中文摘要：

针对新闻文本,将事件片段定义为最小划分单元,提出了一个两阶段的事件片段时序关系识别算法,第一阶段时序映射和第二阶段时序识别。使用真实网上新闻数据进行实验,结果表明,所提方法相比普通的分类器及强规则算法,准确率提高了4%。

英文摘要：

Event fragments as the smallestr unit was defined and a two-stage event fragments temporal relation recognition algorithm was proposed. The first stage is sequential mapping and the second is sequence recognition. The real online data increased accuracy by 4% compared to common classifier and strong rule algorithm.