Reading CJK Names on Business Cards: Order, Splitting and Romanization
· 11 min read · By the BizCardPro.AI team
A business card is the only document that tells you, authoritatively, how a person writes their own name in two scripts at once. It is also the point at which most contact databases quietly break: the surname lands in the given-name field, the characters are dropped, or a generated romanization replaces the one the person actually uses.
This is a reference for getting the name right. It covers name order, the Chinese romanization systems that make one character into five different English words, compound surnames, Japanese readings, Korean passport spellings, and the storage rules that keep all of it intact. For the characters below the name, see the companion reference on CJK job titles and company forms.
Which part of a CJK name is the surname?
Surname-first is the norm in all three languages, and in all three the surname is short.
| Language | Native form | Surname | Given name | Typical lengths |
|---|---|---|---|---|
| Chinese | 陳大文 | 陳 (Chan) | 大文 (Tai Man) | Surname 1 char (rarely 2); given 1–2 |
| Japanese | 山田太郎 | 山田 (Yamada) | 太郎 (Tarō) | Surname 1–4 kanji, usually 2; given 1–3 |
| Korean | 김민준 | 김 (Kim) | 민준 (Min-jun) | Surname 1 syllable (rarely 2); given usually 2 |
The trap is the English side. It is written for a Western reader, so companies routinely reorder it to given-first — and now the card contradicts itself: 陳大文 on one face, Tai Man Chan on the other. A scanner that reads only the English face, or that applies a first-name/last-name assumption to it, stores the wrong field and never flags it.
Reliable tells for finding the surname:
- Read the native side first. The leading character or syllable is the surname, with the compound exceptions below.
- Capitals are a deliberate signal. YAMADA Taro, CHAN Tai Man, KIM Min-jun — many bilingual cards capitalize the surname precisely to remove the ambiguity. So does a comma: Yamada, Taro.
- Check the email address. t.yamada@, taiman.chan@ and mjkim@ usually resolve the order before you have to guess.
- Match the two sides against each other. If one word of the romanization corresponds to the first native character, that word is the surname regardless of where it sits in the line.
Why is the same Chinese surname spelled Chen, Chan and Tan?
Because “Chinese” is a family of languages, and each romanization system transcribes a different one. 陳 is chén in Mandarin, can4 in Cantonese and tân in Hokkien. A Beijing card romanizes the Mandarin reading; a Hong Kong card romanizes the Cantonese one, because Cantonese is what Hong Kong speaks and what its immigration department transcribes. Same character, same family, different spoken language — not a mistake by either party.
On top of that sit four competing Mandarin and Cantonese systems:
| Character | Hanyu Pinyin | Wade-Giles | Tongyong Pinyin | HK Government Cantonese | Jyutping |
|---|---|---|---|---|---|
| 陳 / 陈 | Chen | Ch’en | Chen | Chan | can4 |
| 張 / 张 | Zhang | Chang | Jhang | Cheung | zoeng1 |
| 李 | Li | Li | Li | Lee | lei5 |
| 黃 / 黄 | Huang | Huang | Huang | Wong | wong4 |
| 吳 / 吴 | Wu | Wu | Wu | Ng | ng4 |
| 謝 / 谢 | Xie | Hsieh | Sie | Tse | ze6 |
| 徐 | Xu | Hsü | Syu | Tsui | ceoi4 |
| 蔡 | Cai | Ts’ai | Cai | Choi | coi3 |
| 林 | Lin | Lin | Lin | Lam | lam4 |
| 劉 / 刘 | Liu | Liu | Liou | Lau | lau4 |
Where each one is used:
- Hanyu Pinyin (漢語拼音) — official in mainland China since 1958, in Singapore, and in Taiwan since 2009. The default assumption for a mainland card.
- Wade-Giles — the older British system, dominant for Taiwanese names and places before 2000 and still carried by many individuals. It marks aspiration with an apostrophe (Ch’en, Ts’ai) which is almost always dropped in practice, so Ts’ai is printed Tsai.
- Tongyong Pinyin (通用拼音) — Taiwan’s official system from 2002 to 2008. Recognizable by jh- for pinyin zh- (Jhang), s- for x- (Sie, Syu) and c- for q-.
- Hong Kong Government Cantonese romanization — the system behind Hong Kong ID cards and street signs. No tones, no diacritics: Chan, Cheung, Wong, Lau.
