Constructing Chinese online course learning system based on deep learning

Aiming at the core problems of online education in colleges and universities, such as unidirectional teaching, interactive asynchronism and evaluation linearity, this study proposes a deep learning-driven course recommendation system based on gated loop unit. By constructing a time-sensitive modeling framework for student behavior, the hierarchical attention mechanism and conversational interest extraction technology are integrated systematically and innovatively, breaking through the bottleneck of long-term dependence modeling of traditional recommendation models. At the level of algorithm design, GRU architecture is used to capture the dynamic learning track of cross-semester, and the adaptive integration of long- and short-term interests is realized through the collaborative control of update gate and reset gate. Experiments show that the AUC index of the proposed model is improved by 6.9%-7.5% compared with the traditional collaborative filtering algorithm and deep recommendation model, which verifies the effectiveness of time series modeling for feature extraction of educational data. Through B/S architecture and cognitive load optimization design, the system forms a closed-loop cognitive service system of "data acquisition - feature extraction - dynamic recommendation", which provides an innovative technical path for solving the quality dilemma of online education.