Abstract
Data completeness is an issue in many applications where data is integrated from multiple sources, or created via human-intensive workflows. In consequence it is often not clear whether or which parts of query answers are reliable, and which are not.
In this talk I will show how one can describe complete parts of a database, and how one can develop a modified relational algebra that, given information about the complete parts of the database, computes the parts of query results that can be trusted. I will show the complications that arise when one wants to take into account conclusions possible by inspecting the state of the database, and present implementation techniques. I will also briefly talk about future research interests in the area of statistical completeness assessment of large knowledge bases such as Wikidata or YAGO.