Several systems have been proposed for describing the internal structure of Chinese characters, including their strokes, components, and the stroke order, and the location of each in the character's ideal square. This information is useful for identifying variants of characters that are unified into one code point by Unicode and ISO/IEC 10646, as well as to provide an alternative form of representation for rare characters that do not yet have a standardized encoding in Unicode. Many aim to work for regular script, as well as to provide the character's internal structure which can be used for easier look-up of a character by indexing the character's internal make-up and cross-referencing among similar characters.
Character Description Language (CDL) is an XML-based declarative language co-created by Tom Bishop and Richard Cook for the Wenlin Institute. It defines characters by the arrangement of components, which are not required to reflect the semantic or etymological history of the character. In order for a component to fit into the allotted portion of the whole character's square, a set of fewer than 50 strokes allows one to construct approximately 1,000 components, which may in turn describe tens of thousands of characters.
Chapter 18 of The Unicode Standard (version 15.0) defines the "Ideographic Description Sequences" (IDS) syntax used to describe characters in featural terms, by arrangements of components with code points. Sixteen special characters in the range U+2FF0..U+2FFF act as prefix operators to combine other characters or sequences to form larger characters.
| Character | Unicode Character Number | Full Unicode Name |
|---|---|---|
| ⿰ | U+2FF0 | Ideographic description character left to right |
| ⿱ | U+2FF1 | Ideographic description character above to below |
| ⿲ | U+2FF2 | Ideographic description character left to middle and right |
| ⿳ | U+2FF3 | Ideographic description character above to middle and below |
| ⿴ | U+2FF4 | Ideographic description character full surround |
| ⿵ | U+2FF5 | Ideographic description character surround from above |
| ⿶ | U+2FF6 | Ideographic description character surround from below |
| ⿷ | U+2FF7 | Ideographic description character surround from left |
| | U+2FFC | Ideographic description character surround from right |
| ⿸ | U+2FF8 | Ideographic description character surround from upper left |
| ⿹ | U+2FF9 | Ideographic description character surround from upper right |
| ⿺ | U+2FFA | Ideographic description character surround from lower left |
| | U+2FFD | Ideographic description character surround from lower right |
| ⿻ | U+2FFB | Ideographic description character overlaid |
| | U+2FFE | Ideographic description character horizontal reflection |
| | U+2FFF | Ideographic description character rotation |
Two additional ideographic description characters are scattered in other Unicode blocks. U+303E 〾 IDEOGRAPHIC VARIATION INDICATOR is not officially an ideographic description character, but is sometimes used in ideographic description sequences.
| Character | Code point | Block | Name |
|---|---|---|---|
| 〾 | U+303E | CJK Symbols and Punctuation | Ideographic variation indicator |
| | U+31EF | CJK Strokes | Ideographic description character subtraction |
These sequences are useful in describing to the reader a character that is not directly printable, either because it is absent in a given font, or is absent from the Unicode standard altogether. For example, the sawndip character
encoded in CJK Unified Ideographs Extension F as U+2DA21 𭨡 can be described as ⿰書史. Another use is for dictionary lookup purposes, as a rough input method for queries.
These sequences can be rendered either by keeping the individual characters separately or by parsing the Ideographic Description Sequence and drawing the ideograph so described. They do not, by themselves, provide unambiguous rendering for all characters. For instance, the sequence ⿱十一 represents both ⼟ 'EARTH' with the middle bar being narrower, and ⼠ 'SCHOLAR' with the middle bar being wider.
Unicode's specification for these sequences is based on the characters and syntax of the earlier GBK encoding. Additional symbols are later encoded to fill in the missing combinations.
The IDSgrep free software package by Matthew Skala extends Unicode's IDS syntax to include additional features for dictionary lookup; it is capable of converting KanjiVG's database to its own extended IDS format, or of searching EIDS files generated by the related Tsukurimashou font family.
The composition formula (構字式) describes a character by its components and a set of composition symbols. It is the notation of the Chinese character structure database (漢字構形資料庫), which the Document Processing Laboratory of the Institute of Information Science, Academia Sinica, in Taiwan has built since 1993. Version 1.4 of the database, released on 17 May 2002, introduced the formulas to represent characters missing from character sets. The design rests on the laboratory's observation that characters differing only in the placement of their components mostly have two components, placed left and right, above and below, or outside and inside. The same three relations underlie the root set of National Chiao Tung University: a 1972 master's thesis there by Ni Geng (倪耿) tested more than ten ways of combining parts and found that joining them side by side, joining them vertically and containment greatly simplified the structure of characters, the other ways being rare, and the set's formal grammar builds characters from roots with these three positional symbols.
