Since the last decade, XML has become inevitable for complex data representation. In this paper, we address a problem of measuring the similarity between XML documents and propose a new XML document similarity approach, which considers the asymmetric similarity and the similarity of both semantic content and document structure. Here, we only consider the measurement of similarity between two XML documents based on the same schema. A prototype has been implemented to validate and evaluate the performances of our proposal. We do believe that our method can also be used to evaluate the similarity of other tree-structured complex data.