Twitter's recommendation algorithm selects the top tweets for the For You timeline by using core models and features to extract latent information from tweet, user, and engagement data. The recommendation pipeline consists of candidate sources, real graph ranking, engagements analysis, embedding space approaches, and scoring with a neural network. The goal is to deliver relevant and diverse tweets to users.
Table of contents
How do we choose Tweets?Candidate SourcesRankingHeuristics, Filters, and Product FeaturesMixing and ServingWhat’s Next?19 Impressions