话题演化挖掘研究可以准确完整地获取新闻话题动态演化各个阶段的话题内容，帮助用户理解新闻话题的来龙去脉以及话题内容之间的相关性和差异性，因此在网络新闻检索、网络舆情监控、互联网突发事件检测与应急管理等方面具有十分重要的作用和应用前景。现有工作由于缺乏对话题特征随时间发展而动态演变的深入分析，仅仅采用均值泛化的思想去增量扩充演化中的话题特征，引入大量话题无关信息，影响了话题关联的准确率，从而导致最终话题演化挖掘结果的偏斜。因此，针对以上问题，文中通过引入话题特征演变特性，提出一种针对话题演化的特征计算模型，在此基础上利用已有话题相关文档和最新文档进行话题信息动态增量扩充，通过对话题特征进行正向融合以及逆向过滤完成对特征信息的抗噪处理，提高话题关联的正确率，有效地解决了话题演化的偏斜问题。%The research on the topic evolution mining can obtain the topic information accuratelyand completely at all topic episodes,which is able to help users understand the cause and effect aswell as the correlation and difference of news topic.Thus,it has a very important role in WebNews Search,Network Public Opinion Monitoring,Internet Incident Detection and EmergencyManagement,etc.Owning to lacking the in-depth analysis of the dynamic evolution of topicfeatures over time in the existing work which only uses the mean generalization thought to extendtopic features in the evolution process incrementally,a large number of irrelevant topic information isintroduced into the current work.Meanwhile the low accuracy of the topic associated computationproduced by current work leads to the deviation phenomena of the topic evolution mining.Aimingto deal with this issue,this paper first proposes a feature computation model of the topic-evolution-oriented through introducing the evolution characteristics of the topic feature,and then on this basis,the article conducts the forward fusion and reverse filter under the existing topic-relatedstories and newly arrived stories in order to fulfill the incremental expansion of topic informationand anti-noise processing.The experiment results show that this method improves the topicassociation precision and solve the topic evolution deviation problem effectively.
展开▼
