Behavior-Aware English Reading Article Recommendation System Using Online Distilled Deep Q-Learning

Due to the differences of students' English proficiency and rapid changes in reading interests, online personalized recommendation is a highly challenging problem. Although some works have been proposed address dynamic change recommendation, there are two issues with these methods. First, it only considers whether students read recommended articles. Second, methods often fail capture real-time changing interests users. To above challenges, deep Q-network based framework was proposed. The authors further use user's behavior scores as reward information get more feedback. In addition, adaptive module introduced short-term on fly utilized consistent loss KL divergence distill knowledge from model. Extensive experiments offline dataset IWiLL website demonstrate superior performance method.