Labels and datatypes are together the most important factors in identifying the best correspondences in matching XML Schemas. Besides label matching, datatype matching is also necessary since XML Schema is designed for datatype definitions for XML documents. We here propose a new approach for matching datatypes by taking into account various kinds of relations between datatypes and the mechanisms of definition for user-defined types. All datatypes in two schemas being matched are put into a global hierarchical structure of inheritance and aggregative relations. We show that our approach is flexible enough to deal with additional built-in datatypes and it matches completely or partially user-defined datatypes. Moreover, the experimental results are also promising with 83% correlation to human ratings and 10% superior to other approach (e.g., Cupid''s), which provides us a firm baseline for the next research phase.
展开▼