block by the style sheets. The technique is called “DIV plus CSS”. Based on this observation, a web page is first partitioned into several blocks using DSS_DOM. Secondly importance values are assigned to all the blocks using an evaluation algorithm. The algorithm involves the information of style sheets and the structure of DSS_DOM. The contents in low-importance-value-blocks are not- related-contents. The proposed model is evaluated with a data mining task-web page classification. Experiments show satisfactory results.