Data Skeptic Bypassing the Popularity Bias
23 snips
Oct 15, 2025 Václav Blahut, a machine learning researcher at Seznam.cz, dives into the intricate world of personalized news recommendations. He highlights the challenges of popularity bias, explaining how it can skew user exposure toward trending content while neglecting niche interests. Václav introduces the concept of inverse recommendation, where the focus shifts to finding the right users for less popular items. He discusses strategies to balance diversity and business metrics and emphasizes the importance of adapting user profiles through multiple embeddings for a more personalized experience.
AI Snips
Chapters
Transcript
Popularity Bias Amplifies The Rich-Get-Richer Effect
- Popularity bias means popular items get amplified by recommender models, drowning niche content.
- This causes a rich-get-richer effect that can reduce relevance and harm niche publishers.
Real Production Data Powers Research
- Václav works on the newsfeed at seznam.cz and uses production user interactions as his playground.
- He analyzes clicks, time spent, likes, and hides to train and evaluate recommender models.
Use Bandits And Randomness For Cold Start
- For cold-start users, cluster items and use bandit algorithms to sample from clusters for diversity.
- Also inject random recommendations to collect unbiased data about new content.