Yizhou Sun, an associate professor at the Department of Computer Science at UCLA, alongside Quanguan Gu won this year’s Web Search and Data Mining Test of Time Award. Their paper “Personalized Entity Recommendation: A Heterogeneous Information Network Approach” co-co-authored with Xiao Yu, Xiang Ren, Bradley Sturt, Urvashi Khandelwal, Brandon Norick, and Jiawei Han, was originally published in WSDM 2014.

Recommendation systems previously used matrix factorization of a bipartite graph between users and items. This paper enabled the use of additional information in the form of a heterogeneous information network with multiple types of objects and relations for recommendation tasks for the first time. The User-Item bipartite graph that is traditionally used in recommender systems is just part of such heterogeneous networks. In addition, this paper provides: (1) a personalized recommendation model for different groups of users; and (2) meta-path-based interpretation of why such a recommendation is provided to each group. Note that this paper was inspired by Yizhou’s PathSim paper, which systematically introduced heterogeneous information networks and proposed the concept of meta-path. PathSim paper won the Test of Time Award in VLDB’22 two years ago.

Yizhou and Quanquan’s work has inspired many follow-up papers in both academia and industry and is now commonly accepted that recommendation tasks are an essential link prediction problem in heterogeneous graphs.