多源媒体文本主题演变的可视分析

Visual Analysis of Topic Transition among Different Sources of Text Corpora

传统新闻媒体和新兴社交媒体对同一事件的观点有差异,且互相影响.为了有效地分析不同来源的新闻事件评论主题的异同以及随着时间推移互相影响的转移趋势,提出一套可视分析方法.首先提出一个基于主题分析的信息传递模型,使用LDA模型提取文本主题,通过计算主题的内容相关性和时间相关性得到主题间的传递关系;然后使用基于桑基图和时间线技术相结合的方法对主题传递模型进行可视化,并提供主题层级视图、关键词词距视图和原数据视图帮助用户理解主题;最后研发了可视分析系统原型.使用"韩国部署萨德事件"这一案例,验证了系统的可用性和有效性.

Traditional news media and social media may convey different perspectives and topics may affect each other in the same event.This paper presents a set of visual analysis methods to analyze the similarities and differences in comments of event on multi text sources and the topic transition pattern over time.First,an information transition model based on topic analysis is proposed.LDA model was used to extract topics and the transmission relationships between topics are analyzed by content correlation and temporal correlation.An approach which combine Sankey graph and timeline technique is applied to visualize the topic transition model.The topic hierarchy view,word distance view and raw data view are used to help users comprehensive understanding the topics.Finally,this paper provides a visual analysis system prototype.The event of“South Korea deployed THAAD”is provided to verify the usability and effectiveness of the visual analysis system.