The problem with wacky, unpredictable "offbeat news" is that it's predictable and unwacky. Most of journalism plays by "Man Bites Dog" rules, informing you only of deviations from the norm. Offbeat News is perfectly happy to report Dog Bites Man, so long as the dog is wearing a funny hat. Surely you were expecting another dumb criminal who robs a bank with a banana, or the kitten who thinks he's a lion. What would be really newsworthy would be if the kitten robbed the bank, or the dumb criminal thought he was a lion. But reality is rarely so accommodating. We must take matters into our own hands.

Fortunately, news headlines are quite formulaic. It's now possible to rip them in twain and recombine them to form the real news. So offbeat, it never happened.

Dog Bites Dog

The offbeat news The depressing news you're using the offbeat news to escape 



Want it all? Here you go.


The original headlines are taken every hour from The shuffled headlines are updated every five minutes.

Dog Bites Dog is powered by a couple of Python scripts; one that uses Beautiful Soup to scrape headlines, and one that uses FRELI to figure out where to chop up the headlines.

Several problems can lead to malformed headlines (as oppose to merely nonsensical headlines, which are good clean fun). The most egregious (and easy to fix) is the disagreement between a singular object and a plural predicate, or vice versa. Other headlines prove intractable due to the vagaries of English syntax or words that have multiple meanings; the script usually knows to ignore a headline it can't understand, but sometimes it oversteps the bounds of its circumscribed knowledge without realizing it.

Thanks to Sumana Harihareswara for naming this feature.

This document is part of Crummy, the webspace of Leonard Richardson (contact information). It was last modified on Friday, August 23 2013, 19:13:55 Nowhere Standard Time and last built on Saturday, June 03 2023, 14:20:02 Nowhere Standard Time.

Crummy is © 1996-2023 Leonard Richardson. Unless otherwise noted, all text licensed under a Creative Commons License.

Document tree:
Site Search: