Unbiased Neural Ranking Model for Product Search

Explore an innovative approach to product search ranking in this conference talk from Haystack EU 2022. Dive into Otto's implementation of neural networks for Learning to Rank (LTR), leveraging multiple implicit customer feedbacks like clicks and orders. Learn about their strategy to address position bias using a separate bias-estimator. Discover the architecture of their neural ranking model, including an encoder for generating embeddings of user queries and product descriptions, and a transformer-based scoring function that models competition between products in search results. Understand how the ranking function is embedded in a multitask learning framework to learn from different relevance signals. Gain insights from Laurin Luttman, a working student at Otto with extensive experience in machine learning, as he presents his master thesis research on unbiased neural ranking models for product search.