一种基于主题分类的文本过滤方法及其硬件实现

Text Filtering Algorithm Based on Topic Classification and Realization of DSP Hardware

针对不良文本的过滤问题,提出一种基于主题分类的文本过滤方法,通过对文本信息进行向量化,引人文本特征抽取技术,筛选出针对文本内容的最优的特征项集合,利用SVM分类技术,来判断文本的态度和立场,达到内容审查过滤的目的.并利用DSP在硬件上加以实现,实验表明该方法同传统的过滤方法相比具有较高的准确率和召回率,且过滤时间大幅减少.

Concerning Chinese vicious-topic information text filtering problems, this paper presents an improved text filtering method based on SVM classifier. Through replacing filtering method based on words feature items can be distinguished from different classes effectively. The experimental results indicate that this method is of both higher precision rate and recall rate compared with traditional methods.