Exploiting Auxiliary Information for Better Recommendation
summary
Claims that injecting item-side auxiliary data beats plain collaborative filtering on sparsity and cold start, with published hierarchy and KG models, code in Python 3 and PyTorch, and offline gains on Amazon and MovieLens variants across clothing/electronics/home domains. Category model quantifies vertical and horizontal category influence and enforces complementary vs alternative distances; KG model does path mining + RNN encoding + attention pooling. Experiments show consistent lifts over baselines and partial adoption yields measurable improvement in production recommendations.
AI-written summary · about
chapters (4)
- 0:00Start
- 4:27Category Hierarchy
- 5:25Example of Knowledge Graph Ii in the Movie Domain
- 13:53Knowledge Graph
related talks
23mML-Powered Search & Recommendation 101
43mRecommendation System beyond traditional Collaborative filtering
05mAugmenting Visual Information in Knowledge Graphs for Recommendations
32mLet's talk about Recommendation Systems!
37mA primer on recommendation systems.
21mRecsys Keynote: Improving Recommendation Systems & Search in the Age of LLMs
01h 21mUnlocking Personalization: A Deep Dive into Modern Recommendation Algorithms
41mRecommender Systems: Crafting Effective Personalised Experiences