Book Content Scanner

Version 0.1

A context-aware EPUB scanner identifying objectionable content (profanity, hate speech, violence, sexual content) using AI models and customizable optimization levels.

Features

• Keyword-based initial filtering with context-aware exclusions
• Transformer models for AI analysis:
  • Toxic-BERT (profanity)
  • RoBERTa Hate Speech
  • Twitter-RoBERTa Sentiment
• Phase 3 optimization levels (A/B/C) for 8GB VRAM:
  • Option A: Enhanced analysis + violence detector + caching
  • Option B: Option A + 8-bit quantization + batch processing
  • Option C: Option B + 4-bit quantization + quantized Llama Guard
• Custom violence detector
• Smart caching to avoid redundant AI calls
• Configurable entirely via scanner_settings.py

Installation

1. Clone the repository:

   git clone https://github.com/jree101/book-content-scanner.git
   cd book-content-scanner

2. Install dependencies:

   pip install -r requirements.txt

3. Place EPUB files in the books/ folder.
4. Run the scanner:

   python book_scanner.py

Configuration

Edit scanner_settings.py:

• PHASE3_LEVEL = "A" # Change to "B" or "C" for more optimizations
• Adjust CONFIDENCE_LEVELS thresholds
• Enable/disable models based on hardware

Repository Layout

book-content-scanner/
|-- .gitignore
|-- LICENSE
|-- README.md
|-- requirements.txt
|-- content_config.py
|-- scanner_settings.py
|-- book_scanner_phase2.py
|-- books/            # Place EPUBs here
|-- scan_reports/     # Generated reports

License

This project is licensed under the MIT License. See the LICENSE file for details.