Reproduce “JointMap: Joint Query Intent Understanding For Modeling Intent Hierarchies in E-commerce Search” 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
Ali Ahmadvand, Surya Kallumadi, Faizan Javed, Eugene Agichtein
The JointMap framework, developed by researchers from Emory University and The Home Depot, introduces a joint learning model to simultaneously classify e-commerce search queries as commercial or non-commercial and map commercial queries to relevant product categories. It demonstrates statistically significant performance gains over state-of-the-art baselines on real-world datasets, notably improving Macro-F1 by 6.3% for product category mapping while effectively handling class imbalance.
Key insights
- Simultaneously learning high-level query intent (commercial vs. non-commercial) and fine-grained product category mapping enhances overall performance through beneficial transfer bias.
- Explicitly modeling and accurately identifying non-commercial queries is critical for a holistic e-commerce search experience, moving beyond product discovery alone.
- Focal Loss is a crucial component for robust multi-label classification in highly imbalanced e-commerce datasets, significantly improving performance on minority classes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.