Skip to main content

Answer-Sheet Evaluation

Upload photos or scans of a handwritten answer sheet, let Testify's OCR + AI evaluate each question, then let the teacher review and adjust the AI's scoring before publishing. This is the handwritten-answer counterpart to OMR: students submit photos instead of bubble sheets, and the AI reads free-form written answers.

Answer-sheet evaluation flow

Who Can Use It​

  • Students — upload one or more images of their answer sheet and track processing status.
  • Teachers — review the AI's per-question evaluation, override marks, and publish the final score.
  • Admins — configure the feature flag and credit usage.

How It Works​

The pipeline is four stages:

Upload  →  OCR  →  AI Evaluation  →  Teacher Review  →  Published Result
  1. Upload — the student adds 1–N images (phone camera or flatbed scan). Files are validated, uploaded, and tied to the submission.
  2. OCR — handwriting is transcribed page-by-page. Page images stay on file for the reviewer.
  3. AI Evaluation — each question is matched to its transcribed answer, graded against the rubric, and scored. For each question the AI records: marks awarded, correct/partially correct flag, confidence, feedback, mistakes, improvement suggestions, and the evaluation method.
  4. Teacher Review — a split-view panel shows scanned pages on the left and per-question AI evaluations on the right. The teacher can edit any score or feedback before publishing.

Submission States​

StateWhat the student sees
uploadUpload form
uploading / processingProgress + "evaluating your sheet…"
evaluatedAI score, marked Pending Teacher Approval
reviewedFinal score + teacher feedback, released

Student Workflow​

  1. Open the Answer-Sheet exam from Assignments or My Exams.
  2. Click Upload Answer Sheet.
  3. Add one image per page (JPEG / PNG). Drag to reorder if pages are out of order.
  4. Click Submit. The page shows live processing status.
  5. Once the AI completes, the student sees a provisional score flagged as awaiting teacher approval.
  6. When the teacher publishes, the final score, per-question feedback, and improvement suggestions are released.

Teacher Workflow​

  1. Open the exam in Offline Exams → Answer-Sheet Exams.
  2. Pick a submission from the list.
  3. The OfflineAnswerReview panel opens:
    • Left — scanned pages with page navigation.
    • Right — per-question evaluation card: marks awarded / max, correct flag, AI confidence, feedback, mistakes, improvement suggestions.
  4. For each question, accept the AI score or override: edit marks, correct/partial flag, or rewrite feedback.
  5. Click Save & Publish when done. Marks become the student's final result and counts toward class analytics.

Creating an Answer-Sheet Exam​

  1. Go to Offline Exams → + New and pick Answer Sheet Exam.
  2. The Pre-Create Wizard asks for board, grade, and subjects — this seeds folders and filters.
  3. Build the question paper as usual (or import from a document).
  4. Assign the exam to a class, section, or group.
  5. Publish — the exam now appears on the students' Assignments list with an Upload Answer Sheet CTA.

Settings​

  • AI evaluation model — configured at Admin → AI Management.
  • Max pages per submission — default 20.
  • Image size limit — 10 MB per page.
  • Credit cost — per-page OCR + per-question evaluation; see the AI usage report.

FAQs​

How accurate is the handwriting OCR? Good on neat English, Hindi, and Bengali handwriting. Struggles with heavy cursive, faint pencil, or tilted pages. The split-view review is there exactly because the teacher needs to catch OCR misreads.

Can the student re-upload? Only until the submission is marked evaluated. Once AI evaluation has run, re-uploads require a teacher reset.

What if a page is missing? The teacher can request a re-upload from the review panel. The student is notified and can append missing pages without restarting the evaluation.

Is this the same as OMR? No. OMR is for bubble answer sheets (machine-readable). Answer-sheet evaluation is for handwritten free-form answers and relies on OCR + AI.