An Intelligent Research Assistance System for Plagiarism and Formatting Compliance

Authors:

Mrs. A. Bhanusri, Allu Chanikya, Manda Kameswari Heranmhaee, Lenka Pramodh Department of Information Engineering and Computational Technology

Maharaj Vijayaram Gajapathi Raj College of Engineering (A), Vizianagaram, Andhra Pradesh, India

Abstract—Large language models have recently transformed how academic papers are drafted and evaluated. While these tools assist authors in generating content rapidly, they complicate the peer-review process by blurring the lines of originality. This paper introduces a unified research assistance system designed to help students and reviewers evaluate manuscripts more efficiently. Our approach combines natural language processing and machine learning to analyze document integrity. The system detects semantic plagiarism, flags AI-generated text using a multi-metric scoring algorithm and verifies compliance with standard IEEE formatting. We also implemented a generative feature to help authors clarify their writing while maintaining an objective academic tone. To evaluate the system, we tested it against a custom dataset of 1,000 mixed-source academic papers. The results demonstrate that our combined-metric approach achieves over 94% accuracy in AI detection, outperforming standalone perplexity analyzers, and significantly reduces similarity search latency through local FAISS vector indexing.

Index Terms—Natural Language Processing, Plagiarism De- tection, Semantic Similarity, Machine Learning, Academic For- matting