Student Submission Grader that ingests Canvas discussion posts and peer comments, scores length/structure, referencing, and engagement, then combines and rounds to a 0–100 grade and generates concise feedback. Automates Canvas navigation with Selenium, uses LLM blocks for rubric-aligned judgments, and exports per-student artifacts for auditability. Inputs: Canvas discussion post + comments. Flow: Grade length → Grade referencing → Grade engagement → Combine & round → Generate feedback. Output: numeric grade (0–100) and 1–2 sentence feedback. Features: skip already graded, CSV export, configurable level→grade map, timing/accounting logs, and optional human verification before submit.
Project Purpose:
Automate discussion grading in Canvas using deterministic LLM blocks for consistent, explainable results.
System Diagram:
Inputs:
Canvas discussion submission + comments to other students.
Flow:
Grade on length → Grade on referencing → Grade on engagement → Combine and round grade → Generate feedback
Output:
0–100 grade and two short feedback sentences (one sentence for top scores).
Technical Key Features:
Challenges: