Abstract:In the mining of comments on tourist attractions, horizontal comparison of multiple attractions is often used as the research entry point, for horizontal comparison between attractions and for tourists to choose attractions, but less in-depth analysis of single attractions, for the single element of attractions to improve the service accurately. This article takes the Lingering Garden as an example, constructs clusters according to constituent elements, and conducts sentiment analysis of clustering of whole and sub-elements based on domain dictionary. The results showed that the positive emotions related to the "mountain stone" elements in the Lingering Garden accounted for 66%, which was lower than the average positive emotions of the sub-elements by 78.3%. It can be seen that cluster analysis based on garden constituents can help accurately extract sentiment analysis of Internet reviews. This method provides an easy-to-operate and more accurate theory and method for optimizing and improving gardens and other tourist attractions.