In the introductory series of Computer Science, Section Leaders spend a disproportionate amount of time grading students on style, and that’s a huge bottleneck for how many students the school can handle. We want to investigate how a machine learning approach could be used to automate the process of style grading coding assignments
Our algorithm (run it, look into main.py
) depends on a folder called data at the root level. We omitted the data from the repo. If you are interested, email gdasilva@stanford.edu.
- Feature extraction is still naive. Look into more meaningful information to extract from samples of code
- Look into models that are more appropriate for our context.