Automatic Essay Grading System for Japanese Language Exam using CNN-LSTM

Abstract

This paper discusses the design for the development of an automatic essay grading system (SIMPLEO) using variations of the Convolutional Neural Network (CNN) and hybrid Convolutional Neural Network (CNN)-Long Short-term Memory (LSTM) for the assessment of the Japanese essay exam which is being developed by the Department of Electrical Engineering, University of Indonesia. Of the several variations tested, the most stable model is a model that has CNN-LSTM with kernel sizes of 5, the number of filters 64, pool size of 4, LSTM hidden units of 25, batch size of 50, repeated training of 50 epochs, and the SGD optimizer with a learning rate of 0.01 produces the highest prediction accuracy, which is 70.07%.