Interoperation of heterogeneous and autonomous information systems has traditionally been hampered by semantic differences in their data models. In this paper, we address the problem by defining a methodology called TIME, which is based on an extensible meta model. Its key features are: a set of meta-types which can be used to represent the syntax and the semantics of data modeling concepts, a knowledge base of transformation rules that map a meta-type into other meta-types, and an inference engine which uses the transformation rules to translate schema from source to target models. The extensibility of the meta-model is achieved by organizing the meta-types into a generalization hierarchy that record similarities among modeling concepts. The hierarchy of meta-types allows the reuse of transformation rules during automatic generation of data model translators.