Personalization in IR is one of the hottest topics in the AI-takes-all economy: we should not aim to be "just" semantically relevant, but also tailor results to users' preferences and intent. However, personalization in digital commerce is easier said than done: most shoppers visit a given store no more than twice a year, and bounce rates across verticals show that it is important to personalize as early as possible. In this talk we share effective strategies to tackle the challenge of "in-session personalization" in NLP and IR tasks, starting from the straightforward case of "one-shop" personalization and then generalizing to more. On the business side, we argue using industry benchmarks and data from our network that in-session personalization is a fundamental part of any relevance journey in the hyper-competitive e-commerce market. On the tech side, we build on the latest deep learning trends to show how increasingly sophisticated representations of real-time intent can power personalization in product search. Once dense architectures are in place, we are ready to tackle the challenge of "transfer learning": can shopper's intent be transferred from one shop to another without annotated data? Using insights from lexical learning, sessions from different shops can be projected in the same space to power "zero-shot" personalization for downstream NLP services; finally, we prove with quantitative and qualitative benchmarks that zero-shot predictions are a significant improvement over industry baselines.
12.06.2020 18:20 – 19:00