DBrev: Dreaming of a database revolution

The database community has provided excellent frameworks for efficient querying and online transaction or analytical processing. The main assumption underlying most of these frameworks is that there is no uncertainty regarding the stored data. However, in recent years, many important applications have emerged that need to manage noisy, corrupted, or incomplete data. This includes, e.g., anonymized data, data derived from sensor systems, or data from information extraction and integration systems. For such applications the assumption of logical consistency may not be valid and needs to be revised. In particular, techniques like probabilistic modelling and statistical inference may be necessary to be able to draw meaningful conclusions from the underlying data. This paper presents DBrev, a hypothetical, intelligent database system for managing large quantities of data that involves uncertainty. We explain the main features of DBrev based on the scenario of information extraction and integration. We point out research challenges that need to be tackled and discuss a new set of assumptions that future database management frameworks need to build on.