Lexiguard

A machine learning–powered text spam detection system built for the final NLP project, designed to help the La Alianza Hispana community identify and protect against spam messages.

Lexiguard includes multiple classifiers, a Flask API server, a React-based web client for real-time predictions, and an integrated digital literacy education component.

Deployment

The application is currently deployed on Render: https://lexiguardnlp.onrender.com/

Note: Due to free tier memory constraints (512MB), the deployed version only runs Logistic Regression. For full 4-model functionality, run locally.

Demo Video

https://youtu.be/gk5xpqXjcR0

Features

Spam Detection Models

Model	Description
Logistic Regression	Reliable linear classifier
Naive Bayes	Probabilistic classifier with stemming
Neural Network	Dense feed-forward classifier
Transformer	DistilBERT fine-tuned model (99.28% accuracy)

The Flask API supports all 4 models with dynamic switching via dropdown selector.

Digital Literacy Education

Designed specifically for elderly members of the La Alianza Hispana community, Lexiguard includes:

• Real-time spam indicator detection - Identifies 8 warning signs in messages (urgency, free prizes, ALL CAPS, suspicious links, etc.)
• Interactive education modal - Explains why a message is spam and what specific red flags were detected
• Bilingual support (EN/ES) - Full Spanish translations for accessibility
• Safety tips - Actionable advice to protect against scams (never share passwords, verify sender identity, etc.)
• Visual feedback - Animated text box pulses (green for safe, red for spam) to reinforce learning

The education component directly supports service learning objectives by teaching digital literacy skills to vulnerable community members.

Steps to run:

Prerequisites

If pulling for the first time, install Git LFS to download the Transformer model:

brew install git-lfs
git lfs install
git lfs pull

Environment Setup

Create a .env file in client/ with:

REACT_APP_API_URL=http://localhost:8000

Server

cd server
pip install -r requirements.txt
python -c "import nltk; nltk.download('punkt')"
python app.py

The backend will start on: http://localhost:8000

Client

cd client
npm install
npm start

The frontend will start on: http://localhost:3000

Project Structure

LexiguardNLP/
├── client/               # React frontend
├── server/              # Flask API
│   ├── app.py          # Multi-model API server
│   ├── *.pkl           # Serialized models & vectorizers
│   ├── *.keras         # Neural network model
│   └── transformer_*/  # DistilBERT model & tokenizer
└── models/             # Jupyter notebooks for training