Top-k Queries

Synonyms

Ranking queries; Top-n queries

Definition

A top-k query retrieves up to k query answers with the highest (lowest) score according to some scoring function f, where ties are resolved arbitrarily. Without loss of generality, we assume that we retrieve the results with highest scores. In other words, top-k queries return a subset of the query answers S ⊂ Q such that

$$ \left|S\right|{\le} k\bigwedge \forall t{\in} S\nexists r{\in} \left(Q{-}S\right)\bigwedge F(r){>}F(t) $$

On relational data, two popular variants of top-k queries are defined: (1) top-k join query and (2) top-k aggregate query.

Top-k Join Query. Consider a set of relations R1 . . . Rn. A top-k join query joins R1 . . . Rn and returns the k join results with the largest combined scores. The combined score of each join result is computed according to some function F(p₁, … , p_m), where p₁ , … , p_m are scoring predicates defined over the join results.

An example SQL template for a top-k join query is:

SELECT < select...