Implementasi Algoritma Damerau-Levenshtein Distance dan Pendekatan Deep Neural Network Dalam Analisis dan Perbaikan Kesalahan Eja atau Penulisan Real-Word Error Bahasa Indonesia

Abstract

Effective communication, both verbal and written, is an important aspect of human life to convey information and build harmonious relationships. In written communication, errors in writing, such as non-words (words not found in the dictionary) and real-words (words found in the dictionary but used out of context), can hinder message comprehension. Although technology has made the writing process easier, the challenge of detecting and correcting writing errors, especially those related to context, is still common. This research aims to design a system that is able to identify and correct real-word errors in Indonesian text using the Damerau-Levenshtein Distance algorithm and the Deep Neural Network (DNN) approach. The Damerau-Levenshtein Distance algorithm calculates the edit distance by considering deletion, insertion, substitution, and transposition operations, making it very suitable for overcoming typos. Meanwhile, the Deep Neural Network approach is used to assess the suitability of candidate words based on the overall sentence context, where the candidate with the highest suitability will be selected as the correction result. The system successfully achieved an accuracy rate of 98.56 % and a word correction success rate of 94.17 %. The results of this study prove that the combination of the Damerau-Levenshtein Distance algorithm and the Deep Neural Network approach has a good ability to detect and correct word errors, especially real-word errors