Well, its here. The Universal Translator

[Bisz]

http://www.macrumors.com/2014/05/27/microsoft-demos-skype-translation-tool/

This is exciting news.

They are only starting with a few languages and its through Skype, but I see a not-too-distant future where you have an app on your phone that you can use to communicate with people speaking a different language in real-time. Initially I see it as: you say something in your language, it gets translated and played through the phone's speakers, the person you are talking to replies in their language and that get translated back to yours. With a little it of specialized hardware I don't see why the sentence by sentence process can't be improved to be true real time, word by word.

 

[Metryq]

Another incremental step. The "science" of translating one human language to another was once thought a trivial task for computers, but it turned out more "art" was needed. Lots of text-to-text translators exist, which succeed to varying degrees.

With the addition of voice recognition to mobile devices, it is to be expected that voice-to-voice translators would come along. Of course, voice recognition apps are still far from perfect even when working within one language.

One of the wilder incremental steps, at least to me, is "augmented reality" translators like Word Lens. Word Lens uses the same kind of OCR (optical character recognition) found in text scanners to read signs. The app then substitutes translated text in the mobile device's screen. That is, one can hold up a smartphone to a sign in a foreign language and get a translation on the smartphone screen, as though one were looking through a window at the native text—and the translated text looks like the original sign!

Predictably, Word Lens found its way onto Google Glass. Some day in the future, one may find oneself housebound if one's visor breaks down. No more GPS signs in virtual space, no translations, no way to interface with computers in stores and elsewhere...

 

[DarthTom]

Another incremental step. The "science" of translating one human language to another was once thought a trivial task for computers, but it turned out more "art" was needed. Lots of text-to-text translators exist, which succeed to varying degrees.

A great example of that is the below. I saw the article about Kate's butt picture being published in a German tabloid newspaper today. Below is the headline.

I don't speak German but the Bing translator turns it into the below. For anyone that speaks German is that an accurate translation?

The other problem of course is there is a lot of nuance in speech. Take the word, 'cool,' in English. considerable context is needed sometimes to understand what the speaker is saying.

German Headline:

Danke für dieses
arschgeile Wochenende!

into

Thank you for this
ass horny weekend!



Bild - German newspaper

 

[Bisz]

The other problem of course is there is a lot of nuance in speech. Take the word, 'cool,' in English. considerable context is needed sometimes to understand what the speaker is saying.http://www.bild.de/unterhaltung/leu...dashian-schoene-kehrseiten-36136770.bild.html

The article I posted says that Skype is looking at tapping into it's users conversations to help it understand natural language and get more accurate. Basically "crowd sourcing" language.

 

[DarthTom]

The article I posted says that Skype is looking at tapping into it's users conversations to help it understand natural language and get more accurate. Basically "crowd sourcing" language.

At least for me, this type of device would be most useful on say my first visit to say, Beijing. Everyone I'd interact with would fail to understand my "context," when speaking in vernacular.

Back to my example, I point to the water fountain and say, "this is really cool." Does the unit understand that I'm talking about the the water temperature v. the appearance of the fountain and how does it translate that into Chinese so both parties are communicating?

Don't get me wrong - this would be an amazing piece of software for international business communication, travel, and generally making new friends who don't speak like languages.

But as they say, " I'm from Missouri, you got to show me," before I'll believe it works as seamlessly as they claim it will.

They should name it, 'Babel,' in difference to the biblical story where god allegedly punishes mankind for daring to build a tower to the heavens and gives everyone different languages. In short - suck it god.

 

[FPAlpha]

Another incremental step. The "science" of translating one human language to another was once thought a trivial task for computers, but it turned out more "art" was needed. Lots of text-to-text translators exist, which succeed to varying degrees.

A great example of that is the below. I saw the article about Kate's butt picture being published in a German tabloid newspaper today. Below is the headline.

I don't speak German but the Bing translator turns it into the below. For anyone that speaks German is that an accurate translation?

The other problem of course is there is a lot of nuance in speech. Take the word, 'cool,' in English. considerable context is needed sometimes to understand what the speaker is saying.

