ABScribeX

ABScribeX brings ABScribe's in-place Variation Fields and reusable AI Modifiers to any web editor—be it Gmail, LinkedIn, or Reddit—via a synchronized pop-up interface, enabling rapid, non-linear exploration and organization of multiple writing variants right where you compose.

About

ABScribeX is a Chrome extension that implements the research concepts from the ABScribe paper, which won recognition at CHI 2024. The original ABScribe study showed that this approach significantly reduces writing workload and enhances the revision process when working with AI-generated text variations.

Key Features

• Universal Web Editor Support: Works seamlessly with Gmail, LinkedIn, Reddit, and any web-based text editor
• In-Place Variations: View and compare multiple writing variations without cluttering your document
• AI Modifier Buttons: Convert prompts into reusable buttons for quick text transformations
• Synchronized Interface: Popup window stays in sync with your original content
• Non-Destructive Editing: Explore variations without overwriting your original text
• Context Menu Integration: Right-click any text field to start editing with ABScribeX

How It Works

1. Capture: Right-click on any text field or editor on the web
2. Edit: Use the popup interface to generate and explore writing variations
3. Apply: Choose your preferred variation and apply it back to the original editor

Installation

1. Clone this repository
2. Install dependencies: npm install
3. Build the extension: npm run build
4. Load the extension in Chrome from the dist folder

Development

npm install
npm run dev      # Development mode with hot reload
npm run build    # Production build
npm run test     # Run tests

Research Background

ABScribeX is based on the CHI 2024 paper "ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models" by Mohi Reza, Nathan Laundry, and colleagues. The research demonstrates that this interface approach significantly improves the writing process when working with AI-generated content.

License

This project builds upon the ABScribe research and is intended for educational and research purposes.