This study proposes an intelligent service recommendation system that integrates dynamic planning model and improved label propagation algorithm (OCDSLP). Aiming at the limitations of traditional methods in semantic association mining and cross-cultural interaction adaptation, a three-tier system framework based on B/S architecture is firstly constructed to realize the separation of multi-role authority management and layered business logic. The OCDSLP algorithm is then proposed to optimize the community discovery process by fusing network topology similarity and semantic similarity. The state compression dynamic planning model is introduced to filter the optimal service combination segments. Experiments show that the OCDSLP algorithm achieves semantic matching accuracy of 85.1% and 84.3% on Usedcar DB and IMDB datasets. In the English cultural communication platform application, the proposed model performs optimally over the six control models under different number of Top-K neighbors. In the MAE dimension when the number of neighbors is 20, the error value reaches the lowest, only 18.891, and it is about 15.3% lower than the next best method NMF (RMSE=62.986) at the RMSE dimension threshold of 20. The results of the questionnaire survey show that users recognize more than 80% of the system functionality and experience.
