Dataset · ucirvine
Classic SMS message corpus for spam filtering
ucirvine/sms_spam
A set of 5,574 real English SMS messages labeled as spam or ham for text classification tasks.
What's inside
A public collection of SMS messages gathered for mobile spam research. The dataset contains 5,574 real unencrypted texts in English, each assigned to one of two classes: legitimate message (ham) or spam.
What you can do with it
- Training basic text classifiers for spam detection in messengers and SMS
- Testing and benchmarking lightweight NLP models on binary classification
- Practice and prototyping of short text message processing pipelines
- Tasks
- Text classification
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 9K
- Likes
- 59
Models trained on it
- StrandsAgents/strands-decider-2B-qwen3.5-v1-2610Text classification 24.5K
- StrandsAgents/strands-decider-2B-hobson-v21Text classification 15.6K
- StrandsAgents/strands-decider-2B-hobson-v19Text classification 7K
- StrandsAgents/strands-decider-E4B-gemma4-v1-2610Text classification 1.4K
- StrandsAgents/strands-decider-12B-gemma4-v1-2610Text classification 1.1K
- manjunathshiva/opendecider-nanoZero shot classification 1K
- StrandsAgents/strands-decider-E2B-gemma4-v1-2610Text classification 728
- StrandsAgents/strands-decider-26B-A4B-gemma4-v1-2610Text classification 360
How to load it
from datasets import load_dataset
ds = load_dataset("ucirvine/sms_spam")