基于潜在语义索引和遗传算法的文本特征提取方法

The Method of Text Feature Selection Based on LSI and GA

本文采用潜在语义索引(LSI)和遗传算法(GA)进行文本特征提取。在采用潜在语义索引将语义关系体现在VSM(Vector Space Model)中,通过奇异值分解(SVD,Singular Value De-composition)可以有效地降低向量空间的维数,但通过维数约简后的文本特征仍要保持在数百维左右,因此本文采用遗传算法在此基础上继续降维。实验结果表明,这两种方法结合可以极大的降低文本向量空间的维数,并能提高分类准确率。

This paper selects the features of text by using LSI and GA （Genetic Algorithm）. This paper uses LSI to reflect the relation of words in VSM （Vector Space Model）. The dimension of VSM can be reduced greatly by Singular Value Decomposition. However, after that the text features have still several hundreds dimensions, so this paper continues to reduce the dimension by using GA in this base. The results of exexperiment indicate that combining these two methods can greatly reduce the dimension of VSM and advance precision of text classifying.