BERT-based Sentiment Analyzer

A sentiment analysis project leveraging BERT and PyTorch to classify Google Play Store reviews into negative, neutral, and positive sentiments. The dataset is scraped using the google_play_scraper library. Pre-trained model weights are stored externally (Google Drive) due to size constraints.

---

Project Structure

.
│   README.md
│   LICENSE
│   01_dataset_scraper.ipynb       # Scrape Google Play reviews dataset
│   02_text_preprocessing.ipynb    # Clean & preprocess text
│   03_bert_sentiment_classifier.ipynb  # Train & evaluate the model
│
└───data
│   ├── apps.csv
│   ├── reviews.csv
│
└───models
    ├── model.bin
    ├── model_base_cased_state_842.bin
    ├── model.pth

---

Usage

1. Scrape Dataset Run the notebook:

   01_dataset_scraper.ipynb

2. Preprocess Data Run the notebook:

   02_text_preprocessing.ipynb

3. Train & Evaluate Model Run the notebook:

   03_bert_sentiment_classifier.ipynb

---

Model Performance

Label	Precision	Recall	F1-Score	Support
Negative	0.83	0.80	0.81	257
Neutral	0.75	0.75	0.75	253
Positive	0.85	0.88	0.86	308
Accuracy	-	-	0.81	818
Macro Avg	0.81	0.81	0.81	818
Weighted Avg	0.81	0.81	0.81	818

---

Technologies & Libraries

• BERT (Hugging Face Transformers)
• PyTorch
• Google Play Scraper
• Pandas, NumPy
• Scikit-learn

---

License

This project is licensed under the MIT License.