Predicting E-Learning Resource Quality Based on Multi-Modal Data

While the open resource construction mode brings more autonomy to learners, it is also very likely to lead to the problem of uneven quality of resource content and bring considerable difficulties to resource management. Learners' comments based on e-learning resources are a true reflection of the quality of resource content, and through text analysis using natural language processing technology, we can analyze learners' likes and dislikes of resources and reveal the quality problems of resource content, so as to provide targeted early warning feedback and intervention, and ultimately improve the quality of resource content and management effectiveness. Based on this, this study mainly focuses on several aspects of the research content: the construction of e-learning resources content quality crisis early warning model. On the one hand, the content quality evaluation index system of e-learning resources is established. On the other hand, deep learning algorithms in natural language processing technology are used to construct ALBERT+TextCNN sentiment multi-classification model and ALBERT+Seq2Seq+Attention text multi-label classification model based on learners' comments, respectively, to depict the whole process of early warning of e-learning resources content quality crisis in early warning identification and early warning evaluation.