Topic Visualization Based On Multi-Source Text Corpora

Different groups in society may have different opinions on an event,which influence each other as the event develops.It is of great significance for managers such as government agencies,news media and large enterprises to understand the impact of public opinion communication to explore the changes of different groups of opinion flows on the network and the influence of news media or government agencies on the public.We propose a topic transition model to understand the opinion flow within or across social groups.Firstly,the dynamic topic model is used to analyze the opinion flow between individual groups.Then the cosine similarity method is used to calculate whether there is interaction between groups in different periods from the two aspects of time correlation and content correlation.After the above two steps,we build the theme transformation model.Then,based on agenda setting theory,selective exposure principle and Markov property,we calculate the influence of other groups on a group.We design a visual analysis system to present the data,which is mainly composed of the topic flow graph,the topic transition graph,the forwarding network slice graph and the thematic map.Topic flow graph and topic transition graph provide visual analysis at group granularity level.In order to enable users to reach the granularity level of single users,the system provides a forwarding network slice chart to show the forwarding relationship of individual tweets among multiple social groups in each time window under a single topic.In public opinion events,with the emergence of facts and disputes,personal views may change.Therefore,the forwarding network slice also provides a personal opinion time series corresponding to the time window.In addition,some major events or natural disasters have the characteristics of regional centralized transmission.So we design the thematic map to help analysts to consider the influence of geographical location on public opinion.Finally,two case studies are provided to further verify the effectiveness and availability of the system.