基于本体和加权朴素贝叶斯的网络舆情主题分类  

Topic Classification of Network Public Opinion Based on Ontology and Weighted Naive Bayes

作　　者：丁晟春[1,2] 王小英 刘梦露 Ding Shengchun;Wang Xiaoying;Liu Menglu(School of Economics and Management,Nanjing University of Science and Technology,Nanjing 210094,China;Jiangsu Collaborative Innovation Center of Social Safety Science and Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学经济管理学院,江苏南京210094 [2]江苏省社会公共安全科技协同创新中心,江苏南京210094

出　　处：《现代情报》2018年第8期12-17,34,共7页Journal of Modern Information

基　　金：国家社会科学基金一般项目"基于社会网络分析的网络舆情主题发现研究"(项目编号:15BTQ063)

摘　　要：及时准确地对舆情信息进行主题分类,不仅能实时了解舆情动态变化,还能为预判舆情发展趋势、舆论引导建立基础。本文提出一种基于本体和加权朴素贝叶斯的网络舆情主题分类方法,通过使用本体将领域知识和领域文本特征融入分类过程中。将该方法应用到动物卫生领域舆情主题分类中,分类结果精确度为0.9402,Marco_F1达到0.9339。通过与朴素贝叶斯(NB)和THUCTC两种方法的对比实验,证明本文提出的基于本体和加权朴素贝叶斯的分类方法有效且具有可行性,但是领域本体的概念、关系的完备程度会影响分类的效率。Timely and accurate classification of public opinions can not only understand the dynamic changes of public opinions in real time,but also can establish the foundation for the development trend of public opinions and the guidance of public opinions. In this paper,a topic classification method based on ontology and Weighted Naive Bayes was proposed,which integrated domain knowledge and domain text features into the classification process by using ontology. Applying this method to the topic classification for animal health-related public opinions,and the accuracy and Marco_F1 of experiment were respectively 0. 9402 and 0. 9339. Compared with the two methods of Naive Bayes and THUCTC,it was proved that the proposed classification method based on ontology and weighted naive Bayes was effective and feasible. The completeness of concepts and relationships in domain ontology could affect the efficiency of classification.

关 键 词：网络舆情 主题分类 本体 加权朴素贝叶斯 

分 类 号：G254.1[文化科学—图书馆学]