IF YOU have ever received a document containing off-putting expressions such as “At your desecration” or “Sorry for the incontinence”, then you have witnessed the havoc that can be wreaked by placing unthinking trust in spellcheckers. The problem is widespread enough to have acquired a name: the Cupertino effect.
, is home to computer giants Apple and Hewlett-Packard. It owes the dubious honour of sharing its name with a dumb error to the fact that some early spellcheckers flagged up the word “co-operation” if it was spelled without a hyphen. Type “cooperation”, and they came up with the suggestion “Cupertino”.
That problem was soon fixed, but a quick search of the UN website still turns up evidence of Cupertino carnage. There are references to the “South Asian Association for Regional Cupertino”, “political, economic and trade Cupertino”, and a “presentation on African-German Cupertino” – unfortunately not a delicious new coffee beverage.
Advertisement
Benjamin Zimmer, a linguist at the University of Pennsylvania, has , including an article in The New York Times that gave the first name of American footballer DeMeco Ryans as Demerol, one in The Denver Post that turned Harry Potter villain Voldemort into Voltmeter, and a Reuters story that transformed the Muttahida Quami movement of Pakistan into the Muttonhead Quail movement.
“A Denver Post story turned Harry Potter villain Voldemort into Voltmeter”
So why do spellcheckers offer such bizarre suggestions? The programs recognise four basic types of misspelling: deletion, addition, transposition and substitution (as in infomation, developping, beleive and independant). Then they calculate how many of these changes it takes to go from a word they don’t recognise to one in their dictionary – a concept known as “edit distance” – and suggest those with the shortest edit distance as alternatives.
Early spellcheckers didn’t know what to do with hyphens – hence Cupertino – and ran into problems when the first letter of a word was missing. So Word 97 users who skipped the first letter of “identified” were urged to accept “denitrified”. These faults are now mostly a thing of the past, but uncommon names and novel slang can still trip up the unwary.
More recent tweaks also compensate for human foibles such as mixing up words and letter groups that sound the same, like “quiet” and “quite” or “f” and “gh”. It’s a small step forward for Cupertino – er, cooperation – between people and computers.
