Automated Essay Evaluator using BERT Algorithm and Semantic Analysis

This study presents the development and evaluation of an Automated Essay Evaluator (AEE) system utilizing BERT algorithm and Semantic Analysis to address limitations in traditional essay assessment methods. Using a developmental descriptive research design guided by the CRISP-DM process, the system integrates multiple Natural Language Processing models to assess grammar, structure, semantic relevance, and originality with a high degree of accuracy. Designed for use in educational institutions, the AEE integrates advanced NLP models such as bert-based-uncased, all- MiniLM-L6-v2, spelling-correction-english-base, chatgpt-detector-roberta, and ms-macro-MiniLM-L6-v2 to grammar, structure, semantic relevance, and originality. The AEE system ensures wide-ranging evaluation because it determines if there are semantic relationships between essays, identify grammatical errors, checks for plagiarism and detect if the content is Al generated. The result demonstrates high accuracy when it comes to plagiarism and Al generated detection with a score of 94%. ISO/IEC 25010 quality metrics was utilized getting high ratings where Safety scoring received the highest average weighted mean of 4.86. Flexibility, adaptability and user-friendliness were achieved through the feedback from educators, students, and IT professionals. By combining pedagogical relevance with technical robustness, the AEE demonstrates its potential as a reliable, consistent, and efficient assessment tool that supports learning outcomes while safeguarding academic integrity.