基于多类别相关度的潜在语义挖掘算法

摘要

鉴于当前 web文本分类存在的问题，阐明基于文档和类别相关度的生成局部区域的方法，即 S-LLSA。将各种类别信息应用于语义分析中，研究特征词的局部特征，通过相关分类器求解文本对类别的相关度参数，在此基础上，将其引入到生成局部区域的环节之中。实验结果表明， S-LLSA 能够妥善处理局部区域奇异值分解问题，在很大程度上改善了web文本分类结果，使其潜在语义空间得到有效描述。%For the existing problems of web text classification and representations,a local relevancy latent semantic analysis algo-rithm (S-LLSA)was designed based on the correlation between the document and the categories to generate local area.Category information was introduced in singular value decomposition (SVD),local feature of feature words was analyzed,and classify ca-pability of support vector machine was used to select local area.Experimental results show that S-LLSA algorithm effectively solves the key problem of singular value decomposition,greatly improves the effectiveness of web text classification,and better represents web text latent semantic space.