Learning to Rank at Reddit: A Project Retrospective
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only! Grab it Explore the challenges and lessons learned in implementing Learning to Rank (LTR) for Reddit's site search in this 50-minute conference talk from Haystack US 2024. Gain insights into how Reddit's search team is working to improve the search experience, aiming to make it unnecessary for users to add "Reddit" to their Google searches. Discover the team's journey in transforming relevance into a data-driven, repeatable solution, including their experiences with training data, feature development, the Solr Learning to Rank plugin, and scaling LTR to handle thousands of queries per second. Learn from the speakers' candid sharing of obstacles faced and solutions devised as they progress from a laboratory environment to an assembly line approach for continuous, data-informed improvement. Benefit from the expertise of Doug Turnbull, Chris Fournier, and Cliff Chen as they discuss their roles in enhancing Reddit's search capabilities and share valuable insights applicable to various search optimization projects. Read more