Sentiment Analysis Using NLP 📊

Welcome to the world of Natural Language Processing (NLP)! In this project, we'll explore sentiment analysis from customer reviews using some powerful NLP techniques. Buckle up as we dive into the code, data, and some fascinating insights!

Table of Contents

1. Overview
2. Getting Started
3. Data Preprocessing
4. Sentiment Analysis
5. Results
6. Conclusion

---

Overview

This project aims to classify customer sentiments based on Amazon product reviews. We use NLP tools to preprocess the text data, analyze it, and eventually predict whether reviews are positive or negative.

Libraries Used

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import nltk

---

Getting Started

Dataset

The dataset we are working with is the Amazon Fine Food Reviews dataset. You can find it here.

First, we load the dataset and take a subset of 500 reviews to keep things manageable.

df = pd.read_csv('data/Reviews.csv')  # Reading the reviews data
df = df.head(500)  # Taking a subset of 500 reviews
print(df.shape)  # Prints: (500, 10)

---

Data Preprocessing

Before diving into analysis, we need to clean and preprocess the data. This includes tokenizing the text, removing stop words, and other common NLP tasks.

Tokenizing the Text

We use nltk to tokenize the words and prepare them for analysis.

from nltk.tokenize import word_tokenize

df['tokenized'] = df['Text'].apply(lambda x: word_tokenize(x.lower()))

Removing Stop Words

Stop words (common words like "the", "is", "and") don't contribute much meaning and can be removed.

from nltk.corpus import stopwords

stop_words = set(stopwords.words('english'))
df['filtered_tokens'] = df['tokenized'].apply(lambda x: [word for word in x if word not in stop_words])

---

Sentiment Analysis

Now for the exciting part! We analyze the sentiment of reviews by looking at their textual data.

Word Cloud Visualization

A quick look at the most frequent words in positive and negative reviews:

from wordcloud import WordCloud

# Generate word clouds
positive_reviews = " ".join(df[df['Score'] > 3]['Text'])
wordcloud = WordCloud(width=800, height=400).generate(positive_reviews)

# Display the word cloud
plt.imshow(wordcloud, interpolation='bilinear')
plt.axis("off")
plt.show()

Sentiment Classification

To classify sentiment, we can use basic techniques such as checking for positive or negative keywords.

# Sample code to classify based on score (positive/negative sentiment)
df['sentiment'] = df['Score'].apply(lambda x: 'positive' if x > 3 else 'negative')

---

Results

After analyzing the data, we found some interesting insights. For example, the majority of reviews in the dataset are positive, which is common for product reviews.

Data Visualization

We also took a look at the distribution of review scores:

sns.countplot(x='Score', data=df)
plt.title('Distribution of Review Scores')
plt.show()

---

Conclusion

This project highlights the basics of sentiment analysis using NLP techniques. We used a simple dataset and some basic text-processing techniques to analyze and classify sentiment. While this is just scratching the surface of NLP, it demonstrates how powerful these techniques can be for understanding large-scale textual data.