Eugene Wu, assistant professor of computer science at Columbia Engineering, has won a National Science CAREER Award for his research proposal, "Visual Database Interfaces." This prestigious award, which is one of the top recognitions for young faculty at or near the beginning of their careers, will support Wu's development of scalable techniques to generate visual database interfaces (VDIs) tailored to specific analysis tasks.
As users and companies collect more and more data into massive databases, data analysis has become an increasingly critical tool for millions of users. A major stumbling block, however, has been the complexity of the primary database interface, SQL. Wu's goal is to build VDIs that enable people in industry and the sciences to more easily access, visualize, analyze, and monitor data.
Cumbersome even for experts to use for data analysis, SQL (and databases in general) are inaccessible to most less-technical users. To address this challenge, developers currently build custom visualization interfaces for end-user use cases. While custom-designed interactive visualizations are faster and simpler to use, it is both difficult to find the necessary technical expertise, and expensive to develop these custom visualization interfaces. In addition, the process of both understanding a user's specific analysis needs and then designing and implementing a bespoke interactive interface for each user and task is very expensive and time consuming.
"Just as people have access to specialized apps for any given task (e.g., photo filtering apps like Instagram instead of using a generic application like Photoshop), our VDIs will ensure that there is a customized visual analysis interface for any given data analysis need," said Wu, who joined Columbia Engineering in 2015. "Our work will be far-ranging and useful not only in science but also in business, media, and any domain that relies on data."
Wu's vision for VDIs involves the development of two novel systems. The first system scalably extracts analyses from historical or live query logs and automatically generates interactive interfaces that are customized to each analysis. The second is a human-in-the-loop interface design tool that enables designers to improve the interaction design of VDIs while knowing which optimizations and systems modifications are needed to ensure that interactions remain responsive.
This project builds upon Wu's extensive expertise in both database research and visualization. It is also a crucial step towards his larger project to drastically simplify and automate the process by which new human-data interfaces are created so that everyone can easily use and benefit from data.