Web-Based Writing Learning Application of Basic Hanacaraka Using Convolutional Neural Network Method

Abstract

The Javanese script, known as Hanacaraka, or Carakan, is one of the traditional Indonesian scripts developed and used on the island of Java. The government's efforts to preserve the use of Javanese language and script by making Javanese a compulsory subject of local content at the education level in Central Java and East Java. In the basic competence of writing, the Javanese script has a complicated shape so that students have difficulty writing and recognizing Javanese script writing. Through this research a web-based basic Javanese writing learning application was designed that can recognize handwriting digitally which aims to help learn basic hanacaraka writing for beginners, especially students at the basic education level in Central Java and East Java. Handwriting Recognition is a system that can recognize handwritten characters and convert them into text that can be read and understood by machines or computers. The handwriting recognition process in this study uses the Convolutional Neural Network (CNN) algorithm which has the capability and ability to recognize patterns in images. Based on the tests that have been carried out between the two architectural models that have been made, the performance of the CNN model that will be used from various experiments has an accuracy of 98.29% and a loss of 0.0746 on the training data. As well as producing an average accuracy value of 99.52%, an average error rate of 0.48%, an overall accuracy of 95.03% and an overall error rate of 4.97%.