User Feedback in Probabilistic Integration

Ander de Keijzer, Maurice van Keulen

Abstract: Data integration approaches mostly attempt to resolve semantic uncertainty and conflicts between data sources during the data integration process. In some application areas, this is impractical or even prohibitive. We propose a probabilistic XML approach that allows storage and querying of uncertain data. It requires only minimal user involvement during data integration, because most semantic uncertainty and conflicts can be resolved by exploiting user feedback on query results, thus effectively postponing user involvement to query time when a user already interacts with the system. We show that repeated feedback gradually improves the factual correctness of integrated data.