Research On The Techniques Of Online Public Opinion Monitoring Based On Topics

The rapid development of the Internet has accelerated the informatization process of the society, making the exchange of information and feedback between people more direct, convenient and free, but also provided a shortcut for the rapid spread of false and negative information. Therefore, real-time and effective monitoring of public opinion is needed to improve the strategic decision-making and to prevent the devastating effects of bad information.Based on the study of domestic and foreign technology of public opinion, this paper emphatically introduces the advanced technology of topic detection and sentiment analysis, proposes a topic-based online public opinion monitoring method. The method can crawl topic-related web pages by the restriction of domain rules and URL rules. Besides, the method proposes a novel topic detection method with improved Single-pass clustering algorithm using the time characteristics of news reports. Finally, based on the expansion of the emotion dictionary and the modification effect of different terms, a sentiment analysis method is proposed. Through the information collection and data analysis of actual news portal, this method can capture real-time topic-related news and overall emotional attitude and distribution.This method has the following advantages. Firstly, a topic-based web crawler is realized which can reduce the excessive interference of irrelevant data. Secondly, based on the time characteristics of the news story, an improved online incremental clustering algorithm is proposed to overcome the input sequence sentivity problem of the traditional incremental clustering algorithm. Thirdly, based on the combination of the synonym words and the point mutual information method, a dictionary-based sentiment analysis method is achieved. The method can make full use of the modification effect of different features, and can minimize the misjudgment of sentiment plority by taking the relative position information of features into account. Theoretical analysis and experimental comparison show the feasibility and effectiveness of these methods.