Textual Persuasion Analysis for Identifying Manipulative Languages
R K Bhavana, Sandeep Mathias · International Journal For Multidisciplinary Research · 2025
Manipulative language, often used in political, commercial, and social discourse, can distort public opinion by appealing to emotions, authority, or fallacious reasoning. This paper presents a comprehensive framework for detecting such manipulation using natural language processing. Leveraging a taxonomy of 25 persuasion techniques, we fine-tune the XLM-RoBERTa transformer model to perform multi-label classification of manipulative content. Additionally, we develop a counter-narrative generation module using a T5-based model to suggest ethically persuasive responses. Our results show promising accuracy in classification across varied techniques and demonstrate the effectiveness of automated counter-speech generation. This work contributes to efforts in enhancing digital media literacy and resisting propaganda in online platforms.