Fast latent semantic index using random mapping in text processing

The calculation bottleneck problems of Kohonen self organizing feature map (SOFM) neural network in the high-dimensional vector environment of text processing and problems of input vector spaces had been analyzed in this paper, and then based on the theoretic analysis of RM (random mapping) and LSI (latent semantic indexing) method respectively, a RM-based fast latent semantic indexing method used in text processing was presented. The fast LSI method could greatly emerges original semantic links and settles the above mentioned problems in a low-cost, efficient and controllable way in the experiment. So the size and the calculation cost of Kohonen SOFM neural network were greatly reduced in text processing environment.