New text categorization models using back-propagation neural network (BPNN) and modified back-propagation neural network (MBPNN) are proposed. An efficient feature selection method is used to reduce the dimensionality as well as improve the performance. The basic BPNN learning algorithm has the drawback of slow training speed, so we modify the basic BPNN learning algorithm to accelerate the training speed. The categorization accuracy also has been improved consequently. Traditional word-matching based text categorization system uses vector space model (VSM) to represent the document. However, it needs a high dimensional space to represent the document, and does not take into account the semantic relationship between terms, which can also lead to poor classification accuracy. Latent semantic analysis (LSA) can overcome the problems caused by using statistically derived conceptual indices instead of individual words. It constructs a conceptual vector space in which each term or document is represented as a vector in the space. It not only greatly reduces the dimensionality but also discovers the important associative relationship between terms. We test our categorization models on 20-newsgroup data set, experimental results show that the models using MBPNN outperform than the basic BPNN. And the application of LSA for our system can lead to dramatic dimensionality reduction while achieving good classification results.