DebateGPT: An Real Time Argument Analyzer using Transformer and NLP

Debate and public speaking evaluation traditionally depend on human judgment, which can be subjective, inconsistent, and time-intensive. With advancements in Artificial Intelligence (AI) and Natural Language Processing (NLP), automated systems can now provide structured and objective language analysis. This paper presents DebateGPT, a real-time AI-based debate argument analyzer that integrates speech recognition, transformer-based NLP models, sentiment analysis, and grammatical evaluation into a unified mobile-compatible framework. The proposed system captures spoken input, converts it into text using advanced Speech-to-Text (STT) models, and analyzes argument structure, logical coherence, and emotional tone. The backend is developed using FastAPI and Python, while the frontend is implemented using Kotlin Compose for Android. Transformer models from Hugging Face are utilized for argument mining and sentiment classification, and speech recognition is powered by models such as OpenAI Whisper. The system generates structured feedback reports in real time, enabling debaters to identify strengths and weaknesses immediately. The proposed architecture aims to reduce subjectivity in debate evaluation while enhancing learning efficiency and communication skill development. Key terms Natural Language Processing, Transformer Models, Speech Recognition, Argument Mining, Sentiment Analysis, Real-Time Evaluation.