Streaming Recommendation

Data Science Northeast Netherlands, Meppel, The Netherlands. Oct 13, 2016.

Summary

This presentation examines streaming recommendation systems across multiple content domains, comparing the unique challenges and approaches for recommending music, movies, and news articles in real-time environments. The work highlights domain-specific constraints including feedback mechanisms (implicit vs. explicit), content shelf-life variations, and the cold start problem, particularly for news recommendation where traditional collaborative filtering fails due to rapid content turnover. The research contributes practical insights into how recommendation strategies must be adapted based on content characteristics, user behavior patterns, and temporal dynamics, with particular emphasis on news recommendation systems that process thousands of new articles daily while maintaining personalization and diversity in suggestions.

Slides

📄 View Slides PDF directly

Links

Slides
Meetup