Data visualization is necessary for human cognition as the amount of information we are exposed to increases. One strategy is to build a visual representation of the data as a chart or plot. As part of the Data Visualization course of the online master’s program offered by the Department of Computer Science at Rice University, we built a system to automatically check the correctness of plots submitted by the students. For classes with large enrollments, manually grading these plots requires a significant amount of the instructor’s time. Since these plots might be built using different visualization packages, the plots may be visually equivalent but have different underlying structures due to the fact that each package has a distinct API. The non-unique representation of plots and the variety of plotting packages is problematic both for the machine checker to grade and for students to conceptually understand.

We designed a hierarchical universal representation of plots that accounts for the variety of package-specific structures of plots, a machine checker that compares each student’s plot to the instructor’s plot in the universal representation, and a universal specification that dedicates how the machine checker should work. Together, they form a system for automatically checking students' plots.

We tested this plot checking system on student assignments in the Data Visualization course and reported the results.