Google Reviews Sentiment Analysis with Transformers

Advanced sentiment analysis system using state-of-the-art Transformer models to analyze Google Maps reviews.

Features

• 🤖 Multi-Model Transformer Analysis

  • Sentiment Analysis (DistilBERT)
  • Emotion Detection (RoBERTa)
  • Category Classification (BART Zero-Shot)

• 📊 Comprehensive Insights

  • Sentiment distribution and scoring
  • Emotion detection (joy, anger, sadness, etc.)
  • Automatic category classification
  • Keyword extraction
  • Interactive visualizations

• 💾 Data Management

  • CSV export of raw and analyzed data
  • Organized file structure
  • Easy to integrate with other tools

Installation

1. Clone the repository:

git clone https://github.com/yourusername/google-reviews-sentiment-analysis.git
cd google-reviews-sentiment-analysis

2. Create virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure environment:

cp .env.example .env
# Edit .env with your Google Places API key and Place ID

Usage

Run complete analysis:

python main.py --all

Or run individual steps:

python main.py --fetch      # Fetch reviews from Google
python main.py --analyze    # Analyze with Transformers
python main.py --visualize  # Generate visualizations

Output

• data/raw_reviews.csv - Raw reviews from Google
• data/analyzed_reviews.csv - Reviews with sentiment analysis
• visualizations/sentiment_report.html - Interactive dashboard
• visualizations/summary_report.txt - Text summary

Models Used

• Sentiment: distilbert-base-uncased-finetuned-sst-2-english
• Emotion: j-hartmann/emotion-english-distilroberta-base
• Categories: facebook/bart-large-mnli (zero-shot)

Project Structure

├── config.py              # Configuration
├── fetch_reviews.py       # Google Places API integration
├── analyze_reviews.py     # Transformer analysis
├── visualize_results.py   # Visualization generation
├── main.py               # Main pipeline
├── data/                 # CSV storage
└── visualizations/       # Output reports

Requirements

• Python 3.8+
• Google Places API key
• GPU recommended (but works on CPU)

License

MIT License