Reproduce “Personalized Query Auto-Completion Through a Lightweight Representation of the User Context” 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
Manojkumar Rangasamy Kannadasan, Grigor Aslanyan
eBay Inc. researchers developed a personalized Query Auto-Completion (QAC) method utilizing fastText embeddings to represent short-term user search context. The approach, integrated into a LambdaMART ranker, improved QAC performance on a large-scale e-commerce platform by over 20-30% across various metrics compared to a popularity baseline, and notably surpassed traditional text-based personalization features.
Key insights
- fastText embeddings, trained on user session reformulations, effectively encode syntactic and semantic relationships between queries, making them suitable for representing user context.
- Simple cosine similarity features derived from these embeddings provide interpretable and highly effective signals for personalized QAC, showing clear monotonic relationships with ranking scores.
- Lightweight embedding-based features demonstrate superior performance compared to traditional text-based features and largely capture the information they provide, offering a more efficient and robust personalization approach.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.