Reproduce “Improving Ad matching via Cluster-Adaptive Keyword Expansion and Relevance tuning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Dipanwita Saha, Anis Zaman, Hua Zou, Ning Chen, Xinxin Shu, Nadia Vase, Abraham Bagherjeiran
eBay researchers developed a system that uses cluster-adaptive keyword expansion and relevance tuning to improve ad matching. The system increased ad impressions by 3% and generated a net-positive impact on total ad revenue by connecting more user queries with relevant product listings, while maintaining slot-neutral CTR and reducing RMSE for relevance judgments.
Key insights
- Adaptive, cluster-specific similarity thresholds are critical for effectively balancing precision and recall in semantic keyword expansion across diverse e-commerce concepts.
- Incremental learning with lightweight models can efficiently refine a complex, pre-existing relevance model for new data distributions (expanded keywords) without a full, costly retraining.
- Lightweight post-processing filters for attributes like gender and numerical consistency significantly enhance the precision of semantically expanded keywords in e-commerce contexts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.