---

Smart App - Document Classification Web Application

Smart App is a web application that allows users to upload documents, classify them into predefined categories using a pre-trained machine learning model, and view the classification results. The application is built with a FastAPI backend and a React frontend, leveraging modern tools and libraries for seamless functionality.

---

Features

Backend

• Document Upload: Users can upload .txt , .pdf and .docx files, which are processed and classified by the backend.

• Document Classification: Uses the facebook/bart-large-mnli model for zero-shot classification into categories:

  • Technical Documentation
  • Business Proposal
  • Legal Document
  • Academic Paper
  • General Article
  • Other

• Database Storage: Uploaded documents and their metadata (filename, content, predicted category, confidence scores, and upload timestamp) are stored in a PostgreSQL database.

• RESTful API: Well-documented API endpoints for uploading documents and retrieving classification results.

• Error Handling: Robust error handling for invalid file types, corrupt files, database errors, and model inference failures.

• Logging and Monitoring: Detailed logs for debugging and monitoring.

Frontend

• User-Friendly Interface: Built with React for a responsive and intuitive user experience.
• Drag-and-Drop Upload: Users can upload files via drag-and-drop or file selection using react-dropzone.
• Classification Results: Displays the predicted category and confidence scores for each uploaded document.
• Document List: View all uploaded documents with sorting (by filename, category, upload time, or confidence score) and pagination (9 documents per page).
• Error Handling: Informative error messages for failed uploads or data fetching.
• Responsive Design: Works seamlessly across devices.

---

Technologies Used

Backend

• FastAPI: For building the RESTful API.
• PostgreSQL: For database storage.
• SQLAlchemy: For ORM and database interactions.
• Hugging Face Transformers: For zero-shot classification using the facebook/bart-large-mnli model.
• Pydantic: For data validation.
• PyPDF2 and python-docx: For handling additional file formats.
• Pytest: For testing.

Frontend

• React: For building the user interface.
• Axios: For API communication.
• React-Dropzone: For file uploads.
• CSS Modules: For scoped and maintainable styling.
• Jest: For testing.

---

Installation and Setup

Prerequisites

• Docker and Docker Compose installed.
• Python 3.11+.
• Node.js and npm installed for the frontend.

Steps to Run the Application

1. Clone the Repository:

   git clone https://github.com/asayda01/smart_app.git
   cd smart_app

2. Set Up Environment Variables:

   • Create a .env file in the root directory with the following content:

     POSTGRES_USER = your_username
     POSTGRES_PASSWORD = your_password
     POSTGRES_DB = your_database
     POSTGRES_HOST = your_host
     POSTGRES_PORT = your_port
     

3. Build and Run with Docker:

   docker-compose up --build

   • The backend will be available at http://localhost:8000.
   • The frontend will be available at http://localhost:3000 (if configured).

4. Run the Frontend Locally:

   • Navigate to the frontend directory:

     cd frontend

   • Install dependencies:

     npm install

   • Start the development server:

     npm start

---

API Documentation

Endpoints

• POST /upload/: Upload a document for classification.

  • Request Body: multipart/form-data with a file.
  • Response: Classification results (predicted category and confidence scores).

• GET /documents/: Fetch all uploaded documents.

  • Response: List of documents with metadata.

• DELETE /documents/{doc_id}/: Delete a document by ID.

  • Response: JSONResponse including a message.

---

Database Schema

The documents table stores the following metadata:

• id: Primary key.
• filename: Name of the uploaded file.
• content: Extracted text content.
• predicted_category: Predicted category.
• confidence_scores: Confidence scores for all categories.
• upload_timestamp: Timestamp of upload.

---

Model Choice

The facebook/bart-large-mnli model was chosen for its:

• Accuracy: High performance in zero-shot classification tasks.
• Speed: Efficient inference without the need for fine-tuning.
• Confidence Scores: Provides useful confidence metrics for low-confidence predictions.

Alternatives considered:

• TF-IDF with Logistic Regression: Simpler and faster but less accurate for diverse document types.
• Other Hugging Face Models: Explored newer models such as MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli, google/flan-t5-base and microsoft/deberta-v3-large-mnli but prioritized ease of integration and reliability.

---

Future Improvements

• Enhanced error handling and user feedback.
• Integration of newer ML models for improved accuracy.
• Expanded testing coverage.

---

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository.
2. Create a new branch for your feature or bugfix.
3. Submit a pull request with a detailed description of your changes.

---

License

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

---

Contact

For questions or feedback, please contact:

• GitHub: asayda01

---

-

---

You can download the README.md file by clicking the "Download README.md" button above or by copying the content into a new file. Let me know if you need further assistance!