Yue Wang

Yue Wang


I have completed my PhD degree at U of M and started as an assistant professor at the University of North Carolina at Chapel Hill.
Here's my new homepage!

Computer Science & Engineering
Department of Electrical Engineering & Computer Science
University of Michigan

3332 North Quad, 105 S. State St. Ann Arbor, MI 48109

I am a member of the Foreseer group led by Professor Qiaozhu Mei in the School of Information, University of Michigan. I received both my undergraduate and master's degrees from School of Information Security Engineering, Shanghai Jiao Tong University. I received a dual master's degree from School of Electrical and Computer Engineering, Georgia Institute of Technology.

Research

I am broadly interested in text data mining, including related areas such as machine learning, information retrieval, natural language processing, information and social networks, and health informatics.

My research is focused on interactive machine learning. We see "big data" almost everywhere, but turning massive unlabeled text data into accurate models and reliable knowledge requires significant human effort. We can reduce the effort by enabling machine learning algorithms to interact with humans, and the classical "active learning" is a first step. Why sometimes uncertainty sampling learns even slower than random sampling? What if the data is only accessible via search (e.g. Google's Web index)? What if the interesting class is extremely rare (e.g. e-discovery)? What if the human has rich domain knowledge beyond class labels (e.g. medical domain)?

My thesis develops principled interactive machine learning algorithms. I present a novel game-theoretic framework that unifies passive and active learning algorithms, providing guidance to the design of new interactive learning algorithms (work in progress). I propose algorithms that invites human users to efficiently teach machines in natural and intuitive interaction modes: searching with keywords (SIGIR'14, AMIA'16), labeling features, highlighting rationales (AMIA'17 talk, JAMIA submission), in addition to labeling examples. My work is inspired by and applied to various data mining problems, including high-recall retrieval, literature review, content analysis, federated web search, and clinical natural language processing.

News

Publication

See also my Google Scholar profile.

Teaching

Service

Conference Organization
Conference Review
Journal Review

Mentorship

Internship

Personal