An Online Education Course Recommendation Method Based on Knowledge Graphs and Reinforcement Learning.

This paper presents an in-depth study and analysis of online education course recommendations through a knowledge graph combined with reinforcement learning, and proposes a deep learning-based joint extraction method of course knowledge entities and relations in the education domain. This joint extraction method can extract both course knowledge entities and their relationships from the unstructured text of online courses, thus alleviating the problem of error propagation. On the other hand, since some parameters in the joint model can be shared by the entity identification task and the relationship classification task, this helps the model to capture the interaction between the two subtasks. Similar courses are judged based on the extracted course knowledge points, while course knowledge chains are generated based on the relationships between course knowledge points. In terms of user learning behavior, by analyzing user online learning behavior data, this paper uses five variables, namely the number of learning hours, the number of discussions, the number of visits, the number of task points completed, and the number of learning courses, to judge and cluster user similarity using an information entropy-based learner behavior weight assignment method. Based on the course knowledge map, this paper firstly constructs a learner model with four dimensions of basic learner profile, cognitive level, learning style, and historical learning records. Secondly, it predicts the target knowledge points of learners based on their learning data using the Armorial algorithm and maps them in the knowledge map, then uses natural language processing related techniques to find the conceptual similarity between knowledge points and proposes a deep recommendation strategy based on the knowledge graph correlations. At the same time, the recommended courses based on learners' behavioral data are more relevant and accurate, which greatly improves learners' efficiency and satisfaction in the learning process. [ABSTRACT FROM AUTHOR]