Think you have trouble deciphering social media slang? Try translating it. Microsoft researchers have been studying how to translate social media, and in their efforts they came across a way to teach the company’s upcoming Skype Translator how to speak more like us.
Some researchers think social media could be key to getting computers to better understand humans. Social media experiments are “important examples of a new line of research in computational social science, showing that subtle social meaning can be automatically extracted from speech and text in a complex natural task,” saysDan Jurafsky, an expert in computational linguistics at Stanford, who recently led work on teaching computers about human interactions by listening to speed dating.
The Skype Translator app, set for beta release later this year, translates multilingual conversations over the service as they’re happening. In May, Gurdeep Singh Pall, corporate vice president of Skype and Lync at Microsoft, and a German-speaking colleague demoed the app at the Code Conference, in Rancho Palos Verdes, Calif. As Pall spoke in English, both German and English subtitles scrolled along the bottom of the screen while real-time audio translation accompanied the subtitles. More.
See: IEEE
Related Translation News posts:
- Microsoft unveils real-time audio and text-based language translation via Skype
- Skype Translator vows to translate multilingual voice calls – but is it any good?
- Microsoft readies real-time language translator for Skype
Subscribe to the translation news daily digest here. See more translation news.
Share an article »
ProZ.com blog »
Australia
Local time: 04:32
English to Hungarian
+ ...
Thanks, Maria, I published it mentioning you as my source on:
http://www.scoop.it/t/what-would-you-loose-if-nobody-would-translate
United States
Local time: 14:32
Russian to English
+ ...
Cannot they do something real, or get a real job?
[Edited at 2014-08-23 09:06 GMT]
Japan
Local time: 03:32
Japanese to English
+ ...
Skype's real-time Translator learns how to speak from social media
Oh, you mean it's learning to translate from stuff like this?
Local time: 13:32
German to English
+ ...
Is that for the text box underneath? I know Skype mostly for speech - the stuff you hear rather than read.
China
Local time: 02:32
Chinese to English
Oh, you mean it's learning to translate from stuff like this?
If a machine learned to understand tweets like those, it would be well on the way to learning to translate properly. Twitter could be a goldmine for MT researchers.
Though, looking at those tweets makes me so very glad that I don't do Twitter. Who in their right mind thinks that allowing that drivel into their head makes the world a better place?
United States
Local time: 14:32
English to German
+ ...
Please visit:
http://www.proz.com/forum/translation_news/269908-microsoft_unveils_real_time_audio_and_text_based_language_translation_via_skype.html
I took a look at the output.
Japan
Local time: 03:32
Japanese to English
+ ...
Who in their right mind thinks that allowing that drivel into their head makes the world a better place?
Hundreds of millions of people, it seems, judging by the amount of followers some of these accounts have.
And if machines learn to understand English like that...well, at that point we humans would in all likelihood either be extinct or else enslaved to them.
Japan
Local time: 03:32
Japanese to English
+ ...
Please visit:
http://www.proz.com/forum/translation_news/269908-microsoft_unveils_real_time_audio_and_text_based_language_translation_via_skype.html
I took a look at the output.
Yes, that was an interesting analysis you did. To think, that was a staged, prepped demonstration where both parties were speaking near-perfect English and German.
If anyone thinks that Skype Translator is going to correctly translate speech from someone like Snoop Dogg, then it might be time to see a psychiatrist.
China
Local time: 02:32
Chinese to English
I raise that as an example of stuff you can do with massive participation. Is your translator having trouble understanding certain accents? Release a viral video requiring responses in which people read a certain script. You get a million responses, with all kinds of accents and a known text, and your computer gets a zillion more datapoints to improve its comprehension. Siri must be doing that already.
Once a computer system is genuinely interacting with people in a particular way, I think improvement can be very quick. ▲ Collapse
Japan
Local time: 03:32
Japanese to English
+ ...
I'm actually more optimistic about the possibility of a Skype translator working than you. Do you remember the Google image game? Early on, G had the problem of how to get the computer to understand images, so they developed a game whereby they flashed images up to two people at the same time, and they had to write the words that the image looked like. You got points if both people wrote the same words, so that incentivised bored students to play, and very quickly, G got a remarkably accurate database of keywords reflecting what real people see in images (as opposed to what the image is putatively "of").
I raise that as an example of stuff you can do with massive participation. Is your translator having trouble understanding certain accents? Release a viral video requiring responses in which people read a certain script. You get a million responses, with all kinds of accents and a known text, and your computer gets a zillion more datapoints to improve its comprehension. Siri must be doing that already.
Once a computer system is genuinely interacting with people in a particular way, I think improvement can be very quick.
What you are describing though is recognition. Computers are pretty good at that already, as you can tell from how far voice-recognition technology has come (as well as the image example you describe).
The way I see it, the problem is not recognition, but synthesis. A computer has been taught to recognize the elements of an image, but can it create one from scratch? Similarly, even my lowly laptop can recognize complex English sentences perfectly through voice recognition, but it certainly can't come up with them on its own, or even convert them accurately into another language.
And frankly, it's more than a little scary to imagine what will happen when a machine does actually get to the point where it can synthesize data at the level of a human. Because I can guarantee that it won't stay at that level for long.
[Edited at 2014-08-24 18:18 GMT]
China
Local time: 02:32
Chinese to English
Now, none of the above applies: MMPORGs. You want to get a computer to learn how to give accurate instructions in French, you take over WoW in French, tell it to experiment with its language, count how many people do the required thing. Very quickly, it has functional French.
Once computers interact with us, they will get good at interacting with us.
As to the worry - I'm not worried because of a philosophical belief in the importance of desires to consciousness. Basically, SkyNet won't happen because computers don't want anything, and we don't actually know how to program desires in.
But I take seriously the idea that everyone could be out of a job in 20 years. Us, sure, but everyone else as well.
[Edited at 2014-08-24 18:33 GMT] ▲ Collapse
Japan
Local time: 03:32
Japanese to English
+ ...
As to the worry - I'm not worried because of a philosophical belief in the importance of desires to consciousness. Basically, SkyNet won't happen because computers don't want anything, and we don't actually know how to program desires in.
[Edited at 2014-08-24 18:33 GMT]
I hope you are right...for all of our sakes.
(continues building EMP device)
To report site rules violations or get help, contact a site moderator:
You can also contact site staff by submitting a support request »
This discussion can also be accessed via the ProZ.com forum pages.