The authors, who publish knowledge on the Web related to readable electronic documents on a screen, use the technology of hypertext links or the hypertext writing approach for making their sites more interesting and more attractive, and also to enrich it by information coming from other Web sites. However, hypertext links are not without posing problems for readers and search engines like disorientation and cognitive overload for readers and publicities links for search engines. We are interested in semantics of hypertext links, in terms of extraction and exploitation with the aim of facilitating the search of knowledge on the Web. In this article, we propose an original methodology for the semantic extraction of hypertext. Firstly, we show how we constitute a corpus of documents on the Web, which will be our data set. Then we propose a method of semantic analysis of hypertext links. This method consists in making the semantic analysis of calling context of link and context called by link, and explicit in a formal way the semantic relation between the two contexts.