How Google's Nexus One censors cuss words

The built-in voice-to-text feature on Google's new Nexus One phone replaces rude utterances with hash marks.

Chris Matyszczyk
2 min read

Some of you who have been basking in the beauty of your new Nexus One Googlephone may not have tried out all of its delightful features.

And what I am about to tell you may lead you to utter some naughty words. Please, go ahead. I have heard them all, in several different languages. And I respect the vehemence of the vernacular.

However, your Nexus One will not be so charmed by the vigor of your tongue. It will, dare I utter the word when referring to a product from the newly emancipated Google, censor you.

You see, the pungently polite people at Reuters were playing with their Nexus One when they noticed something about its built-in voice-to-text feature.

Every time they said something naughty into the phone, the naughty word came out as "####"--and not just "f---." It even censored the "S" part of BS.

Reuters immediately called Google and screamed at them: "What the #### are you miserable ############# playing at?"

It's a ####ing good phone, so I'm told. CC Pittaya/Flickr

Oh, perhaps I have stretched the boundaries of possibility with that heartening notion. They probably asked a little more politely, given that they secured a really quite ingenious reply from a Googleperson.

Apparently, the censorship is not because Google is trying to clean up the world and turn it into the nicest parts of Alabama. No, the company is worried about what might be transcribed.

"We filter potentially offensive or inappropriate results because we want to avoid situations whereby we might misrecognize a spoken query and return profanity when, in fact, the user said something completely innocent," Google told Reuters.

Yes, the technology isn't quite perfect, so even the potential of a misplaced curse is being avoided at all costs.

What interests me most is how Google chooses its list of naughty nuances. Is there some poor engineer over at the Googleplex whose sole task was to write software that immediately identifies expletive expressions? Do they take account of those who swear in Spanish, Italian, or, like me, Polish?

And if you say "For crying out loud" a little too quickly, might the transcription come out with a four letter f-word (or rather four hash marks) at the beginning?

One other thing. I have a Croatian friend. If I ever got a Nexus One, I would like to be able to address him by his name. His name is Fuk. Would his name be transcribed, every time, as ###? How sad.