The thirteen symbols are encoded consecutively at 8DF1–8DFD in the user-defined area of Big5. A connector stands between two components, and a repetition symbol comes before the one component it repeats.
| Big5 | Class | Meaning | Example |
|---|---|---|---|
| 8DF2 | Connector | left to right | 順 from 川 and 頁 |
| 8DF1 | above to below | 含 from 今 and 口 | |
| 8DF3 | outside to inside | 圍 from 囗 and 韋 | |
| 8DFC | Component sequence | start | 解 from 角, 刀 and 牛, enclosed in writing order |
| 8DFD | end | ||
| 8DF4 | Repetition | two, one above the other | 炎 from 火 |
| 8DF7 | three, one above the other | ||
| 8DF5 | two, side by side | 朋 from 月 | |
| 8DF8 | three, side by side | ||
| 8DF6 | three in a triangle | 焱 from 火 | |
| 8DFB | four, side by side | ||
| 8DFA | four, one above the other | ||
| 8DF9 | four in a square | 燚 from 火 |
Because a formula follows the shape of the character, it needs no lookup table, and different writers usually arrive at the same formula. Complex characters still admit several formulas, which the database normalizes by expanding each component into its smallest components (字根) and sorting them. By 2009 at least 23 databases of Taiwan's national digital archives program used the structure database for missing characters, among them Academia Sinica's Scripta Sinica full-text database (漢籍電子文獻). The laboratory also drafted a national standard on the notation, later titled 中文字構形索引序列 ("Chinese character structure index sequence"), for looking up rare characters alongside CNS 11643, and sent it to the Bureau of Standards, Metrology and Inspection for review.
Several projects write a character as a formula of components and arithmetic signs. Most of them join side-by-side parts with a plus sign and stacked parts with a slash; CBETA uses an asterisk for side-by-side parts instead.
The Chinese Buddhist Electronic Text Association (CBETA) in Taiwan writes characters missing from its digital Buddhist canon with a general composition formula (一般組字式) of six half-width signs and two kinds of brackets. It chose the notation so that the texts can be read as plain text, without installing fonts of user-defined characters or images. The National Institute of Korean History's Unicode Hanja search (유니코드한자 검색시스템) in the Korean History Database takes formulas for its character-splitting (파자) search, with + for side-by-side parts and * for a combination.
In Japan, Aozora Bunko writes a kanji outside JIS X 0208 as a note that describes its shape, followed by its JIS X 0213 or Unicode code point where it has one, as in ※[#「てへん+劣」、第3水準1-84-77]. The description uses the full-width signs +, −, / and <, and names a part in words where that matches the standard's glyph more closely. The kanji braille of Kawakami Taiichi (川上泰一) spells out a character's shape in a jishiki (字式, "character formula") with + for left and right and / for above and below, as in 木偏+寸 for 村. The Yokohama kanji braille group 横浜漢点字羽化の会 added nine further signs, such as \ for overlapping parts (一\| for 十) and > for a part inside an enclosure (囗構え>玉 for 国).
| Sign | CBETA | Korean History Database | Aozora Bunko | Kanji braille jishiki |
|---|---|---|---|---|
+ | after -, the replacement: 間-日+月 for 閒 | left and right, may be omitted: 氵+飛 | left and right: 金+乙 for 釔 | left and right: 木偏+寸 for 村 |
/ | above and below: 立/日 for 音 | above and below: 髟/目 | above and below: 王/厶 for 丟 | above and below: ウ冠/寸 for 守 |
* | left and right: 日*月 for 明 | combination: (臼*米)/皿 for a variant of 盥 | ||
- | removal: 請-言 for 青 | removal: 石+(毁-殳) | removal: 尓−小 for 𠂉 | |
@ | containment: 囗@大 for 因 | |||
< | 勹<一 for 勺 | the left part at the lower left of the right part: 工<弋 for 式 | ||
? | a component with no character to write it: (?*匕)/月 for 背 | |||
() | grouping: 組-且+((土/(土*土))/兀) for 繞 | evaluated first: 穴/(爿+未) for a variant of 寢 | grouping: 七/(七+七) for 㐂 | grouping, written “ ”: 司 as “一/口”<上かど |
[] | a formula within running text: 羅[目*侯]羅母耶輸陀羅比丘尼 |