- Jyutping (粵拼) — the modern linguistic standard for Cantonese, with tone numbers (can4). It is a study and input tool, not a naming convention; you will not see it printed on a card, but you may meet it in a language database.
Two further complications worth knowing before you deduplicate anything:
Southern-dialect spellings across Southeast Asia. Cards from Singapore, Malaysia and older diaspora communities often romanize Hokkien, Teochew, Hakka or Hainanese readings instead:
| Character | Mandarin | Cantonese | Common Southeast Asian spellings |
|---|---|---|---|
| 陳 | Chen | Chan | Tan (Hokkien, Teochew), Chin (Hakka) |
| 林 | Lin | Lam | Lim |
| 王 | Wang | Wong | Ong |
| 黃 | Huang | Wong | Ng, Wee |
| 吳 | Wu | Ng | Goh |
| 楊 | Yang | Yeung | Yeo, Yeoh |
| 蔡 | Cai | Choi | Chua |
Romanizations collide. Ng is 吳 in Cantonese but 黃 in Hokkien. Wong is both 王 and 黃. Pinyin Xu covers 徐 and 許 — which separate again in Cantonese as Tsui and Hui. Romanized names are therefore never a safe deduplication key on their own; the characters are.
What happens when a Chinese surname is two characters?
Compound surnames (複姓) are uncommon but not rare, and naive splitting always gets them wrong: given 歐陽志明, a splitter that takes one leading character files the person under 歐 with a given name of 陽志明. Both fields are then wrong.
| Compound | Pinyin | Hong Kong spelling |
|---|---|---|
| 歐陽 / 欧阳 | Ouyang | Au-Yeung |
| 司馬 / 司马 | Sima | — |
| 諸葛 / 诸葛 | Zhuge | — |
| 上官 | Shangguan | — |
| 司徒 | Situ | Szeto |
| 夏侯 | Xiahou | — |
| 令狐 | Linghu | — |
| 皇甫 | Huangfu | — |
The list is closed and short, which makes the rule mechanical: if the first two characters match a known compound surname, they belong together. Korea has its own set, inherited from the same tradition — 남궁 (Namgung), 선우 (Seonu), 황보 (Hwangbo), 제갈 (Jegal), 독고 (Dokgo) — and they break the “Korean surnames are one syllable” assumption in exactly the same way.
Length alone proves nothing. A four-character Chinese name may be a one-character surname with a three-character given name, or — in an older Hong Kong and Taiwanese convention for married women — the husband’s surname, then her own surname, then her given name. The romanized side usually settles it, because a compound surname is written as a single unit there.
How do you know how a Japanese name is pronounced?
You mostly do not, from the kanji alone. Japanese surnames carry multiple possible readings, and no rule picks between them — which is why the rōmaji printed on the card, or furigana above the kanji, is the only reliable source.
| Kanji | Possible readings |
|---|---|
| 高橋 | Takahashi |
| 山崎 | Yamazaki or Yamasaki |
| 中田 | Nakata or Nakada |
| 河野 | Kōno or Kawano |
| 東 | Azuma or Higashi |
| 東海林 | Shōji or Tōkairin |
東海林 is the standard illustration: two entirely unrelated readings of the same three characters, both borne by real families. Given names are more variable still. Treat any reading you did not read off the card as provisional.
Hepburn or Kunrei-shiki? Hepburn (ヘボン式) maps kana to English-like spellings and is what appears on cards, passports and signage. Kunrei-shiki (訓令式) is the systematic ISO-aligned form taught in Japanese schools, and occasionally surfaces in academic or older material.
| Kana | Hepburn | Kunrei-shiki |
|---|---|---|
| し | shi | si |
| ち | chi | ti |
| つ | tsu | tu |
| ふ | fu | hu |
| じ | ji | zi |
| しゃ | sha | sya |
So 藤井 is Fujii in Hepburn and Huzii in Kunrei-shiki; 土屋 is Tsuchiya or Tutiya.
Long vowels are where the variants multiply. さとう can be written four defensible ways:
| Native | Hepburn (macron) | Keyboard spelling | Passport / card |
|---|---|---|---|
| 佐藤 | Satō | Satou | Sato, Satoh |
| 伊藤 | Itō | Itou | Ito, Itoh |
| 大野 | Ōno | Oono | Ono, Ohno |
Japanese passports drop the length mark by default but permit an OH spelling on request, which is why Ohno and Itoh exist alongside Ono and Ito. The cost of dropping it is collision: 大野 (Ōno) and 小野 (Ono) become the same string. One more variant to expect — traditional Hepburn writes ん as m before b, m and p, giving Homma for 本間 and Shimbashi for 新橋, where modern Hepburn writes Honma and Shinbashi.
