I’m a software engineer working on knowledge base at Naver Corp.
I’m particularly interested in understanding and finding new insights from unstructured texts by transforming the text to structured form. Ultimately, I would like to develop a simple, robust, and efficient model with an interpretable inference process that can be easily applied to industry.
My most recent works are focused on relation extraction, geometric embedding, graph neural networks, and knowledge graphs. I primarily worked with members at Information Extraction and Synthesis Laboratory (IESL) and IBM Research while receiving a MS in Computer Science at the University of Massachusetts-Amherst.
Revisiting Virtual Nodes in Graph Neural Networks for Link Prediction
Eunjeong Hwang, Veronika Thost, Shib Sankar Dasgupta, Tengfei Ma
(Under Review at ICML 2022)
Event-Event Relation Extraction using Probabilistic Box Embedding
Eunjeong Hwang, Jay-Yoon Lee, Tianyi Yang, Dhruvesh Patel, Dongxu Zhang, Andrew McCallum
ACL 2022 (To Appear)
Interdependency between the stock market and financial news
Eunjeong Hwang, Yong-Hyuk Kim
IEEE BigData2019 workshop
Joint constrained learning using box embedding
Implemented relation extraction model using box embedding. Effectively reduced constraint violations in relations labels by utilizing box’s inherent ability of handling anti-symmetric relations, which indicates the output from the model is more coherent than our baseline vector model.
Virtual node augmented graph neural networks for link prediction
Proposed virtual node augmented graph neural networks by using single virtual node to multiple virtual nodes for link prediction to capture long-range dependencies in graphs. Our model constantly produce better performance on large and dense graphs, such as ddi, collab, and ppa (OGB benchmark datasets).
Question answering on knowledge graph using box embedding
Implemented question answering model by embedding queries as probabilistic boxes using gumbel distribution. Our model was more mathematically explainable than Query2Box model and produced similar performance.
Interdependency between stock market and financial news articles
Analyzed the interdependency between stock market and financial articles using sentiment analysis. Discovered trends that stock prices respond to social issues before the articles do.
Information Extraction and Synthesis Laboratory (IESL), UMass Amherst, 2020.05 - 2020.08
Naver, S. Korea, 2021.03 - Present
Software Engineer at Knowledge Base team
IBM, S. Korea, 2018.01 - 2019.04
IBM, S. Korea, 2017.09 - 2017.12
Application Developer Intern
The Development Factory, Australia, 2017.01 - 2017.02
Software Engineer Intern
University of Massachussets-Amherst, 2021.02 - 2021.05
Grader at CS685 Graduate Natural Language Processing course (Prof. Brendan O’Connor)