基于卡方特征和BTM融合的短文本分类方法  

Short Text Classification Based on Chi-square Feature and BTM

作　　者：李振兴[1] 王松[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2016年第1期36-41,共6页Journal of Lanzhou Jiaotong University

基　　金：中国铁路总公司科技研究开发计划课题(2014X008-F)

摘　　要：针对短文本特征稀疏、上下文依赖而导致的传统文本分类法应用效果不佳的问题,提出一种基于卡方特征和BTM的短文本分类法.首先提取短文本的卡方特征,再利用BTM对短文本建模,获得对应的文档-话题概率特征,最后融合两种特征并基于SVM分类算法实现短文本分类.实验结果表明,相比于常规分类方法,该方法具有较高的Macro-F1值,对短文本的分类具有良好的效果.Aiming at the shortage of traditional text classification method on account of text feature sparse and context dependency,a short text classification method based on Chi-square feature and BTM is proposed.Firstly,Chi-square features of short text are extracted,then it is modeled by BTM to get the corresponding document-topic probability features.Finally,the short text classification is obtained by combining these two features and SVM classification algorithm.Experimental results show that this method has high Macro-F1 value compared to the conventional classification method and verify that the method achieves a better performance in short text classification.

关 键 词：短文本分类 卡方特征 话题模型 BTM 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]