Pannuru/md-validator

Spam Email Detection Using DistilBERT

Project Overview

A DistilBERT-based model that classifies email messages as Spam or Ham.

Dataset

SetFit/enron_spam

• Training samples: 8,000
• Testing samples: 2,000
• Spam label: 1
• Ham label: 0

Technologies

• Python
• PyTorch
• Hugging Face Transformers
• Google Colab
• DistilBERT

Workflow

1. Dataset preparation
2. Tokenization
3. Model training
4. Model evaluation
5. Spam and Ham prediction
6. Spam indicator explanation

Example

Input: Congratulations! You won a free prize.

Prediction: SPAM

Limitation

Spam explanations use predefined keyword indicators and do not establish the model's causal reasoning.