Visual Analysis Of Topic Transition Among Different Sources Of Text Corpora

Government or official agency usually expresses their opinions and attitude via news or other traditional media.For the same event,information from unofficial source such as social media and news comments express different opinions from the official's.The official agency needs to know whether the opinions in the news express effectively.The nongovernmental people want to know whether their opinions are consistent with the official's and whether opinions which different with the official's are responded timely.Word frequency statistics based methods and topic model based methods are widely used in topic extraction.This paper adopts topic model based methods to extract topic.The existing topic evolution analysis works rarely focused on multi sources text.This paper develops different topic analysis strategies and visualization algorithms for different text sources.This paper presents a visual analysis method that shows how the topic is transmitted and developed in different media sources over time.To model topic transition accurately,we develop an information transition model based on topic analyze.A correlated-clustering-based layout is used to visualize the topic transition model.We introduced an improved treemap method to show the multiple levels topics.We allow users to interactively locate the evolution of one topic or one word,and find the corresponding raw data quickly.A case study is provided to demonstrate the usability and effectiveness of the system.