Text Classification on Islamic Jurisprudence using Machine Learning Techniques

Indonesia, which is the world's most populous Muslim-majority nation, Islamic education is not an alternative to secular education but compulsory to any Muslims learn the way of life. The two primary sources of the religion of Islam is the Quran and Hadith. These two are where the majority of the teachings come from. However, when looking for guidance, a Muslim often refers to Islamic scholars to interpret the verses and hadith for them and educate themselves on a topic. However, due to many circumstances, people could not meet and ask a scholar personally regarding Islamic jurisprudence. Chatbot is one of the solutions. Chatbot offers users new opportunities to improve the learning and engagement process, reducing the typical cost of face-to-face consultation. However, to train a chatbot on Islamic Jurisprudence, a text classifier is needed to build a strong knowledge base for the chatbot. This study adopted a standard methodology for building a text classification model. This study used 600 Islamic jurisprudence text data obtained from the books and the web written by an influential Islamic scholar named Ustadz Abdul Somad (UAS). Machine learning algorithms such as Bayesian Network and Naïve Bayes, were employed to classify the text data. Based on the experimental testing results, the Naïve Bayes algorithm is more accurate in all evaluation models, 84.25% using training set, and 76.54% using 10-Fold cross-validation. Meanwhile, the Bayesian Network algorithm is faster in terms of time taken on all evaluation models. Thus, it can be concluded that the text classification model using Naïve Bayes and String to Word Vector filter have the potential to be used effectively but still has plenty of room for improvement.