Navigating Neural Search: Avoiding Common Pitfalls

Explore the intricacies of neural search in this informative conference talk from Haystack EU 2023. Delve into the world of semantic search and learned vector representations as Jo Kristian Bergum, a Distinguished Engineer at Yahoo, demystifies the neural networks behind text embedding models. Gain insights into selecting appropriate open-source text embedding models by considering factors such as language capabilities, task alignment, accuracy, and cost-effectiveness. Examine the challenges of embedding retrieval and vector search, including the potential accuracy degradation caused by approximate vector search. Learn how this degradation can lead to more cost-effective retrieval methods becoming favorable alternatives. Benefit from Bergum's 20 years of experience in deploying search systems at scale as he shares valuable knowledge about navigating neural search and avoiding common pitfalls in this 42-minute presentation.