Evaluating Embedding-Based Retrieval Beyond Historical Search Results

Explore innovative approaches to evaluating embedding-based retrieval (EBR) in e-commerce search beyond traditional historical search results. Delve into Yu Cao's 30-minute conference talk from Haystack EU 2023, where he shares insights on leveraging organizational knowledge to assess query and item embedding quality. Learn about efficient methods for predicting relevance in null and low search result scenarios, addressing the limitations of user-engagement data. Discover how EBR models can improve relevance while expanding recall, particularly for challenging queries. Gain valuable perspectives from Yu Cao, an Applied Researcher at eBay with a background in linguistics, as he discusses the intersection of formal theories and learning models in solving real-world e-commerce search problems.