Why do two Koreans with the same surname spell it differently?
Because the spelling is chosen, not derived. South Koreans select the Latin form of their name when they first apply for a passport, and both official romanization systems explicitly leave personal names as an exception. Siblings routinely end up spelled differently.
| Hangul | Revised Romanization (2000) | McCune-Reischauer | Common real spellings |
|---|---|---|---|
| 김 | Gim | Kim | Kim |
| 이 | I | Yi | Lee, Yi, Rhee, Ri |
| 박 | Bak | Pak | Park, Pak, Bahk |
| 최 | Choe | Ch’oe | Choi, Choe |
| 정 | Jeong | Chŏng | Jung, Chung, Jeong |
| 강 | Gang | Kang | Kang |
| 조 | Jo | Cho | Cho, Joe |
| 노 | No | No | Noh, Roh |
| 오 | O | O | Oh |
Notice how often the third column wins. Revised Romanization is the official system — it replaced McCune-Reischauer for South Korean place names in 2000 and dropped its breves and apostrophes — yet almost nobody named 김 writes Gim, and almost nobody named 박 writes Bak. The Lee spelling of 이 preserves an older initial ㄹ that South Korean orthography drops but North Korean keeps (Ri).
Given names add a formatting axis: 민준 appears as Min-jun, Minjun, Min Jun and MinJun. All four are the same name; only one is the person’s own. Kim, Lee and Park together account for close to half of South Korea, so the given name, company and title do most of the identifying work in a Korean archive — and they have to be captured precisely. The myeongham scanning guide covers the capture side.
How should CJK names be stored in a contact database?
Six rules, in order of how much damage they prevent:
- Store native script and romanization in separate fields. Not one derived from the other at display time — both, persisted. You will search “Kim” from a Latin keyboard on Tuesday and paste 김민준 into a greeting on Thursday.
- Split family and given names into their own fields, in both scripts. A single full-name string forces every downstream system to re-guess the order, and they will not all guess the same way.
- Never overwrite the card’s romanization with a generated one. The printed spelling is the person’s legal and preferred form; a generated one is a hypothesis about pronunciation. If you generate a reading because the card has none, flag it as generated.
- Record the script, not just the language. Traditional and Simplified Chinese are not interchangeable — rewriting 陳大文 as 陈大文 changes how your contact writes their own name. The Chinese card etiquette guide covers which markets use which.
- Keep the original card image attached to the record. It is the only way to settle a doubtful reading later without contacting the person.
- Use UTF-8 end to end. Most “lost characters” are an encoding failure on export or import, not an OCR failure — see the guide on names that turn into question marks in Excel.
What does BizCardPro.AI do with the name?
BizCardPro.AI keeps the native script exactly as printed — 陳大文, 山田太郎, 김민준 — and puts a romanized reading beside it rather than in place of it, so the characters and the spelling the person chose both survive the scan. It splits family and given names into separate fields using surname-first order, including compound surnames like 歐陽 and 司馬, and treats Traditional and Simplified as distinct scripts rather than normalizing one into the other. The photo of the card stays attached to every contact, which is what makes rule five above practical rather than aspirational. For the characters under the name, the CJK title reference is the companion to this page.
The underlying principle is narrow and boring: a name is data about a person, not a string to be normalized. 陳 is Chen and Chan and Tan; さとう is Satō and Satou and Sato; 이 is Lee and Yi and Rhee. The card in your hand already contains the correct answer for the person who gave it to you. The only real failure mode is throwing that answer away and computing a new one.
Frequently asked questions
Is Chen or Chan correct for 陳?
Both are correct — they romanize the same character from different languages. Chen is Mandarin (Hanyu Pinyin, used in mainland China, Taiwan and Singapore). Chan is Cantonese, the spelling used in Hong Kong government romanization. In Hokkien and Teochew communities in Singapore and Malaysia the same surname is usually Tan, and in Hakka families Chin. Wade-Giles renders the Mandarin reading as Ch’en. None of these is a mistake: use whichever spelling the person prints on their own card, because that is the one on their passport, their email address and their published byline.
Which name is the surname on a Japanese business card?
