基于双层模型的维吾尔语突发事件因果关系抽取

Causal Relation Extraction of Uyghur Emergency Events Based on Cascaded Model

针对传统事件因果关系识别覆盖范围小和人工标注代价高等不足,提出了一种基于双层模型的维吾尔语突发事件因果关系抽取方法.该方法采用分治思想,将因果关系抽取问题转化为对事件序列的两次模式识别标注.采用Bootstrapping算法,在第一次模式识别时,标注因果关系的语义角色,并将标注的语义角色标签作为新的特征传递给第二层模式识别,用于因果关系边界标注.该方法用于维吾尔语突发事件显式因果关系的抽取准确率为85.39%,召回率为77.53%,证明了本文提出的方法在维吾尔语主题突发事件因果关系抽取上的有效性和实用性.

Because the traditional events causal relation has the disadvantages of small recognition coverage and the labeling cost is high, a method for causal relation extraction of Uyghur emergency events is presented based on cascaded model. Utilizing the divide-and-conquer strategy, it converts the problem of causal relation extraction to two pattern recognition labeling of event sequence. By applying the bootstrapping algorithm, the method labels the semantic role of causal relation in the first layer of pattern recognition, then utilizes the semantic role label as a new feature and transfers it to the second layer of pattern recognition for labeling causal relation boundary. This method has been used in the explicit causal relation extraction of Uyghur emergency events, and the results have shown that the precision rate and the recall rate can reach 85.39 % and 77.53 %, indicating the efficiency and practicability of the method of causal relation extraction of Uyghur topic emergency events.