摘要
本文结合网络虚拟社会中舆情检索的实际需求,提出了一种面向博客群的主题倾向性分析模型.针对博客主题评论篇幅长短不一的结构特点,模型采用不同的文本倾向性处理方法:对于较长篇幅评论文本,分别统计目标评论中赞同、反对字符的倾向字符权重及其分布密度;对于拥有少量文字的主题评论,通过计算评论中字符倾向权重之和,实现评论倾向性评估.实验中通过构建面向"网络文化"的博客主题测试集,对模型的主题评论倾向性计算方法以及语义检索能力进行验证评估.实验结果表明模型具有较好的文本倾向性识别能力.
Considering the actual requirements of public opinion retrieval in virtual network society, a topic opinion analysis model for blogosphere is proposed. Oriented to the different length of blog comment, two calculation methods for long text and short text respectively are used in the model. For long comment text, the opinion weight of each character and the distribution density of opinion characters in target text are evaluated. For short comment text, the text opinion is calculated by summing the opinion weight of each character. On the other hand, the amount of topics, reviews, comments and the semantic opinion of each comment are analyzed, and evaluates the reputation of each blogger in virtual network society are analyzed. The experiments on the data corpus about ＂Network Culture＂ demonstrate that the model has higher precision of text opinion analysis.
出处
《中央民族大学学报（自然科学版）》
2014年第3期33-37,2+97,共5页
Journal of Minzu University of China(Natural Sciences Edition)
基金
国家"十二五"科技支撑计划"少数民族网络舆情综合分析与云服务关键技术及应用示范"(No.2014BAK10B03)阶段性成果
