Research On Network Public Opinion Monitoring Based On Topic Model

The development of the Internet makes our world has become a global village, and the convenience of the network makes people can obtain information from the network at anytime. With the application of Web2.0 Technology, people no longer simply to obtain information on Internet, but also as the creator of the network information, people write their views and ideas, express their feelings on the Internet. The information on the Internet is exploding, in the face of so much information people may be at a loss what to do, at the same time, it also presents a challenge for our government. Because of the real-time network, speed of information propagation is very fast, one thing that will soon be able to be a wide range of Internet users know, especially the event that about the livelihood of the people, will be more aroused the attention of visitors. Some have an ulterior motive, deliberately spread false, sensitive information in the network, large scope spread by Internet marketer, is very easy to become network public opinion. Often, these information are bad for social harmony and stability. So, it is very necessary to be able to monitoring the information on the network effectively, and grasp the network public opinion information in a real-time.Both native and abroad has research about network public opinion. In abroad, attention been paid to network public opinion earlier, and have mature monitoring system and mechanism now. But started late in our country, it caused the attention of the government in recent years, and became a research hotspot. The domestic research has also made some achievements, some monitoring systems of public opinion on the Internet for real-time monitoring have been designed. There are two main ways used in these systems, one is based on sensitive word matching, anther is based on word frequency statistics method. Both of these two ways has achieved a certain effect, But there are certain defects, both of two ways are not based on the understanding of the semantic information of the text, result in deficiency about accurate detection of public opinion information. So in this article, do a depth study on how to improve the accuracy of public opinion information detection, try to apply the topic model to public opinion monitoring system, use the advantages of topic models for semantic mining to make up for the deficiencies of traditional methods in the public opinion monitoring. The main content of this paper are:1) A detailed analysis of technologies that used by traditional public opinion monitoring system, such as information acquisition technology, information preprocessing, text segmentation technology, text representation model, text clustering and so on. At the same time, researched the deficiency of public opinion analysis based on word frequency statistics.2) Propose a method of network public opinion based on LDA topic model. By modeling on text set using the LDA, find the latent topics of the texts, improving the accuracy of public opinion analysis.3) Based on the theoretical analysis, design and implement a network public opinion monitoring system of campus. The system has functions about public opinion search, daily public opinion, hot public opinion information and so on.