Chinese presents unique challenges for automated PII detection. Unlike alphabetic languages, Chinese uses logographic characters without word boundaries, requiring sophisticated segmentation before entity recognition can occur. Names follow different structural patterns than Western names, with single-character surnames preceding one or two-character given names. Address formats use a hierarchical structure from large to small geographical units. Phone numbers and identification documents vary significantly across Mainland China, Taiwan, Hong Kong, and other Chinese-speaking regions.
Our Chinese language processing addresses these challenges with native NLP models trained specifically for Chinese text. We support both Simplified Chinese (简体中文) used in Mainland China and Singapore, and Traditional Chinese (繁體中文) used in Taiwan, Hong Kong, and Macau. The system handles mixed-script documents, code-switching between Chinese and English, and regional variations in identifier formats and terminology.
Chinese Name Detection
Chinese names require specialized detection approaches:
Name Structure: Chinese names typically consist of a one or two-character surname (姓) followed by a one or two-character given name (名). Unlike Western names, the surname comes first. Common patterns include:
- Single surname + two-character given name: 李明华 (Lǐ Mínghuá)
- Single surname + single-character given name: 王强 (Wáng Qiáng)
- Compound surname + given name: 欧阳修 (Ōuyáng Xiū)
- Four-character names: 司马相如 (Sīmǎ Xiàngrú)
Surname Recognition: We maintain comprehensive surname dictionaries covering:
- The "Hundred Family Surnames" (百家姓) traditional list
- Modern surname frequency data from census records
- Rare and regional surnames
- Compound surnames (复姓) like 欧阳, 司马, 上官, 诸葛
- Ethnic minority surnames from various Chinese ethnic groups
Disambiguation: Many Chinese characters serve as both surnames and common words. Context analysis distinguishes names from regular text:
// Name vs. common word disambiguation
"王先生来了" → 王 detected as surname (Mr. Wang came)
"国王的王冠" → 王 not a name (The king's crown)
Chinese Identification Numbers
Different Chinese-speaking regions use distinct ID formats:
China Resident Identity Card (居民身份证号码):
18-digit number encoding region, birthdate, sequence, and checksum:
Format: RRRRRRYYYYMMDDSSSC
- RRRRRR: 6-digit region code (province, city, district)
- YYYYMMDD: 8-digit birthdate
- SSS: 3-digit sequence number (odd=male, even=female)
- C: Check digit (0-9 or X)
Example: 110105199003071234
- 110105: Beijing, Chaoyang District
- 19900307: Born March 7, 1990
- 123: Sequence number
- 4: Check digit
Validation: Weighted sum modulo 11 checksum
Taiwan National ID (國民身分證統一編號):
Format: LSNNNNNNNNC
- L: Letter indicating registration location
- S: Gender digit (1=male, 2=female)
- NNNNNNNN: 8-digit serial number
- C: Check digit
Example: A123456789
Validation: Weighted checksum algorithm
Hong Kong Identity Card (香港身份證):
Format: L(L)NNNNNN(C)
- L: 1-2 letter prefix
- NNNNNN: 6-digit number
- C: Check digit in parentheses
Example: A123456(7)
Validation: Modulo 11 checksum
Macau BIR Number (澳門居民身份證):
Format: N(NNNNNNN)N
- First digit: ID type
- 7-digit serial number
- Check digit
Example: 1234567(8)9
Chinese Phone Numbers
Phone number formats vary by region:
Mainland China:
Mobile: 1XX-XXXX-XXXX (11 digits, starting with 1)
- 13X, 14X, 15X, 16X, 17X, 18X, 19X prefixes
- Various carrier prefixes: 移动, 联通, 电信
Landline: (0XXX) XXXX-XXXX
- Area code in parentheses or with hyphen
- Beijing: 010, Shanghai: 021, etc.
