Based on the study of a text classification technique, we propose a new text classification method which improves the Vector Space Model and Naive Bayesian classifier by using a weight adjustment measure, implement an experimental text classification system SECTCS (Smart English and Chinese Text Classification System), and make a comparison between various text classification approaches by using SECTCS. Compared with many commercial text classification systems, the behavior of SECTCS is excellent. We introduce its framework, function and running environment, give our experimental results, and discuss a few important technical issues involved in the system to get some valuable conclusions. We also describe how to improve the Vector Space Model and Naive Bayesian classifier in detail.