Thai Address Splitter

split long address string(in Thai) to 'name', 'phone number', 'subdistrict', 'district', 'province', 'zipcode'

Architecture

This library uses a class-based architecture with separated concerns:

• ThaiAddressSplitter - Main orchestrator class
• TextPreprocessor - Text cleaning and normalization
• LocationMatcher - Subdistrict database matching and scoring
• EntityExtractor - Name, phone, and address extraction
• Constants - Centralized patterns, keywords, and mappings

How It Works

flowchart LR
    A[Thai Address String] --> B[Clean Text]
    B --> C[Extract Thai Words]
    C --> D[Find Location Match]
    D --> E[Extract Entities]
    E --> F[Structured Output]
    
    subgraph "Core Steps"
        B
        C
        D
        E
    end
    
    subgraph "Data Sources"
        DB[(subdistricts.json)]
        PATTERNS[Regex Patterns]
    end
    
    D -.-> DB
    E -.-> PATTERNS

Processing Steps

1. Text Preprocessing:
   • Removes administrative prefixes (เขต, แขวง, จังหวัด, อำเภอ, ตำบล, etc.)
   • Normalizes whitespace and cleans input
2. Word Extraction: Splits text and filters Thai words with at least 2 characters
3. Location Matching:
   • Searches each word against the subdistricts database
   • Calculates frequency scores for location confidence
4. Entity Extraction:
   • Removes location words from text (zipcode → province → district → subdistrict)
   • Extracts names with title prefixes (นาย, นาง, คุณ, etc.) or without titles
   • Extracts phone numbers (starting with 06, 08, 09)
   • Remaining text becomes the address field

Example

basic split

const Splitter = require('../splitter');

(async () => {
    const input = 'คุณดอกฝ้าย สายเขียว 799/11 ถนนจักรแก้ว แขวงวังบูรพาภิรมย์ เขตพระนคร  กรุงเทพ 10200 เบอร์ 0911222333';
    const result = Splitter.split(input);
    console.log('result :', { input, result });
})();

Tests

pnpm run test