公共卫生类突发事件的抽取研究

Research on Event Extraction for Public Health Emergencies

[目的/意义]对突发事件信息进行结构化抽取,建立突发事件典型案例库是提升突发事件应对能力的有效手段。[方法/过程]文章提出一个描述突发事件案例的四元特征向量模型,并基于这一特征向量模型构建带有继承机制的突发事件案例信息的抽取框架;然后针对抽取框架的不同侧面的特点,将侧面分为事件时间、事件地点、事件其他侧面三类,并分别对每一类信息的抽取方法和模式进行详细说明。[结果/结论]经过在公共卫生类突发事件语料上进行实验,结果表明本文提出的抽取框架是可行的,对于突发事件要素抽取达到较高的准确率和召回率。

[Purpose / significance] Extracting structured information from emergencies and constructing emergency typical case base are effective for improving the ability to cope with emergencies. [Method / progress] This paper presents a quaternary feature vector for emergency case description. Meanwhile,emergency case information extraction framework with inheritance mechanism is constructed based on this feature vector model. Then,for different aspects features,aspects of the framework are divided into 3 categories： event time,event location,and other aspects. Extraction methods and patterns of each category are discussed in detail.[Result/conclusion]Experiments on public sanitation emergency audio material shows that the extracting model is feasible,which achieves a high precision and recall rate.