Detecting AI-Generated Text: A Bi-GRU with Linguistic Features Approach

The advances in artificial intelligence (AI) technology can transform education. However, the growing infusion of AI technologies into academic environments raises important ethical issues that are essential for safeguarding academic integrity and quality. This paper proposes a detection framework that utilizes linguistic features and a Bidirectional Gated Recurrent Unit (Bi-GRU) model to identify AI-generated texts. The framework extracts perplexity values, readability measures, syntactic complexity metrics, and lexical diversity indicators, which are fed into a Bi-GRU classifier. Trained on an extended version of the DAIGT dataset and evaluated on the Deepfake dataset, the model achieved an accuracy of 98 % with F1 score of 97 % when tested on the DAIGT dataset. It also achieved an accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{7 2 \%}$</tex> and an F1 score of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{7 9 \%}$</tex> on the Deepfake dataset, outperforming state-of-the-art methodologies in these datasets. These findings highlight the potential of combining linguistic feature analysis with deep learning to develop efficient, interpretable, and domain-adaptive systems for AI text detection, addressing the critical need for automating authenticity and maintaining integrity in content generation.