Sentiment Analysis for Fashion Brands

Overview

This project conducts a comprehensive sentiment analysis on customer reviews for two major fashion brands: Adidas and Nike. Utilizing a lexicon-based approach, we leverage three popular sentiment analysis tools—TextBlob, SentiWordNet, and VADER (Valence Aware Dictionary and sEntiment Reasoner)—to extract and compare customer sentiment from textual reviews. The goal is not only to measure and compare brand perception but also to explore how natural language processing (NLP) techniques can categorize feedback based on the type of product purchased.

In today’s fast-paced fashion industry, consumer preferences evolve rapidly, prompting top brands like Nike and Adidas to adapt their product designs, branding, and marketing strategies accordingly. While both brands are known for their high-performance products, not all customers express satisfaction—highlighting the need to better understand consumer sentiments. This project leverages sentiment analysis (opinion mining) to extract and categorize emotions from customer feedback, providing insights into brand perception and satisfaction levels. Using lexicon-based tools such as VADER, SentiWordNet, and TextBlob, the study analyzes customer reviews that were web-scraped from brand websites, with similar datasets also available on Kaggle. By comparing the outputs of each sentiment tool on the same dataset, the project offers a data-driven approach to evaluating customer sentiment across both fashion giants.

A detailed documentation of the entire project workflow, including background, methodology, visualizations, and key insights, is available here.

Table of Contents

• Project Objectives
• Data
• Code Structure
• Setup
• Usage
• Results
• Contributing
• License

Project Objectives

The core goals of this project are as follows:

1. Apply and compare three lexicon-based sentiment analysis methods—TextBlob, SentiWordNet, and VADER—to customer reviews associated with Nike and Adidas.
2. Evaluate the polarity and subjectivity of reviews to identify patterns in customer sentiment.
3. Determine the preferred fashion brand based on comparative sentiment metrics including positive, neutral, and negative sentiments.
4. Explore the application of NLP tools to categorize reviews based on the specific items purchased (e.g., shoes, apparel, accessories).
5. Visualize sentiment distributions to provide a data-driven understanding of brand perception in the fashion retail industry.

Data

The dataset used in this analysis was obtained from Kaggle and contains a rich set of customer reviews, including textual feedback, product categories, and brand labels for both Adidas and Nike. The dataset underwent preprocessing and cleaning before being used for sentiment classification and visualization.

Code Structure

The project is organized into the following main directories and components:

• Data Wrangling & Exploration/:

  • Jupyter Notebooks dedicated to initial data cleaning, transformation, and exploratory data analysis (EDA).
  • Techniques such as tokenization, stopword removal, and word cloud generation are included.

• Model Building/:

  • Scripts and notebooks focused on implementing and comparing TextBlob, SentiWordNet, and VADER.
  • Includes visualizations like sentiment polarity histograms and brand-wise sentiment comparisons.

• Sentiment Analysis App/:

  • A set of Python scripts and modules used to run sentiment analysis on new inputs.
  • Can be extended into a web app for interactive exploration of results.

Setup

To run this project locally, ensure that Python 3.x is installed on your system. The required libraries include:

• pandas – for data manipulation
• nltk – for natural language processing
• textblob – for sentiment analysis
• seaborn and matplotlib – for data visualization
• vaderSentiment – for rule-based sentiment scoring

Install all dependencies using the following command:

pip install pandas nltk textblob seaborn matplotlib vaderSentiment

You may also need to download NLTK corpora:

import nltk
nltk.download('all')

Usage

Once the environment is set up, you can:

1. Load and preprocess the dataset.
2. Run sentiment classification using each tool (TextBlob, SentiWordNet, and VADER).
3. Generate plots to visualize sentiment trends.
4. Optionally, test the sentiment analysis on custom input reviews using the provided script.

Results

The analysis reveals valuable insights into customer perceptions of Adidas and Nike. Key takeaways include:

• Comparative polarity distributions across both brands.
• Analysis of product categories contributing to positive or negative feedback.
• Identification of sentiment strengths and weaknesses for each tool used.

These results can inform marketing strategies, product development, and customer engagement practices for fashion retailers.

Contributing

Contributions to this project are welcome! If you have suggestions for additional features, improvements in NLP modeling, or ideas for visualization enhancements, feel free to open a pull request or submit an issue.

License

This project is open-source and available under the MIT License. Feel free to use and adapt it with attribution.