Abstract: To quickly obtain information of theme and sentiment in network texts, the concept of sentiment summarization is introduced in order to meet the practical needs, and while a method of extracting the sentiment summarization is proposed based on the Conditional Random Fields model. Firstly, extract the basic features, such as sentence length, tips word and sentiment words in the sentence, while the Latent Dirichlet Allocation model is used to analyze the underlying themes of the text and extract theme features. Then these two kinds of features are used in the Conditional Random Fields model and sentiment summarization is obtained. Experimental results show that this approach has better efficiency because it takes into account the information of themes in the text, and it can meet the user's needs.