AI-Based Handwritten Student Answer Evaluation and Performance Prediction System Using NLP, Computer Vision OCR, and Machine Learning

Author

Pruthvi H.M. , Merline Paul , Mouli Krishna K , Sai Charan Reddy , Varun Teja M


Abstract

 Evaluating handwritten university examinations manually remains an exhausting and frequently inconsistent process. Standard optical character recognition tools routinely stumble over idiosyncratic handwriting, while conventional automated grading tools depend too heavily on superficial keyword counts or opaque deep learning representations. We built AI Examiner, an assessment architecture engineered specifically to decouple handwriting legibility from academic evaluation. Our pipeline pairs OpenCV document filtering—employing adaptive contrast equalization and bilateral smoothing—with a consonant-skeleton lexical reconstruction technique that recovers phonetically degraded handwritten tokens. Conceptual assessment relies on dense 384-dimensional sentence representations (MiniLM) mapped across discrete rubric propositions, augmented by an adversarial filter that denies partial credit to irrelevant filler text. To catch academic struggles early, we deploy a three-tier machine learning ensemble (Random Forest, Gradient Boosting, and Logistic Regression) that correlates assessment marks, semantic coverage, and student attendance records. Field trials conducted on university test batches in Parallel Computing and Deep Learning showed that the engine resolved previously observed scoring drifts (eliminating an initial 2.9-mark discrepancy between identical answers) and achieved a 0.72 mean absolute error against seasoned faculty examiners, while detecting off-topic answers with 100% precision.


Keywords

Automated grading; handwritten text recognition; sentence embeddings; rubric proposition scoring; academic risk forecasting; educational data mining.



Full Text:

Download Paper PDF


References


  • Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32.
  • Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of NAACL-HLT (pp. 4171–4186).
  • Friedman, J. H. (2001). Greedy function approximation: a gradient boosting machine. Annals of Statistics, 29(5), 1189–1232.
  • Mettam, G. R., & Adams, L. B. (1999). How to prepare an electronic version of your article. In B. S. Jones & R. Z. Smith (Eds.), Introduction to the electronic age (pp. 281–304). New York: E-Publishing Inc.
  • Reimers, N., & Gurevych, I. (2019). Sentence-BERT: Sentence embeddings using siamese BERT-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing (EMNLP-IJCNLP) (pp. 3982–3992).
  • Smith, R. (2007). An overview of the Tesseract OCR engine. In Ninth International Conference on Document Analysis and Recognition (ICDAR 2007) (Vol. 2, pp. 629–633). IEEE.
  • Strunk, W., Jr., & White, E. B. (1979). The elements of style (3rd ed.). New York: MacMillan.
  • Van der Geer, J., Hanraads, J. A. J., & Lupton, R. A. (2000). The art of writing a scientific article. Journal of Science Communication, 163, 51–59.

Share your valuable work from Social Media Buttons