Researchers have developed a personalized video search system for Apple TV that enhances search result relevance by combining text and ID-based embeddings. The system uses a hybrid approach, learning separate embedding spaces for semantic understanding and collaborative filtering, and integrates these into an XGBoost ranker. Evaluations showed significant improvements in metrics like NDCG@10 and MRR, particularly for ambiguous prefix queries and users with extensive watch histories. An online experiment confirmed these gains with increased tap-through and conversion rates. AI
IMPACT Improves relevance and user experience in video search through advanced embedding techniques.
RANK_REASON Academic paper detailing a new personalization system for video search. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
