Published Paper Details:

--- The increasing digitalization of academic workflows has intensified the demand for intelligent systems that can evaluate handwritten assignments with speed, accuracy, and transparency. Traditional plagiarism detection tools are designed for typed text and fail to handle scanned handwritten documents, while manual grading remains subjective, inconsistent, and time-consuming. To address these limitations, this paper presents SmartEval, an integrated deep learning-driven framework for automated plagiarism detection and assignment grading. The system employs Optical Character Recognition (OCR) based on CNN-BiLSTM models with Connectionist Temporal Classification (CTC) loss to convert handwritten text into machine-readable form. The extracted text is processed through a hybrid similarity engine combining TF-IDF, cosine similarity, and contextual embeddings derived from Sentence-BERT to detect paraphrased, semantic, and cross-document plagiarism. Automated grading is performed using a rubric-weighted scoring model enhanced by semantic relevance, coherence evaluation, grammar assessment, and keyword density analysis. Experimental evaluation demonstrates high accuracy in OCR transcription, strong plagiarism detection performance, and grading outcomes that closely correlate with human evaluators. SmartEval offers a comprehensive, scalable, and objective mechanism for evaluating handwritten academic submissions, enhancing academic integrity and reducing educator workload