NewSense

The product our team introduced for the NLP Dumbathon 2025 at UC Santa Cruz organized by Cal Blanco

A text transformation and redaction application with NLP-powered keyphrase extraction that makes focused reading impossible.

Video Demo

Features

• LLM-powered text transformation and paraphrasing with various modes (brainrot, argumentative, emoji, etc.)
• Intelligent keyphrase extraction and progressive redaction
• Interactive canvas-based UI

Setup

Prerequisites

• Python 3.11+
• Virtual environment (included in dumb/)
• A Gemini API Key from your own account

Installation

Pre setup

Git clone repo. Go to gemini AI studio and go to Getting Started -> Dashboard -> API Keys -> Create API Key

then paste your API Key into a file in the main folder called super_top_secret.txt the file will just simply be the key. No quotes, nothing else, just something like Aasfoiwjoeirjoias

1. Activate your virtual environment

cd /Users/chelseyelmore/Documents/NLPDumbathon-2025
source dumb/bin/activate

2. Install dependencies

pip install -r requirements.txt
python -m spacy download en_core_web_sm

Running the App

You need to run two API servers and then open the frontend.

Terminal 1: Main API (Text Transformation)

source dumb/bin/activate
uvicorn main:app --reload --port 8000

This starts the text transformation API at http://localhost:8000

Terminal 2: Keyphrase Extraction API (Redaction)

source dumb/bin/activate
cd api
uvicorn server:app --reload --port 8001

This starts the redaction/keyphrase API at http://localhost:8001

Open the Frontend

open index.html

Or simply double-click index.html in Finder.

Quick Start

# Terminal 1
source dumb/bin/activate
uvicorn main:app --reload --port 8000

# Terminal 2 (new terminal)
source dumb/bin/activate
cd api
uvicorn server:app --reload --port 8001

# Then open canvas.html in your browser

API Endpoints

Main API (Port 8000)

• GET / - Health check
• POST /transform - Transform text
  • Body: {"text": "your text", "mode": "brainrot"}
• GET /health - Health check

Keyphrase Extraction API (Port 8001)

• GET / - Health check
• POST /extract - Extract keyphrases
  • Body: {"text": "your text", "p": 0.3}
  • Returns: {"keyphrases": [[start, end], ...]}
• GET /docs - Interactive API documentation

Project Structure

.
├── main.py              # Main FastAPI server (text transformation)
├── model.py             # Text transformation logic
├── canvas.html          # Frontend UI
├── requirements.txt     # Python dependencies
├── api/                 # Keyphrase extraction API
│   ├── server.py        # FastAPI server
│   ├── keyphrase_extractor.py
│   ├── config.py
│   ├── utils.py
│   └── README.md        # Detailed API documentation
├── assets/              # Images and assets
└── dumb/                # Virtual environment

Troubleshooting

Port already in use:

# Find process using the port
lsof -i :8000  # or :8001

# Kill the process
kill <PID>

Model not found:

python -m spacy download en_core_web_sm

Import errors:

pip install -r requirements.txt

Development

Both servers run with --reload flag, so they automatically restart when you make code changes.

For more details on the keyphrase extraction API, see api/README.md.

To restate

once you have everything setup, in one terminal run uvicorn api.server:app --reload --port 8001 in the other terminal run fastapi dev main.py then open in finder the index.html (via a web browser)