A Possible World Approach to Uncertain Relational Data

Abstract

Data exchange between embedded systems and other small or large computing device increases. Since data in different data sources may refer to the same real world objects, data cannot simply be merged. Furthermore, in many situations, conflicts in data about the same real world objects need to be resolved without interference from a user. We report on an attempt to make a RDBMS probabilistic, i.e., data in a relation represents all possible views on the real world, in order to achieve unattended data integration. We define a probabilistic relational data model and review standard SQL query primitives in the light of probabilistic data. It appears that thinking in terms of 'possible worlds' is powerful in determining the proper semantics of these query primitives.