[Purpose/significance] This study aims to explore the methods on extracting and visualizing users’ emotions from unstructured user-generated content, analyze user-generated content from a perceptual level, and discuss the related application prospects. [Method/process] The research took book reviews of Douban as analysis object. Emotional dictionary in Chinese domain and LDA latent topic model were used to refine the fine-grained emotional elements. And further, visualization techniques helped to analyze the emotional elements reflected in the review content. [Result/conclusion] The study found that both latent topic model and emotion dictionary can effectively extract the user emotion elements in the content of the review, even though some difference still exists, such as the emotional topic model can provide more exquisite results. By fine-tuning the application scenario, the methods used in this study can be applied to various forms of perceived utility mining tasks about reviews, like experience-based products recommendation.