Data integration technology becomes essential in processing many types of data. This paper proposes a new method called Semi-Automatic XML schema Mapping (SAXM). Semantic similarity is firstly used to separate unmatchable nodes. Structural similarity and data type compatibility are then applied to ascertain the matching of two nodes. The output of SAXM is a set of similarity relationship values between two XML schemas elements. Moreover, a user can verify the mapping to provide a more accurate result. Experimental results show that our proposed method provides the highest recall and comparable precision among well known approaches.