Automated grading using natural language processing and semantic analysis.

Area of Science:

• Natural Language Processing (NLP)
• Artificial Intelligence (AI)
• Educational Technology

Background:

• Educational grading systems are increasingly automating through NLP and AI advancements.
• Traditional grading methods are being augmented by sophisticated computational approaches.
• Existing automated systems often rely on keyword matching or simpler NLP techniques.

Purpose of the Study:

• To explore and analyze automated grading systems, including traditional and contemporary methods.
• To implement and evaluate a novel NLP-based automatic grading system.
• To assess the efficacy, limitations, and future prospects of automated grading in education.

Main Methods:

• Conducted an extensive literature review of automated grading techniques.
• Developed a hybrid NLP system combining similarity measures (edit, cosine, Jaccard, word count) with semantic analysis (Universal Sentence Encoder).
• Implemented a weighted scoring mechanism and a rule-based final scoring layer for evaluation.

Main Results:

• The novel NLP system demonstrated consistent and accurate grading capabilities.
• The system effectively identified student responses requiring further human evaluation.
• Experimental results on 14 student responses validated the hybrid approach's effectiveness.

Conclusions:

• Automated grading systems offer significant potential for enhancing educational assessment.
• Hybrid approaches combining surface-level and semantic analysis provide comprehensive evaluation.
• The developed system offers a promising tool for educators, balancing automation with the need for human oversight.