In our research work, we consider that access to semi-structured documents is carried out by a data-oriented query. With different users and a same query, the returned results are always the same although users’ characteristics (interests, preferences, etc.) may be different. In order to solve this problem and to offer a personalized access to semi-structured documents, our objective is to improve this type of query in order to adapt the result to each user according to his characteristics. On the one hand, we suggest to reorder the results according to user’s interests. On the other hand, we also suggest to establish user’s interests implicitly from his queries.