AI-Driven Web-Based Video Conferencing System with Real-Time Meeting Intelligence and Analytics

Virtual meeting platforms have become critical infrastructure for modern enterprises and educational institutions, yet existing solutions lack intelligent mechanisms to assess meeting quality, participation equity, and content relevance. This paper presents an AI-Driven Web-Based Video Conferencing System that integrates real-time communication with an embedded artificial intelligence analytics pipeline. The system leverages WebRTC for peer-to-peer audio/video transport, a Spring Boot REST backend for session orchestration, and a Python-based AI microservice layer that performs Automatic Speech Recognition (ASR) using OpenAI Whisper, extractive and abstractive Natural Language Processing (NLP) summarisation using the BART-large-CNN transformer, sentiment analysis using the VADER lexicon combined with a fine-tuned BERT classifier, and engagement scoring through a weighted behavioural metric model. A React-based interactive analytics dashboard renders live transcripts, sentiment trend charts, speaking-time distribution graphs, and auto-generated structured meeting summaries. Experimental evaluation across 30 simulated meetings demonstrates 91.4% ASR word accuracy, 94.2% sentiment classification accuracy, and average transcript latency of 1.8 seconds, confirming the system’s suitability for real-world deployment.