AnswerIQ

AnswerIQ: AI-Powered Answer Evaluation System

Explainable, automated grading of student answer sheets.

Category
AI Systems · Full-stack
  • OCR
  • Sentence Transformers
  • Gemini
  • FastAPI
  • Next.js
  • TypeScript
  • Supabase

Abstract preview - not a screenshot

01 / Problem

Grading student answer sheets by hand is slow. AnswerIQ automates it with OCR, semantic analysis and LLM-assisted scoring, and keeps every score explainable.

02 / Approach

  1. 01Hybrid, explainable score: concept coverage 50% · semantic similarity 30% · keyword matching 20%.
  2. 02Question-normalization and answer-mapping pipeline that aligns answers despite inconsistent numbering (Q1 / Q-1 / Question 1) and submission order.
  3. 03Role-based faculty/student workflows.

03 / Architecture

AnswerIQ system architectureAnswer sheet, then OCR, then Question normalization, then Answer mapping, then Hybrid scoring (Concepts, Semantics, Keywords), then LLM-assisted scoring, then Explainable score.FASTAPI BACKEND01Answer sheetSCANNED STUDENT ANSWER PAGES02OCRPAGE IMAGES → RAW TEXT03Question normalizationQ1 · Q-1 · QUESTION 1 → Q104Answer mappingALIGNED TO THE KEY, IN ANY ORDER05ConceptsCOVERAGE · 50%SemanticsSIMILARITY · 30%KeywordsMATCH · 20%06LLM-assisted scoringGEMINI REVIEWS AND JUSTIFIES07Explainable scorePER-QUESTION BREAKDOWN
Fig. 01 - System architectureSimplified

04 / Key details

  • Automates grading of student answer sheets with OCR, semantic analysis and LLM-assisted scoring.
  • Hybrid, explainable score: concept coverage 50% · semantic similarity 30% · keyword matching 20%.
  • Aligns answers despite inconsistent numbering (Q1 / Q-1 / Question 1) and submission order.
  • Role-based faculty/student workflows.

05 / Results & signals

50/30/20
Hybrid scoring
OCR→LLM
Pipeline
Next+FastAPI
Stack
Supabase
Data & auth

Next entry →

Mini-RAG

Ask your documents. Get answers with inline citations.