Does my message leave my device?

No — the bench runs entirely in your browser via Pyodide. Nothing is sent to a server.

How is the model trained?

Naive Bayes with Laplace smoothing (k=1) over the 5,572-row SMS dataset — the same data.csv from the repo, loaded live.

What does Weak filtered mean?

Tokens with P(spam|token) near 0.5 (weak signal, within 0.12 of 0.5) are filtered before the verdict — only strong red/green tokens count.

Hariom Lohar — spam_classifier · Shipped · live demo

A Naive Bayes spam filter you can see working. Type any message and it tokenises, scores each token P(spam|token) with Laplace smoothing, filters weak signals, then stamps SPAM or HAM with P(spam|message).

Trained live in your browser on 5,572 SMS messages (data.csv) via Pyodide + pandas — no mock, no server. The same Python that trained it scores you, token by token, in real time. Open the live bench and try WINNER!! free money now.

Highlights

Tokenisation and scoring you can see
Weak signals filtered, strong signals kept
Shows P(spam|message) with real math
Runs entirely in the browser (Pyodide)

Naive Bayes NLP Pyodide Pandas live

Python 72% · HTML 28%

Dataset

data.csv

Model

Naive Bayes

Runtime

browser