Making GenAI help humans grade better. This is a simplified self-contained version of the code. It is mean to grade a single assignment in a single course at a time.
-
Library function files:
simpleHelper.py: Contains all functions needed to retrieve the CSV and submission data, build a prompt, and get gradings from ChatGPT.
Contains functions primarily focused with the analysis and viewing of the data after getting grades from ChatGPT. -
Main Notebook:
GraderGPT_SimpleNotebook.ipynb: Notebook : Consists of 3 parts, Pre-processing, Grading, and Analysis. Runs in single processing only.
Use requirements.txt to install the needed Python libraries for running the notebook.
Run the command:
pip install -r requirements.txt
Based on your operating system, you also need to install pandoc, following instructions from here: https://pandoc.org/installing.html
If you are using custom assigment decription data extracted from the assignment_description column in the xxxx_assignments.csv:
You will need to add a cleaned_description column that contains this custom data.
If you have assignments that use submission data from a different assignment, you must specify the source assignment_id in the xxxx_assignments.csv:
You will need to add a file_submission_source column that contains this additional data.
If using this column to specify the source assignment_id, any blank cells in the column will cause that assignment to be skipped for grading.
Hence, for assignments whose submissions are present within the same assignment, the file_submission_source will simply be its own assignment_id.