Research On Application Of Short Text Classification In Network Public Opinion Analysis

With the development of the Internet,the network has become an important platform for people to exchange ideas and express opinions.Because of the occurrence of network public opinion and the composition are complex,which has an undeniable impact on social stability and national security,it is even more important to analyze online public opinions.Internet new media tools have become an important tool for the party,government and enterprises to manage public opinion guidance and promote innovation in social governance.Due to the complex and diverse content on the network,only using the classification can satisfy users' needs for network public opinion analysis.Public opinion information on the Internet mostly appears in the form of texts.In the process of analyzing the public opinion on the Internet,text categorization is an important technology.There are two types of text categorization: long text categorization and short text categorization.At present,there are more mature algorithms for long text classification,but short texts have the characteristics of short space and sparse features.The traditional text classification method is ineffective,and there are few researches on short text classification and no real suitable Short text classification method.Therefore,the study of short text classification technology has important practical significance.This article is based on the short text classification technology in the network of public opinion analysis of the practical application as the main research content.By analyzing the problems of high vector dimension and matrix sparseness when the short text is expressed in the traditional vector space model,aiming at the improvement of text representation in short text classification.Firstly,the corpus is modeled using the LDA topic model,and the Gibbs sampling algorithm is used to reason the parameters,the potential topics of the short texts are fully tapped,and then the texts are replaced by the topic words instead of short texts,which achieves the goal of dimension reduction and better The document features are presented.Then the SVM classification algorithm is trained on the obtained "document-topic" distribution matrix,and the classifier is constructed for classification.Finally,the experiment environment is set up and the comparison experiments are designed,and experimental results are analyzed.The results show that using the LDA topic model is feasible for the improvement of text representation in short text classification and the classification accuracy is obviously improved,which ver-ifies the validity of the proposed method.Based on the above research,this paper designs the network public opinion analysis system,establishes the framework of public opinion analysis system,designs the system function module and finally realizes the application of short text classification in the network public opinion analysis.At the end,the main function interface of the system is concentrated,and the performance test of the system is carried out with specific test cases.The result of performance test shows that the system has good practicability and interactivity.