Title: Entity Resolution on Web-scale Knowledge Graphs

Speaker: Mayank Kejriwal

Abstract

Entity Resolution (ER) is the algorithmic problem of determining when two or more entities refer to the same underlying entity. The problem has existed for more than 50 years in the computational community, including patient record linkage, census linkage, and more recently, knowledge graphs (KGs). While many techniques for ER have been developed over the years, particularly for structured datasets, robust and low-supervision solutions for Web-scale knowledge graphs in myriad domains continue to present a challenge. In this talk, I will provide an overview of the landscape of ER research, and identify four challenges that make ER difficult for Web-scale KGs. I will then discuss the multi-community progress that has been made in the last decade on addressing these challenges in joint architectures. I will also identify some open problems and issues that are likely to be promising avenues of research in the upcoming decade.

Slides

Bio

Dr. Mayank Kejriwal is a research lead at the University of Southern California Information Sciences Institute, and a research assistant professor in the USC Department of Industrial and Systems Engineering. Dr. Kejriwal's research is primarily focused on knowledge graphs (KGs), including construction, inference and applications, such as entity resolution and link prediction. Dr. Kejriwal's research has been funded by DARPA, IARPA, corporate grants and philanthropic foundations. His work on KG-centric solutions for human trafficking has been transitioned and recognized on a national level. He is co-author of Knowledge Graphs: Fundamentals, Techniques and Applications (MIT Press, March 2021).