A Framework for Adaptive Learning and Context-Aware Knowledge Enhancement: An Art Education System Based on DeepSeek

Digital platforms have overtaken galleries in art dissemination, yet static formats limit personalization and engagement across diverse audiences. This paper presents an DeepSeek-based conversational system integrating dynamic user profiling and knowledge-enhanced dialogue generation. It features a profile extraction module for real-time interest tracking via multimodal fusion and an LLM-driven (DeepSeek-driven) module balancing personalization and fidelity through dual-constraint optimization. Employing neural differential equations, hierarchical attention, and three-phase joint training, the system fosters bidirectional cognitive interaction, surpassing traditional recommendation logic. Experiments on the Art_GenEvalGPT dataset (13,870 dialogues) show superior performance (BLEU-4: 0.38 vs. 0.31, session turns: 9.4 vs. 5.8, p $\lt 0.01$). A $1,260-$ participant study confirms enhanced engagement ($83 \%$ report deeper thinking) and adaptability (profile matching: 0.82). This framework transforms art dissemination into active meaningmaking, supporting digital cultural preservation, though “information cocoons” and young user challenges suggest future refinements.