Clothing Review Analysis

This project is a web-based application that performs sentiment analysis on clothing reviews. Users can input review text, and the app predicts whether the review is Recommended or Not Recommended using two models:

1. TF-IDF + Random Forest
2. BERT (Bidirectional Encoder Representations from Transformers)

The app provides side-by-side predictions and confidence scores for comparison.

clothing-review-analysis.mp4

---

Features

• Two Models for Prediction:

  • TF-IDF + Random Forest: Lightweight, interpretable traditional machine learning model.
  • BERT: Advanced deep learning-based NLP model for high accuracy.

• Streamlit Interface:

  • Simple web UI to enter reviews for prediction.
  • Display results for both models with confidence scores.

---

Tech Stack

• Backend:

  • Python
  • Hugging Face transformers
  • Scikit-learn

• Web Framework:

  • Streamlit

• Machine Learning Models:

  • Pretrained BERT model fine-tuned for sentiment classification.
  • Random Forest classifier trained on TF-IDF features.

---

Setup Instructions

1. Copy and extract the zip file

2. Install Dependencies

pip install -r requirements.txt

3. Prepare Model Files

• Place the TF-IDF + Random Forest model and vectorizer in the root directory:

  • tfidf_rf_model.pkl
  • tfidf_vectorizer.pkl

• Save the fine-tuned BERT model into the directory bert_sentiment_model/:

  • config.json
  • pytorch_model.bin
  • tokenizer_config.json
  • vocab.txt

---

Run the App

To start the Streamlit app, use the following command:

streamlit run app.py

Once running, open the link provided in the terminal to access the web app in your browser.

---

Usage

1. Enter Review Text: In the text area provided, input a review you want to analyze.
2. Get Predictions: Click on the "Predict" button.
3. View Results: The app displays:
   • Predicted sentiment (Recommended or Not Recommended).
   • Confidence scores from both the TF-IDF + Random Forest and BERT models.

---

Directory Structure

sentiment-analysis-app/
├── app.py                # Main Streamlit app script
├── tfidf_rf_model.pkl    # Trained Random Forest model
├── tfidf_vectorizer.pkl  # Trained TF-IDF vectorizer
├── bert_sentiment_model/ # Directory containing the saved BERT model
│   ├── config.json
│   ├── pytorch_model.bin
│   ├── tokenizer_config.json
│   ├── vocab.txt
├── requirements.txt      # Python dependencies
└── README.md             # Project documentation

---

Model Details

TF-IDF + Random Forest

• Traditional machine learning pipeline.
• Uses TF-IDF for feature extraction and a Random Forest classifier for predictions.

BERT

• The bert-base-multilingual-uncased-sentiment model from Hugging Face's transformers.
• Tokenizer processes the review text, and the model predicts sentiment.

---

Future Enhancements

1. Batch Predictions: Add support for analyzing multiple reviews via file upload.
2. Custom Threshold: Allow users to set a confidence threshold for predictions.
3. Visualization: Include charts or graphs for a better understanding of model outputs.

---

How to Train Models

Run the train.py script to train the TF-IDF + Random Forest model and prepare the BERT pipeline:

python train.py

This will save the trained models (tfidf.pkl, random_forest.pkl) for the TF-IDF + Random Forest method.

---

Acknowledgements

• Hugging Face Transformers for pre-trained BERT models.
• Kaggle for the dataset: "Women's Clothing E-Commerce Reviews".

---

Contributors

• Ernitia Paramasari
  Data Scientist and Machine Learning Engineer

Feel free to contribute to this project by submitting issues or pull requests!

---

License

This project is licensed under the MIT License.

Enjoy analyzing clothing reviews with cutting-edge NLP models! 🎉