网络舆情事件的话题演化分析——以成都女司机为例

A Subtopic Detection Method of Specific Events for Network Public Opinion:Taking News about a Female Driver in Chengdu as Example

[目的/意义]基于网络舆情演化具有的衍生性和动态性的特点,为更好地跟踪具体舆情事件的子话题衍生现象,分析网络舆情事件的话题演化过程。[方法/过程]建立话题演化的全生命周期模型和基于层次的话题表示模型,在分析特征词分布特点的基础上,建立特征词与话题关联度的判别函数,同时改进了single-pass聚类算法,建立子话题识别的两个阶段,并以成都女司机为例,通过子话题识别分析其话题演化过程。[结果/结论]实验结果表明,该方法能保持较高准确率,与实际情况较吻合,而从全生命周期模型来分析话题演化更能体现一般舆情事件的发展历程。

[Purpose / Significance] The aim of this thesis is to improve the performance of tracking the derived phenomenon of specific network public opinion subtopic. Based on the derivative features and dynamic characteristics in the evolution of a topic. [Method /Process]This paper establishes a full life- cycle model of the topic evolution and a hierarchy- based topic representation model. On the basis of the distribution characteristics of feature words,we establish discriminant function to obtain degree of relevance between feature words and topics. The single- pass clustering algorithm is improved at the same time. As a result,we propose a tw o- stage detecting subtopics method. The news about a female driver in Chengdu is taken as an example to verify the method. [Result / Conclusion]Results of the experiment are highly consistent with the actual situation which show s that the proposed method can maintain a high accuracy rate. Besides,using the full life- cycle model to analyze the topic evolution process can better reflect the evolution of public opinion in general events.