Semantic Interoperation Via Intelligent Mediation

in the same representation, and hence their results can be readily merged. To compile the result of query (5) into a set of cargo entities in the ight database, we need a rule: mapping the rst attribute domain to the other attribute domains. These functions are interpretations of the fact that attribute a 1 is the key of entity O. Applying this rule, every pair in the result of query (5) is bound to a cargo entity by the rule instance: rule-instance Compile-Cargo params (cargo(c#: C#, type: Type): cargo-schema, (c; t): C# Type, T: C# 7 ! Type) compile (c; t): T(c) = t =) o(c; t): Cargo Finally the result of compilation is merged with the result of query (2), and returned to users of the ight database. 4 Conclusion We proposed an approach to the semantic interop-eration via intelligent mediation. The key idea of our approach is the separation of heterogeneity in the semantics of data from heterogeneity in the representation of data. Instead of a rich multidatabase language, we advocate a language with minimal representational bias. The semantics of heterogeneous data is communicated independent of representation, and is compiled into homogeneous representation schemes. Our approach directly addresses the critical issues in the semantic interoperation of heterogeneous data. The semantic mediator guarantees that access to heterogeneous data is semantically meaningful. Autonomy is respected and preserved, since users of individual databases are not aaected in the way they view the real world or the way they represent the data. Semantic interoperation is automated; hence, users of individual databases need not be aaected by changes of semantics or representation in other databases. De-coupling semantics and representation makes reasoning with heterogeneous data independent of speciic representation schemes, reducing the space as well as the complexity of potential heterogeneity, and improving the eeciency of semantic interoperation. With this approach, it becomes possible to share the semantics of data without having to share the representation of data. 4 an intelligent communication module capable of interpreting heterogeneous data in the language based on knowledge about the semantic relationships between them, and an intelligent compilation module capable of binding homogeneous data into representation schemes based on knowledge about the semantics of various representation constructs. The language is minimized in terms of represen-tational bias, which is crucial in separating semantics from representation. Communication and compilation are intelligent, meaning that access is possible without users' knowledge about …