Navigation and information retrieval on the Web are not easy tasks; the challenge is to extract information from the large amount of data available. Most of this data is unstructured, which makes the application of existing data mining techniques to the Web very difficult. However, new semantic structures which improve the results of Web Mining are currently being developed in the Web. This paper presents how one of these semantic structures - XML topic maps – can be exploited to help users find relevant information in the Web. This paper is organised as follows: first, we introduce XML topic maps in the context of Tim Berners-Lee’s Semantic Web vision. Then, we show how topic maps allow to characterise and “clean” Web data through the definition of a profile; this is achieved by the analysis of a lattice generated by a classification algorithm - called Galois algorithm. This profile may be used to evaluate the relevance of a web site with regard to a specific request on a traditional search engine. We finally explain how data on the Web can be clustered, organised and visualised in different ways so as to enhance users’ navigation and understanding of these documents.