STATEMENT

One focus of my prior research is to understand documents text jointly with their visual layouts. From manuscripts earlier than A.D. 200 to corporate reports in the 21st century, documents have always been typeset with intricate layouts that optimizes for human readability while being difficult for machine perception. I design methods and tools that helps parse these complex documents, some of the represent work include LayoutParser and VILA.

Recently, I am also interested in Human-AI Collaboration. From interfaces (e.g., PAWLS) to algorithms like OLALA, my goal is to enable better communication between humans and machine learning models, and use AI to empower humans in work and creative endeavors.

EXPERIENCE

2020 Oct

Allen Institue for AI (Semantic Scholar)

Predoctoral Young Investigator

2019 Sep

Harvard University (IQSS)

Data Science Fellow

2019 May

NEWS

2021 Oct

The LayoutParser project is featured in Standford Digital Economy Seminar Series.

2021 Sep

Give a talk on LayoutParser and Document Intelligence at Online Seminar in Economics + Data Science @ ETH Zurich

2021 Sep

The LayoutParser talk video is available on Youtube.

2021 Jun

Please check our new paper on Scientific Document Parsing - Incorporating Visual Layout Structures for Scientific Text Classification.

2021 Jun

The LayoutParser paper is accepted as an oral paper at ICDAR 2021.

TALKS

2021 Sep

A Unified Toolkit for Deep Learning Based Document Image Analysis

Online Seminar in Economics + Data Science @ ETH Zurich
2020 Mar

Information Retrieval over Historical Scans with Non-trivial Layouts

Kaggle Days Meetup Boston
2019 Feb

Deep Learning-based Framework for Automatic Damage Detection in Aircraft Engine Borescope Inspection

ICNC 2019, Honolulu, Hawaii