Designed and implemented an advanced NLP-powered call analysis platform that extracts actionable insights from thousands of voice calls daily, enabling data-driven decision making for sales and customer service operations.
🚀 Core Contributions
🧠 Voice Analytics & NLP Engine
Keyphrase Extraction System: Built unsupervised ML engine processing thousands of calls daily
Algorithm Implementation: Deployed ensemble of Topicrank, Textrank, Yake, and Autophrase with custom pre/post-processing
Speech-to-Text Integration: Leveraged Google's API to convert voice calls into analyzable text
Call Segmentation: Developed sophisticated models to identify distinct conversation components
🔍 Sentiment Analysis & Classification
Multi-model Approach: Combined Google Natural Language API and VADER for comprehensive sentiment scoring
BERT Implementation: Trained custom deep learning models for enhanced sentiment analysis
Student Text Classification: Researched and prototyped state-of-the-art techniques including SVM, LSTM, and CNN
Model Interpretability: Employed ELI5 and LIME to provide transparent explanations of model predictions