摘要
事件检测是文本挖掘的一个重要研究方向,以微博文本的突发地震事件检测为例做了深入研究。首先分别运用三种经典的分类算法来实现突发地震事件检测,将检测结果进行比较,选择出一种最优的分类算法和最适合的特征数。在此基础上提出关键词过滤和时间关系识别的方法将错分的实例进行再分类来提高检测结果。实验表明该方法的检测结果与仅采用经典分类算法相比F_1值提高了5.3%。
Event detection is one of the most important research fields in text mining,and an in-depth stiidy is carried out in the case of micro-blog. First, used three kinds of classical classification algorithms to detection the sudden earthquake events,and the results were compared, selected an optimal classification algorithm and the most suitable number of features. Then, basis of this, put forward the metliod of keyword filtering and temporal relation recognition to classify the wrong examples to improve the detection results. Experimental results show that the proposed method can improve the I value by 5. 3% compared with the classical classification algorithm.
出处
《微型机与应用》
2017年第19期58-61,65,共5页
Microcomputer & Its Applications
关键词
事件检测
文本分类
关键字过滤
时间关系识别关键词
event detection
text categorization
keyword filtering
temporal relation recognition
