In this paper, we propose a topic selection method about web documents. The idea of our approach is to utilize an ontology structure and TF (Term Frequency) values of each term. For improving the performance of documents clustering, our research is strongly demanded. We process web documents for keywords acquisition using TF values and relevancy values between terms using relations defined in WordNet. And then, we proposed the topic selection formula as we consider three kinds of cases during the topic selection. In conclusion, we demonstrate that our approach is very useful for the topic selection of documents.