Researchers have developed SPEAR, a novel framework for community search that enhances query reformulation and retrieval effectiveness. SPEAR addresses the misalignment in current systems by integrating three components: a dual-embedding backbone to protect recall-side semantics, a multiplicative gating aggregator to prevent generic-word shortcuts, and a dynamic rewrite selector for adaptive calibration. Deployed in Dewu's community search platform since 2025, SPEAR has demonstrated significant improvements in offline evaluations, including a +18.2 increase in rewrite semantic similarity and a +99.5 boost in click recall. Online A/B testing confirmed its efficacy, showing a +0.259 increase in query-view CTR and a +0.733 increase in average reading depth. AI
IMPACT Improves e-commerce search by enhancing query understanding and retrieval accuracy, potentially leading to better user engagement and conversion rates.
RANK_REASON The item is a research paper detailing a new framework for information retrieval and search. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
- alphaXiv
- CatalyzeX
- Connected Papers
- DagsHub
- Dewu Township
- Gotit.pub
- Hugging Face
- Litmaps
- ScienceCast
- scite Smart Citations
- SPEAR
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
