With the rapid growth of Internet consumption, the various product comments' form and redundant information are not convenient for the customers to grasp the hot opinions of the historical comments. In view of this, this paper studies the hot opinions of the products' comments and takes the hotel comments data as the main research objects. We filter the comment data from the length of the comments and the feature selection aspect by analyzing the characteristics of hotel comment data. We construct the mathematical model for the processed data and then adopt the affinity propagation clustering algorithm to extract the final hot opinions. Compared with the original comments, the experiment results of the hot opinion extraction are more concise and clear expressed.