📌 AI Decision Intelligence – BERT Sentiment Analysis

An end-to-end AI Decision Intelligence system that uses BERT-based NLP to analyze customer reviews and generate business-level recommendations from sentiment insights.

This project demonstrates how machine learning outputs can be transformed into actionable business decisions, following industry-standard modular architecture.

🚀 Project Overview

Uses BERT (DistilBERT) for sentiment analysis

Processes real-world Amazon customer reviews

Converts raw predictions into decision intelligence

Designed with clean, modular Python structure

Suitable for data science, NLP, and AI roles

🧠 System Architecture

AI Decision Intelligence/ │ ├── Data/ │ └── amazon_reviews.csv │ ├── src/ │ ├── data_loader.py # Data loading & preprocessing │ ├── bert_sentiment.py # BERT sentiment inference │ ├── decision_engine.py # Business decision logic │ ├── requirements.txt ├── README.md

⚙️ Technologies Used

Python 3

Hugging Face Transformers

BERT (DistilBERT fine-tuned on SST-2)

Pandas

PyTorch (backend)

📊 How It Works

Loads customer reviews from CSV

Samples reviews for efficient BERT inference

Applies pre-trained BERT sentiment model

Maps model output to human-readable labels

Generates business recommendations based on sentiment distribution

📈 Sample Output

📊 DECISION SUMMARY: Total reviews: 20 Positive reviews: 10 (50.0%) Negative reviews: 10 (50.0%)

Overall Recommendation: Mixed customer feedback. Conduct deeper analysis before scaling.

▶️ How to Run

1️⃣ Install dependencies pip install -r requirements.txt

2️⃣ Run sentiment analysis python src/bert_sentiment.py

💡 Business Use Cases

Product feedback analysis

Customer satisfaction monitoring

Market launch decision support

Brand sentiment tracking

AI-driven strategy planning

📌 Future Enhancements

Streamlit dashboard for real-time insights

Aspect-based sentiment analysis

Database integration

Model fine-tuning on domain data

Deployment as an API

👩‍💻 Author Nitisha Sharma Aspiring Data Scientist | NLP & AI Enthusiast

• 🎓 Computer Science Student
• 📊 Interested in Data Analytics & Decision Intelligence
• 🤖 Working with BERT, NLP, and Machine Learning
• 🌱 Currently building end-to-end AI projects

🔌 API / Modules Overview

This project is organized into modular Python scripts:

• data_loader.py
  Loads and preprocesses the Amazon reviews dataset.

• bert_sentiment.py
  Uses a pre-trained BERT model (distilbert-base-uncased-finetuned-sst-2-english)
  to classify reviews into positive and negative sentiments.

• decision_engine.py
  Converts sentiment counts into business-level recommendations.

Each module can be reused independently or extended for real-world applications.