In this project, I worked on developing a Speech-to-Text sentiment analysis solution that analyzes recorded conversations between our customer executives and customers. The goal was to enhance communication, detect fraud, and identify top performers based on multiple criteria.

1. Conversation Analysis: The system analyzes conversations to highlight hits and misses, providing real-time feedback and suggestions to executives on their screen. These suggestions are based on the customer’s interaction history, helping improve communication.
2. Fraud Detection: The model flags potential fraud from the executive’s side, ensuring integrity in customer interactions.
3. Performance Evaluation: Using insights from focus group discussions with Telecalling unit leaders, we identified 18 key performance factors. These factors were weighted and used to evaluate the performance of executives, identifying top performers.
4. Training the Model: The conversation data, converted to JSON, was trained to assign scores. The final model scores each conversation on both the total communication quality and the individual performance factors.

This project allowed me to integrate AI-powered sentiment analysis with real-time feedback and performance metrics, improving both customer experience and internal performance monitoring.