Managing Probabilistic Duplicates in Databases