摘要
通过构建新闻专题的事件脉络,可以辅助读者识别事件演化发展阶段,把控事件的全局性信息。随着专题事件的演化发展,相关新闻会持续不断出现对事件进行报道。为了保障事件脉络的完整性和时效性,需要从最新新闻数据流中追踪相关新闻,对事件脉络进行持续跟踪更新。提出了一种面向新闻的专题事件脉络持续跟踪构建方法,采用K-means聚类和基于凝聚式的层次聚类方法检测事件发展阶段,构建以时间为主线、各发展阶段为分支的事件发展演化脉络,保障事件脉络的完整性和连续性;综合实体、关键词和文本3个维度的相似度特征从新闻数据流中持续跟踪与专题事件相关的新闻数据,将追踪到的新闻事件同时更新到事件文本向量和已构建的事件脉络中,实现对事件脉络的持续跟踪构建。
Constructing the story line of topic news can help readers identify the evolution and development stages of events and control the global information of events.With the evolution and development of special events,relevant news will continue to appear to report the events.In order to ensure the integrity and timeliness of the event context,it is necessary to track relevant news in the latest news data stream and continuously track and update the story line.Therefore,a news-oriented method to continuously track and construct the story line for hot events is proposed.K-means clustering and agglomeration-based hierarchical clustering are adopted to detect event development stages,and the development and evolution context of events with time as the main line and each development stage as the branch is constructed to ensure the integrity and continuity of event context.Then,the similarity features of entity,keyword and text are integrated to continuously track the news data related to special events from the news data stream,and the tracked news events are simultaneously updated to the event text vector and the constructed event context to realize the continuous tracking and construction of event context.
作者
欧伟明
翟利志
路瑜亮
周云
苌军红
韩彦忠
OU Weiming;ZHAI Lizhi;LUYuliang;ZHOU Yun;CHANG Junhong;HAN Yanzhong(The 54th Research Institute of CETC,Shijiazhuang 050081,China;The First Military Representative Office Stationed in Shijiazhuang,Military Representative Bureau of Army Equipment Development Department Stationed in Beijing,Shijiazhuang 050050,China)
出处
《计算机与网络》
2022年第20期61-68,共8页
Computer & Network
