Unicode gives each Han character it encodes a code point, but recording how the character is read is a separate job, often left undone for rare ones. We counted the gap in Unicode's own Han database, Unihan 18.0. Its source tags trace 951 characters to name and ID lists from China's Ministry of Public Security. Unihan gives a Mandarin reading for 157 of them. All 622 characters in Extension I carry the tag for the ministry's 2023 ID system, and none of the 622 has an English definition. Unihan holds only one Chinese source tag per character, so 951 is a floor rather than a full count.
Almost three-quarters of Unicode's unified ideographs sit outside the Basic Multilingual Plane
Counting the code points that Unicode 18.0, dated 16 September 2026, marks as unified ideographs, we get 101,997 CJK unified ideographs, drawn from Japanese, Korean and Vietnamese sources as well as Chinese. The total passed 100,000 in September 2025, when Unicode 17.0 added Extension J.
| Block | Range | Unified ideographs |
|---|---|---|
| Original block | 4E00–9FFF | 20,992 |
| Extension A | 3400–4DBF | 6,592 |
| Compatibility block | F900–FAFF | 12 |
| Extension B | 20000–2A6DF | 42,720 |
| Extension C | 2A700–2B73F | 4,160 |
| Extension D | 2B740–2B81F | 223 |
| Extension E | 2B820–2CEAF | 5,774 |
| Extension F | 2CEB0–2EBEF | 7,473 |
| Extension I | 2EBF0–2EE5F | 622 |
| Extension G | 30000–3134F | 4,939 |
| Extension H | 31350–323AF | 4,192 |
| Extension J | 323B0–3347F | 4,298 |
Of the 101,997, 74,401 (72.9%) lie beyond the Basic Multilingual Plane and take four bytes each in UTF-8. How banks and forms store them is covered in our guide to the Chinese characters a bank computer cannot store.
More than half of the ideographs have no Mandarin, Cantonese or English entry in Unihan
A Mandarin reading here means a value in any of four Unihan fields: kMandarin, kHanyuPinyin, kXHC1983 or kTGHZ2013. That covers 44,362 characters (43.5%). kCantonese covers 29,915 and an English kDefinition 23,014. For 55,111 characters (54.0%), Unihan records none of the three.
The gap widens in the extensions. The original block has a Mandarin reading for 20,924 of its 20,992 characters, Extension B for 14,618 of 42,720. Extension J has no reading field of any kind.
Even where a reading exists, UAX #38, revision 41, says kMandarin is "targeted specifically for use by CLDR collation and transliteration" and "may not in all cases track the preferred use in some dictionaries".
Every Extension I character carries a 2023 ID-system tag and lacks an English definition
Unicode 15.1, dated 12 September 2023, describes almost all of its 627 additions as "urgently needed CJK ideographs, synchronized with planned additions to the Chinese national standard, GB 18030". The 622 that form Extension I all carry the source tag GIDC23, which UAX #38 defines as "ID system of the Ministry of Public Security of China, 2023". The tag names neither ID cards nor personal names. Unihan's kMandarin field reads 99 of the 622, and none has a kDefinition.
Characters from the ministry's name and ID lists go without a Mandarin reading in 83.5% of cases
Two other tags name the same ministry. GXM marks "Characters for use in personal names in China", from its Public Order Administration: 282 characters across nine blocks, 216 of them in Extension H and the rest in blocks as old as Extensions A and B. GIDC marks 47 from the ministry's "Supplementary Characters of the Public Security Population Information Special Font".
With Extension I, that makes 951. Unihan has a Mandarin reading for 157 (16.5%) and an English definition for 17, which leaves 794 with no Mandarin reading, by our arithmetic. A name character first encoded from an older GB standard carries that GB tag instead, hence the floor.
An open-source pinyin library comes up short for 98.8% of Extension A
We ran pinyin-pro 3.28.1, the library our rare Chinese character reader uses, over all 101,997 ideographs. It reads 21,131 (20.7%), including 79 of Extension A's 6,592, and with no data it hands back the character itself. The Cantonese library, to-jyutping, reads 6,582 of Extension A.
Until 3 November 2026 our reader printed that glyph in its Pinyin row and rejected Extensions G, H and J, 13,429 characters, with "No Chinese characters detected". It showed a Mandarin reading for one of the 951 name and ID-list characters.
The revision of that date accepts all 101,997, asks the libraries first and fills gaps from Unihan, labelled as such. It gives Mandarin for 44,362 characters, 5,788 of them in Extension A. Stroke counts missing from hanzi-writer come from Unihan's kTotalStrokes, counted "according to the stroke-counting conventions of the IRG". Paste 𮯴 (U+2EBF4) and the card shows ní and ngai4, both tagged Unihan; 𮯰 (U+2EBF0) shows four strokes and "No reading recorded in Unihan". A box instead of a glyph is a font gap, covered in our guide to text that comes out as empty boxes. For a person's name, ask the person.
Readings most often fail to match at the third level of China's standard character table
The State Council issued the 通用规范汉字表 (Table of General Standard Chinese Characters) by notice 国发〔2013〕23号 of 5 June 2013. The notice says general use in society "应以《通用规范汉字表》为准", should follow the table. UAX #38 gives it 8,105 characters in three levels, the third "used for personal names, place names, and for science and technology". We did not open the table itself, a scan of about 100 MB, so we say nothing about its rules on names.
We compared pinyin-pro's single reading with Unihan's transcription of the 2013 通用规范汉字字典, held in kTGHZ2013.
| Level | Characters | Library reading not among the dictionary's |
|---|---|---|
| First | 3,500 | 8 (0.23%) |
| Second | 3,000 | 61 (2.0%) |
| Third | 1,605 | 78 (4.9%) |
The third-level rate is about 21 times the first, on our arithmetic. The library gives 佴 as èr where the dictionary has nài, and 妧 as wàn where it has yuán. A mismatch means only that the reading is absent there; a character can carry readings a dictionary leaves out.
Taiwan's Name Act will not register a Chinese name in characters its listed dictionaries lack
Article 2 of Taiwan's 姓名條例 (Name Act), last amended on 29 May 2024, covers household registration, naturalisation and passports. A Chinese name must use characters listed in 「辭源、辭海、康熙等通用字典或教育部編訂之國語辭典」 (the Ciyuan, the Cihai, the Kangxi dictionary or the Education Ministry's Mandarin dictionary). A name in other characters is 「不予登記」: it will not be registered (text from 全國法規資料庫). The same article lets indigenous peoples use indigenous scripts for traditional names, and Article 4 lets an indigenous-script or romanised form be registered alongside a Chinese name.
We found no Hong Kong rule on name characters, though we did not search the Registration of Persons Ordinance. The page "What is Hong Kong Supplementary Character Set (HKSCS)?" on the Hong Kong government's Common Chinese Language Interface website, read on 3 November 2026, traces the set to the 1995 Government Common Character Set of "locally specific characters required for use on computers by government departments", and lists proper names, "such as names of persons, places or companies", among the broad categories of HKSCS-2016's characters. The latest of four characters approved since HKSCS-2016, 㗩 (U+35E9), was approved by the Chinese Language Interface Advisory Committee on 18 February 2025.
View author profile → · Editorial policy
