I worked with Professor Adam Blank on a project to help students ask better questions using the online office hours question tool. I developed two models that filtered questions on whether they were an actual question ("What is a Linked List" vs "pls help") and whether or not the question was answerable. I created a data labeling interface label_data.py and a model training interface train_models.py to aid in this task. question_funcs.py holds the functions that return T/F for a given question for the given model.
CV Accuracy of actual question model: ~91%
CV Accuracy of answerable model: ~85%