Transformer-Enhanced Contextual Discriminator for Robust Detection of AI-Generated Texts

View at publisher Summarise with VIRAT
Authors: C. Sushama, Poolakunta Manasa, Muggari Vishnupriya, Srirama Harshitha, Puli Sujith
Cited by: 0
References: 25

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The rapid progression of models in large language has made differentiate between human-written and AIgenerated text and it increasingly challenges. This paper proposed a Transformer-Enhanced Contextual Discriminator (TECD) based on a hybrid deep learning framework which integrates with transformer-based contextual embeddings and graph-aware relational learning for sturdy AI-text detection. The suggested model obtains both semantic coherence and global contextual depended by combining Bidirectional Transformer Encoders with a Graph Convolutional Network (GCN) layer. The system includes an attention method that helps to model focus on the most important linguistic cues that influence by it decision. This feature also makes the model more understandable, the text led to a particular classification result. Experimental validates on multiple benchmark datasets shows that TECD achieves superior accuracy and generalizes compared to existing Bi-LSTM and attention-based models, while maintaining the computational efficiency. The results signify that influencing from top-down contextual representations substantially improves the reliability of automated systems in identifying AI-generated content overall varied text domains.

Fields of Study & Subjects

Authorship Attribution and Profiling Topic Modeling Digital Media Forensic Detection

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1861–1865
Publication date
21 January 2026
Language
EN