Chinese Comments Classification and Hotspot Mining Algorithm based on Natural Language Processing

With the popularity of the Internet, more and more people start to leave comments in the comment sections of major platforms to reflect their thoughts, and these comments are characterized by large amounts of information and uneven information quality. In order to distinguish the comment categories more effectively and obtain the hotspot information of comments, this paper adopts the TF-IDF method to vectorize the comment text, and uses the SVD algorithm to reduce dimension of the TF-IDF matrix. We use the LinearSVC model to classify comments, and construct a comment hotspot mining algorithm, which is based on the K-Means algorithm to cluster the comments. The topic words of the clustering results are extracted using the LDA topic model, and the hot words are analyzed by TOPSIS algorithm based on hotspot evaluation indicators. The experimental results show that the accuracy of comment classification can reach more than 90% and can identify the hotspot information of comments.