Hanacaraka Javanese Handwriting Detection Using Recurrent Neural Network (RNN)

Abstract

Hanacaraka Javanese script is a valuable Indonesian cultural heritage, but its use has declined due to a lack of knowledge and ability to read and write the script. The main challenge in detecting and recognizing handwritten Javanese script is the variation of its shape and writing style. This research aims to train computers to recognize Javanese script. Prior to this research, there have been several similar studies with different recognition methods. In this research, the Recurrent Neural Network (RNN) method is used. The process of detection and recognition of Javanese script letters is divided into three parts: input image of script images used as a dataset of 500 images for training data and 8 pictures for prediction test data, the process of creating a Recurrent Neural Network model, and the output of this design is the performance of the Recurrent Neural Network model. The test results show that the model has an overall accuracy of 96%, with an average precision, recall, and F1 score of 96% each. Sentences such as "Dhahara", "Jawanagara", "Malaca", and "Ramayana" were successfully detected completely correctly, although some sentences such as "Jayabaya", "Nyala", and "Palawa" experienced prediction errors.

Keywords

• javanese script, handwriting detection, hanacaraka, deep learning, recurrent neural network (rnn)