German Headline:

Danke für dieses
arschgeile Wochenende!

into

Thank you for this
ass horny weekend!



Bild - German newspaper

Yeah well.. it's a very literal translation

And this is the problem with online translations so far.. these programs do not understand syntax, grammar and the different meanings of words.

The german word "geil" (a colloquialism) means cool, awesome etc but it also means horny as in "He/she's horny and wants to have sex"

Programs do not take this into account and go for the most direct route without regard to context. In that regard i always know when i get emails from my overseas customers who are bad english speakers/writers and who constantly use Google translate.

Often enough i had to confirm multiple times by email to get a picture of what they want because the garbled mess that Google sometimes produces is wholy inadequate for business communication where real money is at stake when two people misunderstand each other due to the language barrier.

When programs reach the state where they can accurately distinguish between cool/cold and cool/awesome then we're on the right track.

How that would work in real time i don't know.. many languages have totally different grammar and sentence structure so how would a program know what you want to say so it will be able to translate without delay? It would have to be clairvoyant.

A Star Trek style universal translator will probably never work like it did in the show but i'd settle for a small delay until i finish the sentence for an accurate translation. it would not be fluid like a normal conversation but it should do the trick.

 

[Metryq]

i'd settle for a small delay until i finish the sentence for an accurate translation.

There would have to be a delay, as syntax is not the same in all languages. "Backwards talks Yoda."

 

[Candlelight]

Will it be able to understand Australian though?

 

[Metryq]

Will it be able to understand Australian though?

Oh, come on now! You're just asking for the impossible! I worked for an Australian boss. After some months, the employees figured we had finally gotten the hang of his accent...

...then we found a tape made many months back and discovered that it was the boss's accent that had been adapting to his US surroundings.

 

[Enterprise1701]

I don't think perfect universal translation will never be possible, much less in real-time.
Consider for example how many languages aren't subject-verb-object word order like English. Klingon, for example: https://en.wikipedia.org/wiki/Klingon_language#Grammar

Or if in the same conversation you need to refer to "LOL" the textspeak phrase and "L.O.L." the acronym for some company. And the person you're speaking to doesn't use the English/Latin alphabet.

Or if two languages don't have the same number of terms or number of usages for certain things. For example both "He is a soldier." and "It is a soldier." in English translate to "Es un soldado." in Spanish. Suppose that there are multiple subjects in the conversation, making "he" or "it" relevant. Or what if you're translating to a language that doesn't have grammatical gender? And what if you try to account for formal/familiar etiquette which differs between languages?

 

[QuarkforNagus]

But you need to know the other language in order for this translator to work.

In ST, the translators seemed to work after hearing (as best a machine can be said to hear) only a few sentences of some completely novel language.

 

[DarthTom]

I don't think perfect universal translation will never be possible, much less in real-time.
Consider for example how many languages aren't subject-verb-object word order like English. Klingon, for example: https://en.wikipedia.org/wiki/Klingon_language#Grammar

I chose the Bing translator because it's a Microsoft product and presumably they'd use the same translation algorithms that Bing uses to do the audio language conversions.

But as the German speaking poster up thread pointed out in my example even Bing translating a newspaper headline wasn't entirely accurate.

 

[publiusr]

All together now, say this quickly:

Recognize speech
Wreck a nice beach

from Beyond 2000

 

[Captrek]

They should name it, 'Babel,' in difference to the biblical story where god allegedly punishes mankind for daring to build a tower to the heavens and gives everyone different languages. In short - suck it god.

Alta Vista used to have a translation service called Babel Fish. It was then acquired by Yahoo!, who shut it down in 2012, replacing it with Bing Translate. I think using the Babel name for the Bing product would be rather confusing.

I chose the Bing translator because it's a Microsoft product and presumably they'd use the same translation algorithms that Bing uses to do the audio language conversions.

Your presumption is incorrect, according to the technology demo shared on Trek Today. The demonstrator points out that you might expect it to be a daisy-chaining of speech-to-text, translation, and text-to-speech services, but it doesn't work that way. He doesn't go into a lot of detail about how it does work, stating only that it's a neural net that does all of it at once, and whose inner workings are somewhat mysterious even to its own engineers.