Graph Neural Network Based Content Recall and Popularity Bias Mitigation for Alibaba's Home‑Decor Platform
The paper presents Alibaba’s home‑decor platform solution that combines graph‑neural‑network side‑information mining and a multi‑view GNN framework with the DICE causal embedding approach to alleviate sparse user behavior and popularity bias, achieving higher recall accuracy and diversity as demonstrated by offline metrics and online A/B test improvements.