Query Mapping and Transformation Techniques for Problem Solvingwith Multiple Knowledge Servers 1

This paper describes a technique for information mediation when multiple heterogeneous knowledge and data servers are to be accessed during query processing. One problem is building an intelligent interface between each knowledge server (KS) and its processor (KP); and the second is to provide interoper-ability among multiple KP/KS so that a query may be answered using information from multiple sources. We present example scenarios which highlight these problems and then outline query mapping and transformation techniques that are applicable. The techniques for solving the interoperability problems involve representations in some canonical form. This includes a canonical representation (CR) corresponding to each KP/KS pair and a merged CR (MCR) to represent the mapping among the CRs. The MCR and CRs include relevant information obtained from a source query, and heterogeneous mapping (het-map) information, for all possible mappings among the multiple servers. The knowledge in the canonical form must be represented so that it can be easily accessed during query transformation. We use an example of translating queries from an object schema to a relational schema to illustrate typical knowledge that must be represented in some canonical form. We use a high level logical language, F-logic, to represent the heterogeneous mapping (het-map) and query transformation information as a set of declarative rules, in the canonical form.