AI Summarizer: Interactive Multi-Modal Processing for Lectures, Meetings and Text Documents
Rahul Dhamdhere, Manthan Dhawale, Satyajeet Jagtap, Harsh Memane, Shashank Lahane, Sneha Shrikant Salvekar · International Journal For Multidisciplinary Research · 2025
This paper introduces an AI-powered summarization system that processes both text and audio content—such as lectures and meetings—to improve productivity. It integrates OpenAI Whisper for transcription, Nomic embeddings for extractive summarization, and DeepSeek’s language model (via Ollama) for generating refined summaries and enabling chatbot interaction. The system runs locally using a Flask backend and HTML/JavaScript frontend. Whisper achieves a Word Error Rate (WER) of ~10%, and the system’s summarization accuracy averages 77.46%, as evaluated by Grok. Designed for students and professionals, future enhancements will include real-time processing and optional cloud integration with privacy safeguards.