<![CDATA[Manual registration processes for migrant workers present significant operational challenges, requiring 10-15 minutes per person with data error rates reaching 15%, hindering service efficiency in protection organizations. This research addresses these challenges by developing a domain-specific Optical Character Recognition (OCR) system optimized through multiple pattern matching algorithms tailored for Indonesian identity documents. Unlike general-purpose OCR approaches, the system implements six pattern variations for RT/RW field extraction and three hybrid strategies (direct, fuzzy, and contextual matching) for occupation fields, specifically designed to handle format inconsistencies in KTP and KK documents. Testing with 50 document samples achieved variable accuracy rates ranging from 75-95% across different field types, with the multiple pattern approach demonstrating 30.8% improvement over single-pattern methods for RT/RW fields and 20% improvement for occupation fields. Real-world deployment at Migrant Care Jember produced measurable operational improvements: 67% time reduction (12 to 4 minutes), 80% error reduction (15% to 3%), and threefold service capacity increase without additional personnel. The integrated confidence level system with visual indicators (green/yellow/red) enables non-technical users to identify fields requiring verification, enhancing practical usability. This study demonstrates that domain-specific pattern matching optimization can effectively bridge the gap between theoretical OCR advancements and practical implementation challenges in resource-constrained organizational settings, with direct implications for migrant worker protection services.]]>
Copyrights © 2026
