A Study Of Event Evolution And Public Opinion Mining Based On News Topics

News topic public opinion mining is to analyze large-scale news events relationship using data mining techniques and to display by visualization techniques.Through public opinion mining,we can provide a convenient way for users to know the development of news topics and the public opinion shifting.The news topic public opinion mining includes three research fields:event evolution,event ranking and event sentiment analysis.While reading news about a specific topic,people are interested in the whole picture of a topic's evolution.It is a hotspot of current researches about how to describe the development of a news topic.We analyze the relationship between events from three different aspects,including Co-occurrence Dependence Analysis,Event Migration Analysis and Time Distance Analysis,then generate the Event Evolution Graph through combining three kinds of relationships.Event Ranking is to evaluate the importance weight of events related to a topic.Generally,a hot topic maybe consist of many events.For a reader,he/she pays more attentions to those important events which is related to the development of topic or has a significant impact on public opinion.We propose a new method to evaluate the importance weight of events automatically.Based on the results of event evolution,combining the public attention and the media attention,we use HITS to calculate the importance weight of events.News readers may be interested in not only the event itself,but also the public opinion.In the past,most studies analyze sentiment on the sentence granularity.However,a news review often refer to more than one name entities,and the sentiments may be different.Therefore,this paper presents an entity-based approach to do sentiment analysis.What's more,we analyze the sentiment shifting for each entity of an event based on entity-based sentiment analysis.Through event sentiment shifting analysis,we discover some special events to supplement the Event Evolution Graph,including public opinion reverse event,one-sided event and disputes event.We conduct experiments to show that our proposed method can bring convenience for people to know the development of the whole news topics and track the public opinion shifting.