Abstract : Extraction of information from the semi–structured documents constitutes an important domain of recent research in the computer science. In this thesis we propose two approaches which permit an evaluation of queries expressed in XPath on the XML documents. We have two objectives in mind: using compressed documents directly — without passing by their binary representations, evaluation of queries on the documents subjected to the control access policy. According to the context, we employ: automata, transitions systems, and rewriting. The results that we obtain show that these classical techniques are appropriate to the treatment of the queries on XML. If we restrict our consideration only to the descendant fragment of XPath, all the approaches introduced in this work can be adapted to the case of compressed documents represented by the dags.