The schema matching is required before performing processes that facilitate interoperability such as (exchange, integration, transformation). It is a technique that discovers the semantic correspondences between elements and attributes of schemas. The main problem in achieving this technique is bound up with the heterogeneity between the schemas. In this work we are interested specifically in the problem of XML schemas matching. The approach that we advocate is a methodological approach that can be adapted in searching documents on the WEB. In our approach, the proposed system interacts with the user through the "schema generator” to produce a source schema that represents the user's personal request. The latter is compared to target schemas using a matching algorithm. The latter is either selected among a battery of matchers or described through a design pattern. We propose so a system of semantic correspondence discovering, which permits to reassemble or reuse existing approaches in order to solve this problem in a flexible and concrete way. The matching result is then, a set of correspondences of schemas. This set is used to determine the similarity between schemas. Finally, the correspondences of schemas are classified according to their similarity value before being proposed to the user.