Construction of intelligent learning systems based on multimodal data fusion and adaptive optimization of personalized learning paths

This research creates a framework to solve the problems surrounding the imprecise fusion of multimodal data in intelligent learning systems, as well as the lack of dynamic adaptability in learning pathways within an intelligent learning system. The proposed framework is based on two nature of Collaboration and Knowledge-Enhanced Multimodal Alignment Network (KMAN) along with Hierarchical Curriculum Reinforcement Learning (HCRL), and provides guidance in the construction of the intelligent learning frameworks. In the construction of the KMAN, domain knowledge is embedded into the model via a “knowledge graph” as part of the feature alignment between multiple modalities and to guide the creation of user-defined states. The empirical evidence obtained from the application of this research indicates that the final score of the learners was 89.5% with a retention of knowledge for 91.8% and a mastery time of only 10.4 min for each knowledge point. Overall, the intelligent learning system constructed in this study supports dynamic adaptive knowledge structure coverage through the integration, alignment, and utilization of multimodal data sources. It should be noted that the current framework has been validated in a structured dataset environment; its generalizability to real-world classrooms and diverse learning contexts warrants further investigation.