A model construction of personalized teaching of English writing based on reinforcement learning

Abstract

By constructing a deep knowledge tracking model based on reinforcement learning theory, students can choose more suitable courses or exercises when selecting learning resources. In view of this, the article proposes a temporal convolutional knowledge tracking model (TCKT-FI) based on reinforcement learning, which obtains students' knowledge mastery state through temporal convolutional network, solves the long sequence dependency problem by using causal convolution and inflationary convolution, and solves the gradient vanishing and exploding problems by using residual network to deal with the deep network structure. Following that, a recommendation algorithm framework based on Deep Reinforcement Learning (DRR) is proposed, and better recommendation strategies (Actor) and value functions (Critic) are explored. Finally, the model's English writing knowledge tracking effect is tested and its performance is verified through experiments. The knowledge tracking effect test experiment shows that at moments 1.1~5, 5.8-9.5, 10.5-15, when students did not learn the knowledge point, the TCKT-FI model shows a significant downward trend, indicating that students have forgotten the knowledge point, but the comparison model fails to simulate the forgetting process of students well. This confirms the accuracy of the model and the efficiency of its calculation. It has practical value in the personalized teaching of English writing.