37RanText & sentiment

SMS spam

I turned 5,572 text messages into word counts, trained multinomial naive Bayes on two thirds of them, and checked one message I typed, which came back as spam.

Original tutorial by Aman Kharwal: Spam Detection (opens in a new tab)

What it printed in 2022

Numbers read from the saved notebook outputs. Cell numbers count every cell in the notebook, Markdown included, starting at 1.
Output	Value	Cell #
Label for one typed message	spam	9

Cell # counts every cell from the top of the notebook, Markdown included, starting at 1.

No accuracy, precision or recall was printed.

The data

The SMS Spam Collection (UCI, CC BY 4.0), read from a re-upload.

What I'd do differently

Report spam precision and recall next to the 87% you get by calling everything ham.
Fit the vectoriser inside each cross-validation fold.

Re-checked in the audits

37
InteractiveHolds up
Spam detection holds up
Yes. It scores 98.4% against 86.1% for 'always ham', catching 96% of spam with 93% precision. Ten-fold cross-validation agrees.
Read the audit

What it printed in 2022

The data

What I'd do differently

Re-checked in the audits

Spam detection holds up

What it printed in 2022

The data

What I'd do differently

Re-checked in the audits

Spam detection holds up