This paper describes two important information mediation prob-lems that arise when multiple knowledge and data servers are accessed for
problem solving. The rst problem is building an intelligent interface be-
tween a knowledge server (KS) and a processor (KP) so that queries may
be answered intelligently and completely. The second problem is to provide
interoperability among multiple KP/KS pairs so that a query may be an-
swered using information from multiple sources. We present example scenar-
ios which highlight each of these problems and then outline techniques and
tasks that are applied towards obtaining a solution. These techniques draw
upon disciplines such as query and transaction processing with knowledge
bases and machine understanding and translation of natural language. The
techniques for solving the information mediation problems described involve
parameterized canonical representations (CR) for the KP/KS pairs. The CR
will represent possible mappings between a query and the knowledge model
and will be used during query transformation to produce a query which will
provide more complete and correct answers. Parameterization will help in
merging CRs to support interoperability for distributed problem solving.
| 