Aiming at the problems available in the traditional method of cross-language text clustering, a Chinese-English cross-language text clustering algorithm based on Latent Semantic Analysis is put forward. [Method] With the method of Latent Semantic Analysis, Singular Value Decomposition of characteristic word-text matrix is carried out. The bilingual latent semantic space in Chinese-English is constructed to realize cross-language latent semantic association so as to reduce dimension and noise. The K-means algorithm which chooses the initial cluster center on the basis of the minimum similarity is adopted to avoid the effect of random selection of the initial cluster centers on the clustering effect. [Results] Experiment results show that the number of reserved characteristic words of each text sand the selection of the spatial dimension value k have certain impacts on the clustering result. When each text retains the top 15 characteristic words and k=200, the F-measure can be optimal. Compared to CLTC, 13.96 percentage points can be improved. [Conclusions] This method has greatly reduced the dimension of text space and improved the cross-language text clustering quality effectively. The clustering effect is better than CLTC.