My first research aim is building robust Natural Language Processing (NLP) models that can automatically understand and learn from large-scale texts of various domains and genres, and produce summaries of high quality. My group studies domains including newswire services, patent documents, dialogues, and social media data. Additionally, I also endeavor to create novel applications based on neural text generation techniques, e.g., generating persuasive arguments.
The second research direction of my group is modeling social interaction via the lens of argument mining and conversation understanding. In this direction, I study conversations and debates via content, language style, and user interaction modeling. I also design argument mining models to characterize the structure and elements of peer reviews, to allow the understanding of what factors make reviews constructive and how to promote effective peer reviewing.
I am also passionate about designing customized NLP algorithms to solve interdisciplinary challenges. Along with collaborators from political science, network science, and economics, I study how media bias is formed and expressed in news outlets, and how misinformation is shared and can be automatically detected. I also work with visualization researchers to design novel algorithms to visualize text with complicated structures, e.g., debates.