Uncertainty Annotated Databases - A Lightweight Approach for Dealing with Uncertainty

Incomplete and probabilistic data models have been proposed to deal with the uncertainty inherent in many real world data collection and management tasks. However, query evaluation over such models is heavy-weight both in terms of computational complexity as well as usability. We introduce UA-databases (UA-DBs), a light-weight model of uncertainty where tuples from a single possible world are annotated with uncertainty information. UA-databases can be derived from commonly used incomplete and probabilistic data models. We present a query semantics for UA-DBs that is compatible with deterministic query processing, as well as many data models expressible as K-relations. Furthermore, we guarantee that tuples that are marked as certain in a UA-DB query result are guaranteed to be certain answers. We implement UA-DBs on top of a DBMS and experimentally demonstrate that this approach is efficient. PVLDB Reference Format: Su Feng, Aaron Huber, Boris Glavic, Oliver Kennedy. Uncertainty Annotated Databases A Lightweight Approach for Dealing with Uncertainty. PVLDB, 11 (4): xxxx-yyyy, 2017. DOI: https://doi.org/TBD