Method in ML and information retrieval where the system selects the k most relevant or highest-scoring items from a larger set of predictions or results.
Generality: 625
