Content-aware cloud-based personalized recommendation for digital cultural services in rural revitalization

Digital cultural services are playing a growing role in enhancing cultural access and supporting sustainable rural revitalization. However, delivering effective personalized recommendations in such environments remains difficult. User-service interactions are often very sparse, cultural services carry rich but heterogeneous semantic information, and the system must scale well in cloud settings. To tackle these challenges, we propose CACR - a content-aware cloud-based personalized recommendation framework. The model jointly learns from user interaction signals and the semantic features of cultural services within a unified cloud architecture. Specifically, we introduce a content encoding module to capture service semantics and design an adaptive fusion mechanism that dynamically combines content representations with collaborative preference learning. We further employ a pairwise ranking objective to optimize recommendation quality. Experiments on the Yelp Open Dataset demonstrate that CACR consistently outperforms BPRMF, NeuMF, DeepFM, and LightGCN across Precision@10, Recall@10, HR@10, and NDCG@10. These gains suggest that incorporating service-side semantic features can meaningfully compensate for sparse interactions. Moreover, the additional computational overhead from content encoding and fusion stays moderate relative to the accuracy improvements, indicating practical feasibility for cloud deployment in rural digital cultural service scenarios.