Automating the detection of plagiarism in educational content using Natural language processing

Plagiarism detection has also grown more challenging with the rise of AI-generated content and multilingual content sources. This paper proposes a system for real-time plagiarism detection based on three cutting-edge transformer-based models: multi-qa-mpnet-base-dot-v1, roberta-large-openai-detector, and xlm-roberta-large. These models collaborate to detect semantic plagiarism, AI-generated plagiarism, and multilingual plagiarism with great accuracy. In contrast to the traditional method based on lexical similarity, our method seeks contextual meaning, hence being more efficient against paraphrased and AI-generated content. The multi-qa-mpnet-base-dot-v1 model improves semantic search functionality, allowing meaning-based plagiarism detection. The roberta-large-openai-detector excels in detecting AI-generated content, solving issues of content authenticity. At the same time, the xlm-roberta-large model supports strong multilingual plagiarism detection functionality, allowing the system to operate across languages. Through rigorous testing, our system attains 95% accuracy in real-time detection, outperforming traditional methods in efficiency and accuracy. This work identifies the importance of deep learning in plagiarism detection and shows how transformer models can be used to improve academic and content integrity. The proposed system offers a scalable, flexible solution for educational institutions, publishers, and online platforms to fight plagiarism effectively in the era of AI-generated content.