Downloads · 30 days
6
22% of all-time downloads
sms-spam-detector-from-scratch
ausjahk/sms-spam-detector-from-scratch
sms-spam-detector-from-scratch is a text classification model from ausjahk. Use it when you need a label for a piece of text.
Custom Bidirectional LSTM + Embedding Model trained entirely from scratch (random weight initialization, no pre-trained BERT) on the Kaggle SMS Spam Collection dataset.
All-time downloads
27
Public
Repo size
1.6 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin1.6 MB · 94%
At a glance
- Task
- Text Classification
- Access
- Public
- Created
- Sep 3, 2026
- Updated
- Sep 3, 2026
- SHA
- c345c1cd
Try a prompt
Model card
From the Hugging Face model README
SMS Spam Detector (Trained From Scratch)
Overview
Custom Bidirectional LSTM + Embedding Model trained entirely from scratch (random weight initialization, no pre-trained BERT) on the Kaggle SMS Spam Collection dataset.
Architecture
- Custom Tokenizer & Vocab (5002 words)
- Embedding Layer (dim=64)
- Bidirectional LSTM (hidden_dim=64)
- Dense Linear Layer (32 -> 2 classes)
- Total Parameters: 390,882 (All random initialization)
Dataset & Performance
- Source: Kaggle SMS Spam Collection
- Test Accuracy: 97.58%
- Labels: 0 = HAM (Legit) ✅ | 1 = SPAM (Scam) 🚨
- Task
- Text Classification
- Created
- Sep 3, 2026
- Updated
- Sep 3, 2026