LATENT SEMANTIC ANALYSIS BASED ON SPACE INTEGRATION
Latent Semantic Analysis (LSA) is a technology which is used to analyze the latent concepts.LSA is based on the Vector Space Model (VSM) and statistics,and it usually takes the Singular Value Decomposition (SVD) as the kernel algorithm.Always,LSA increases the scale of the training data to improve system performance.However,as it needs many extra operations,and it also generates too much co-occurrence paths which are unreasonable between the different features,the problem of noise will be a serious disadvantage.This paper proposes a new method which is called augmented space model to optimize the latent semantic space model.Besides,it is also suggested in this paper that multiple models can be combined with integration technology to improve system performance.Through integration technology and space optimization,the models may describe the latent semantic structure more exactly.At the same time,to some extent,the probability of generating noise co-occurrence is reduced.As shown from comparative experiments,the system accuracy is higher after adopting integration technology and space optimization technology.
Latent semantic analysis Augmented space model Space integration Space optimization Text categorization K-nearest neighbor
Dongfeng Cai Liwei Chang Duo Ji
Knowledge Engineering Research Center,Shen Yang Aerospace University,Shenyang 110136,China
国际会议
杭州
英文
1912-1916
2012-10-30(万方平台首次上网日期,不代表论文的发表时间)