Incremental data integration for federated databases

Abstract

The paper demonstrates how the experimental database language Proto derives and manipulates partial information representations and how these derivations may be used to merge information from a federation of databases. Partial information representations allow queries to be answered even in the event that ambiguity about relationships between the data from clients in a database federation cannot be resolved using available information. The language Proto is an implemented deductive extension to Common Lisp that allows the independent derivation of true and possible answers to queries, together with null values, including non-monotone introspective options that allow the inference system to use heuristics or defaults.<>