Read the original document by opening this link in a new tab.
Table of Contents
1. Introduction
2. Methodology
3. Results
4. Summary and Future Work
Summary
This paper addresses the issue of personalization in ecommerce search by developing personalized ranking features that utilize in-session context to enhance a generic ranker optimized for conversion and relevance. The authors combine latent features learned from item co-clicks in historic sessions with content-based features using item title and price. The study shows a significant improvement in ranker performance without requiring an explicit re-ranking step or complex modeling of user profiles. The proposed technique outperforms a generic ranker in terms of Mean Reciprocal Rank (MRR) and demonstrates the effectiveness of combining content-based and content-agnostic features. The study highlights the application of these techniques in the ecommerce domain and suggests avenues for future research.