Facebook Plunges Again
Tumblr 101: Tips And Tricks To Get Started
Google censors 'Lolita' but finds bestiality acceptable
It seems that Google Voice (or, more specifically Voice when running on Android devices) replaces censored words with a rather bland and completely useless '####' when transcribing your calls and searches. Back in January it came to light that Android running on the Nexus One was censoring offensive terms, but most of us thought the censorship was was a feature and not actual censorship. Google says that the technology is meant to prevent the misrecognizing of offensive words -- you might be saying 'country', but the phone might interpret it as... something much more malevolent.
Now it turns out that the list of naughty words might be a lot more arbitrary. Is 'Lolita' really similar to another offensive term, or easily misunderstood? 'Whorehouse' is also blocked -- I guess it sounds a bit like, er... boar house? 'Beer house' said by a drunk? Who knows...
I'm sure Google will be quick to correct it, if there is in fact a problem -- they're one of the biggest advocates of free speech, after all.
Comments
9
Subscribe to commentsjfjbFeb 1st 2010 10:39AM
.... hence the usefulness of knowing two or three different languages. It requires -- later, if ever -- to cross-reference the results.
Profanity, for example, has a flexible definition for it concerns the impact it has on people, not machines. People program machines, not the other way around yet. People have an agenda based on religion, politics, PR, marketing, you name it. Not machines yet. Long story short, if '*' are used for security to display passwords, '#' can display those words of flexible meaning.
Only things is to know where to start and end the list of those words. Our parents had an equivalent approach, and we survived our childhood.
Where has gone our imagination? You-know-wata-mean. And if not, replace it with '####'. Just kidding.
My two-cents, of course. There are bigger fish in the pond.
Ciao.
Au revoir.
Auf Wiedersehen.
¡Adiós!
See yah!
Sebastian AnthonyFeb 1st 2010 11:02AM
Multiple languages are great, but let's get speech-recognition and stuff working for English first... then we can work on the rest :)
JPFeb 1st 2010 10:57AM
Warehouse.
Also, I've seen the results Google voice gives when transcribing someone who doesn't speak loudly or enunciate well: Transcripts are very useful for getting the gist of a message but they often miss a few words at a time, which end up looking nothing like they were intended to sound. It's safer for them to not have some furious person asking why Google put an offensive word in their voicemail, (or worse, their kid's voicemail) accidentally or no.
jfjbFeb 1st 2010 12:18PM
@ sebastian
I've been using ViaVoice and then Dragon Naturally Speaking for years without major problems for my French accent to be recognized in English, Spanish and German at times of need. I a surprised a corporation the size of Google cannot come up with at least 60 to 70% voice rendering and /or transcription quality.
I know 'devices' may not have cristal clear microphone, I know people speak and do not dictate on the phone, but still, aren't we facing more a debate of intentions rather than a quality assurance process?
C'est ce que je pense, j'ai peut-être tout faux.
Oops.
That's what I think, I may be wrong.
jfjbFeb 1st 2010 12:20PM
Oops, I made a typo, moi?
"I a surprised" should read "I am surprised.
Sebastian AnthonyFeb 1st 2010 12:22PM
I don't think either of us are wrong... just opinionated!
English and French both have one thing in common... free speech :)
jfjbFeb 1st 2010 12:44PM
you sound like my kind of guy
or my brother
|;-)))
EvenioFeb 1st 2010 7:03PM
And no, your "easter egg" has not escaped us. :3
Sebastian AnthonyFeb 1st 2010 7:11PM
I think you might be the only one to notice it...!