Slashdot Mirror
Baidu Shows Off Its Instant Pocket Translator (technologyreview.com)
MIT Technology Review: Baidu showed off the speed of its pocket translator for the first time in the United States during an afternoon presentation at MIT Technology Review's EmTech Digital conference in San Francisco. The Chinese Internet giant has made significant strides improving machine language translation since 2015, using an advanced form of artificial intelligence known as deep learning, said Hua Wu, the company's chief scientist focused on natural-language processing. On stage, the Internet-connected device was able to almost instantly translate a short conversation between Wu and senior editor Will Knight. It easily rendered Knight's questions -- including "Where can I buy this device?" and "When will machines replace humans?" -- into Mandarin, and relayed Wu's responses in clear, if machine-inflected, English.
Will this be able to translate "help me get out of this pre-nup and I'll give up the butt" from Slovenian to English? Asking for a friend.
You are welcome on my lawn.
I hate it when people say "internet connected" instead of using it's proper name: IoT-linked SaaS big-data cloud-hosted service.
If it's not scripted, then this is very impressive. He should have tried "Could you tell me more about what happened on Tiananmen Square?", though.
From which language was your comment translated?
By a strange coincidence, I discovered that Google Translate app on Android has this feature. It sucks and I failed to make it work, but it's there, unmistakably for the subject purpose.
I do not believe in karma. "Funny"=-6. Do good and forbid evil. Yours, Oft-Offtopic Flamebaiting Troll.
Digital Dirty Hungarian Phrasebook
* Actually four.
Right now voice recognition can't even accurately capture what I say, not a fucking hope it'll also be able to translate it to another language.
Example - the above sentence captured using MS speech recognition:
By now voice recognition come even actually capture what I say was that England can also be able to translate into another language
..and in Google Docs:
Right Now voice recognition can't even accurately capture what I say and I f****** hope it also go to translate it to another language
(yeah, Google add the *s)
Google's getting close, although it struggles a lot more at times, but even that is going to sound garbled as fuck when translated - and that's with a quality microphone and a silent room, no background noise, no other people speaking, no traffic, no shitty phone microphone.
China is at least 50 yrs behind and has not even caught up with the translators on Star Trek...
We've been failing at speech recognition so long, maybe it's time to re-train humans to speak in a non-ambiguous language designed with clear sounds that machines can better understand.
Good luck with that; we could call it something like "Esperanto" ...
I completed the Duolingo Esperanto course in less than 2 months. I'm not fluent- but know enough that I'm reading a novel written completely in Esperanto now. Took two months before I started reading my first book. Show me any other language which is possible to casually learn in 2 months.
There is no reason why schools couldn't teach one year of Esperanto and have the whole world communicating effectively. (now, I'm sure it's easier to learn Esperanto if you're coming from a European language because almost all of it's words have European roots- but it's actually the Chinese who seem to be promoting it more in school- and it is easier to learn than most languages due to its predictability and minimal root words. China even has all it's official news translated into Esperanto).
I doubt I'll ever get to have a conversation with anyone in Esperanto (besides my kids who also learned it, but not as thoroughly in the same time); however the advantage is, tests show knowing Esperanto halves the time it takes to learn additional new languages- which is the real reason I learnt it in the first place.
"That's the way to do it" - Punch
When they make it look like a small fish that you can stick in your ear, it'll sell like crazy.
When they make one that can translate a teenager, I'll be truly impressed. I suppose they could analyze the speech patterns and use that to decide what they actually mean when they use a term like "literally". Still, listening to two teenage girls talk when no adult is in the conversation is like comparing math to accounting. They both use a lot of the same terms, but they have completely different meanings.
But when they make one that translates what wives actually mean when they say something, they'll become the most valuable company in the world.
Localized language variants will always be any translator's bugaboo. I remember living back in Connecticut there were some local quarrymen's whose English sounded like another language entirely *LOL*. Apple's speech-to-text can barely deal with English, and routinely *makes up words* that have no counterpart anywhere in our language. Very cool that so much work is being done, though!
--I do what I can, I work in the dark.