AP Lang FRQ Grader

"AP Essay Grader: AI-Powered Feedback for Rhetorical and Argumentative Essay Mastery." Instant, reliable scoring and feedback based on College Board standards.

Comment

Inspiration

As a high school student who took AP Lang this year, I noticed how hard it is to get consistent, fast, and accurate feedback on essay practice especially for AP English Language’s FRQs. Grading rubrics are long and subjective, and peer feedback is often unavailable or inconsistent. We wanted to build a cost-effective, accessible alternative to tutoring: a tool that helps any student, anywhere, improve their writing with AI-powered scoring aligned to the official College Board rubric.

What it does

AP Lang FRQ Grader uses AI to automatically evaluate rhetorical and argumentative AP Lang essays using the official College Board rubric (0–6 scale across 2 categories: Rhetorical Essay and Synthesis Essay). Students can select an essay type (Rhetorical or Argument), paste their prompt, passage, and essay, and get instant scores along with rubric-based feedback.

How we built it

Frontend: Built with vanilla JavaScript, HTML, and Tailwind CSS, deployed via Node Backend: We used a fine-tuned GPT-4 model trained on real College Board samples with rubrics to emulate AP readers. Data: We formatted sample student essays and official grader feedback into JSON chat format for fine-tuning. OpenAI API: The frontend securely sends essay data to our fine-tuned model via a protected HTTP endpoint.

Challenges we ran into

Most College Board essay samples are scanned and handwritten, making them hard to use for AI training. We had to carefully clean and format the data to match OpenAI’s fine-tuning structure. Crafting a clear interface for submitting prompts, passages, and essays required multiple iterations.

Accomplishments that we're proud of

We fine-tuned a custom GPT-4 model that replicates AP-level essay scoring with nuanced commentary. The grading ability of our custom fine-tuned GPT-4 model is highly accurate: in testing, it only showed a 1-point margin of error in the Evidence and Commentary section, proving its strong alignment with official AP scoring standards.

What we learned

How to fine-tune large language models with custom JSON chat data. How powerful AI can be in democratizing educational feedback when aligned with real-world rubrics.

What's next for AP Lang FRQ Grader

Add support for Synthesis essays once we acquire more training data. Add rubric explanations for each score to help users understand their feedback better. Build analytics features so users can track their writing progress over time. Launch as a free tool for AP classrooms to use during practice sessions or homework.

Built With

Share this project:

Updates