Chinese folk music, originating from real life and grassroots communities, often features rich forms and humorous content, possessing invaluable historical, cultural, and artistic value. This study aims to promote the inheritance and development of Chinese folk music by designing a user needs survey questionnaire for a digital learning resource repository of Chinese folk music, summarizing its current development status and the challenges it faces. Given the diversity and irregularity of Chinese folk music entities, a music named entity recognition model is constructed with a main structure comprising a representation layer, a TDCNN encoding layer, a recurrent network layer, and a prediction layer, specifically for entity recognition in the field of Chinese folk music. Additionally, a link prediction model based on feature mapping and bidirectional convolution is established to achieve mapping conversion between entities and relationships, thereby uncovering their subtle connections. By integrating the two models, a folk music relationship extraction model is proposed to provide technical support for the establishment of a Chinese folk digital learning resource repository. In comparative experiments with various similar models, this model demonstrated the highest accuracy rate (99.08%) and F1 score (97.17%), showing high compatibility with the requirements of Chinese folk music relationship extraction tasks.
