Inspiration

The inspiration for this project stems from the growing prevalence of AI-generated content and the challenges it poses in distinguishing machine-generated text from human-authored content. With the rapid advancement of natural language processing (NLP) models, there is a need for tools that ensure transparency and trust in digital content.

What it Does

The AI-Generated Text Detection Extension analyzes text in web pages, PDFs, or user-inputted text to determine whether it is machine-generated or human-authored. It provides the following features:

Scans entire web pages for AI-generated content.
Analyzes selected text or text copied to the clipboard.
Highlights AI-generated content with user-defined or default color coding.
Integrates a user-friendly interface, including a global button for easy access and detailed result visualization.

How We Built It

Frameworks and Technologies: The extension was built using the Plasmo framework for browser extension development, along with TypeScript and React for the frontend.
Architecture: The backend leverages Django and RESTful APIs to handle text analysis, model integration, and result computation.
AI Models: Pre-trained AI text detection models, including those based on GPT-2 and GPT-3, were utilized and hosted in containerized environments using Docker.
Key Components:
- Manifest: Configuration for metadata and permissions.
- Background Script: Handles event management and API calls.
- Content Script: Retrieves and processes text from web pages or PDFs.
- Model Hub: A platform for uploading, managing, and integrating detection models.

Challenges We Ran Into

Accuracy of Detection Models: Balancing speed and accuracy in detecting AI-generated content while ensuring reliable predictions.
PDF Analysis: Segmenting and analyzing content in PDFs proved challenging due to variations in text formatting and sentence structures.
Performance: Ensuring real-time analysis without significant delays, particularly for large web pages or complex documents.
Usability: Designing an intuitive and accessible interface for users of varying technical proficiency.

Accomplishments That We’re Proud Of

Successfully developed a functional browser extension that offers multiple text analysis methods.
Created a flexible architecture that supports the integration of additional AI models through the Model Hub.
Achieved high usability scores during testing, with users praising the clarity and effectiveness of the interface.
Implemented a robust caching mechanism to optimize performance and reduce redundant computations.

What We Learned

Technical Skills: Gained expertise in browser extension development, integrating APIs, and leveraging Docker for containerized deployments.
Model Integration: Learned how to adapt pre-trained AI models for specific use cases and ensure compatibility with the backend.
User-Centric Design: Understood the importance of usability testing and iterative design to meet user expectations.
Scalability: Realized the value of designing a modular and scalable system to accommodate future enhancements and use cases.

What’s Next for AI-Generated Text Detection Extension-Model

Enhanced PDF Analysis: Develop advanced segmentation techniques for more reliable paragraph-based analysis.
Model Expansion: Integrate new and improved AI detection models to handle multilingual content.
UI Redesign: Optimize the user interface for better accessibility and visual appeal.
Commercialization: Explore freemium models and licensing opportunities to make the extension widely available.
Real-Time Processing: Incorporate machine learning optimizations to achieve faster, real-time text analysis.
Broader Functionality: Expand features to include sentiment analysis, text originality checks, and other NLP-based utilities.