On the Japanese side, the surname comes first: in 山田太郎, 山田 (Yamada) is the family name and 太郎 (Tarō) is the given name. Japanese surnames are usually two kanji, but they range from one (東) to four (勅使河原). On the English side the order may be flipped to Taro Yamada, or the surname may be printed in capitals (YAMADA Taro) to remove the ambiguity. Since 2020 the Japanese government has written Japanese names family-name-first in its own English-language documents, so both orders are now in circulation.
Why do two Koreans with the same surname spell it differently?
Because South Koreans choose the Latin spelling of their own name when they apply for a passport, and that choice is personal rather than algorithmic. 이 is officially I under Revised Romanization and Yi under McCune-Reischauer, but real people write Lee, Yi, Rhee or Ri. 박 is officially Bak but is almost always written Park. Siblings can and do end up with different spellings of the same surname. The spelling printed on the card is the person’s legal and preferred one; never replace it with a generated one.
How do I know if a Chinese name has a two-character surname?
Compound surnames (複姓) are a closed, short list — 歐陽, 司馬, 諸葛, 上官, 司徒, 夏侯, 令狐, 皇甫 and a handful more. If the first two characters of a name match one of them, they are the surname together. A four-character name is not proof of a compound surname: it may be a one-character surname with a three-character given name, or, in older Hong Kong and Taiwanese convention for married women, the husband’s surname followed by her own and then the given name. Check the romanized side, where 歐陽 appears as one unit (Ouyang or Au-Yeung).
Should I use the romanization printed on the card, or generate my own?
Always keep the printed one. A romanization on a business card is not a transliteration the person would accept an alternative to — it is how their passport, their email address, their LinkedIn profile and their published work spell their name. A generated romanization is a guess about pronunciation; the printed one is a fact about identity. Store the card’s spelling verbatim, add a generated reading only when the card has none, and mark which is which.
What is the difference between Hanyu Pinyin and Wade-Giles?
Both romanize Mandarin, but with different letter conventions. Hanyu Pinyin (official in mainland China since 1958, and in Taiwan since 2009) writes 張 as Zhang and 謝 as Xie. Wade-Giles, the older British system, writes them Chang and Hsieh, and marks aspiration with an apostrophe — 陳 is Ch’en, 蔡 is Ts’ai. In practice the apostrophes are dropped on cards and in passports, which is why Ts’ai appears as Tsai. Taiwan also used Tongyong Pinyin between 2002 and 2008, which writes 張 as Jhang and 謝 as Sie.
Is 佐藤 spelled Sato, Satou or Satō?
All three represent the same name, satō, with a long final vowel. Satō uses a macron (Hepburn, standard in scholarship); Satou reflects the kana spelling さとう as typed on a Japanese keyboard; Sato simply drops the length, which is what Japanese passports and most business cards do. Satoh is a fourth variant, and Japanese passport rules permit an OH spelling for a long o on request. Record whichever form the card prints — and be aware that dropping the macron creates collisions, since 大野 (Ōno) and 小野 (Ono) both flatten to Ono.
Why does the English side of a bilingual card show the given name first?
Because it is written for a Western reader, and the company assumes Western order. The native side of a Chinese, Japanese or Korean card is surname-first; the English side is frequently reordered to given-first, so the two faces of the same card disagree about which word is the family name. This is the single most common way a contact database ends up filing someone under their given name — the scan reads the English side, applies a first-name/last-name assumption, and stores it silently.
How should CJK names be stored in a contact database?
In separate fields: native family name, native given name, romanized family name, romanized given name, plus the script (Traditional Chinese, Simplified Chinese, kanji, hangul). Keep the card’s own romanization as printed and never overwrite it with a generated one. Store the file in UTF-8, and keep the original card image attached to the record so a doubtful reading can be checked against the printed source rather than re-guessed.
Keep reading
· 10 min read
How to Import Scanned Business Cards into Outlook (vCard and CSV)
No Outlook plugin needed. The two file routes into Outlook contacts, a column-by-column field mapping table, and the encoding and delimiter traps to avoid.
Read the guide →· 8 min read
Vertical Japanese Business Cards (縦書き): Why Scanners Break on Them
Vertical 縦書き meishi read top-to-bottom, right to left — the opposite of what OCR assumes. What breaks, the 縦中横 trap, and how to capture them cleanly.
Read the guide →· 8 min read
How to Scan Business Cards into SAP C4C and S/4HANA (No Integration Needed)
No SAP connector needed. The CSV route from a stack of trade-fair cards to SAP C4C and S/4HANA contacts: export columns, field mapping and encoding traps.
Read the guide →