Development of Text Classification Based on Difficulty Level in Adaptive Learning System using Convolutional Neural Network

Text classification is advantageous and necessary in a variety of fields. The adaptive learning system is one of the sectors that requires text classification. Adaptive learning is a type of online learning that makes learning recommendations based on the needs of the students. The learning module and the testing module are the two aspects of adaptive learning. The purpose of the learning module is to give recommendations of learning materials. The testing module, on the other hand, is used to assess and test the student's progress. The Learning Module requires text classification based on difficulty level to ensure that students have access to learning resources that are appropriate for their ability level. The dataset used in this study is the ebook of Science for junior high school, and we also crawling from RuangGuru (one of the E-learning platforms in Indonesia). The method that used in this study is CNN that combines with word embedding techniques. The dimension of Glove and Word2Vec are 100 dimensions. CNN-Word2Vec performs better than CNN-Glove with 96% accuracy and 61.20 seconds of computing time. Based on this, CNN-Word2Vec is appropriately applied to Adaptive learning that requires accuracy and fast computing time.