Microblog Text Classification System Based on TextCNN and LSA Model

Abstract

With the development of the internet technology, kinds of short text information are growing explosively. Information explosion has become an urgent problem. To find the target information accurately from a large number of short texts and to find the short text with the same semantic meaning of the target information, we propose and build a microblog text classification system. Take the official microblog of Peking University as an example; the TextCNN model based on Convolutional Neural Network (CNN) is used for the classification of the text of Peking University's Weibo. The text is pre-processed accordingly and converted into word vectors through unsupervised learning. TextCNN is used for training to realize the classification of Weibo text. Finally, the paper sets up a web server to interact with the users, and uses the Latent Semantic Analysis (LSA) algorithm to recommend relevant categories of Weibo based on user search content. The experimental results show that the classification accuracy of the TextCNN model used in the article is 85.94%. This paper realizes the function of microblog content classification and accurate search. We help users quickly and accurately find the content they want, reducing the time of browsing useless information.