International Journal of Scientific Research and Engineering Development

The proliferation of politically biased content on social media platforms such as X (formerly Twitter), Reddit, Threads, and digital news portals represents a significant challenge to democratic discourse. This paper presents STANCE.AI, an AI-based political stance detection framework that integrates Natural Language Processing (NLP) preprocessing with transformer-based large language models (LLMs) to automatically classify online political content as Left, Right, or Neutral. The proposed pipeline employs SpaCy for tokenization and lemmatization, NLTK for stopword removal, and regular expressions for emoji and hashtag normalization, producing clean label-encoded input (0=Left, 1=Neutral, 2=Right) for downstream inference. The active inference engine uses Meta's LLaMA 3-70B model served through the Groq API, achieving 87.6% accuracy, 88.1% precision, 86.9% recall, and an F1-score of 87.5% on a 500-article Media Bias Fact Check sample. Planned extensions include fine-tuned RoBERTa-base (Cardiff NLP) and BERTweet-base targeting over 90% accuracy on benchmark datasets. A real-time web dashboard implemented with React.js and FastAPI provides interactive stance visualization across four functional modules: Analyze, Dashboard, Live Feed, and Models. Data is collected from verified Indian news outlets including NDTV, The Wire, Republic World, The Hindu, and Indian Express, as well as social media platforms via NewsAPI. Experimental results demonstrate that the system accurately identifies ideological orientation in both formal news text and informal social media discourse, achieving zero false-acceptances on cross-domain tests. The framework promotes media transparency and critical awareness in digital political communication.

Tanish P. Toradmalle,"STANCE.AI: An AI-Based Political Stance Detection Framework Using NLP, Transformer Models, and Real-Time Social Media Analysis" International Journal of Scientific Research and Engineering Development, V9(2): Page(1384-1391) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.