The success achieved by deep learning algorithms in image recognition
and speech recognition has brought AI closer to our lives, but these systems
are still limited by their ability to learn via imitation instead of learning
by understanding the context. The key to unleashing the potential of AI technologies
lies in how to develop their "cognitive" problem solving skills. Zhongyuan
Wang is helping machines understand human language by building knowledge graphs
that map complex relationships among billions of entities.
“At present, the best way to resolve the problem is to build a
knowledge graph which contains common sense knowledge, encyclopedia knowledge, and the relationships between things, and then construct an inference system upon this
knowledge graph, so that machines can be more intelligent and will be able to leverage
knowledge for different scenarios,” Wang said.
Wang joined Microsoft Research Asia in 2010 and was responsible
for developing the Microsoft Concept Graph. This is one of the representative pioneering
large-scale knowledge graph projects in the world. Wang’s original research
achievements include the invention of an innovative text understanding approach
called Conceptualization. His approach is the first to use a probabilistic common-sense
knowledge graph that is as rich as our mental world in terms of the concepts
(of worldly facts) it contains. This work significantly improved text
understanding and won the ICDE 2015 Best Paper Award. In 2016, Wang joined
Facebook as a Research Scientist, and led a team to build the world’s largest
entity linking production service, which was one of the most important NLP
services in Facebook.
In 2018, Wang joined Meituan-Dianping — a Chinese leading
e-commerce platform for location-based food delivery services, broad lifestyle, and travel services. Wang is
building a world-class NLP team as head of the Search and NLP Department of the
company. Based on NLP, deep learning, and knowledge graph techniques, his team
handles billions of reviews about restaurants, travel destinations,
entertainment options, and other location-based services. They are constructing
the world’s largest food and entertainment knowledge graph called Meituan Brain.
On top of this, they are building intelligent services to empower Meituan-Dianping’s
business and help people eat better and live better. For example, Meituan Brain
helps to significantly improve Meituan’s search and recommendation system. It also
helps restaurant owners make better decisions on how to win more customers. In
financial risk management, it has been able to help financial teams to detect
fraud rings.
Wang believes that building knowledge graphs will make artificial
intelligence go further. In the near future, intelligent services will cover
more business scenarios than what we see today.