Lee aims to design systems that provide automated recommendations in visual data exploration and machine learning
2020 Facebook Fellowships were awarded to 36 Ph.D. students out of more than 1800 applicants, and Doris Lee, a School of Information student working with Assistant Professor Aditya Parameswaran, was selected as a recipient.
Fellowship winners receive two years of paid tuition and fees and a $42,000 annual stipend to cover living and conference travel costs. Lee’s fellowship will support her research in developing automated systems to help data scientists identify fruitful approaches to the analysis and visualization of big data sets. Her work aims to provide automated guidance of visualizations, model decisions, and analyses to help users arrive at their high-level discovery goals more quickly. The goal is to enable them to spend less time doing tedious manual searching, reduce the possibility of becoming overwhelmed or getting lost in the data, and avoid suboptimal or even erroneous model or parameter choices.
“My work is about building tools that enable people to more easily access and learn something new from their data,” Lee said. “Typically, when people in data science do exploratory analysis, they want to get their feet wet in the data and get an overview of what things look like, what are the charts and graphs they might be able to generate. But to do this currently is a very tedious process.”
Lee’s research — she calls it a ‘visualization recommendation system’ — seeks to make the process easier for data scientists and others. “Think of it as similar to what Netflix does for its users,” Lee said. “If you want to watch a movie but aren’t sure which one, Netflix can make a recommendation to you based on your history or your interests.” Lee’s work can help data scientists sift through their data in a similar fashion.
Her advisor, Assistant Professor Parameswaran, said that because of Lee’s undergraduate degree in physics and astrophysics (UC Berkeley ’16), “Doris brings a keen sensibility for the data science problems faced by domain scientists. She then combines that with technical software system building chops to develop tailored as well as general purpose tools for making data exploration easy."
Lee considers the I School the perfect setting for this research, as close collaboration with users is an essential part of her work, and the school provides ample opportunity for engagement with other researchers as well as the practitioners who may benefit from the technology.