Sentiment Analysis on Amazon Reviews (NLTK + Python).

This project demonstrates how to perform sentiment analysis on Amazon product reviews using Natural Language Toolkit (NLTK) and scikit-learn.

Features. Loads Amazon review dataset (20,000+ samples)

Preprocesses text using: Tokenization

Stopword removal

Lemmatization

Applies the preprocessing pipeline to clean review text

Trains and evaluates sentiment classification model

Dataset. Amazon reviews dataset is loaded from PyCaret’s GitHub.

It contains: reviewText: customer review Positive: sentiment label (1 = positive, 0 = negative)

🛠️ Preprocessing Steps.

Tokenization – split reviews into words

Stopword Removal – remove common English words (e.g., the, and, is)

Lemmatization – reduce words to base form (running → run)

Rejoin Tokens – convert processed tokens back into a string

📊 Model Evaluation. Confusion Matrix

[[ 1131 3636]

[ 576 14657]]

Classification Report. Class Precision Recall F1-Score Support

0 (Negative) 0.66 0.24 0.35 4767

1 (Positive) 0.80 0.96 0.87 15233

Overall Accuracy: 79%. Model performs well on positive reviews, but struggles with recall for negative reviews.

✅ This project builds a full pipeline from text preprocessing → feature extraction → sentiment classification → evaluation, and can be extended with advanced ML/DL models for better performance.