Public digital cultural resources have the characteristics of large amount, complicated classification, and strong homogeneity. It is difficult for users to efficiently find resources of real interest in massive resources. It is a key technology to solve the above problems that personalized recommendation can capture user's interest and actively recommend favorite resources. This paper addresses the problem of high sparseness of user cultural behavior data and the rapid changes in user cultural interest encountered in the traditional collaborative filtering approach in the public digital culture sharing service. Based on the characteristics of the semantic analysis of public digital cultural resources and the characteristics of the recommended algorithm, two optimization methods for collaborative filtering recommendation are proposed. The effectiveness of the proposed method to solve the above problems is verified by experiments.