基于语义倾向性分析的不良文本检测模型研究

摘要

着眼于因特网的内容安全,在分析传统语义倾向性判别的类型及不足的基础上,本文提出一种改进的语义倾向性分析模型.该模型以词组为单元,同时引入几种常见的句式结构.通过对这些句式结构的判别,确定出每句话语义倾向的侧重点,再参照JRBPO准则予以定性.最后,利用各单句与上下文逻辑间的关联性,得出整个文本的语义倾向.随机抽取的测试样本表明,本方法对不良文本检测具有较高的处理速率和检测准确率.%To focus on the content security of the Internct,an improved semantic orientation analysis model was put forward after researching the traditional types and their shortcomings.This model used the phrase as a unit,meanwhile,introduced several common sentence structures.By judging these sentence structures,it could determine each sentence orientation and which is the key point.Then a qualitative was obtained according to the JRBPO guidelines.Finally,the whole text tendency was obtained based on the correlation between every sentence orientation and context logic.Random test sample show that this method of harmful text detection has high processing speed and accuracy.