How Search & Recommendation Technologies Evolve: Insights from Suning’s 2018 Conference

On December 15, 2018, Suning Search & Recommendation Technology Conference was held at Suning headquarters in Nanjing, gathering over 400 search and recommendation experts from major internet companies.

Keynote speaker Jia Hongyuan, head of search R&D at Suning.com, emphasized that search and recommendation systems act as a bridge between users and information, stressing stability, intelligence, and the need for open communication and collaboration.

Guo Chaowei from Suning presented the evolution of Suning’s product search engine, describing its transition from open‑source solutions to custom‑developed engines with extensive optimizations in functionality, recall, ranking, and operations.

Liu Shangkun of Alibaba’s Youku algorithm team shared how they redefined multimedia video retrieval for the video industry, leveraging knowledge graphs, entity extraction, content‑level understanding, entity linking, and intent analysis to enhance video distribution depth and breadth.

Professor Dai Xinyu of Nanjing University analyzed the opportunities and challenges of multi‑source data fusion for recommendation systems, highlighting recent advances in neural‑network‑based fusion models and outlining future research directions.

Zhang Dongxu of iFlytek discussed the evolution from traditional display ads to interactive and AI‑driven advertising, presenting iFlytek’s AI capabilities, recommendation system construction ideas, and solutions to optimization challenges.

Zhang Huaquan of 360 introduced recent explorations in query understanding, including semantic parsing, query normalization, and query rewriting techniques.

Chen Wei of Yuewen Group described recommendation systems as filtering mechanisms that help users quickly find needed content and enable merchants to precisely target users, thereby enhancing user‑merchant interaction.

Sun XinXin of Suning.com explained the evolution of algorithm models in Suning’s recommendation system, comparing recall and ranking models, their advantages and disadvantages, and showing performance improvements after each iteration.

Attendees were invited to submit any questions about the conference content via the “SuningTech” WeChat public account, where editors would consult experts and provide feedback.