📊 EmotionLens: Sentiment & Emotion Analysis of Amazon Reviews

🔍 Overview

EmotionLens is an NLP project analyzing Amazon product reviews.
It combines EDA, sentiment analysis (VADER), emotion detection (HuggingFace), and keyword extraction (TF-IDF) to uncover hidden insights in customer feedback.

🚀 Features

• Clean & preprocess customer reviews
• Perform sentiment analysis using VADER
• Detect emotions using HuggingFace DistilRoBERTa model
• Extract top keywords with TF-IDF
• Generate visualizations (rating distribution, wordclouds, correlation heatmap, etc.)
• Provide business insights from customer feedback

📂 Dataset

A sample Amazon Reviews dataset was used (CSV format).
👉 For Big Data scaling, this project can be extended to the Amazon Review Dataset (millions of rows) using PySpark/Dask.

📊 Key Insights

• ⭐ 5-star reviews → Joy with keywords like love, amazing, exceeded.
• ❌ 1-star reviews → Sadness/Fear with keywords like broke, disappointed, terrible.
• 📈 Strong correlation (0.82) between star ratings and sentiment scores.
• ✍️ Review length does not strongly impact sentiment.
• 💡 Businesses can leverage this to quickly detect unhappy customers.

📸 Visual Highlights

⭐ Ratings Distribution

☁️ Wordclouds

Positive reviews highlight words like amazing, love, exceeded,
while negative reviews emphasize terrible, broke, disappointed.

🎭 Emotion by Rating

5⭐ reviews are dominated by Joy,
while 1⭐ reviews show mostly Sadness and Fear.

✍️ Average Review Length by Rating

❤️ Distribution of Dominant Emotions

📈 Sentiment Score by Rating

🔥 Correlation Heatmap

⚙️ Installation

Clone the repo and install dependencies:

git clone https://github.com/Krish5986/EmotionLens.git
cd EmotionLens
pip install -r requirements.txt