Probabilistic Evaluation of Expressive Queries on Bounded-Treewidth Instances

Though data uncertainty naturally appears in many real-life situations, traditional database theory and systems tend to assume that the data is reliable and complete. The reason is that of complexity and performance: on arbitrary relational database instances annotated with probabilities, performing exact probabilistic query evaluation is hard. However, a criterion on the shape of the database has been shown in recent work to be sufficient and in some sense necessary to the tractability of this task. Databases whose treewidth is bounded by a constant k are exactly those that can be tractably queried, with respect to quantitative uncertainty estimation. But this is a data complexity result, that does not take into account the cost in terms of the query or of k -- in many cases, this cost is too high for real-world applications. The aim of our PhD research is to study in which circumstances the overall complexity of probabilistic query evaluation can become tractable, aiming at both theoretical and practical results.