Analysis Of The Comparison Between Jaro-Winkler And Levenshtein Distance Algorithms For Indonesian Language Error Checking In Theses Of Politeknik Negeri Ketapang Students
Novi Indah Pradasari, Darmanto, Ar-Razy Muhammad

Politeknik Negeri Ketapang

Abstract

This research analyzes the performance comparison between the Jaro-Winkler and Levenshtein Distance algorithms in detecting spelling errors in Indonesian-language theses written by students of Politeknik Negeri Ketapang. The growing need for automated language error-checking systems, particularly in academic writing, drives the exploration of these algorithms^ effectiveness in identifying misspellings. The study utilizes a dataset of student theses containing various types of annotated spelling errors to assess both algorithms. Key performance indicators include accuracy, speed, and sensitivity to different error types. The Jaro-Winkler algorithm emphasizes phonetic similarity, particularly for errors occurring at the beginning of words, making it suitable for detecting errors in words that are phonetically similar but incorrectly spelled. Meanwhile, the Levenshtein Distance algorithm calculates the minimum edit distance between words, allowing it to excel in identifying typographical errors. Experimental results show that each algorithm has specific strengths: Jaro-Winkler is more effective for phonetically-based errors, while Levenshtein performs better for minor typographical errors. This comparison provides insights into the potential integration of both algorithms into Indonesian language error-checking tools to improve the accuracy of automated systems for academic writing correction.

Keywords: Please Just Try Jaro-Winkler, Levenshtein Distance, spelling errors, algorithm comparison, Indonesian language, academic writingto Submit This Sample Abstract

Topic: Artificial Intelligence (AI)