Querying with Vague Quantifiers Using Probabilistic Semantics

Many realistic scenarios call for answers to questions involving vague expressions like almost all, about half, or at least about a third. We present a modular extension of classical first-order queries over relational databases, with binary, proportional, semi-fuzzy quantifiers modeling such expressions via random sampling. The extended query language has an intuitive semantics and allows one to pose natural queries with probabilistic answers. This is also demonstrated by experiments with an implementation involving the (geographical) MONDIAL data set.