How Alibaba’s AI Team Boosted Search Ranking by 21% to Win CIKM Cup

Recently, Alibaba's Search Division Natural Artificial Intelligence team won the CIKM Cup international data mining competition, improving the benchmark search ranking metric by 21.28% using deep learning, machine learning, and ensemble methods.

CIKM, founded by the ACM in 1992, is a top academic conference in information retrieval, knowledge management, and databases, gathering leading researchers to discuss challenges and set future research directions. The CIKM Cup is an international data mining contest organized around the conference theme.

The team’s leader, Si Luo, noted that search ranking and personalization are increasingly vital in e‑commerce, impacting user experience and GMV, and that Alibaba heavily applies AI technologies to its search system.

Competition winners Wu Chen and Yan Ming explained that they employed ensemble learning, extensive model tuning, and feature selection, using a Learning‑to‑Match approach from natural language processing to represent Learning‑to‑Rank, and combined features with deep neural networks to capture strong retrieval intent in user queries.

Si Luo added that the Natural Artificial Intelligence team focuses on NLP, machine learning, and deep learning research, and that participating in the CIKM Cup validates innovative algorithms and fosters exchange with peers.

In 2015, during the Double‑Eleven shopping festival, the Search Division’s personalized recommendation technology earned Alibaba’s top “CEO Award,” highlighting the commercial success of AI. Recent years have seen Alibaba recruit global AI experts such as Zhou Jingren, Jin Rong, and Hua Xiansheng; Si Luo himself was a tenured professor at Purdue before joining Alibaba in 2014.