基于信息单元融合的新闻原子事件抽取

Atomic Event Extraction Based on Information Unit Fusion

原子事件抽取是将非结构化文本进行结构化表示的重要方法.针对新闻语料,本文提出了一种基于信息单元融合的原子事件抽取方法.在中文分词、词性标注、命名实体识别等自然语言处理技术的基础上,利用语言规则将信息单元标识出来并进行融合,达到浅层句法分析的效果,通过原子事件抽取算法将原子事件从经信息单元融合后的语料中抽取出来.基于信息单元融合的原子事件抽取方法不仅对文本长度没有严格限制,并且不受事件类型的约束;实验结果表明,基于信息单元融合的原子事件抽取方法是有效的.

Atomic event extraction is an important means to represent the unstructured text structurally. This paper proposes an information basic unit fusion approach to extract atomic event from the news. On the basis of Chinese word segmentation,part of speech tagging and named entity recognition,information basic units can be marked and fused according to linguistic rules. And then,the atomic events can be extracted from information basic unit fused texts by the atomic event extraction algorithm. This approach does not restrict the length of texts and the types of atomic events. The experiment results demonstrate the effectiveness and feasibility of the atomic event extraction approach based on information unit fusion.