Name: SEA: November - Recommendations
Start: 2025-11-28T16:00:00+01:00
End: 2025-11-28T17:00:00+01:00
Location: Lab 42

**Location:** Room L3.36, Lab42, Science Park Amsterdam

And here is the **zoom [link](https://uva-live.zoom.us/j/69777529870).**

We’re excited to announce our upcoming **SEA Talk** on *Recommendations*, featuring two speakers:

**Speaker #1**: **Binyam Gebre**

**Title**: Embedding-Based Personalization without Real-Time Inference: The Bol Approach
**Abstract**: In personalized recommender systems, embeddings are widely used to represent customers and items, enabling retrieval through approximate nearest neighbour search. However, this approach faces two key challenges: 1) a single customer embedding often collapses diverse interests, and 2) maintaining fresh customer embeddings requires costly real-time infrastructure. In this talk, I present a method that addresses both issues by representing customers through their recent item clicks and purchases and retrieving candidates from precomputed nearest neighbours in embedding space. This eliminates the need for real-time model inference while preserving diversity (through multi-query retrieval) and real-time responsiveness (through on-the-fly profile construction). We deployed this approach to personalize promotional items at Bol, one of the largest e-commerce platforms in the Netherlands, achieving a 4.9% uplift in conversions. The method demonstrates that scalable personalization can be achieved without expensive real-time infrastructure when exhaustive precomputation is feasible.
**Bio**: Binyam Gebre is a Lead Data Scientist at Bol, where he develops deep learning solutions for large-scale recommender, search, and advertising systems. He previously worked as a Deep Learning Scientist at Philips Research in Eindhoven. He earned his PhD from Radboud University Nijmegen in 2015.

**Speaker #2**: **Thorsten Krause**

**Title**: A Non-Parametric Choice Model That Learns How Users Choose Between Recommended Options (RecSys'25)
**Abstract**: Choice models predict which items users choose from presented options. In recommendation settings, they can infer user preferences while countering exposure bias. In contrast with traditional univariate recommendation models, choice models consider which competitors appeared with the chosen item. This ability allows them to distinguish whether a user chose an item due to preference, i.e., they liked it; or competition, i.e., it was the best available option. Each choice model assumes specific user behavior, e.g., the multinomial logit model. However, it is currently unclear how accurately these assumptions capture actual user behavior, how wrong assumptions impact inference, and whether better models exist.
In this work, we propose the learned choice model for recommendation (LCM4Rec), a non-parametric method for estimating the choice model. By applying kernel density estimation, LCM4Rec infers the most likely error distribution that describes the effect of inter-item cannibalization and thereby characterizes the users’ choice model. Thus, it simultaneously infers what users prefer and how they make choices. Our experimental results indicate that our method (i) can accurately recover the choice model underlying a dataset; (ii) provides robust user preference inference, in contrast with existing choice models that are only effective when their assumptions match user behavior; and (iii) is more resistant against exposure bias than existing choice models. Thereby, we show that learning choice models, instead of assuming them, can produce more robust predictions. We believe this work provides an important step towards better understanding users’ choice behavior.
**Bio**: Thorsten Krause is an external PhD student at Radboud University under the supervision of Harrie Oosterhuis and Data Scientist in the search ranking team at OTTO. His research spans choice modelling for and exposure bias in recommender systems.

**Counter:** SEA Talks #293 and #294.

Maarten de Rijke

Yubao Tang

SEA: Search Engines Amsterdam

Community & Environment

Science

Apache Solr

Technology

Information Architecture

Elasticsearch

Big Data

Web Technology

Software Engineering

Apache Lucene

SEO (Search Engine Optimization)

SEA: November - Recommendations

Artificial Intelligence

Information Retrieval

Lab 42

Delen

SEA: Search Engines Amsterdam

SEA: November - Recommendations

SEA: Search Engines Amsterdam

Details

Leden zijn ook geïnteresseerd in