Candidate generation Recommender Systems
noun phrase
Definition: A retrieval-stage process that produces a manageable set of promising candidates (items, documents, actions) for downstream ranking or scoring [Google ML Glossary].
Example in context: “We also defined recall accuracy of MESE to be the percentage of ground truth items that appear among the 500 generated candidates in the candidate generation phase and ranking accuracy to be the percentage of items that appear in the top k positions in the candidate ranking phase.” [Yang et al. 2022]
Related terms: retrieval stage, candidate retrieval, first-stage ranking