How Alibaba’s Search & Recommendation Evolved: From Rules to Cognitive AI

Introduction

On September 28, 2018, Alibaba’s e‑commerce search division held a technical summit titled “Search·Recommend the Future.” The talk reviews the past evolution of search and recommendation algorithms and looks ahead to future developments, aiming to combine machine intelligence with human cognition for new interactive experiences.

1. Characteristics of Taobao Search

Taobao hosts billions of items across thousands of leaf categories and hundreds of top‑level categories. The primary challenge is helping users find items that match their intent.

While the overall architecture resembles traditional search engines—data collection, analysis, indexing, retrieval, relevance scoring, ranking, and feedback mechanisms—Taobao’s commercial nature introduces unique technical aspects.

Rapid data updates : New products are added daily, and existing listings constantly change (titles, descriptions, prices, images, availability). These changes must be reflected in search results in near real‑time, unlike the slower update cycles of web search.

Image‑driven relevance : Product images play a crucial role in user research and purchase decisions, requiring effective utilization of visual information alongside text.

Full‑link behavior : Search, comparison, and purchase all occur within Taobao, providing rich pre‑search, in‑search, and post‑search user behavior data that can be leveraged for better ranking algorithms.

Ecosystem considerations : Search ranking must balance user experience with commercial rules to ensure fairness, traffic distribution, and support the livelihoods of countless merchants who depend on Taobao.