A Deep Learning Framework Based on LSTM and Transformer for Enhancing Cultural Adaptation in Ethnic Music Education
Kai Song · 2025
National music education faces challenges such as single content design and insufficient cultural adaptability in the field of cultural inheritance and personalized teaching. Particularly in educational scenarios involving multi-ethnic and multi-cultural backgrounds, traditional teaching methods and generalized technologies struggle to effectively meet students' diverse needs. To address these issues, this study proposes a dynamic teaching content optimization framework based on deep learning, combining Long Short-Term Memory (LSTM) and Transformer models to model students' learning behaviors and cultural background characteristics. By dynamically generating teaching content tailored to individual needs and cultural contexts, the framework aims to improve students' learning outcomes, cultural identity, and classroom participation. The experiment selected college students from Sichuan and Yunnan provinces, focusing on key ethnic groups such as Tibetan, Qiang, Yi, and Miao, with a total of 200 students involved. The results show that compared to traditional teaching methods and deep learning approaches without cultural adaptation, students in the framework group demonstrated an average improvement of 38% in academic performance, a cultural identity score of 4.8 (out of 5), and a 52% increase in classroom participation.