Research on knowledge query network model with self-attention mechanism
Gang Wu, Yan Cheng · 2022 3rd International Conference on Education, Knowledge and Information Management (ICEKIM) · 2022
As an important research area for providing a personalized education platform to students, knowledge tracking (KT) aims to model the learner's knowledge state and predict the probability of the learner's answer to the next question. In recent years, KT models based on deep learning outperformed all traditional KT models because of their ability to capture complex representations of human learning. After combing through the literature on the knowledge tracking model based on deep learning, we combine the advantages of knowledge query network (KQN) and self-attention knowledge tracking (SAKT) to propose a deep knowledge tracking model of a self-attention knowledge query network (SAKQN) that introduces a self-attention mechanism. The introduction of the self-attention mechanism in KQN can not only retain the ability to model sequences, but also enhance the correlation of different positions of a single sequence to calculate the representation of the sequence to obtain more accurate key characteristics of the internal key characteristics of the learner's historical question record. On four public datasets, we have carried out experiments to show that the SAKQN model proposed in this paper has the highest average AUC value on the 4 datasets of 86.39%, which is 6.18% higher than the KQN model and 1.93% higher than the SAKT model.