Abstract: 【Purpose/significance】Explore the topic of policy text in specific fields, and reveal the field of policy deployment and fu⁃ture development trend in China.【Method/process】This paper proposes a time series modeling and visualization method of public poli⁃cy text, use DTM model to cut the time sequence and model the topic of policy text, and skip gram word embedding technology in deep learning word2vec algorithm is used to effectively predict the context vocabulary, so as to enhance its semantic expression and policy interpretation, and more accurately reveal the deployment focus of public policy in China.【Result/conclusion】Experiments show that the method proposed in this paper has better knowledge extraction and semantic expression ability for public policy topic recognition and policy text quantification, and has good revelation for public policy mining and information disclosure in China.【Innovation/limita⁃
tion】This paper proposes a temporal modeling method of public policy texts which integrates word vector semantic enhancement and DTM model, and improves the topic semantic expression of policy texts to a certain extent. In the future, deep learning technologies such as LSTM algorithm and Bert model will be considered to identify domain knowledge units and syntax structures in policies.