Back to projects
Mar 03, 2024
3 min read

Automated Essay Scoring System - AI-powered feedback system for student essays

Automated essay scoring (AES) systems, leveraging LLM and NLP, offer a solution by efficiently analyzing essays and providing objective feedback aligned with predefined rubrics.

Automated Essay Scoring System (AESS)

Overview:

This project tackles the challenges of traditional essay grading by creating an AI-powered system that provides efficient, scalable, and objective feedback to students.

Tech Stack:

  • Frontend: Next.js, React
  • Backend: FastAPI (Python)
  • Natural Language Processing (NLP): spaCy, Language Tools
  • Large Language Model (LLM): Gemini
  • Database: MongoDB
  • Additional Services: Firebase

Project Description:

  • Students securely login using social accounts (e.g., Google).
  • Essays can be uploaded (PDF, DOCX, TXT) or directly typed.
  • The system analyzes essays using NLP for grammar, spelling, and readability.
  • A Retrieval-Augmented Generation (RAG) model retrieves relevant rubrics (currently hardcoded).
  • A Large Language Model (LLM) scores the essay based on rubrics and generates personalized feedback.
  • Students receive feedback reports with overall scores, rubric-specific breakdowns, suggestions for improvement, and readability analysis.

Project Objectives & Scope:

  • Automate essay grading to save teacher time and effort.
  • Provide objective feedback based on pre-defined rubrics.
  • Reduce subjectivity and bias in grading.
  • Offer students personalized suggestions for improvement.
  • Enhance the efficiency and effectiveness of essay evaluation.

Technical Challenges:

  • Ensuring accurate and unbiased scoring through NLP and LLM.
  • Developing a robust RAG model for dynamic rubric selection (future work).
  • Integrating the system with existing learning platforms (future work).

What Was Improved:

  • Essay grading efficiency and scalability.
  • Objectivity and consistency of feedback.
  • Student understanding of strengths and weaknesses in writing.
  • Identification and correction of grammatical errors.

How It Works:

  1. Student logs in securely using Google account.
  2. Essay is uploaded or directly typed.
  3. System analyzes essay using NLP and LLM.
  4. RAG model retrieves relevant rubrics (currently hardcoded).
  5. LLM scores the essay based on rubrics and generates feedback.
  6. Student receives a comprehensive feedback report.

Customer Satisfaction (Future Work):

  • Collect student feedback on the clarity, helpfulness, and accuracy of the feedback reports.
  • Analyze feedback to identify areas for improvement.

Screenshots

Login Page AESS Login Page

Essay Upload Page AESS Essay Upload Page blank AESS Essay Upload Page upload file popup AESS Essay Upload Page filled

Feedback Report Page AESS Feedback Report Page AESS Feedback Report Page percentage popover AESS Feedback Report Page spelling/grammatical mistakes

Metrics Page AESS Metrics Page

Suggestions Page AESS Suggestions Page

Feedback Page AESS Feedback Page

Essay Statistics Page AESS Essay Statistics Page


Demo Video