A recommendation system predicts which products, videos, articles or actions may be useful to a person. It can use item features, behavior from similar users, the current session and an explicit query to score and rank available options.
Recommendations influence what people see, so evaluation should measure more than clicks. Relevance, diversity, novelty, fairness and long-term satisfaction matter. A personalized system also needs limits on data collection and a way to avoid reinforcing one mistaken assumption.
Acronyms and aliases
recommendation engine synonymrecommender system synonym
Related terms
Frequently asked questions
How does a recommendation system work?
It represents users, items or session context and predicts a relevance score for candidate items. The highest-scoring eligible options are then ranked or displayed.
What is the difference between search and recommendations?
Search responds primarily to an explicit query. Recommendations can use broader context and behavior to surface items even when the user has not named them directly.
