Research And Implementation On Key Techniques Of Online Public Opinion Analysis

Nowadays, internet has characteristics of time-sensitive and interactive, which make it an important way to get information for people. Internet therefore has a tremendous ability to lead public opinion and to influence the audience. Online public opinion has become an important part of social public opinion and plays a more and more important role in society. The Internet has characteristics of virtual, liberty, interactivity, timeliness real-time, leading to online public opinion has characteristics of direct, busty, deflection. We face a vast amount of web pages which grow tremendously every day. It is hard to identify online public opinion artificially. In order to get online public opinion more quickly and accurately, this is the reason why much attention has paid to online public opinion analysis.This paper first analyzes the research situation of online public opinion analysis in the domestic and foreign, and then introduces the related theory of online analysis. On the basis of the above work, this paper studies the key technologies of online public opinion analysis and the research work includes the following five areas:(1)Based on the analysis of various news representation models, we using vector space model as our news representation model, and in the experiment we tested the relationship between the word characteristic dimension and the accuracy and time consumption of topic detection.(2)We propose our algorithm selection strategy for topic detection in the internet, and on the basis of the analysis of various clustering algorithm, we use topic detection model based on BIRCH clustering algorithm. The experimental show that when we use the topic detection model, its precision and false positive meets the application requirement.(3)We propose a topic tracking model based on multiple features. Only each of multiple features is more than the threshold, we consider the topics are the same topic. The topic tracking model can validly distinguish the same topic from the similar topic.(4)We analyze the characteristics and the development cycle of the hot topics, and then calculate the heat of topics according to the continuous day, the news documents in the topic, the number of users, the number of comments. The experiments show that the method can effectively analyze the hot topics in the internet.(5) On the basis of the research work, we design and implement the online public opinion analysis system. Our system meets the application requirement. It can accurately detect and track the topics in the internet.