Javanese Script Single Letters Classification using GoogLeNet Architecture and Adam Optimizer Based on Convolutional Neural Networks (CNN)

This study examines the decline in the usage of the Javanese language, which has experienced a decrease in the number of speakers from approximately 82 million in 2007 to 68.2 million in 2015. The convergence of the Javanese script and Optical Character Recognition (OCR) technology is proposed as a solution, allowing for the preservation and accessibility of the Javanese script in the digital age. The integration of Convolutional Neural Networks (CNNs) in Javanese script classification achieved a high accuracy rate of 92.95% in identifying positive and negative cases. The dataset used for training consisted of 8440 sample images, which were divided into 20 subfolders for training and testing. The results presented in Table IV demonstrate the successful implementation of the classification model, achieving a 98.87% sensitivity, 100% precision, and 98.88% specificity. This research contributes to the preservation and understanding of the culturally significant Javanese script while addressing the decline in its usage.