%0 Journal Article %T 一种基于多类别信息的局部潜在语义分析算法研究<br>Local relevancy latent semantic analysis algorithm based on multi-category information %A 陈珂 %A 柯文德 %A 刘美 %A 张良均 %J 南京邮电大学学报(自然科学版) %D 2016 %X 为了有效解决现有Web文本分类方法普遍存在的分类效果不佳、性能低下等问题,文中基于局部潜在语义分析的理论原理,利用支持向量机分类优势,设计出一种基于文档与类别之间相关度的生成局部区域的算法,即S-LLSA。该算法在奇异值分解过程中引入不同类别信息,分析特征词的局部特征,使用支持向量机分类器计算文本对类别的相关度参数,并应用于局部区域生成过程。通过实验表明,S-LLSA算法有效解决了局部区域如何进行局部奇异值分解问题,有效地提高并优化了Web文本分类效果,更好地表示了Web文本潜在语义空间。<br>Based on the theoretical principles of latent semantic analysis,and combined with support vector machine (SVM) classifier performance,a local relevancy latent semantic analysis algorithm (S-LLSA) is designed to solve multiple problems about Web text categorization and representation.The category information is introduced in singular value decomposition (SVD),analysed local feature of feature words are analyzed and classify capability of support vector machine are used to select local area by the algorithm.The experiment shows that S-LLSA algorithm solves the key problem of singular value decomposition,improves the effectiveness of Web text classification and represents Web text latent semantic space %K 文本分类 局部潜在语义分析 支持向量机 奇异值分解 S-LLSA< %K br> %K text classification local latent semantic analysis support vector machine(SVM) singular value decomposition(SVD) S-LLSA %U http://nyzr.njupt.edu.cn/ch/reader/view_abstract.aspx?file_no=201601018&flag=1