Implementation of the Jaro-Winkler Algorithm for Automatic Grading of Short Answer Questions in the Kifayatul Azhar Elementary School Exam

Abstract

Kifayatul Azhar Primary School faces significant obstacles in examination administration. High logistical costs and inefficient social media coordination hamper physical exam distribution. Existing digital tools, such as Google Forms, cannot adapt to minor typographical errors in short-answer evaluations. This research develops a web-based automated grading system using the Jaro-Winkler algorithm and the Rational Unified Process (RUP) framework. The RUP methodology manages iterative development through four stages: Inception, Elaboration, Construction, and Transition. The system engine performs text preprocessing tasks such as case folding and tokenization prior to calculating Jaro-Winkler similarity scores. Experimental results indicate that d_w≥0.8 serves as the optimal threshold for differentiating acceptable typographical variants from substantive errors. The system validation process included 26 integration test cases and 76 User Acceptance Test (UAT) cases, encompassing four distinct user roles: Admin, Teacher, Mentor, and Parent. All tests demonstrated a complete functional success rate. This system provides a centralized, automated environment that reduces administrative workload while maintaining grading objectivity.