Skip to content

REALSCORE is a privacy-focused, browser-based automated text scoring system that provides real-time feedback on writing quality. Using TensorFlow.js, it analyzes text directly in your browser without sending data to any server.

Notifications You must be signed in to change notification settings

Divuzki/realscore

Repository files navigation

REALSCORE - Automated Text Scoring System

License: MIT

REALSCORE is a privacy-focused, browser-based automated text scoring system that provides real-time feedback on writing quality. Using TensorFlow.js, it analyzes text directly in your browser without sending data to any server.

🚀 Features

  • Real-time Analysis: Get instant feedback as you write
  • Complete Privacy: All processing happens in your browser
  • Multiple Scoring Dimensions:
    • Grammar and syntax
    • Content coherence
    • Vocabulary richness
    • Topic relevance
    • Overall quality score
  • Rich Text Editor: Full formatting capabilities
  • Data Persistence: Auto-save and local storage
  • Export Options: Save your work in multiple formats
  • Responsive Design: Works on desktop and mobile
  • Offline Support: Continue working without internet

🛠️ Technology Stack

  • Frontend: React 18, TypeScript, Tailwind CSS
  • Text Processing: TensorFlow.js, Universal Sentence Encoder
  • State Management: React Context API
  • Storage: LocalForage
  • Text Editor: React Quill
  • Visualization: Chart.js with React Chartjs 2
  • Icons: Lucide React
  • Build Tool: Vite

📋 Requirements

  • Node.js 18.0.0 or higher
  • Modern web browser (Chrome, Firefox, Safari, Edge)
  • 4GB RAM minimum (8GB recommended)
  • 500MB free disk space for model storage

🚀 Getting Started

  1. Clone the repository:

    git clone https://github.com/divuzki/realscore.git
    cd realscore
  2. Install dependencies:

    npm install
  3. Start the development server:

    npm run dev
  4. Open your browser and navigate to http://localhost:5173

📝 Usage

  1. Create a New Document:

    • Click the "Create New Document" button
    • Start typing in the rich text editor
  2. View Analysis:

    • See real-time scores in the sidebar
    • Review detailed breakdowns for each metric
    • Get suggestions for improvement
  3. Save and Export:

    • Documents auto-save to local storage
    • Export to PDF, DOCX, or TXT formats
    • Access previous documents from the dropdown menu

🔍 Scoring Metrics

Grammar & Style (0-10)

  • Sentence structure
  • Punctuation
  • Subject-verb agreement
  • Writing mechanics

Coherence & Organization (0-10)

  • Logical flow
  • Paragraph structure
  • Transitions
  • Overall organization

Vocabulary Usage (0-10)

  • Word choice
  • Vocabulary diversity
  • Language sophistication
  • Context appropriateness

Topic Relevance (0-10)

  • Content focus
  • Main idea development
  • Supporting details
  • Topic consistency

🔒 Privacy Features

  • All text analysis happens client-side
  • No data sent to external servers
  • Local storage encryption
  • Optional data purging
  • No analytics or tracking

🛠️ Development

Project Structure

src/
├── components/     # React components
├── hooks/         # Custom React hooks
├── types/         # TypeScript definitions
├── utils/         # Utility functions
└── main.tsx       # Application entry point

Key Components

  • TextEditor: Rich text editing interface
  • ScoreSidebar: Real-time scoring display
  • ScoreRadarChart: Score visualization
  • ModelLoader: ML model initialization

Custom Hooks

  • useTensorflowModels: ML model management
  • useDocumentStorage: Local storage operations

🤝 Contributing

  1. Fork the repository
  2. Create a feature branch
  3. Commit your changes
  4. Push to the branch
  5. Open a Pull Request

📄 License

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

🙏 Acknowledgments

  • TensorFlow.js team for browser ML capabilities
  • React team for the excellent framework
  • Open source community for various tools and libraries

⚠️ Known Limitations

  • Initial model loading may take time
  • Heavy CPU usage during analysis
  • Limited to modern browsers
  • Requires JavaScript enabled

Built with ❤️ using React and TensorFlow.js

About

REALSCORE is a privacy-focused, browser-based automated text scoring system that provides real-time feedback on writing quality. Using TensorFlow.js, it analyzes text directly in your browser without sending data to any server.

Resources

Stars

Watchers

Forks