Interactive demo

SMS bulk-style classifier

Trained on SMS spam / not-spam labels. The score is P(bulk-style). How much the text resembles bulk or automated messaging in the dataset, not whether a message is dangerous, fraudulent, or “malicious.”

Common misconceptions & disclaimer
01

Bulk-style is not the same as a scam

Spam here means mass/automated wording, not “scam.” 2FA codes and shipping pings can read “bulk” and still be fine.

02

A high score is not a danger rating

High P(bulk-style) is not a danger score. Low P is not proof of safety—use judgement on links and payments.

Disclaimer

Coursework demo. Not for production or safety decisions.

Failure case (example)

Representative error

Legitimate reminder with “Reply STOP to unsubscribe” scored 95.7% spam—opt-out wording is a strong spam cue in training, so real senders can false-positive without sender context.

Checking inference API…
user@spam-detector:~$ sms_input.txt
Message
Course report (PDF)

Full write-up: course report (PDF). Coursework demo. P(bulk-style) measures wording similarity to bulk SMS in training—not scams or safety.