Examples:
13812345678 or 138-1234-5678
(010) 8765-4321 or 010-87654321
Taiwan:
Mobile: 09XX-XXX-XXX (10 digits)
Landline: (0X) XXXX-XXXX
Examples:
0912-345-678
(02) 2345-6789
Hong Kong:
Mobile/Landline: XXXX XXXX (8 digits)
- Mobile: 5, 6, 7, 9 prefixes
- Landline: 2, 3 prefixes
Examples:
9123 4567
2345 6789
Chinese Address Detection
Chinese addresses follow a distinctive hierarchical pattern:
Address Structure (Large to Small):
省/自治区 → 市/地区 → 区/县 → 街道/镇 → 路/街 → 号/弄 → 室/单元
Example:
北京市朝阳区建国路93号万达广场A座1502室
- 北京市: Beijing City
- 朝阳区: Chaoyang District
- 建国路93号: 93 Jianguo Road
- 万达广场A座: Wanda Plaza Building A
- 1502室: Room 1502
Address Variations:
- Formal addresses with full administrative hierarchy
- Abbreviated addresses omitting province for well-known cities
- Traditional address formats in Taiwan and Hong Kong
- Mixed Chinese-English addresses common in international business
Common Address Components:
Administrative: 省, 市, 区, 县, 镇, 乡, 村
Street types: 路, 街, 道, 巷, 弄, 里, 胡同
Building types: 大厦, 广场, 中心, 花园, 小区
Unit indicators: 栋, 幢, 座, 号, 室, 单元, 楼
Simplified vs Traditional Chinese
Our system handles both character sets with automatic detection:
Character Set Detection:
Simplified indicators: 国, 银, 发, 对, 业, 学
Traditional equivalents: 國, 銀, 發, 對, 業, 學
The system analyzes character frequency to determine:
- Purely Simplified text
- Purely Traditional text
- Mixed text (common in some contexts)
Cross-Reference Processing: PII patterns learned in one character set apply to the other. A name pattern detected in Simplified text will also be recognized in Traditional form.
Regional Terminology: Beyond character differences, terminology varies:
- Mainland: 身份证 | Taiwan: 身分證
- Mainland: 手机 | Taiwan: 手機 | HK: 手提電話
- Address terms, official titles, and institutions differ by region
Chinese Word Segmentation
Unlike alphabetic languages, Chinese lacks word boundaries:
Segmentation Challenges:
Input: 北京市长江大桥
Segmentation options:
- 北京市 / 长江大桥 (Beijing city / Yangtze River Bridge)
- 北京 / 市长 / 江大桥 (Beijing / mayor / Jiang Daqiao - a name)
Context determines correct segmentation
Our Approach: We use neural segmentation models trained on large Chinese corpora, combined with domain-specific dictionaries for PII-related vocabulary. This ensures accurate segmentation particularly around names, addresses, and identifiers.
Financial Identifiers
Chinese financial documents contain specific identifiers:
Bank Account Numbers:
Mainland China: 16-19 digits
- Major banks have specific formats
- ICBC, CCB, ABC, BOC, etc.
Taiwan: Various formats by bank
Hong Kong: 9-12 digits typically
Tax Identification:
Mainland China:
- Individual: Uses ID card number
- Business: 统一社会信用代码 (18-character)
Taiwan:
- Individual: 身分證統一編號
- Business: 統一編號 (8 digits)
Regional Compliance
Chinese-speaking regions have distinct privacy frameworks:
Mainland China:
- Personal Information Protection Law (PIPL) - 个人信息保护法
- Cybersecurity Law - 网络安全法
- Data Security Law - 数据安全法
- Cross-border data transfer restrictions
Taiwan:
- Personal Data Protection Act (PDPA) - 個人資料保護法
- Requirements for data collection consent
- Cross-border transfer notifications
Hong Kong:
- Personal Data (Privacy) Ordinance (PDPO)
- Data Protection Principles
- Cross-border data flow considerations
Mixed Language Processing
Business documents often mix Chinese and English:
Example mixed text:
"请联系John Smith先生，电话：+86 138-1234-5678，
邮箱：[email protected]，
地址：北京市朝阳区CBD核心区Building A, Suite 1502"
Detected PII:
- English name: John Smith
- Chinese phone: +86 138-1234-5678
- Email: [email protected]
- Mixed address: 北京市朝阳区CBD核心区Building A, Suite 1502
Our system seamlessly processes mixed content, applying appropriate detection rules for each language while maintaining context across the document.
API Usage
Specify Chinese language processing in API calls:
POST /v1/redact
{
"text": "客户姓名：李明华，身份证号：110105199003071234",
"language": "zh",
"region": "CN",
"redaction_types": ["name", "national_id", "phone", "address"]
}
Response:
{
"redacted_text": "客户姓名：[NAME]，身份证号：[NATIONAL_ID]",
"detections": [
{
"type": "name",
"value": "李明华",
"script": "simplified",
"confidence": 0.96
},
{
"type": "national_id",
"value": "110105199003071234",
"format": "china_resident_id",
"valid_checksum": true,
"confidence": 0.99
}
]
}
