💬 NLP – Sentiment Analysis Assignment 🧐

✍️ Author: Mirellys Arteta Davila

This project is part of the Natural Language Processing (NLP) module from the KeepCoding AI Bootcamp.
The objective is to build and evaluate a sentiment analysis pipeline using Amazon product reviews.

---

📁 Structure

• EDA → download + data exploration.
• Preprocessing → Text cleaning pipeline.
• Modeling → Training & comparing 2 Machine Learning and 1 Deep Learning sentiment classifiers.
• Reports → Metrics, evaluation and conclusions.

---

📦 Dataset

• Small subsets for experimentation Amazon Reviews → Data 2018 5-core

  • Amazon Fashion - 3176 reviews

  • Appliances - 2277 reviews

  • Software - 12805 reviews

• Format: one-review-per-line in JSON.

Sample Review:

{
"image": ["https://images-na.ssl-images-amazon.com/images/I/71eG75FTJJL._SY88.jpg"], 
"overall": 5.0, 			   // rating of the product
"vote": "2",                   // helpful votes of the review
"verified": True, 
"reviewTime": "01 1, 2018",    // time of the review (raw)
"reviewerID": "AUI6WTTT0QZYS", // ID of the reviewer
"asin": "5120053084",          // ID of the product
"style": {                     // Dictionary of the product metadata
	"Size:": "Large", 
	"Color:": "Charcoal"
	},                          
"reviewerName": "Abbey",       // Name of the reviewer 
"reviewText": "I now have 4 of the 5 available colors of this shirt... ", 
"summary": "Comfy, flattering, discreet--highly recommended!", 
"unixReviewTime": 1514764800   // time of the review (unix time)
}

• Preprocessing: only reviews with ratings (1–5)
• Sentiment labels created from star ratings:

Stars	Sentiment Label	Category
⭐️	0	Negative
⭐️⭐️	0	Negative
⭐️⭐️⭐️	removed	Neutral
⭐️⭐️⭐️⭐️	1	Positive
⭐️⭐️⭐️⭐️⭐️	1	Positive

---

🧹 Preprocessing

• Custom pipeline: normalization, tokenization, stopword removal, optional stemming
• Duplicates removed based on cleaned token sequences
• TF-IDF vectorization for model input

---

🤖 Models

• ✅ Logistic Regression** (with GridSearchCV)

• 🟡 Multinomial Naive Bayes** (with GridSearchCV)

• 🔵 Deep Learning model (LSTM)

---

📊 Evaluation

• Accuracy, precision, recall, F1-score
• Confusion matrix and classification report
• Threshold-based analysis using precision-recall curves
• Visual performance vs regularization plots for each ML model

🛠️ Tools & Libraries

🧪 Data & Preprocessing

• pandas, numpy → Data handling and manipulation
• string, re, unicodedata → Text normalization and cleanup
• random, pickle, collections.Counter → Utilities and storage
• nltk → Tokenization, stopword removal, stemming, n-grams

📊 Visualization

• matplotlib.pyplot, seaborn → Data plots, metrics, and trends
• WordCloud → Visualize most frequent terms
• FreqDist (nltk) → Token frequency analysis

🤖 Machine Learning

• scikit-learn:
  • TfidfVectorizer → Bag-of-Words (TF-IDF) encoding
  • LogisticRegression, MultinomialNB → Classifiers
  • GridSearchCV, train_test_split → Model selection and evaluation
  • chi2 → Feature selection
  • classification_report, confusion_matrix, accuracy_score, precision_recall_curve, roc_curve → Evaluation metrics

---

📝 NLTK Resources

• stopwords → English stopword list (downloaded at runtime)