Knowledge Graph Embedding Evolution for COVID-19

Guest post by Dr. Steven Skiena, Stony Brook University

This Success Story is a report on the results of the Northeast Big Data Innovation Hub’s 2020 Seed Fund program.

Steven Skiena and Xingzhi Guo presented this research to the COVID Information Commons (CIC) community as part of the February 2022 Lightning Talks and Research webinar. A recording of this presentation is available via the CIC website.

With this award, the researchers at Stony Brook University built and made available to the research community a time-evolving knowledge graph associated with COVID-19 articles in Wikipedia, tracking changes since the beginning of the pandemic outbreak. With this dynamic visualization tool, researchers can observe how scientific understanding of the COVID-19 pandemic evolved between 2020 and 2022. Moreover, this model has potential future applications, as it can be used to demonstrate how knowledge is created and modified in broader contexts. This powers fundamental research into how embeddings evolve with time. The approach  is described In the paper “Subset Node Representation Learning over Large Dynamic Graphs” authored by Xingzhi Guo, Baojian Zhou, and Steven Skiena, and available on arXiv. As detailed in the paper, the team proposes a new method, namely Dynamic Personalized PageRank Embedding (\textsc{DynamicPPE}) for learning a target subset of node representations over large-scale dynamic networks. Additional information about the project, including datasets and coding, can be found on GitHub. This project was developed with the support of Xingzhi Guo, a Computer Science Ph.D. student at Stony Brook University. 

Conference Presentations:

Subset Node Anomaly Tracking over Large Dynamic Graphs (with X. Guo and B. Zhou), 28th ACM SIGKDD Conf. Knowledge Discovery and Data Mining (KDD 2022), Washington DC, August 14-18, 2022.

Lead PI: Steven Skiena (Stony Brook University)

Steven Skiena is a Distinguished Teaching Professor of Computer Science at Stony Brook University.

He was co-founder and the Chief Science Officer of General Sentiment, a social media and news analytics company. His research interests include algorithm design, data science and their applications to biology.

Skiena is the author of several popular books in the fields of algorithms, programming, and data science. The Algorithm Design Manual is widely used as an undergraduate text in algorithms and within the tech industry for job interview preparation.