Text Classification Model for Public Opinion Management in Colleges and Universities Based on Improved CNN

Abstract

In recent years, the incidence of online public opinion in universities has been high. Monitoring, forecasting and responding to public opinion in colleges and universities have increasingly become the work that education management departments at all levels and universities attach great importance to. The purpose of this article is to build a network public opinion management text classification model based on a convolutional neural network, mainly by analyzing its internal mechanisms and models in an attempt to find new laws of public opinion generation and dissemination. The main methods used in this paper are three-layer Bayesian model and convolutional neural network on LDA text, and describe the text classification process. The experiments used are mainly through the existing Weibo review data in the public data set as training and verification data and the Weibo data crawled in real time in the web version of Weibo as raw data. Finally, the data is preprocessed to sort out the text review information. After analyzing the data, it is concluded that the LDA + improved TextCNN model proposed in this paper is improved overall than the original TextCNN model, and the LDA + improved TextCNN model in food has decreased compared with the TextCNN model.