Probabilistic Methods For Querying Global Information Systems

As databases undergo a paradigm shift from stand-alone applications to an architecture that supports data sharing on a global scale, they have to increasingly cope up with a fundamental problem: that of handling uncertainty and lack of information. Users often have to formulate queries over unfamiliar data sources: quite often they don’t understand their schema, or their particular representation of data values. Probabilistic methods have been used for a long time in the AI community to model uncertainties in knowledge bases, but a similar treatment is missing in databases and database queries. The goal of this report is to propose the use of queries on probabilistic databases to cope with uncertainties resulting from the schema and data heterogeneity in global data sharing and exchange.