一种新闻事件演化建模方法

摘要

A new method is proposed for modeling the news event evolution to precisely present the relationships between events.This method utilized the events timestamp,events content similarity,and events dependence between features to build a new event evolution model,and defined five different event evolution patterns to identify the seminal events,the intermediary and ending events.Ultimately,an event evolution graph model was constructed to present the underlying events relationship.Experiments were conducted,confirming that the proposed method is efficient for detecting event evolution,and improves performance of system.%为了准确地发现话题中事件间的潜在关系,提出一种新闻事件演化建模方法.该方法利用事件的时间关系、内容相似性、命名实体关联信息构建新的演化关系模型,并通过定义事件的五种演化模式,识别出演化过程中的开始、中间、结束事件,最后根据新演化模型及演化模式建立事件演化的有向无环图模型,揭示事件发展的潜在脉络结构.实验结果表明,本文方法能够有效检测事件演化,提升系统性能.