Probabilistic models for uncertain data

Uncertainty is ubiquitous in the outcome of many automatic processes (such as information extraction, natural language analysis, machine learning, data integration) or for all tasks that involve human judgment, contradicting information, or measurement errors. This uncertainty can be captured by probabilistic models -- probabilistic information can now be stored, queried, updated, aggregated in a well-founded manner. This talk will provide concrete motivation for probabilistic data management, review some of the most important models for probabilistic data (tables, trees) and present some of the important results in this research area, both theoretical and applied. A concrete example of the use of a probabilistic data management system will also be demonstrated.