Proposal of cascade neural network model for text document space dimension reduction by latent semantic indexing

The paper describes the neural network model which in the information retrieval process solves the document set dimension reduction for representation of text documents in Slovak language. This model comes out of the vector space model, which for document set uses the full index representation. To decrease the matrix dimension for document set representation the Latent Semantic Model is used. Main advantage of latent semantic model in relation to the vector space model is the great reduction of the matrix dimension for document set representation. Described approach is performed by cascade neural network.