Research On Analysis And Monitoring Of Internet Public Opinion About Food Safety

Food safety is a social problem which related to public safety. In recent years, food safety events happen almost every year and these events have attracted the public’s attention to food safety since they spread quickly and affect many people. Food safety incidents triggered a large number of Internet public opinion, so it is important to do the research of public opinion of food safety events. At present, data on the Internet increase rapidly, the Internet is becoming more and more diversified and complicated. It’s necessary to improve the speed of topic detection without improving miss and false detection and monitor topic combining multiple data sources. The research contents in this paper are:the improvement of topic detection and tracking, the research of topic monitoring, the improvement of sentiment analysis of topics and the design and implement of the Internet public opinion monitoring system about food safety. The details are as follows:1. In the process of topic detection based on Single-Pass, clustering speed and precision may decrease rapidly for growing topics. In order to improve the speed and accuracy of topic detection, this paper improved Single-Pass combing food safety. The improvements are:set clustering strategy, improve text similarity measurement and add a topic obsolete strategy.2. At present, topic detection and monitoring usually based on single data sources, this paper monitor food safety topics in the perspective of multiple data sources. In order to combine news data and microblog data effectively and provide data support for topic warning, the IRI Internet public opinion index system is researched.3. Microblog sentiment analysis is an important part in public opinion about food safety. Microblog vectors are relatively sparse after vectorization, which will lead to poor result for sentiment analysis approaches based on machine learning. In order to reduce the effect of spares vector and improve the result of microblog sentiment analysis, semantic-based sentiment analysis is researched. In semantic-based sentiment analysis, semantic eventually embodies in word similarity measurement. In order to improve the accuracy of word similarity computation, word similarity computation method based on the Latent Dirichlet Allocation (LDA) model is researched.4. The topic detection and tracking method, topic monitoring method and semantic-based sentiment analysis above are applied to the food safety Internet public opinion monitoring system. A food safety Internet public opinion monitoring system is built, the system’s run result on real-world web data proved its effectiveness.