I received a skeptical query in response to my recent article “When Robots Write the News“:

C’mon, those were in spam? They were too grammatical. The ones I get are usually the first several words of one sentence and the next several words of another, without regard to grammar.

All of the text I quoted was lifted from spam, and all phrases placed next to each other in a sentence actually appeared in that order in the spam.  However, I did remove a few extraneous words and edit the punctuation to improve the effect.  Sentences that are together in a paragraph came from the same message but did not necessarily appear in that order.  There’s one exception: A woman found buried in sand in a bathtub has clinched the playoff. and “This man needs my handbags like trading one menacing existence for another,” she said. came from two separate messages, but the combination was just irresistable.

I do receive spam in which the random text is much more choppy and ungrammatical, but many of the spams I’ve been getting at work lately use longer phrases and string them together in a way that comes strangely close to making sense.