Reproduce “AliCoCo: Alibaba E-commerce Cognitive Concept Net” 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
Xusheng Luo, Luxin Liu, Yonghua Yang, Le Bo, Yuanpeng Cao, Jinhang Wu, Qiang Li, Keping Yang and 1 more
AliCoCo is a large-scale e-commerce Cognitive Concept Net designed to bridge the semantic gap between user shopping needs and platform item organization, comprising over 5 million e-commerce concepts and 2.8 million primitive concepts linked to billions of items. This infrastructure notably improves search relevance and powers cognitive recommendation systems, covering over 75% of actual daily shopping needs.
Key insights
- Formalizing "e-commerce concepts" as explicit nodes representing high-level user needs and shopping scenarios effectively bridges the semantic gap between user intent and platform item organization.
- A semi-automatic, multi-layered knowledge graph construction approach, combining machine learning with targeted human expertise and active learning, is crucial for building and maintaining a scalable and high-quality knowledge base in a dynamic e-commerce environment.
- Integrating external common sense knowledge significantly enhances models' ability to understand abstract or short concepts and improves semantic matching between user needs and items, particularly where direct textual similarity is low.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.