,168 have none (our count from Unihan's own GB 2312 and Big5 mappings). 寬 is 3021.6, but 宽 is plain 3021, so a search for 3021.6 cannot say whether 宽 belongs; this page lists such characters apart rather than dropping them. What is left is down to stroke count and the reader's eye, which is why lists put GB 2312 and Big5 characters first and fewer strokes before more.

Where you still meet it

The system was devised by Wang Yunwu of the Commercial Press, as 字通 records it, and made public in 1926, according to Hu Shi's preface to it. A revised version, with these ten shapes and the fifth corner, existed by February 1930, when Hu Shi wrote a postscript describing it.

Unicode's own description says the system is now rarely used for typing but continues in indexing — in academic studies, reference works and some Chinese dictionaries. It is also a way to look up a character you can see but cannot read: you do not need its pronunciation or its radical, only its corners.

For learners in Singapore and Malaysia

Students here usually meet characters through Hanyu Pinyin, which only helps once you know how a character is said. The four-corner code works from the shape alone, which makes it a useful way to find an unfamiliar character in an older dictionary or catalogue that is indexed this way.