iQurNet: A Deep Convolutional Neural Network for Text Classification on the Indonesian Holy Quran Translation

The holy book of Muslims has been translated into more than a hundred different languages. In Indonesia, the translation appeared around the Walisongo era or the 15th century and was widely spread for Islamic teaching. However, the Indonesian translation has not been grouped by the main content of the verses. In this research, a deep learning model is proposed for text classification called iQurNet. Our model incorporates an embedding layer, several layers of one-dimensional convolutional layers, and is fully connected with 3 hidden layers. The dataset was collected from the website of the Ministry of Religion of the Republic of Indonesia. Prior to computing, we conduct preprocessing to remove any trivial words, special characters, and weigh the output class. We also compare some machine learning models namely Naïve Bayes, Decision Tree, Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and Neural Network (MLP). The results, in Naïve Bayes with accuracy at 66%, Decision Tree at 59%, SVM at 71%, KNN at 68%, MLP at 73 %, and iQurNet at 84%. In conclusion, our model become the best model in performance for its accuracy.