This paper reports on experiments into the semantics of nouns in documents through WordNet. Because WordNet is founded on semantic relations, we can compare every two nouns of a document with their common categories to discover whether they have semantic relations and then group them into a cluster if they do. In this project, generally we just verify three semantic relations of WordNet, namely hypernym, meronym and synonym.  Assume once we have the clusters, we can tell the basic meanings of a document. Based on this assumption, we can compare those documents by comparing their clusters to see their closure. This project is designed on the above purpose and implemented under Unix system using the Scheme language.